Welcome to MilkyWay@home

GPU Requirements [OLD]

Message boards : Number crunching : GPU Requirements [OLD]
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 10 · 11 · 12 · 13 · 14 · 15 · 16 . . . 18 · Next

AuthorMessage
KWSN - Xfer.cns

Send message
Joined: 23 Jan 10
Posts: 3
Credit: 1,788,406
RAC: 0
Message 46244 - Posted: 15 Feb 2011, 4:15:12 UTC - in response to Message 46230.  

Any idea where I can get a Win7_64 AMD SSE3 optimized app?


I'd be after that too as it would got well with my AMD :)
ID: 46244 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Bernt
Avatar

Send message
Joined: 12 Feb 11
Posts: 2
Credit: 61,763
RAC: 0
Message 46246 - Posted: 15 Feb 2011, 15:04:40 UTC

Hi, Can anyone please advise a good ATI Graphical card with the possibility for double precision math to be used for MW app.

Thanks
ID: 46246 · Rating: 0 · rate: Rate + / Rate - Report as offensive
europa

Send message
Joined: 29 Oct 10
Posts: 89
Credit: 39,246,947
RAC: 0
Message 46247 - Posted: 15 Feb 2011, 15:38:44 UTC - in response to Message 46246.  

Does it have to be an ATI? I'm running MSI Twin-Frozr Fermi cards on my machines and a "junior Fermi" as a second card on one all under Linux with good results.

Regards,
Steve
ID: 46247 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile Beyond
Avatar

Send message
Joined: 15 Jul 08
Posts: 383
Credit: 729,293,740
RAC: 0
Message 46248 - Posted: 15 Feb 2011, 17:43:31 UTC
Last modified: 15 Feb 2011, 17:44:05 UTC

For crunching here the ATI/AMD cards are far faster. I'd look at the 5850 and 6950 as good values for fast cards.
ID: 46248 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Bernt
Avatar

Send message
Joined: 12 Feb 11
Posts: 2
Credit: 61,763
RAC: 0
Message 46250 - Posted: 15 Feb 2011, 20:54:59 UTC - in response to Message 46246.  

Thanks a lot for your advice.

Bernt
ID: 46250 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Matt

Send message
Joined: 3 Nov 09
Posts: 1
Credit: 2,204,838
RAC: 0
Message 46281 - Posted: 18 Feb 2011, 16:43:18 UTC

Hey everyone, just wanted to add that I've just installed a Nvidia Geforce 560ti to my pc, and I'm seeing 7 minutes per work unit. What an incredible speed up compared to CPU.
ID: 46281 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile The Gas Giant
Avatar

Send message
Joined: 24 Dec 07
Posts: 1947
Credit: 240,884,648
RAC: 0
Message 46297 - Posted: 19 Feb 2011, 12:14:06 UTC - in response to Message 46281.  

Hey everyone, just wanted to add that I've just installed a Nvidia Geforce 560ti to my pc, and I'm seeing 7 minutes per work unit. What an incredible speed up compared to CPU.

You are quite correct. GPU is significantly better than CPU. You just need to try a recent AMD GPU and wu's will complete in 60's....LOL.
ID: 46297 · Rating: 0 · rate: Rate + / Rate - Report as offensive
John Clark

Send message
Joined: 4 Oct 08
Posts: 1734
Credit: 64,228,409
RAC: 0
Message 46302 - Posted: 19 Feb 2011, 16:38:02 UTC
Last modified: 19 Feb 2011, 16:40:00 UTC

I support TGG's point.

I am only using my old HD3850 between Milkyway and Collatz, dictated by BOINC Manager and the MW servers. This old generation ATI card (now on HD6950) crunches MW units in 9 minutes and 43 seconds.

My HD5970, currently on DNETC, would finish them in less than 90 seconds 2 or 4 at a time.
Go away, I was asleep


ID: 46302 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Arunas

Send message
Joined: 12 Jan 08
Posts: 1
Credit: 570,648
RAC: 0
Message 46465 - Posted: 6 Mar 2011, 20:34:21 UTC

I bought AMD HD 6850 expecting to speedup some boinc projects. This double precition limitation is a big disappointment! Would it be possible to emulate double precision with multiple single precition operations? Otherwise it makes little sense to support AMD at all, as almost all of HD 6000 series have single precision...
ID: 46465 · Rating: 0 · rate: Rate + / Rate - Report as offensive
John Clark

Send message
Joined: 4 Oct 08
Posts: 1734
Credit: 64,228,409
RAC: 0
Message 46466 - Posted: 6 Mar 2011, 20:40:00 UTC - in response to Message 46465.  

as almost all of HD 6000 series have single precision...



As did the HD5xxx series before it and the HD4xxx series before that again.

It was only the top end cards in each series that supported double precision.

I.E:

HD3870 and HD3850

HD4890, HD4870 and HD4850;

HD5970, HD5870 and HD5850;

The HD6990, HD6970 and HD6950.

All cards below those mentioned (with I am sure some exceptions) only support single precision.
Go away, I was asleep


ID: 46466 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Haris Dublas

Send message
Joined: 25 Feb 10
Posts: 49
Credit: 10,137,837
RAC: 0
Message 46480 - Posted: 7 Mar 2011, 12:34:46 UTC - in response to Message 46466.  

as almost all of HD 6000 series have single precision...



As did the HD5xxx series before it and the HD4xxx series before that again.

It was only the top end cards in each series that supported double precision.

I.E:

HD3870 and HD3850

HD4890, HD4870 and HD4850;

HD5970, HD5870 and HD5850;

The HD6990, HD6970 and HD6950.

All cards below those mentioned (with I am sure some exceptions) only support single precision.


You forgot the 4700 series.
ID: 46480 · Rating: 0 · rate: Rate + / Rate - Report as offensive
John Clark

Send message
Joined: 4 Oct 08
Posts: 1734
Credit: 64,228,409
RAC: 0
Message 46496 - Posted: 7 Mar 2011, 21:25:46 UTC - in response to Message 46480.  
Last modified: 7 Mar 2011, 21:26:22 UTC

You forgot the 4700 series.


I am not conversant with the HD47xx series card. I may be wrong, but I assume these were the predecessors of the HD57xx cards, and the HD57xx were only single precision capable.

But you may know the HD47xx series, in some forms, were capable of double precision maths.
Go away, I was asleep


ID: 46496 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile arkayn
Avatar

Send message
Joined: 14 Feb 09
Posts: 999
Credit: 74,932,619
RAC: 0
Message 46501 - Posted: 8 Mar 2011, 0:00:23 UTC

Technically, you forgot the 4770, 4830 and 5830.
ID: 46501 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Derrick Knight

Send message
Joined: 16 Mar 11
Posts: 4
Credit: 8,635
RAC: 0
Message 46580 - Posted: 17 Mar 2011, 2:37:24 UTC
Last modified: 17 Mar 2011, 3:24:47 UTC

I just downloaded BOINC today and joined SETI and MilkyWay. I joined MilkyWay because I thought my GPU would be useful and it is being done at RPI (alumnus). Anyway, I think all my GPU computations went bad. When I opened up my GPU for work, 10 projects began and finished within a few seconds. Their status is "Computation error" and they have not updated in several hours.

My GPU is nVIDIA Quadro FX 5000M (mobile). It is a Fermi card and supposedly is double precision. My drivers come from HP since this is a workstation (8740w). I'm hoping it is not a driver problem because I am not going to stray from the HP provided ISV certified drivers.

Any help would be appreciated.

EDIT: I saw the error files get replaced by new files...those did the same thing.
ID: 46580 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile arkayn
Avatar

Send message
Joined: 14 Feb 09
Posts: 999
Credit: 74,932,619
RAC: 0
Message 46584 - Posted: 17 Mar 2011, 7:06:45 UTC - in response to Message 46580.  

I just downloaded BOINC today and joined SETI and MilkyWay. I joined MilkyWay because I thought my GPU would be useful and it is being done at RPI (alumnus). Anyway, I think all my GPU computations went bad. When I opened up my GPU for work, 10 projects began and finished within a few seconds. Their status is "Computation error" and they have not updated in several hours.

My GPU is nVIDIA Quadro FX 5000M (mobile). It is a Fermi card and supposedly is double precision. My drivers come from HP since this is a workstation (8740w). I'm hoping it is not a driver problem because I am not going to stray from the HP provided ISV certified drivers.

Any help would be appreciated.

EDIT: I saw the error files get replaced by new files...those did the same thing.


First thing I would try would be the Fermi Cuda app instead of the Open_CL version, your drivers might not support the correct version of Open_CL.
http://www.arkayn.us/forum/index.php?action=tpmod;dl=item23
ID: 46584 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Derrick Knight

Send message
Joined: 16 Mar 11
Posts: 4
Credit: 8,635
RAC: 0
Message 46586 - Posted: 17 Mar 2011, 13:44:39 UTC - in response to Message 46584.  



First thing I would try would be the Fermi Cuda app instead of the Open_CL version, your drivers might not support the correct version of Open_CL.
http://www.arkayn.us/forum/index.php?action=tpmod;dl=item23


Thanks. I downloaded, unzipped and ran the program. It appears I need some instructions on how to use it. I found a txt file called stderr. Here is what is reports from my last attempted run.

08:39:11 (5616): Can't set up shared mem: -1. Will run in standalone mode.
No device was specified on the command line
Looking for a Double Precision capable NVIDIA GPU
Found 1 CUDA cards
Found a Quadro 5000M
Device can be used it has double precision support
Chose device Quadro 5000M
Couldn't find input file [astronomy_parameters.txt] to read astronomy parameters.
APP: error reading astronomy parameters: 1
08:39:11 (5616): called boinc_finish


ID: 46586 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile Crunch3r
Volunteer developer
Avatar

Send message
Joined: 17 Feb 08
Posts: 363
Credit: 258,227,990
RAC: 0
Message 46589 - Posted: 17 Mar 2011, 14:36:57 UTC - in response to Message 46586.  



Thanks. I downloaded, unzipped and ran the program. It appears I need some instructions on how to use it. I found a txt file called stderr. Here is what is reports from my last attempted run.



1. shut down boinc completely
2. open the following folder "C:\ProgramData\BOINC\projects\milkyway.cs.rpi.edu_milkyway"
3. copy all files that you've extracted from the zip file into that folder
4. start boinc again



Join Support science! Joinc Team BOINC United now!
ID: 46589 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Derrick Knight

Send message
Joined: 16 Mar 11
Posts: 4
Credit: 8,635
RAC: 0
Message 46593 - Posted: 17 Mar 2011, 19:46:12 UTC - in response to Message 46589.  



Thanks. I downloaded, unzipped and ran the program. It appears I need some instructions on how to use it. I found a txt file called stderr. Here is what is reports from my last attempted run.



1. shut down boinc completely
2. open the following folder "C:\ProgramData\BOINC\projects\milkyway.cs.rpi.edu_milkyway"
3. copy all files that you've extracted from the zip file into that folder
4. start boinc again



Either Boinc doesn't quit when it says it quits, or it just required a full restart. In either case it is now working. Thanks for your help!
ID: 46593 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile arkayn
Avatar

Send message
Joined: 14 Feb 09
Posts: 999
Credit: 74,932,619
RAC: 0
Message 46595 - Posted: 17 Mar 2011, 21:22:08 UTC - in response to Message 46593.  

There are 2 different parts to BOINC, the manager is what you see but there is also the underlying process that does the real work.

To shut down BOINC, you go to the advanced menu and select Shut down connected client, click ok and then cancel. Then you can quit out of the manager.
ID: 46595 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Derrick Knight

Send message
Joined: 16 Mar 11
Posts: 4
Credit: 8,635
RAC: 0
Message 46597 - Posted: 17 Mar 2011, 22:40:15 UTC - in response to Message 46595.  

There are 2 different parts to BOINC, the manager is what you see but there is also the underlying process that does the real work.

To shut down BOINC, you go to the advanced menu and select Shut down connected client, click ok and then cancel. Then you can quit out of the manager.


I greatly appreciate your help. Thanks again. (now kindly stop being so useful so I can stop spamming this board!)
ID: 46597 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Previous · 1 . . . 10 · 11 · 12 · 13 · 14 · 15 · 16 . . . 18 · Next

Message boards : Number crunching : GPU Requirements [OLD]

©2024 Astroinformatics Group