You might want to mention that most of the NVidia GeForce cards are poor for double precision tasks, especially the Fermi cards.

May I suggest you add the GeForce GTX 465, GTX 550 Ti, GTX 560, GTX 560 Ti, GTX 570 and GTX 590, and if you are going to include the GeForce GT 430, you might as well include the GT 545 versions. Not sure about the GT 530 though, is it worth including?

While you are at it, and due to their 1/2 FP32 performance you might want to add the the extremely expensive,
Tesla's:
Tesla C2050 (515 gigaflops of double precision performance) 3GB
Tesla C2070 (Workstation) same as the C2050 but 6GB
These are similar to the GTX470 cards; 448:56:40, but without as much disabled, with more GDDR5 and with ECC - which when on further reduces performance.

and the even more expensive,
DataCentre Tesla's:
Tesla M2090 (665 DP Gigaflops)
Tesla M2070, Tesla M2050 and S2050 (515 DP GFlops)

Oh, and the,
Quadro Desktop's:
Quadro 6000, 5000, 4000 should all work.

Relative SP to DP performance's:
Tesla series: DP performance is 1/2 GF100 SP performance,
Quadro 4000-6000: DP performance is 1/2 SP
GTX 5xx: DP is 1/8 SP
Quadro 600-2000: DP is 1/12 SP

Then there are all those OEM and Notebook GPU's ;)

I see that HD4730 it's not listed here, nor is listet anywhere as non DP capable.

So my question is: Does it support DP or not? Did anyone tested it?

And yes, I googled it few times and did not found anything.

Thank you.

Update: onik1318 said that it does support it. You should add it to the list!

You might want to mention that most of the NVidia GeForce cards are poor for double precision tasks, especially the Fermi cards.

Tesla's:
Tesla C2050 (515 gigaflops of double precision performance) 3GB

I'd rewrite that to say the top 20 companies that need heavy double precision use and don't depend on function libraries written elsewhere. Nvidia still has the best selection of function libraries written elsewhere, especially for CUDA.

No wonder that the top 20 computers all use AMD/ATI GPUs.

You might want to mention that most of the NVidia GeForce cards are poor for double precision tasks, especially the Fermi cards.

Yes, he should have just said....

You might want to mention that most of the NVidia cards are poor for double precision tasks.
____________

"You might want to mention that most of the NVidia GeForce cards are poor for double precision tasks, especially the Fermi cards".

I dribbled on and lost the ball...

As suggested, Except the Fermi cards would have been closer to the mark, if you consider the GeForce cards not to be full fat Fermi's, and "most of the NVidia GeForce cards are poor for double precision tasks" might have sufficed - there is probably not many Tesla's here and those that forked out for one would probably have read the manual.

Anyway, Matt has added some compatible cards to the main list.

Cheers guys,

i have been getting messages saying my GPU lacks the needed compute capability.
then the message says i need 1.3, required for double precision math.
what would you suggest i have installed to so i dont get this message, without costing too much. thanks
____________

NVIDIA:
- Requires Compute Capability 1.3 and Above.
What is meant by 1.3 and above?
I have NVIDIA
geforce 220 installed.
thanks.
____________

Your card is Compute Capability 1.2. The messages you are getting are saying you must have Compute Capability 1.3 inclusive or above to run the application. That will - as you have already guessed - mean a new card in order to run that application.

Compute Capability
The term indicates at a technical level what the card is capable of doing, it applies to NVidia Cards only, AMD cards have different terminology. The details of that do not matter to us mere mortels, but it is important to wrap the head around what card is capable of what. There is a good Wikipedia Summary on it - "read over" technical details, and absorb the overview bits in particular the "Supported GPUs" table (scroll down the wikipedia article two pages and you'll see it.

http://en.wikipedia.org/wiki/CUDA#Version_features_and_specifications

Any card in that table with Compute Capability 1.2 inclusive or below will not work for the application your card is trying to run. In the table the cards of 1.3 and above will run it. As always the table is "dynamic" and a couple of cards may be missing from the list - but it will be fine for what you need.

Scroll further down and you see the technical differences - worth doing even if you are non-tech as it will give an overall feel for it, and show that its more than "tech speak", there are solid everyday differences.

http://developer.nvidia.com/cuda-gpus - is a straightfoward Card - Compute Capability list on the NVidia Site

New Card Selection
Thats a doozy :) It depends on a lot - what you are going to do with the card, what your overall priorities are(eg, BOINC, Gaming, other non-BOINC Tasks), finance and current overall PC hardware spec. There unfortunately is no short cut to a simple statement "go buy XXX card", at least not if you are to avoid wasting money.

Part of that discussion is also whether you buy NVidia, or AMD. In a crude sense, outside BOINC for most people, there is not really much difference to everyday useage as the use of various NVidia or AMD specific languages doesnt come into play (eg for gaming) - easy to create such a list just for the sake of it, but in the broad scheme of things forget it, its just a question of the card power/cost ratio. Inside BOINC its a whole new ball game, and there are Big differences AMD Versus NVidia Versus Project Application, could write a book on it - some do rofl :)

Suggested Way Ahead
Go to the Number Crunching Forum, start a new thread "Suggestions for a New Graphics Card". In first post list what you want to use the card for, both inside and outside BOINC, the spec of the PC you want to run it on, and the budget level you are prepared to pay for it. That will keep detailed discussion off this thread.

This will probably seem long winded, but with card selection long term, unless you focus on this and get suggestions and reasons, you *will* end up with the wrong card and waste money long term.

Regards
Zy

Compute Capable 1.2 cards have no double precision capabilities. These stripped down 40nm cards use to be good entry level cards elsewhere, being efficient for single precision jobs, but lacking DP they were never any use here; hence they are not on Matt's list of compatible cards.

... and the AMD/ATI HD cards that don't do DP (well enough):

6850
6790
6670
6570
6450

5770
5750
5670
5570
5470

4770
4670
4650
4550
4350

... and the AMD/ATI HD cards that don't do DP (well enough)

wrong ... the 4770 can do DP...

Good you spotted that, though Matt had included the 4770 on his list of good GPUs.
I just listed the non DP cards from the AMD pdf link which suggested the 4770 could not do DP.

Good you spotted that, though Matt had included the 4770 on his list of good GPUs.
I just listed the non DP cards from the AMD pdf link which suggested the 4770 could not do DP.

The HD4770 has a RV740 GPU this is an RV770 with 8 Shader-Clusters and in 40 nm Technologie (RV770 uses 55 nm).
So it does DP as a lot of particpations demonstrate every day ;-)

Thanks heaps, i will look into it amd get a decent card that will do the job with room to spare.
much appriciated.

The machine I'm typing this message on is a stock Dell Dimension E520, with Pent D 2.8 processor, 5Gb RAM, and Win 7 Ultimate 64bit. It has an HD4770 card in it running Milkyway at 94% and 73C, and CPU's running at 100%. The whole shebang draws 220W from the wall socket via my powermeter, which is well withing the 305W stock PSU.

I'm getting virtually the same RAC from it as my 4850 cards with less heat and less power needed. In terms of bang for buck I can't beat a 4770 for a lower end DP card.
____________
Don't drink water, that's the stuff that rusts pipes

Hey! Wanted to mention AGP bus cards. I am currently running an ATi HD3850 on an AGP bus with the most recent CCC, drivers, and AMD/ATi AGP bus patches.

Though this card does not * officially* support Stream or Open CL on an AGP form factor (the 3850 PCIe does), it seems to perform well.

I am currently running MW@Home on my AGP rig (2.8Ghz Pentium Dual core, Radeon HD3850 AGP) and am currently running out an average credit of ~10,000 with a turnaround time of 0.09days. This is using BOINC. I am unsure if its using ATi Stream or Open CL.

Joined MW@H 4 days ago, and have already racked up 123,000 credits on my AGP computer.

Long and short of it:

Please add the HD38XX AGP series as a supported GPU.

Also, if you have an older AGP number cruncher, and want to add some punch, now you know what to pick up lol. :)


http://milkyway.cs.rpi.edu/milkyway/show_host_detail.php?hostid=298446

It would appear that you are using your three i7 computers' CPU's as well as your 3850.
While the 3850 is doing about twice as much work as one i7 (all 8 threads), together the three i7's could do about 1.5 times (ball park) as much as your 3850:

63434273 43112228 2 Jul 2011 | 17:25:34 UTC 2 Jul 2011 | 19:37:23 UTC Completed, validation inconclusive 857.24 32.78 pending MilkyWay@Home v0.82 (ati14)
63404084 43091741 2 Jul 2011 | 16:20:48 UTC 2 Jul 2011 | 18:37:33 UTC Completed, validation inconclusive 722.74 41.12 pending MilkyWay@Home v0.82 (ati14)
63377855 43074532 2 Jul 2011 | 15:26:52 UTC 14 Jul 2011 | 15:26:52 UTC In progress --- --- --- MilkyWay@Home v0.88
63317894 43034981 2 Jul 2011 | 13:27:38 UTC 2 Jul 2011 | 15:40:05 UTC Completed, validation inconclusive 427.71 20.36 pending MilkyWay@Home v0.82 (ati14)
63268920 43002552 2 Jul 2011 | 12:05:42 UTC 14 Jul 2011 | 12:05:42 UTC In progress --- --- --- MilkyWay@Home N-Body Simulation v0.66 (mt)
63098450 42875711 2 Jul 2011 | 5:41:11 UTC 14 Jul 2011 | 5:41:11 UTC In progress --- --- --- MilkyWay@Home v0.88
63090193 42885939 2 Jul 2011 | 5:40:04 UTC 14 Jul 2011 | 5:40:04 UTC In progress --- --- --- MilkyWay@Home N-Body Simulation v0.66 (mt)
63083446 42881359 2 Jul 2011 | 5:33:56 UTC 2 Jul 2011 | 19:13:30 UTC Completed, validation inconclusive 5,919.47 10,948.85 pending MilkyWay@Home N-Body Simulation v0.66 (mt)

My HD 5850 run times vary from 212sec down to 91sec, making it about 5 times as fast as an HD 3850 or roughly 20 times as fast as an i7.
The Enthusiast (Cayman) cards are about 2 times as fast as an HD 5850 - Just a bit of perspective.

PS. If I remember correctly the 3870 should also work.

Well, 3 computers' CPU's (no idea where I got 3 i7's from).
Furl, while I'm glad to see your 3850 works, why are you using your HD3850 but not really using your HD 4700/4800?

By the way, I would suggest you use Boinc 6.10.33 rather than 6.10.26.

@Furl Hawkens
and as I read you are a very new cruncher: There is an option in the
Assistant/Settings Menue:

"Use GPU during computer is used" This makes sure if you are doing some other work on the PC (Mail, Office, WEB-Browsing, etc) the GPU chruches.

Wenn you need the GPU for gaming or video etc., you can pause the GPU working for BOINC in the Quick- Taskbar.

This makes sure if you are doing some other work on the PC (Mail, Office, WEB-Browsing, etc) the GPU chruches.

This makes sure if you are doing some other work on the PC (Mail, Office, WEB-Browsing, etc) the GPU chruches.

And you need to be aware that having MW crunch using the GPU all the time can make using IE8, MS/Word, Excel, or even scrolling through a long list of files in a folder a painfully slow experience. With an HD5870 crunching, I'm frequently reaching for the Snooze GPU option just to do basic work.

This makes sure if you are doing some other work on the PC (Mail, Office, WEB-Browsing, etc) the GPU chruches.

And you need to be aware that having MW crunch using the GPU all the time can make using IE8, MS/Word, Excel, or even scrolling through a long list of files in a folder a painfully slow experience. With an HD5870 crunching, I'm frequently reaching for the Snooze GPU option just to do basic work.

In case you're not aware of it yet, few if any of the Nvidia GT 3xx series boards are likely to have double precision, since most of them were essentially older designs with new part numbers.

I'm not seeing that on my machines (3850/4870/4870/5970).

The HD5870 is running under XP SP3.

Thanks for the info on running the GPU always, though I had already set it. I have been crunching with Boinc for a few years now, just new to MW@H lol.


As for why I'm not using my 4890, only the 3850, its cuz the 4890 overheats. It always has unfortunately,

Its an XFX Radeon HD4890 XXX edition, with a "lifetime warranty". When running stress tests, or distributed computing projects (ie anything that actually pushes it) it will completely overheat (ie pass 100c and lock the comp). It does ok for games, but still averages in the 95c mark when playing things like crysis and GTA-IV.


When I contacted XFX they actually told me that the card was never designed to be pushed the way that furrmark or DC projects do. They said that unless it started to fail from overheating with normal PC games that they refused to warranty it out. They straight out told me not to run the programs that are overheating it lmao...

This was when the card was brand new, still has the issues, still holding me back, and they still refuse to warranty it...

All this was through email, so I still have it in writing, if anyone here wants to see.

My advice, avoid XFX.

As for why I'm not using my 4890, only the 3850, its cuz the 4890 overheats. It always has unfortunately,

As for why I'm not using my 4890, only the 3850, its cuz the 4890 overheats. It always has unfortunately,

MW@H is very low dependant on memory-clock; that means, you can 'underclock' your card without significant loss of credit, but remarkable loss of heat. I use 'Afterburner' for that.

There were several discussions in the past. Many crunchers tested that, it works.

Is that model designed to have a fan that pushes air past the GPU, then out of the case, instead of circulating it within the case? When I switched from a model circulating the hot air within the case to a model using about the same power but blowing the hot air out of the case, the GPU temperature dropped about 30 degrees C.

I think we are spaming again this thread.


I invite you to go on here here

I have a question for the GPU nVidia crunchers,

I have a pc with a 425W power supply so only a GeForce GTX 550Ti
will work. According to this threat it will run MW.
However when I go to the nVidia site, there is no information available about this card. What does that mean?
And which brand do you suggest?

Thanks.
____________
Greetings from,
TJ

Hi TJ I have a similar problem. I see that ATI 4870 is on the list but it is aborting 100% of the WU and then it says computation error.

I run MW@home stock app 0.82 and I have the latest drivers....
So I am at a loss.

I have 100% failure rate for last 3 days and I dont know why.

Martin

It abort it then it keeps crunching until it reaches 100% then it say computaion errors?

Any help anyone?

ATI 4870 driver version 8.861.0.0
Boinc 6.12.33

Martin

Martin, I have a HD 4770 and had similar problems when I updated the ATI drivers. I kept going back to 10.12 which worked consistently. I finally got 11.6 to work by first completely uninstalling the old drivers and rebooting before the new install.

This may or may not work for you, but it's worth a try if you didn't do this already.

I have a question for the GPU nVidia crunchers,

I have a pc with a 425W power supply so only a GeForce GTX 550Ti
will work. According to this threat it will run MW.
However when I go to the nVidia site, there is no information available about this card. What does that mean?
And which brand do you suggest?

Thanks.

It abort it then it keeps crunching until it reaches 100% then it say computaion errors?

Any help anyone?

ATI 4870 driver version 8.861.0.0
Boinc 6.12.33

Martin

http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=2468&nowrap=true#49994

I have a question for the GPU nVidia crunchers,

I have a pc with a 425W power supply so only a GeForce GTX 550Ti
will work. According to this threat it will run MW.
However when I go to the nVidia site, there is no information available about this card. What does that mean?
And which brand do you suggest?

Thanks.

When I went to this Nvidia site, I had no trouble finding a link to suitable information:

http://www.nvidia.com/object/geforce_family.html

Strange - for me, the results were somewhat slow to appear, but no problems otherwise.

If you'd like to reach that web page more directly:

http://www.nvidia.com/object/product-geforce-gtx-560ti-us.html
http://www.geforce.com/#/Hardware/GPUs/geforce-gtx-550ti/overview

If you'd like to reach that web page more directly:

http://www.nvidia.com/object/product-geforce-gtx-560ti-us.html
http://www.geforce.com/#/Hardware/GPUs/geforce-gtx-550ti/overview

Under IE, I get the same results you do for perhaps 20 seconds. Then, it finally comes up. Depending on what it's doing during those 20 seconds, it might happen slower on slower connections.

The links are then clickable, but the one for Specifications gives much of the text fairly dark gray on a black background.

One of the items visible during those 20 seconds is a rotating arrow, like those before some videos.

Guru3d did a good review of the 550ti, summary is at (worth reading the full review):

http://www.guru3d.com/article/palit-geforce-gtx-550-ti-sonic-review/20

They were not overly keen on it, but the version tested was not a GForce version, it was a Palit factory o/c. There is another review at Guru3d, worth looking up in their hardware reviews.

They seem to say it was fine, didnt complain as such, but implicitly recommended the 560 if the budget stretches.

Worth cycling through a few of the card reviews there, they always do a power section to them with a recommended power supply. Just be aware their test rigs are such that they draw unusual power levels - which they do point out - and crudely, a PSU around 50w less than the formal recommend would still do the job. Its all well laid out in the PSU section(s)

Regards
Zy

Every MW WU I have crunched lately is in error or invalid. I have no idea why. My several PCs use GTX275 and ATI 4850, 4870. Formerly, before something entirely unknown occurred, I was happily processing for MW. Sure would like to again. Here's an example of a CPU WU (this PC has GTX275 which also is erring):
http://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=48449026

And here, on a different PC, an ATI 4870 WU:
http://milkyway.cs.rpi.edu/milkyway/result.php?resultid=66604679

Can someone steer me in the right direction?

BTW: I am happily crunching for several other projects.

Thanks Art
____________

You need to update your mw apps.

You need to update your mw apps.

If you're still using that method, and not certain methods of setting parameters, the apps should automatically be updated to whatever ther workunits were designed to use.

If you're still using that method, and not certain methods of setting parameters, the apps should automatically be updated to whatever ther workunits were designed to use.

Using an appinfo.xml file is one of the methods of setting parameters that stops automatic downloads of new app versions, so expect to have to do it manually for as long as you use that file.

I haven't found an instruction file either, but here's a thread relevant to how to do it for another project:

The optimized apps are the same as the stock apps, only with the addition of an app_info (and an description?).
As long as you don't have any problems with the stock app, you be ok.
If you get the max elapsed time exceeded error, try with the app_info.xml.

OK Thanks to all. Will just observe for now, see if I get completed and credited WU. I have a Nvidia and a ATI PC working solely on MW now, should have results soon.

I use KWSN apps on SETI but had not known optimized apps existed at MW...thanks for the help.

Regards
Art
____________

If you get the max elapsed time exceeded error, try with the app_info.xml.

You can add an AMD HD4970 to the list.

REgards

If you get the max elapsed time exceeded error, try with the app_info.xml.

..And that's exactly the result. All WU still error out.
http://milkyway.cs.rpi.edu/milkyway/result.php?resultid=71860152

By "..try with the app_info.xml" do you mean from this site?
http://www.mpits.net/opt_mw.php

Edit: I installed the optimized app and appinfo.xml from http://www.mpits.net/opt_mw.php, the SSE4.1 version which CPUZ says my Q9650 will do. The GPU is GTX275, driver 275.33. MW downloaded only separation apps. Every one completed in one second with this:
...

Does anyone know if the FX3800 is compatible, I dont see it on the list, but wanted to see before I spent time trying to congfigure it to work.

The FX 3800 is an Nvidia Quadro design.

http://www.funtech.com/NVIDIA-Quadro-FX-3800

Its spec sheet doesn't mention a Fermi GPU chip.

Unclear if a Windows 7 version of a driver for it is available.

To me, it looks like an older design unlikely to be suitable for Milkyway@Home, but you should wait for an opinion from someone more familiar with it before you give up entirely.

Nvidia's spec sheet for the FX 3800:
http://www.nvidia.com/object/product_quadro_fx_3800_us.html
Mentions that it was designed for professionals (and therefore probably originally rather expensive)

A list of legacy products that mentions it:
http://www.nvidia.com/page/quadrofx_family.html

Seems that a Windows 7 driver IS available for it:
http://www.nvidia.com/object/io_1241025430517.html

Some comparison info:
http://www.nvidia.com/content/PDF/product-comparison/product-comparison-master-revised.pdf

Still no information about whether it has the main feature required for Milkyway@Home - hardware double precision support.

When you start BOINC, do the first few lines of the Messages tab of the Advanced view mention finding its GPU and what computing capability it has?

For example, here are the first few lines of that tab for my desktop:

7/16/2011 7:40:56 PM Starting BOINC client version 6.10.58 for windows_x86_64
7/16/2011 7:40:56 PM log flags: file_xfer, sched_ops, task
7/16/2011 7:40:56 PM Libraries: libcurl/7.19.7 OpenSSL/0.9.8l zlib/1.2.3
7/16/2011 7:40:56 PM Data directory: C:\ProgramData\BOINC
7/16/2011 7:40:56 PM Running under account Bobby
7/16/2011 7:40:57 PM Processor: 4 GenuineIntel Intel(R) Core(TM)2 Quad CPU Q9650 @ 3.00GHz [Family 6 Model 23 Stepping 10]
7/16/2011 7:40:57 PM Processor: 6.00 MB cache
7/16/2011 7:40:57 PM Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 cx16 sse4_1 syscall nx lm vmx smx tm2 pbe
7/16/2011 7:40:57 PM OS: Microsoft Windows Vista: Home Premium x64 Edition, Service Pack 2, (06.00.6002.00)
7/16/2011 7:40:57 PM Memory: 8.00 GB physical, 15.74 GB virtual
7/16/2011 7:40:57 PM Disk: 919.67 GB total, 540.84 GB free
7/16/2011 7:40:57 PM Local time is UTC -5 hours
7/16/2011 7:40:57 PM NVIDIA GPU 0: GeForce GTS 450 (driver version 27533, CUDA version 4000, compute capability 2.1, 962MB, 476 GFLOPS peak)

The last line of this group is the one to look for.

If you get the max elapsed time exceeded error, try with the app_info.xml.

..And that's exactly the result. All WU still error out.
http://milkyway.cs.rpi.edu/milkyway/result.php?resultid=71860152

By "..try with the app_info.xml" do you mean from this site?
http://www.mpits.net/opt_mw.php

Edit: I installed the optimized app and appinfo.xml from http://www.mpits.net/opt_mw.php, the SSE4.1 version which CPUZ says my Q9650 will do. The GPU is GTX275, driver 275.33. MW downloaded only separation apps. Every one completed in one second with this:
...

Sorry, those apps are for CPU, not for GPU!

Snip

Win XP might need a little treaking of the app_info to keep the system responsible. But let's get your GPUs working first.

So thanks for the help on the FX3800, below is what I am seeing from the start of the event log:


7/19/2011 2:54:21 PM | | Starting BOINC client version 6.12.33 for windows_x86_64
7/19/2011 2:54:21 PM | | log flags: file_xfer, sched_ops, task
7/19/2011 2:54:21 PM | | Libraries: libcurl/7.19.7 OpenSSL/0.9.8l zlib/1.2.5
7/19/2011 2:54:21 PM | | Data directory: C:\ProgramData\BOINC
7/19/2011 2:54:21 PM | | Running under account kroseler
7/19/2011 2:54:21 PM | | Processor: 8 GenuineIntel Intel(R) Xeon(R) CPU W3550 @ 3.07GHz [Family 6 Model 26 Stepping 5]
7/19/2011 2:54:21 PM | | Processor: 256.00 KB cache
7/19/2011 2:54:21 PM | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 cx16 sse4_1 sse4_2 syscall nx lm vmx tm2 dca popcnt pbe
7/19/2011 2:54:21 PM | | OS: Microsoft Windows 7: Enterprise x64 Edition, (06.01.7600.00)
7/19/2011 2:54:21 PM | | Memory: 15.99 GB physical, 31.98 GB virtual
7/19/2011 2:54:21 PM | | Disk: 232.88 GB total, 197.27 GB free
7/19/2011 2:54:21 PM | | Local time is UTC -5 hours
7/19/2011 2:54:21 PM | | No usable GPUs found
7/19/2011 2:54:21 PM | Milkyway@home | Application uses missing NVIDIA GPU


I am running the nvidia driver rev of 8.17.12.7081. Not sure if that helps.

Has anyone had a chance to see if the new AMD Llano APUs have DP Graphics and any experience with getting MW@H to run on them? I know AMD has put a lot into marketing these chips as having integrated direct computer capabilities, but it wont help us here if the double-precision ability is not there, otherwise it looks like a very power-efficient alternative.

So thanks for the help on the FX3800, below is what I am seeing from the start of the event log:


7/19/2011 2:54:21 PM | | Starting BOINC client version 6.12.33 for windows_x86_64
7/19/2011 2:54:21 PM | | log flags: file_xfer, sched_ops, task
7/19/2011 2:54:21 PM | | Libraries: libcurl/7.19.7 OpenSSL/0.9.8l zlib/1.2.5
7/19/2011 2:54:21 PM | | Data directory: C:\ProgramData\BOINC
7/19/2011 2:54:21 PM | | Running under account kroseler
7/19/2011 2:54:21 PM | | Processor: 8 GenuineIntel Intel(R) Xeon(R) CPU W3550 @ 3.07GHz [Family 6 Model 26 Stepping 5]
7/19/2011 2:54:21 PM | | Processor: 256.00 KB cache
7/19/2011 2:54:21 PM | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 cx16 sse4_1 sse4_2 syscall nx lm vmx tm2 dca popcnt pbe
7/19/2011 2:54:21 PM | | OS: Microsoft Windows 7: Enterprise x64 Edition, (06.01.7600.00)
7/19/2011 2:54:21 PM | | Memory: 15.99 GB physical, 31.98 GB virtual
7/19/2011 2:54:21 PM | | Disk: 232.88 GB total, 197.27 GB free
7/19/2011 2:54:21 PM | | Local time is UTC -5 hours
7/19/2011 2:54:21 PM | | No usable GPUs found
7/19/2011 2:54:21 PM | Milkyway@home | Application uses missing NVIDIA GPU


I am running the nvidia driver rev of 8.17.12.7081. Not sure if that helps.

Has anyone had a chance to see if the new AMD Llano APUs have DP Graphics and any experience with getting MW@H to run on them? I know AMD has put a lot into marketing these chips as having integrated direct computer capabilities, but it wont help us here if the double-precision ability is not there, otherwise it looks like a very power-efficient alternative.

So thanks for the help on the FX3800, below is what I am seeing from the start of the event log:
[snip]
7/19/2011 2:54:21 PM | | No usable GPUs found
7/19/2011 2:54:21 PM | Milkyway@home | Application uses missing NVIDIA GPU


I am running the nvidia driver rev of 8.17.12.7081. Not sure if that helps.

I found MW would not work with this filename in the xml (see above):

<name>mmilkyway_separation_0.82_windows_intelx86__ati14.exe</name>

***Note the name MMilkyway_separation...etc.

My XP 32 bit ATI 4870 PC would not process until I removed the duplicate M. Once that was done things seem to work.

Is that what Len LE/GE means by:

"Win XP might need a little treaking of the app_info to keep the system responsible. But let's get your GPUs working first."

And could this be the reason 32 bit ATI xml's need "a little tweaking"?? It was for me.

Removing the double M from the xml all is well. On my 4870 WU's are taking about 2.5 minutes.

Regards
Art

I've run MW in the past using an app_info.xml to process multiple WUs simultaneously with no problem. During the recent "break" I saw that the app numbers had changed so I modified the app_info.xml file accordingly (from .52 to .82, for example). Now, although I don't get any error messages, I also can't process multiple WUs simultaneously anymore. It assigns the whole GPU to the WU. All my GPUs are Nvidia.


I'm having this problem even with my new NVIDIA GTX 550TI card with 2 GB of video ram. For this card, I even went and installed the latest Nvidia drivers (280.13) but no joy. Ubuntu 11.04 is currently with something like 270.19.

Am I missing something or are the WUs so demanding for video ram on the GPUs that multiple simultaneous WUs are no longer possible?

Thanks for any help that you can provide.

Regards,
Steve

I've run MW in the past using an app_info.xml to process multiple WUs simultaneously with no problem. During the recent "break" I saw that the app numbers had changed so I modified the app_info.xml file accordingly (from .52 to .82, for example). Now, although I don't get any error messages, I also can't process multiple WUs simultaneously anymore. It assigns the whole GPU to the WU. All my GPUs are Nvidia.


I'm having this problem even with my new NVIDIA GTX 550TI card with 2 GB of video ram. For this card, I even went and installed the latest Nvidia drivers (280.13) but no joy. Ubuntu 11.04 is currently with something like 270.19.

Am I missing something or are the WUs so demanding for video ram on the GPUs that multiple simultaneous WUs are no longer possible?

Thanks for any help that you can provide.

Regards,
Steve

Thanks for the suggestion. However, it "sounds" like it's for a Windows app_info.xml. I'm on Ubuntu. :)

However, I will see if I can work it in.

Regards,
Steve

P.S. This is interesting. When I re-start Boinc as a service and look at the initial messages, I see a line where it sees the E@H app_info.xml but there is no mention of the one for MW@H. If it doesn't even see the app_info file that would explain why no multiple WU's but then why doesn't it see the file?

Remember, when BOINC is started as a service, it's generally unable to use any GPU at all, regardless of what GPU you have.

Oh, Boinc is seeing the GPUs no problem. The problem is that MW (but not E@H) either does not see the app_info.xml at start up or ignores it and only runs one WU/GPU instead of 2 as it has in the past with the entry set to .5.

Since this is affecting all of my machines and started after the downtime last week, I'm becoming convinced that the problem is not at my end.

Regards,
Steve

Can you show a listing of the files in your MW folder and also post the app_info.xml text.
____________

Hmm. Hey guys.
I'm just wondering about this.
What would be the most efficient way to crunch MW ?
I have GTX560Ti currently which is munching away GPUGrid.
My cpu is mainly running WCG but this month I'm helpin my team OCN boost it's points in malariacontrol.
I'm not interested in the points. Just about the what would be better..
I guess it's my 560Ti ?

Probably the wrong thread but figured pointless to create a new thread so I'll just post this here.

Any of the high-end AMD/ATI GPUs that can handle double precision are best. Scroll back to the first post in this thread to see most of them listed.

Any Fermi-based card, such as your 560Ti, or any GTX2xx series card from GTX260 up, come in second.

Using your CPU cores comes in a distant third.

Yeah well I know that already, but thanks for answering! :)
I may have shaped my question wrong.
What I was meaning is the efficiency, meaning how much power used to complete certain amount of project data in what time or something like that.

Thanks for the help.

Here is the information that you requested. Everything runs fine, EXCEPT that no matter what I do, the Fermi's only process one WU at a time each.

It's a problem on all 3 machines. They're all running under Ubuntu 11.04, AMD multi-cores with Fermi cards and at least 8GB of either DDDR2 or DDR3 ram.

Just as a test, I re-started Collatz and it's running 3WU's simultaneously per Fermi so I don't think it's HW or Boinc. E@H was also running fine.

I've also gone back and forth with file permissions with the app_info.xml and the executables but that didn't make any difference.

Here's the list of files.

/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/app_info.xml
/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/histogram
/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.62_x86_64-pc-linux-gnu__mt
/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/milkyway_separation_0.82_x86_64-pc-linux-gnu__cuda_opencl
/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/milkyway_separation_0.88_x86_64-pc-linux-gnu
/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/orphan_test_2model.lua
/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/p-17-3s-fix2.txt
/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/parameters-10-3s-fix2.txt
/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/parameters-10-3s-fix-new2.txt
/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/parameters-13.txt
/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/parameters-13-3s-fix-new2.txt
/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/parameters-82-2s.txt
/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/single_t82_0.txt
/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/single_t82_1.txt
/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/single_t82_2.txt
/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/single_t82_3.txt
/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/single_t82_4.txt
/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/stars-10-good.txt
/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/stars-13.txt
/var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway/stars-17.txt

and here's the app_info.xml.

<file_ref>
<file_name>
milkyway_nbody_0.62_x86_64-pc-linux-gnu__mt
</file_name>
<main_program/>
</file_ref>
</app_version>
</app_info>

<app>
<name>milkyway</name>
<user_friendly_name>Milkyway@home</user_friendly_name>
</app>

<file_info>
<name>milkyway_0.88_x86_64-pc-linux-gnu</name>
<executable/>
</file_info>

<app_version>
<app_name>milkyway</app_name>
<version_num>88</version_num>
<avg_ncpus>0.05</avg_ncpus>
<max_ncpus>0.05</max_ncpus>
<flops>1.0e11</flops>
<file_ref>
<file_name>milkyway_0.88_x86_64-pc-linux-gnu</file_name>
<main_program/>
</file_ref>
</app_version>

<app_info>
<app>
<name>milkyway</name>
<user_friendly_name>Milkyway@home</user_friendly_name>
</app>

<file_info>
<name>milkyway_0.88_x86_64-pc-linux-gnu__sse2</name>
<executable/>
</file_info>

<app_version>
<app_name>milkyway</app_name>
<version_num>88</version_num>
<plan_class>sse2</plan_class>
<avg_ncpus>0.05</avg_ncpus>
<max_ncpus>0.05</max_ncpus>
<flops>1.0e11</flops>

<file_ref>
<file_name>milkyway_0.88_x86_64-pc-linux-gnu__sse2</file_name>
<main_program/>
</file_ref>
</app_version>
<app_info>

Thanks again for the help.

Regards,
Steve

Yeah well I know that already, but thanks for answering! :)
I may have shaped my question wrong.
What I was meaning is the efficiency, meaning how much power used to complete certain amount of project data in what time or something like that.

I think I see the problem, the app_info is definitely out of order and is getting closed prematurely. It also does not list the correct app for CUDA processing.

Try pasting this one into the app_info and saving.

Thanks so much for the help.

Right off, this solved the problem of Boinc not seeing the app_info.xml.

At first, I was still having problems getting MW to run and more and more it just felt like some sort of file ownership/permissions problem even though I had been back and forth on this for days.

Finally, it occurred to me to see how Collatz was handling things since it ran from the very start with no fuss. Of course, when their app installs it also installs a properly configured app_info.xml file and spares the user all the needless troubleshooting grief.

I then changed all of the MW file ownerships/permissions to match those of Collatz, and bingo, everything works great.

So, thanks again for the help. It's great to have things not only working but working at extra efficiency with the GPUs.

Regards,
Steve

I just received a message regarding the GPU: Notice from server: An ATI GPU supporting double precision math is required.

If I don't upgrade hardware, is this message telling me that I cannot run any milky way tasks and I should delete the project until I do upgrade? Or does it just use the CPU? I have some milky way credits, but maybe tasks that completed some time ago did not have a Double Precision arithmetic requirement. Please clarify. Thanks much.

BTW - I recently rejoined after some time away, so I am not current on issues.

I just received a message regarding the GPU: Notice from server: An ATI GPU supporting double precision math is required.

If I don't upgrade hardware, is this message telling me that I cannot run any milky way tasks and I should delete the project until I do upgrade? Or does it just use the CPU? I have some milky way credits, but maybe tasks that completed some time ago did not have a Double Precision arithmetic requirement. Please clarify. Thanks much.

BTW - I recently rejoined after some time away, so I am not current on issues.

That's what I thought, so thanks for letting me know. Appreciate the response!

I just received a message regarding the GPU: Notice from server: An ATI GPU supporting double precision math is required.

If I don't upgrade hardware, is this message telling me that I cannot run any milky way tasks and I should delete the project until I do upgrade? Or does it just use the CPU? I have some milky way credits, but maybe tasks that completed some time ago did not have a Double Precision arithmetic requirement. Please clarify. Thanks much.

BTW - I recently rejoined after some time away, so I am not current on issues.

Milkyway has applications for the CPU, ATI GPU's and Nvidia GPU's. Only problem is all GPU's have to have Double Precision to be used here. Your 58x0 GPU does not meet that requirement. You can still use the CPU to process though.

I think you mean her HD 5x00 GPU. If she had a 58x0 GPU, she would be golden.

Has anyone had a chance to see if the new AMD Llano APUs have DP Graphics and any experience with getting MW@H to run on them? I know AMD has put a lot into marketing these chips as having integrated direct computer capabilities, but it wont help us here if the double-precision ability is not there, otherwise it looks like a very power-efficient alternative.

I recently built a system using the A8-3850 and it does not support DP on the GPU. Figured I would post my experiences in case anyone goes looking through these threads. The other projects that support AMD/ATI GPUs that do not require DP math will recognize and use the A8-3850's integrated GPU, however. The IGP only runs at 600MHz by default and, while there are BIOS out there (I'm looking at you, A75M-UD2H) that offer you an option to overclock the GPU only, it does not work. The only way to overclock the IGP currently is to raise the BCLK for the entire CPU. I was able to get the IGP from 600MHz to 876MHz with 146 x 6 but it came at significant cost of power and heat.

What completion times are you getting on other projects (Collatz, Moo!, Primegrid)(OCed and stock speeds)?

. The other projects that support AMD/ATI GPUs that do not require DP math will recognize and use the A8-3850's integrated GPU, however.

What completion times are you getting on other projects (Collatz, Moo!, Primegrid)(OCed and stock speeds)?

Based on the numbers above, the A8-3850's 6550D (600MHz core, 400 stream processors) is approximately 1/4 the speed of the HIS 6950 (840MHz core, 1408 stream processors) for number crunching purposes.

The IGP is 25 times faster than a single core on the A8-3850 for the same kind of Collatz and Primegrid tasks. The HIS 6950 is 101 times faster.

I have the HD 5570.

Based on the numbers above, the A8-3850's 6550D (600MHz core, 400 stream processors) is approximately 1/4 the speed of the HIS 6950 (840MHz core, 1408 stream processors) for number crunching purposes.

The IGP is 25 times faster than a single core on the A8-3850 for the same kind of Collatz and Primegrid tasks. The HIS 6950 is 101 times faster.

Thanks much for posting this, pretty interesting. It seems the A8 IGP is much faster than anything previously integrated with the CPU. Not a bad crunching option for corporate users and those not wanting to buy expensive GPUs. One last question, any idea of the power draw (total watts: GPU running WUs, not running WUs).

I just hooked an ASUS GeForce GT 440 in my server. So far, it works with MW, Einstein, Collatz, PrimeGrid, GPUGRID, DNETC and DistrRTgen

You can add the Nvidia Quadro 600 (do not confuse with 6000).

Tested and 100% working.

It's a little bit slow, but it's better than nothing :)

Why does that document, "AMD_APP_SDK_Getting_Started_Guide_v2.4" say that the ATI FirePro V5800 card supports double-precision if it doesn't?

seems that ther is a mistake:

wiki:
http://en.wikipedia.org/wiki/Comparison_of_AMD_graphics_processing_units#FirePro_3D_series
FirePro 3D V5800[96] April 26, 2010 Juniper XT(RV840)
But the V5900 has a Cayman proc. so this must have DP

I still have a 'JUNIPER' or HD5770, which is a Single Precision GPU.
Excellent and fast card, not to be used in Milkyway, but Collatz C. or SETI (Bêta),
etc.
A few days ago, I started gettin new work, which goes amazingly fast, every 1 to 2 minutes 2 WUs, are finished and UPloaded and another 2 are downloaded.
And maybe this Double Precision and High Load, 99% drives those 5870s
to its edge, the whole system, with an I7-2600 CPU (HT=ON), heats up the room with
550W/hour. (Even an AC is needed, 3 QUADs and 480; 470 and GTS250 GPUs).

But 'one has to have something to do' , besides work, living a life, etc., etc.
(Why am I telling this?). ;-)

And about using an app_info.xml file, can be usefull, if you want to run run 2
WUs on each card, but check if it works OK and give more efficiency by doing 2, in
less then double time, when doing 1!

<coproc>
<type>ATI</type>
<count>1</count>(0.5)(0.33)
</coproc>
<cmdline>1(2)(3) instances_per_device</cmdline>
[/color[

(I used to run 3 SETI MB WUs on my GTX480 and X9650CPU, no problems!)(Only heat)

Only, no typos in these app_info.xml file, all stated in this file, has to be
present or, it starts raining errors...........(I once saw about a thousand WUs dissapear
cause of such an error, tried to run 2 Einstein WU for GPU per GPU.
One typo...........error...........

Now only 1 per GPU, sometimes, I do turn GPU; Voltage and frequency down, Memory only lower frequency, or an auxillary or regular "office" Fan, wirh 1 side of the case is open.
____________

[color=red]Knight Who says Ni

Why does that document, "AMD_APP_SDK_Getting_Started_Guide_v2.4" say that the ATI FirePro V5800 card supports double-precision if it doesn't?

Thanks FruehwF for clarifying that.

I had no problem using my Q6600 cpu an ATI HD4380 here with Win7.
Now that I am using Ubuntu 11.10 the GPU does nothing.
Is there something I need to do in addition to running bonic and adding Milkyway as a project?

(I hope this is posted to the correct place.)

Thanks

I doubt if standard Ubuntu installations include the driver that BOINC needs to make that card usable, so check the BOINC log file to see if it even recognized the presence of that card. Scroll to about a dozen lines from the top and see if a line near the CPU description offers a a GPU description as well.

Mention what driver version number is shown, if any; some versions can cause problems.

Also check if your preferences for BOINC tell it that it's allowed to use ATI GPUs - if not, it will ignore the card.

The Einstein@Home project is also astronomy-related and does not require the hardware double precision support that Milkyway@Home requires, so you might want to try running a few GPU workunits for them (to avoid confusion, do this with all types of their CPU workunits disabled).

http://einstein.phys.uwm.edu/

If none of those help, you'll have to wait for someone more familiar with Ubuntu to offer more help.

As far as my knowledge Einstein@Home does not support ATI cards

I just checked, and you're right; they don't.

PrimeGrid does, however, for one of their sieve applications, so they should be adequate for a test of whether the board will even work under BOINC.

http://www.primegrid.com/

Thanks for the help. You are right, Ubuntu doesn't even see my ATI card.
I'll do some more research on that.

Primegrids PPSsieve on ATI/AMD-GPUs uses OpenCL. You need to install either the APP-driver or the Stream-SDK.
OpenCL for the entire HD4000 series is marked as Beta and be aware, OpenCL is like Java. Runs on many platforms but fast is different...
____________

As rroonnaalldd says - Milkyway is best with ATI/AMD GPUs, PrimeGrid (the PPSsieving) is best with nVidia GeForce GPUs.

The latter can crunch using PCI (not express) 9500GTs OK, even if these are a bit slow now. These used to give an individual GPU RAC of 30,000, but this will now top out at 20,000. But the normal PCI expansion slots (4) should give an output just under 100K with a P3 rig.
____________
Go away, I was asleep

Ok, i have a AMD Radeon HD 6570M/5700 Series(Dell). Why am i not getting any GPU WU's? I am getting WU's from other projects. Any cure for it?

Ok, i have a AMD Radeon HD 6570M/5700 Series(Dell). Why am i not getting any GPU WU's? I am getting WU's from other projects. Any cure for it?

Thank you very much. So, the problem remains. Its a shame honestly, i don't want to do prime because i don't think encryption tables help humanity. Oh well, its a tie.

You may want to consider Einstein@Home, then, which is at least related to astronomy:

http://einstein.phys.uwm.edu/

Or, if you prefer, watch for Poem@Home to start offering GPU workunits related to protein research, and therefore with some connection to medical research:

http://boinc.fzk.de/poem/

Question for you - where do I get the info-app.xml file for milkyway?

I have one for Seti, but pretty sure it won't be something I can just copy over....

On Seti, I used it to set my GPU to .5, so I can process two tasks on the GPU at a time - can I do that for milkyway as well?

Genuine Intel Intel(R) Core(TM) i7 CPU 960 @ 3.20GHz [Family 6 Model 26 Stepping 5](8 processors
NVIDIA GeForce GTX 460 SE (1024MB) driver: 2856

Wikipedia says this is a compute number of 2.1, so I assume it will work.

Thank you very much. So, the problem remains. Its a shame honestly, i don't want to do prime because i don't think encryption tables help humanity. Oh well, its a tie.

I recently joined the Milkyway@Home Project and the system reports I have the following GPU:

AMD ATI Radeon HD 4350/4550 (R710) (1024MB) driver: 1.4.1385

I do not see it in the list of ATI Radeon models suitable for double-precision math.

Does this mean I am unable to participate in this project and as such should I remove it from my computer? I certainly don't want to continue to receive and attempt to process tasks if it isn't providing the accuracy required by the project.

____________

simply as you can see from the task you're processing, you're doing your jobs on the cpu. In particular, you're doing jobs not solvable by the gpu at the moment, so don't worry, you're not wasting your time here!

Hi Peter, ideally set your preferences to:

Run only the selected applications:
MilkyWay@Home: no
MilkyWay@Home N-Body Simulation: yes

You can also set the following if you like:

Use CPU, Enforced by version 6.10+: yes
Use AMD (ATI) GPU, Enforced by version 6.10+: no
Use NVIDIA GPU, Enforced by version 6.10+: no

That way you'll only be doing the WUs that GPUs are unable to run.
Your GPU will work fine at Collatz, just set your preferences there to use only the GPU.

Peter, it should also work fine on PrimeGrid. Personally, I prefer finding prime numbers to working on the Collatz problem - I know of more uses for prime numbers that any results from the Collatz problem.

You may also want to look at POEM@Home; their protein work is related to medical research enough that I give them computer time also. They are already using some of the cards in the HD xxxx series, and would like to try on more to help them develop a list of which of them are suitable.


Something else for the rest of this thread: My desktop with the GT 440 occasionally gives me Milkyway@Home messages saying that I do not have the necessary double precision. Since the GT 440 uses a Fermi chip, I thought it had double precision built in. Could the developers check why I get such a message anyway, and either eliminate that message or make it clearer just what is missing?

64-bit Windows 7 Professional SP1
16 GB memory
GT 440 graphics board, 285.62 driver
i7-2600 CPU
BOINC 7.0.3


On another item for this thread: The AMD HD 7970 became available in the last few weeks, and the HD 7950 is expected to become available around the end of this month. Any word yet on whether they will work for Milkyway@Home? Any word on the board length and the power requirements (about the most important factors in whether I'll get one)? One BOINC project has already mentioned that they know it works on some other projects, but will need some changes in their application to work under their project.

GPUGRID has announced that that the HD 7xxx series of boards are the first AMD/ATI boards with enough of the right architectural features that they might be able to use them. So expect some competition for BOINC use of those boards, although not immediately.

7970s .... see: http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=2741 but read all the thread

Size, full specs and full review(s): http://www.guru3d.com/category/Videocards/

Regards
Zy

That was enough to determine that the HD 7970 is not suitable for my computers.

Close enough to hope that the HD 7950 might be, though.

I installed a bare hd 5850 oem vendor id ati(1002), and preceeded to download and install the most current Catalyst from ATI. When all was said and done, winxp 64-bit saw it fine but BOINC found no usable gpu! I use my gpus strictly for BOINC and don't play any games but I assume that the card's performance is up to the specs. These cards appear to have been retailed through Comp USA, Tiger Direct, and likely other sources, but never seem to come with a driver disk, just adapters and crossfire hardware. Does anyone have one of these things up and crunching on a winxp 64-bit box? I'm thinking since the drivers are for win 7 and Vista that something just doesn't come out right. I suppose placing the card in a win 7 system would likely solve the problem, but that is not an option at the moment.
Any comments would be appreciated.
Thank you

An HD5850 card is double precision and should work fine at Milkyway. I run one on Win 7 64 bit with driver CCC 11.10 without any problems. As far as I know Win 7 64 drivers wont work on Win XP 64.

Bill

Unload current drivers and load these :

http://sites.amd.com/us/game/downloads/Pages/radeon_xp-64.aspx

Release Notes:
http://support.amd.com/us/kbarticles/Pages/AMDCatalystSoftwareSuiteVersion1112.aspx

Note in particular:

"The AMD Catalyst™ Control Center / AMD Vision™ Engine Control Center requires that the Microsoft® .NET Framework SP1 be installed for Windows XP and Windows Vista. Without .NET SP1 installed, the AMD Catalyst™ Control Center / AMD Vision™ Engine Control Center will not launch properly and the user will see an error message. "

Post again in Number Crunching Forum as a new thread if you still get issues

Regards
Zy

I got the same problem, running on win 7 64 bit. Everything was workning fine until i downloaded and installed AMD latest driver(12-1_vista_win7_64_dd_ccc.exe). Now, boinc say that no double precision hardware is present.

I went back to the old driver as a test (11-12_vista64_win7_64_dd_ccc_ocl.exe) and everything is working again.

Looks like AMD changed something in the card identifier that your application cannot recognize, therefore idenfying the GPU as incompatible.

I have just updated my 4850 card from Catalyst 11.5 to 12.1 and now Milkyway@home reports errors. Which driver version should I use?

I have just updated my 4850 card from Catalyst 11.5 to 12.1 and now Milkyway@home reports errors. Which driver version should I use?

11.9

Help!
I have two ATI FirePro 3D V8700 gpus which just got knocked out of the project because of the changes in work units. I don't believe it's the changes per se which are the cause of this, but the inability of BOINC to correctly identify the gpu.
Ever since I installed them over a year ago, BOINC keeps reading them as HD 4700/4800 gpus. I am still plagued by messages from WUprop@home saying would you please identify your gpus because of this failure. This has happened with the last three versions of BOINC I have installed, choices are limited because I run 64-bit versions of win7 and xp.
With the changes implemented to the new wu's I have now started to receive the following message from mw@home: "Catalyst driver version is not okay for OpenCL application with this GPU."
The only upgrade for OpenCL for this card is 1.1 from 2010. I updated the Catalyst Center to the most current version, which installs version 1.1 by default. Nothing further is available until they come out with version 1.2, scheduled in August.
I think if BOINC would accurately identify the GPU everything would work fine. These were both exceptional cards completing a wu in around 3 minutes prior to the change.
Any assistance would be appreciated.

Thank you

It looks like all of the 7000 series will be able to crunch here as even the 7750 has a double precision floating point.

AMD App Acceleration2

Supports OpenCL™ 1.2, DirectCompute 11 & Microsoft C++ AMP
Double Precision Floating Point

Yes they can crunch here, but the HD 7750 and HD 7770 models will do so relatively slowly. Double precision FP values of 51.2 and 80 for HD 7750 and HD 7770 are lower than a HD 3850.

You are probably aware of this arkayn, I was just pointing it out in case anyone decided to rush out and buy a 7750 or 7770 for MilkyWay.

I am waiting for the 7850 anyway.
____________

Yes the 78xx models should have excellent performance/watt on single precision projects. MilkyWay too if double precision is one quarter of single precision like 79xx models.

I haven't been following developments so I don't know if Pitcairn 78xx models will have double precision of one quarter or one sixteenth like the 77xx models. If it is only one sixteenth, then possible future Tahiti LE model (7890?) would be the better choice for MilkyWay.

Don't throw away your 5800's just yet; the 7800's FP64 (single precision) is 1/16th FP32 (double precision) >:((

AMD Radeon HD 7870 GHz Edition & Radeon HD 7850 Review: Rounding Out Southern Islands, by Ryan Smith.
____________

Hello, I have GeForce GTX 580 card. Few weeks ago I updated nVidia display driver to ver 295.73. Since then I've experienced a lot of problems crunching fast GPU tasks. "Computation error" - that's the NORMAL message I see last weeks. When will you update your GPU tasks to fit the new driver requirements? Thank you.

It is not the app but the driver causing the problem, when the monitor goes to sleep it is also putting the CUDA card asleep and causing the tasks to error out because there are no capable CUDA cards.

You will either need to turn off display sleep and manually turn off the monitor or go to an older driver.

Primegrid has posted it as a news item.
http://www.primegrid.com/forum_thread.php?id=4122
____________

It is not the app but the driver causing the problem, when the monitor goes to sleep it is also putting the CUDA card asleep and causing the tasks to error out because there are no capable CUDA cards.

You will either need to turn off display sleep and manually turn off the monitor or go to an older driver.

Primegrid has posted it as a news item.
http://www.primegrid.com/forum_thread.php?id=4122

I have been trying to get tasks for my GTX 295s but nothing is sent. Is there no work for cuda?
____________



What you do today you will have to live with tonight

It would seem that this project is not seeing my GPUs I have reset the project to no avail, I have detached and reattached with the same results nothing. so I guess I can not run this project.
____________



What you do today you will have to live with tonight

It would seem that this project is not seeing my GPUs I have reset the project to no avail, I have detached and reattached with the same results nothing. so I guess I can not run this project.

This is what I am still getting when I try to get work.

3/8/2012 10:02:04 PM | Milkyway@Home | update requested by user
3/8/2012 10:02:07 PM | Milkyway@Home | Sending scheduler request: Requested by user.
3/8/2012 10:02:07 PM | Milkyway@Home | Requesting new tasks for NVIDIA
3/8/2012 10:02:08 PM | Milkyway@Home | Scheduler request completed: got 0 new tasks
3/8/2012 10:02:08 PM | Milkyway@Home | No tasks sent
3/8/2012 10:02:08 PM | Milkyway@Home | Tasks for CPU are available, but your preferences are set to not accept them

here is my preferences page.

Resource share 230
Use CPU Enforced by version 6.10+ no
Use AMD (ATI) GPU Enforced by version 6.10+ no
Use NVIDIA GPU Enforced by version 6.10+ yes
Is it OK for MilkyWay@Home and your team (if any) to email you? yes
Should MilkyWay@Home show your computers on its web site? yes
Default computer location ---
Maximum CPU % for graphics
0 ... 100 100
Run only the selected applications (all applications)
If no work for selected applications is available, accept work from other applications? yes
Frequency (in Hz) that should try to complete individual work chunks. Higher numbers may run slower but will provide a more responsive system. Lower may be faster but more laggy.
default 60.0 (corresponds to 60 fps) 0

I was running BOINC 7.0.3 but upgraded to 7.0.20 to see if that would work Nope it did not.
I have never had any trouble getting GPU work before.
____________



What you do today you will have to live with tonight

Stealth Eagle,

Looks like you may need to check your preferences at the Milkyway@Home end of the connection as well. Many of the BOINC projects allow settings there to limit what kinds of workunits you will be sent.

I'm having no trouble getting workunits under 7.0.3.

Have you tried?:

Setting "Run only the selected applications MilkyWay@Home: yes
MilkyWay@Home N-Body Simulation: no"

Setting "If no work for selected applications is available, accept work from other applications?" to No

Aborting any nbody tasks in your cache

Suspending DistrRTgen

Increasing your "Minimum work buffer" in BOINC Manager, Tools>Computing preferences>network usage tab

so for example (btw it's not my main computer anymore) my NIVDIA GEFORCE GT220 (http://www.geforce.com/Hardware/GPUs/geforce-gt-220/specifications) won't work?

so for example (btw it's not my main computer anymore) my NIVDIA GEFORCE GT220 (http://www.geforce.com/Hardware/GPUs/geforce-gt-220/specifications) won't work?

NVidia outdo AMD, with their inability to do FP64!
I thought the HD 7770 (1/16) was miserable but that was a mid range card. NVidia's new 'Top' card, the GTX680 only offers 1/24th the FP32 performance.
For here, Avoid the GTX680!
____________

Zy,
I just bought a new EVGA GT 520 after my GT 8500 yoasted itself. The GT 520 has a Compute Capability of 2.1 and my card has 2 GB of memory yet I have never seen any BOINC application use the GPU from the GT 520. I particularly noticed with Milkyway since I have had several dozen Milkyway tasks that require 2 CPU/GPU units to process. But every time one executes, it uses my two CPU's but never uses the GPU to process the application. I had no problem with using the GT 8500 GPU so I am really puzzled why it won't use the GPU on the GT 520. After all it has plenty of memory with 2 GB and the speed of the GPU seems to be fast enough at 810 MHZ. The GPU can be overclocked with EVGA's supplied software.
So is this card supported for GPU use with BOINC or do I have to move to a higher card? EVGA does offer a swap facility so I can get full purchase credit if I do have to move up to a more powerfull card but I only have 90 days to do so.

Chi

Zydor now frequents Donate@Home
____________
Go away, I was asleep

Zy,
I just bought a new EVGA GT 520 after my GT 8500 yoasted itself. The GT 520 has a Compute Capability of 2.1 and my card has 2 GB of memory yet I have never seen any BOINC application use the GPU from the GT 520. I particularly noticed with Milkyway since I have had several dozen Milkyway tasks that require 2 CPU/GPU units to process. But every time one executes, it uses my two CPU's but never uses the GPU to process the application. I had no problem with using the GT 8500 GPU so I am really puzzled why it won't use the GPU on the GT 520. After all it has plenty of memory with 2 GB and the speed of the GPU seems to be fast enough at 810 MHZ. The GPU can be overclocked with EVGA's supplied software.
So is this card supported for GPU use with BOINC or do I have to move to a higher card? EVGA does offer a swap facility so I can get full purchase credit if I do have to move up to a more powerfull card but I only have 90 days to do so.

After a lot of 'Trojan and viral', infections, I had to clean out 3 hosts,
2 of them now have work.
The I7-2600 + 2x EAH5870 GPU's now run Milkyway and do very well, I think.
With the I7-2600 free to load and unload the GPU's, they do a WU in 60 to 65 seconds, which I think is good.
I run WINDOWS 7 64bit (OEM) Pro and BOINC 7.0.25, even with only the GPU's used, the system gets very hot! And need additional cooling, using cold outside air.
I started on wednesday, very early, about 01:00, but found the system shutted down, this morning. Thermal shut-down!
And I noticed the 4 cores, with HT turned on, have a ~20% average load.
The DDR 3 1600MHz, also gets hot. Also use a C(ompact)F(lash), as
Ready-Boost-Cache, making the HDD's, 5TByte total, faster on reading and writing.
It uses (a staggering) 550Watt per hour, so that will be 'My Donation", as
I'm (very) low on income.............

Milkyway, Einstein, SETI, bêta too, and GPUgrid are now my main projects.
(Still have to install optimized app.'s for SETI MB and Astropulse.).

____________

Knight Who says Ni

Fred J. Verster,

Have you tried the Tthrottle add-on for BOINC? I have found it good at slowing down CPU workunits on my laptop to control heating, even though it does not seem to have much ability to slow down GPU workunits.

It's also able to display the current temperature for each CPU core that has a suitable device for measuring the temperature, and also each GPU that has such a device (most of the recent CPUs and GPUs have them).

http://efmer.eu/boinc/

http://www.efmer.eu/boinc/tthrottle_manual.html

Likely to be suitable only for Windows versions of BOINC, though.

My desktops use it only for displaying the temperatures, though - they came with adequate cooling.

My laptop needs a little more to perform well, though: Tell it to use only two of the following items - its two CPU cores and its GPU.

Note that it it takes extra effort to tell it how to control non-BOINC programs that often try to run at night, such as some antivirus programs.

Some computers come with programs that can speed up their fans, but such programs often make them noisy as well.

Watch what can block any of the vents on your computers - one of my previous computers gave a scorched smell and became unbootable soon after the side with a vent for letting air in was slid against the side of the cabinet holding it.

Note that BOINC seems to have problems with adjusting its requests for workunits to handle such slowdowns properly, so you may need to limit its queue of workunits waiting for execution so that few workunits will be found in the queue at any one time.

I would like to get back into milkyway@home but at the moment I could only do it using a low profile gpu.
I am aware of a few options with nvidia stock but I would much rather stick with radeon's for this project.
Is anyone here using a milkyway compatible ATI low profile gpu.

____________

I would like to get back into milkyway@home but at the moment I could only do it using a low profile gpu.
I am aware of a few options with nvidia stock but I would much rather stick with radeon's for this project.
Is anyone here using a milkyway compatible ATI low profile gpu.

I find it amazing that AMD have not got any products in this sector, oh well, thanks anyway arkayn.

I'll have to figure out which nvidia low profile gpu's have the best output.


____________

Don't the new 7750s and 7770s support DP float? Is I read it the dp-float-performance is only 1/16th of sp-float, but I think that's not too unusual for low-profile cards.
Anyway, the "Cape Verde" chips should support 64 bit fp and come with very little power consumption.
Happy crunching!

Don't the new 7750s and 7770s support DP float? Is I read it the dp-float-performance is only 1/16th of sp-float, but I think that's not too unusual for low-profile cards.
Anyway, the "Cape Verde" chips should support 64 bit fp and come with very little power consumption.
Happy crunching!

Ah, ok; sorry 'bout that. What do you mean by "low-profile" then? I thought single-slot would count as low profile :-)

Ah, ok; sorry 'bout that. What do you mean by "low-profile" then? I thought single-slot would count as low profile :-)

Well the best solution I can find for a low profile gpu is the GTS450 -


http://www.palit.biz/palit/vgapro.php?id=1392

192 cores with a Cuda compute capability 2.1

Is this is good as it gets at the moment?, anyone know of anything in the pipeline?
____________

Hm, ok - I see the passively cooled card that arkayn posted is a single slot with a slim PCB. But Halz's GTS450? That's a standard dual-slot, isn't it?

Hm, ok - I see the passively cooled card that arkayn posted is a single slot with a slim PCB. But Halz's GTS450? That's a standard dual-slot, isn't it?

Hm, ok - I see the passively cooled card that arkayn posted is a single slot with a slim PCB. But Halz's GTS450? That's a standard dual-slot, isn't it?

You might want to mention that most of the NVidia GeForce cards are poor for double precision tasks, especially the Fermi cards.

Tesla's:
Tesla C2050 (515 gigaflops of double precision performance) 3GB

The Tesla C2050 costs something like EUR 2.000+ (not sure how much that in US Doller is) and in comparison to the latest ATI/AMD cards has relatively poor DP performance.

For example:
AMD 6950 costs something like EUR 200 and delivers 563 Gflops in DP!

source: http://www.legitreviews.com/article/1488/1/

No wonder that the top 20 computers all use AMD/ATI GPUs.

quote]You might want to mention that most of the NVidia GeForce cards are poor for double precision tasks, especially the Fermi cards.

Wow thanks for that 2 year old review. LMAO!

NVIDIA is the best![/quote]

That post was 1 year ago.
____________
Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.

You might want to mention that most of the NVidia GeForce cards are poor for double precision tasks, especially the Fermi cards.

Tesla's:
Tesla C2050 (515 gigaflops of double precision performance) 3GB

The Tesla C2050 costs something like EUR 2.000+ (not sure how much that in US Doller is) and in comparison to the latest ATI/AMD cards has relatively poor DP performance.

For example:
AMD 6950 costs something like EUR 200 and delivers 563 Gflops in DP!

source: http://www.legitreviews.com/article/1488/1/

No wonder that the top 20 computers all use AMD/ATI GPUs.

Wow thanks for that 2 year old review. LMAO!

NVIDIA is the best!

Hello,

There is a new chart document with the new gpu supported :

http://developer.amd.com/sdks/amdappsdk/assets/AMD_APP_SDK_Getting_Started_Guide_v2.7.pdf

And this one confirm that i can seat on my 6870 for double-precision :(...

Enjoy ;)

Seat on? Not a phrase I've seen before.

Section 4 appears to say that HD 6870 does NOT have double precision, though.

HD 79nn cards appear to generate more heat than the cooling for my computer room can handle, so I'd like to see more about the relative performance of the cards with DP but lower in the HD 7nnn series.

Notice from server: An ATI GPU supporting double precision math is required

Look, my dear MilkyWay server, I KNOW !!!
I know it already, so why do YOU keep telling me this ???

I have even UN-checked both GPU options in preferences of all MilkyWay profiles home, school, work ... default !

: Use NVidia GPU and use AMD ATI GPU - it is DISABLED.

So, please, why are You (my dear MilkyWay SERVER) stil telling me this? :(

Someone please HELP me, or i will have to KILL the server :D

Thanx a lot
____________

I notice the Nvidia cutoff is the GTX 260. I have GTS 250. I am looking at the specifications and while the GTX 260 has certain features like more cores, higher memory interface width and bandwidth it is slower in other areas to the GTS 250 and the 260 only supports OpenGL 2.1 while the 250 supports 3.0. I have included the specification links form Nvidia are you sure the GTS 250 doesn't qualify?

http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-260/specifications

http://www.geforce.com/hardware/desktop-gpus/geforce-gts250/specifications

I want to be be able to help the project as best as I can. I have fairly powerful machine with Dual Core, Dual CPU's with Hyperthreading for 8 CPU's. Maximum memory of 16 gigs and all the hard drives are U320 15,000 RPM 300 GB SCSI Drives with OS/cache/operations each on a different drive to take advantage of the multitasking multithreading options of the SCSI drives.

James


____________

Joined MilkyWay@Home in 2012
Online since ArpNET days
First activity on Honeywell 1648
Series Mainframe in 1975 at age 12.

I believe the important difference is something that Nvidia didn't bother to put in the specifications, either Compute Capability (usually abbreviated CC) or the ability to do hardware double precision calculations.

If you'd like a different BOINC project related to astronomy but more likely to be able to use that card, try:

http://einstein.phys.uwm.edu/
Note that they will have some shutdowns this weekend.

SETI@Home
http://setiathome.ssl.berkeley.edu/
some shutdowns this weekend likely for them too

If you also want one able to use the CPU but not a graphics card, try:

http://www.cosmologyathome.org/

So turning off the GPU option in settings area would not let me do MilkyWay WU's?

And as second thing. I forced a driver update to the latest Nvidia drivers and I gained OpenCL dll's and I also now have an option on the card in the 3D settings that can be set to be program specific called "extensions limit" which according to the description since it defaults to "Off" says the extension string has been trimmed with the option "On" it will pass longer strings. I think that is talking about precision do you know if that is the case? I remember double precision from my Fortran days 30 years ago.

If neither of those will help I am looking at getting a 295 GTX which is the upper card limit I can put on this motherboard.

I've been on SETI for 11 years, I signed up for all the astronomy ones I could on the 8th when I some from a post on SETI that the BOINC app is supposed resource time share so while I could set profiles for the other machines to different settings I adjust my design station to share time with SETI + 6 other projects.


____________

Joined MilkyWay@Home in 2012
Online since ArpNET days
First activity on Honeywell 1648
Series Mainframe in 1975 at age 12.

If it's under 3D settings, it's almost certainly unrelated to double precision.

Milkyway@Home needs HARDWARE double precision support; the lack of that cannot be fixed by a new driver.

Turning off GPU workunits might allow you to get some CPU workunits; try it and see. However, I've seen BOINC refuse to download any CPU workunits until it already has at least one GPU workunit (not necessarily from the same BOINC project).

Also, you might want to measure the length of the space your computer has for graphics cards, unless you wouldn't mind buying a card so long it won't fit in your computer.

I ended up figuring this would be a problem with the card. The GPU on it has worked great with SETI for several years on the GeForce 1GB 250 GTS so I should meet the need for having done something with a GPU and I did get 10 WU's on the 2nd day just nothing since I'll leave the setting with the GPU options turned off and see if I get anything.

But I went ahead and ordered the highest end card I can put in this motherboard and OS which is the GeForce 1.74GB 295 GTX which is on the approved list and the added benefit to me is that has dual GPU's on it each with 240 cores while my current one has only 120 with 1 GPU. Plus the 295 GTX has a whole bunch of other features on it that aren't in the 250 GTS.

James

____________

Joined MilkyWay@Home in 2012
Online since ArpNET days
First activity on Honeywell 1648
Series Mainframe in 1975 at age 12.

Can someone direct me to where I can download earlier versions of the Catalyst software? Running Windows7 x64 and updated Catalyst today to 12.1 and MW is all aborting. Read earlier posts and someone said 11.9 is good. Tried to find it but had no luck. :(
____________

4870 GPU
4870 GPU