ATI Catalyst™ 9.3 Folding@Home Performance Enhancements

['Cuda340]

Just seen this in the Catalyst™ Software Suite
Version 9.3 Release Notes:

Using ATI Stream™ technology, owners of the ATI Radeon™ HD 4000 Series, ATI
Radeon™ HD 3000 Series, and ATI Radeon™ HD 2000 Series of graphics
accelerators will see a performance boost in their Folding@Home scores when
running the latest Folding@Home client and upcoming cores in conjunction with
ATI Catalyst™ 9.3

[orion456]

It's about time...

[SniperXX]

Anyone on here switch to that and see a difference?

[McTimson]

I just installed it, and there's not much of a noticeable difference on my 4850, if any difference.

But, I think this may have something to do with it:
Using ATI Stream™ technology, owners of the ATI Radeon™ HD 4000 Series, ATI Radeon™ HD 3000 Series, and ATI Radeon™ HD 2000 Series of graphics accelerators will see a performance boost in their Folding@Home scores when running the latest Folding@Home client and upcoming cores in conjunction with ATI Catalyst™ 9.3

Currently, this WU is using fahcore_11, so the beta core probably has some sort of enhancements.

[Mr.Guvernment]

It's about time...

it has never been ATI's fault the performance was bad, F@H GPU clients are poorly coded for ATI cards, it is Standfords fault.

['Cuda340]

Yup the "upcoming cores" is the key piece of info.

[turbohans]

it has never been ATI's fault the performance was bad, F@H GPU clients are poorly coded for ATI cards, it is Standfords fault.

I agree, just like with games these clients can be written to inherently benifit from difrent things.

[orion456]

Installed it. On a 5747 WU its taking exactly the same time to completion....so much for that hope. :bang head

Now we wait again for supposed cores coming out...sounds like the A2 core that ended up only working with Linux.

['Cuda340]

According to mhouston:

There will be an updated core needed that is still in beta testing to see increases.

Who knows how long that will be .............:shrug:

[Bon3thugz43v3r]

Can't wait for my HD4870 to fold at full power.

[wingless]

Do we have an ETA on the new, enhanced WUs?

[oakstave]

Can't wait for my HD4870 to fold at full power.

+1. 3700 PPD just won't do.

[orion456]

+1. 3700 PPD just won't do.

:bang head

['Cuda340]

ATI v1.24 Core available

I have not had a chance to read up on this, but here you go:

http://foldingforum.org/viewtopic.php?f=51&t=9162

[Bon3thugz43v3r]

Interesting...looks like I got it.


[20:19:40] Trying to unzip core FahCore_11.exe
[20:19:40] Decompressed FahCore_11.exe (3203072 bytes) successfully
[20:19:45] + Core successfully engaged
[20:19:50]
[20:19:50] + Processing work unit
[20:19:50] Core required: FahCore_11.exe
[20:19:50] Core found.
[20:19:50] Working on queue slot 07 [March 24 20:19:50 UTC]
[20:19:50] + Working ...
[20:19:50]
[20:19:50] *------------------------------*
[20:19:50] Folding@Home GPU Core - Beta
[20:19:50] Version 1.24 (Mon Feb 9 11:00:12 PST 2009)
[20:19:50]
[20:19:50] Compiler : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.762 for 80x86
[20:19:50] Build host: amoeba
[20:19:50] Board Type: AMD
[20:19:50] Core :
[20:19:50] Preparing to commence simulation
[20:19:50] - Looking at optimizations...
[20:19:50] - Created dyn
[20:19:50] - Files status OK
[20:19:50] Error: Missing work file=<>


Edit: Wow with this new core my cpu usage is now 50% WTF

Edit 2: Well I did a little bit of research. It looks like you have to add two commands to your environment variables. CAL_NO_FLUSH 1 and BROOK_YIELD 2

Edit 3: Ok I have no idea what I am talking about CAL_NO_FLUSH 1 Results in 0% CPU USAGE and only 14% GPU USAGE!!!! This is not good obviously as we want gpu usage at 100%.

[ChasR]

it has never been ATI's fault the performance was bad, F@H GPU clients are poorly coded for ATI cards, it is Standfords fault.

ATi and Pande Group worked together in the creation of the GPU2 client so you can't simply blame it on "Stanford".

[Alacritan]

Interesting...looks like I got it.


[20:19:40] Trying to unzip core FahCore_11.exe
[20:19:40] Decompressed FahCore_11.exe (3203072 bytes) successfully
[20:19:45] + Core successfully engaged
[20:19:50]
[20:19:50] + Processing work unit
[20:19:50] Core required: FahCore_11.exe
[20:19:50] Core found.
[20:19:50] Working on queue slot 07 [March 24 20:19:50 UTC]
[20:19:50] + Working ...
[20:19:50]
[20:19:50] *------------------------------*
[20:19:50] Folding@Home GPU Core - Beta
[20:19:50] Version 1.24 (Mon Feb 9 11:00:12 PST 2009)
[20:19:50]
[20:19:50] Compiler : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.762 for 80x86
[20:19:50] Build host: amoeba
[20:19:50] Board Type: AMD
[20:19:50] Core :
[20:19:50] Preparing to commence simulation
[20:19:50] - Looking at optimizations...
[20:19:50] - Created dyn
[20:19:50] - Files status OK
[20:19:50] Error: Missing work file=<>


Edit: Wow with this new core my cpu usage is now 50% WTF

Edit 2: Well I did a little bit of research. It looks like you have to add two commands to your environment variables. CAL_NO_FLUSH 1 and BROOK_YIELD 2

Edit 3: Ok I have no idea what I am talking about CAL_NO_FLUSH 1 Results in 0% CPU USAGE and only 14% GPU USAGE!!!! This is not good obviously as we want gpu usage at 100%.

I got it too and it also utilizes a whole CPU core, and what's odd is that even though F@H is supposed to be a low priority process, it causes my system run kinda laggy, though not as bad as with the SMP client. I haven't tried using the environmental variables though.