I could be wrong but I think the automatically downloaded Windows 32-bit 0.23 version will need the file to be called "brook32a_ati.dll" because the brook32a_ati.dll file is downloaded and copied and used as brook32.dll:
<file_ref>
<file_name>brook32a_ati.dll</file_name>
<open_name>brook32.dll</open_name>
<copy_file/>

I don't have the automatically downloaded Windows 32-bit v0.23 currently installed but going by the previous version the above code or something like it will be found in the "sched_request_milkyway.cs.rpi.edu_milkyway.xml" file in the BOINC folder.

I'm not certain why the _ati was added to the filename, but it was possibly to distinguish it from planned _amd versions of the brook file (for Cat 8.12 and Cat 9.1). The auto _amd versions ended up not being deployed at MilkyWay. Just a guess but the extra "a" in the auto downloaded brook file probably denotes the recently updated brook version that includes the Cypress fix.

You're exactly right. The boinc client should download brook32a_ati.dll from the server (which is at http://milkyway.cs.rpi.edu/milkyway/download/brook32a_ati.dll) and then rename it to brook32.dll

I think I might email the dev lists to see if there's some bug that would make this not overwrite a previous brook32.dll

Maybe this isn't a bug. If a MW WU was running during the download then the "old" DLL was in use and so it couldn't be replaced by the "new" one.

How I understand the BOINC documentation, that copy flag causes the dll to be copied (and renamed) to the slot directory for that WU. It does not matter if the old brook##.dll was in use during download.

Can anybody check if the copy to the slot directory really happens? The standard procedure without the copy flag would be to create a symbolic link (just look at the file size to determine if it is the real file or not) which simply does not work for dll files.

If it does, reduce the memory down to as low as you can get it - memory speed is irrelevant at MW (I am running a 5970 on 300 memory with no issues, others I have seen running 4xxx below 200 with no issues). A low memory setting will save you a bunch on power, run cooler and/or allow a higher gpu clocks without overvolting.

If all that holds true ..... a check on PSU capacity/output/power draw maybe a further check once you get going.

Regards
Zy

Is memory speed also irrelevant at Collatz? In other words, will an ATI card set up for MW work well for Collatz also or does it require more VRAM clock?

The card will work, but veeeery slowly if its set for low memory (aka 200/300 etc)

Collatz needs the memory bandwidth, Milkyway does not - its the latter thats the reason that a high memory setting is a waste of time at Milkyway.

Collatz needs the bandwidth to cope with the datasets being tossed around, the higher the memory the better as more ends up being sent in one chunk. If the memory is left low as for Milkyway, Collatz needs several sessions to pass the same dataset.

When running Collatz I usually run it at 925clocks /1350 memory on a 5970.

Thanx much... this explains what I'd thought to be an oddity between two 4850 equipped machines (one has ram clock that will not go below 993 in ccc, other has been running with VRAM clock at 650, both with GPU clock at ~650).

Thanx much... this explains what I'd thought to be an oddity between two 4850 equipped machines (one has ram clock that will not go below 993 in ccc, other has been running with VRAM clock at 650, both with GPU clock at ~650).

UPDATE: I guess I expected a couple seconds difference not a couple MINUTES!

On a AMD 9850, windozeXP machine w/ 2GB RAM with 4850 (1GB) I took the timings from it running last night at the lower vram clock (forgot I'd set it down from 650 to 475 yesterday). I then ran about 5 MW WUs and 5 Collatz WUs with the higher vram clock setting.

850 core/475 vram vs 850/950.

MW no measurable difference.

Collatz avg 19min 44secs vs 16min 52secs!!

---

- da shu @ HeliOS,
"A child's exposure to technology should never be predicated on an ability to afford it."

Thanx much... this explains what I'd thought to be an oddity between two 4850 equipped machines (one has ram clock that will not go below 993 in ccc, other has been running with VRAM clock at 650, both with GPU clock at ~650).

UPDATE: I guess I expected a couple seconds difference not a couple MINUTES!

On a AMD 9850, windozeXP machine w/ 2GB RAM with 4850 (1GB) I took the timings from it running last night at the lower vram clock (forgot I'd set it down from 650 to 475 yesterday). I then ran about 5 MW WUs and 5 Collatz WUs with the higher vram clock setting.

850 core/475 vram vs 850/950.

MW no measurable difference.

Collatz avg 19min 44secs vs 16min 52secs!!

Yet another update:

On a 5830 that was clocked 855 / 855 changing it to 855/1000 and running some Colatz WUs resulted in a decrease in average runtime of 38.5 seconds / WU. And again no change in MW WUs... well with all that proven well 'nuff for me... it's back to low vrams and MW workunits.

Glad to hear, that this is solved, cause those High-End 5700/5800 &5900 can deliver a lot of (Whetstone)MIPS, also 'called FLOPS'.

My ATI host:Q6600(stock),DRAM 800MHz; 4,4,4,13-2T;HD5770 &EAH4850.
(Switched this & other CUDA host off, due to heat, it's 33C, atm. outside, inside a nice 24C (AC=ON) and UPLoads are/were still 'stuck' and also ran out of SETI MB, just as SETI Bêta has to little AP tasks and most of them need 3,4,5,6, even 7 wingmen.
>50% of my AP-tasks have a 3,4,5,6,7 number at the _end.

My C.C. tasks take ~11-12 minutes on the 5770 and 17 min on 4850, M.W. only runs on 4850 (Double Precision).
When running only C.C. on both cards, I need more then 10 WU's an hour....

---

Knight Who says Ni

any hope of getting said ati version under linux64?

Boy do I wish...

Will this update be automatically applied now when I attach to the project?

Is there a possibility to test a linux 32 or 64 bit gpu client app for this project ?
I have an Ati Radeon HD5870 which crunch on Dnetc.
I would prefer to crunch on Milkyway@Home if a linux gpu app has been provided with.

Thanks.

Is there a possibility to test a linux 32 or 64 bit gpu client app for this project ?
I have an Ati Radeon HD5870 which crunch on Dnetc.
I would prefer to crunch on Milkyway@Home if a linux gpu app has been provided with.

I converted one of my XP64 machines over to Ubuntu v10.04 to try DNETC on the ATI card and it works... but I had to use the Berkeley variant install instead of the Debian/Ubuntu package due to a permissions bug in x64 package with ATI driver... it's an open bug on launchpad I was trying to narrow down but it's beyond my ability to pursue further till I get some help from the package devs.

---

- da shu @ HeliOS,
"A child's exposure to technology should never be predicated on an ability to afford it."

Glad that this is solved, now have a Q6600 host with an HDD5770 & EAH4850, BOINC (6.10.58) reports them as 2 CAL ATI 5770, but just upgraded my XP64 host,
(QX9650+GTX470+GTX480), also reported as 2 NVIDIA 480 C.C.2.0 )

But I'm going to buy an 5850 or 5870, replace for the 4850.
Can't use the 5700 on MW, cause it's s.P. (Only Collatz C.)

---

Knight Who says Ni

Does anyone run MW on a Radeon 6850 successfully?
just upgraded and wanted to know if the ATI 58xx fix also comply to this family of cards, as i cant run the Gpu vers of MW (just SSE2)

Collatz is running smoothly though but perhaps they use other techs to get things running!?

---

Does anyone run MW on a Radeon 6850 successfully?
just upgraded and wanted to know if the ATI 58xx fix also comply to this family of cards, as i cant run the Gpu vers of MW (just SSE2)

Collatz is running smoothly though but perhaps they use other techs to get things running!?

ATi 6850 and 6870 does not have DOUBLE precision as required by this boinc project. In the 6xxx family only the 69xx will have it!

Does anyone run MW on a Radeon 6850 successfully?
just upgraded and wanted to know if the ATI 58xx fix also comply to this family of cards, as i cant run the Gpu vers of MW (just SSE2)

Collatz is running smoothly though but perhaps they use other techs to get things running!?

ATi 6850 and 6870 does not have DOUBLE precision as required by this boinc project. In the 6xxx family only the 69xx will have it!

ah.....oh....well, didnt ckeck out on that before buying :)
guess there was a reason why it was that cheap.

for knowledge check this : http://en.wikipedia.org/wiki/Comparison_of_AMD_graphics_processing_units#Northern_Islands_.28HD_6xxx.29_series

thx for reply

---

I converted one of my XP64 machines over to Ubuntu v10.04 to try DNETC on the ATI card and it works... but I had to use the Berkeley variant install instead of the Debian/Ubuntu package due to a permissions bug in x64 package with ATI driver... it's an open bug on launchpad I was trying to narrow down but it's beyond my ability to pursue further till I get some help from the package devs.

Oh yea, a good workaround was found to this via xhost command in the /etc/init.d/boinc-client script. Not on that machine at the moment or I'd included it but if somebody still needs it reply here.

---

- da shu @ HeliOS,
"A child's exposure to technology should never be predicated on an ability to afford it."

I think we don't need this one as sticky any more.

MrS

---

Scanning for our furry friends since Jan 2002

Does anyone run MW on a Radeon 6850 successfully?
just upgraded and wanted to know if the ATI 58xx fix also comply to this family of cards, as i cant run the Gpu vers of MW (just SSE2)

ATi 6850 and 6870 does not have DOUBLE precision as required by this boinc project. In the 6xxx family only the 69xx will have it!

ah.....oh....well, didnt ckeck out on that before buying :)
guess there was a reason why it was that cheap.

The 68xx series basically replaced the 57xx which is also without DP. Seems that the reasoning for the naming was that the 68xx is as fast or faster in most games than the 58xx. Games generally don't use DP. The 69xx replaced the 58xx and has DP and in addition to high end gaming is also aimed at the science app field, CAL and OpenCL.