weird ati unified shader count

[Athlonite]

ok i have a weird problem it's that when i start up F@H gpu2 my ati hd2600xt reports itself as only having 80 unified shaders but when F@H isn't running it reports 120 unified shaders err ??WTF?? why is it doin that

[sykozis]

Because F@H isn't capable of using all of ATI's shaders....least that's my theory....could be wrong.

[happy]

Originally Posted by sykozis

Because F@H isn't capable of using all of ATI's shaders....least that's my theory....could be wrong.

+1 as well

[Athlonite]

well that just suks i hope they're gonna fix it soon

[Zebatshu]

Well i found some tweeks one can set in windows registry to enable F@H to use the gfx card better, i got about 500ppd extra from them, + it lowered my CPU usage by a few %. Will post the links when i get home. They are working on a new client, one can only hope it will use the ATI cards to their full potential.

[Zebatshu]

Ok it's not really registry tweeks, just system variables.

If you right click on my computer > advanced > environment variables > system variables > new
then add the following:
CAL_NO_FLUSH=1
BROOK_YIELD=2
FLUSH_INTERVAL=128
CAL_PRE_FLUSH=1

The flush_interval can be set higher or lower depending on you gpu. This has lowered my cpu usage while running the gpu client, also gave me a extra boost in ppd. My FI is set to 256 and it work perfect on my setup. Sorry if this have been posted before.

Here is the definitions of the system variables:

FLUSH_INTERVAL is the one that is going to effect GFX performance. Basically, it corresponds to the number of functions submitted in one shot to the GPU. The GPU will not do anything else, including UI updates, until that submission completes. A lower value will basically reduce the timeslice folding@home gets on the GPU, but UI and graphics responsiveness will improve. However, as the value gets smaller, the CPU/OS/Driver overheads increase, so there is a tradeoff between folding@home performance and UI 'snappiness'.

CAL_NO_FLUSH and CAL_PRE_FLUSH, which were poor name choices by me, change how things are submitted to the hardware. CAL_NO_FLUSH changes how we build up packets of work to submit to the hardware. CAL_PRE_FLUSH basically 'double buffers' the building of command buffers so one can be processed while another is being built.

BROOK_YIELD has several modes, 0/1/2. 0 will spin the CPU giving the lowest latency response to the GPU as possible. 1 will yield the CPU when waiting on the GPU to complete to any process of the same priority. 2 will yield the CPU to any process. Now, for really small flush intervals and small proteins, there is a chance the GPU was almost done when the CPU yields. You have to wait to be rescheduled, which could be up to a millisecond. A fast GPU completes many of the kernels in <100 microseconds, so this can have a large impact. If you have a larger flush interval, you can build up several milliseconds of work so missing by a bit is less of an issue.

Also the link to the form where i collected this info:
http://foldingforum.org/viewtopic.php?p=94584#p94584

[sykozis]

If they would switch to OpenCL, performance would improve as a result. Instead, they will probably stick with CUDA for nV and Stream for ATI.

[deltatux]

Originally Posted by Athlonite

ok i have a weird problem it's that when i start up F@H gpu2 my ati hd2600xt reports itself as only having 80 unified shaders but when F@H isn't running it reports 120 unified shaders err ??WTF?? why is it doin that

Where do you see this info?

Originally Posted by sykozis

If they would switch to OpenCL, performance would improve as a result. Instead, they will probably stick with CUDA for nV and Stream for ATI.

I heard that they were working on an OpenCL version of Folding@Home...

deltatux

[sykozis]

Originally Posted by deltatux

Where do you see this info?



I heard that they were working on an OpenCL version of Folding@Home...

deltatux

No offense...but as with everything else, I'll believe it when I see it. Software developers rarely like to deviate from proven software architectures.

[Athlonite]

@ Deltatux i use everest v5.30

@ sykozis apparently the new GPU3 core thats comming up early 2010 will use opencl so that other video cards like intel gma and S3/via chrome will also be able to fold aswell but your right i'll believe it too when i see it

@ deltatux part 2 look at the blue highlighted txt
Field Value
Graphics Processor Properties with F@H running
Video Adapter Asus EAH2600 XT
BIOS Date 05/1
GPU Code Name RV630
Part Number Uªvé
PCI Device 1002-9588 / 1043-0222 (Rev 00)
Transistors 390 million
Process Technology 65 nm
Die Size 153 mm2
Bus Type PCI Express 2.0 x16 @ x16
Memory Size 256 MB
GPU Clock 810 MHz
RAMDAC Clock 400 MHz
Pixel Pipelines 4
TMU Per Pipeline 2
Unified Shaders 80 (v4.0)
DirectX Hardware Support DirectX v10
Pixel Fillrate 3240 MPixel/s

Memory Bus Properties
Bus Width 128-bit
Real Clock 2214 MHz
Effective Clock 2214 MHz
Bandwidth 34.6 GB/s

Miscellaneous
Utilization 99%
Driver Cat 9.11
Field Value
Graphics Processor Properties without F@H running
Video Adapter Asus EAH2600 XT
BIOS Date 05/1
GPU Code Name RV630
Part Number Uªvé
PCI Device 1002-9588 / 1043-0222 (Rev 00)
Transistors 390 million
Process Technology 65 nm
Die Size 153 mm2
Bus Type PCI Express 2.0 x16 @ x16
Memory Size 256 MB
GPU Clock 810 MHz
RAMDAC Clock 400 MHz
Pixel Pipelines 4
TMU Per Pipeline 2
Unified Shaders 120 (v4.0)
DirectX Hardware Support DirectX v10
Pixel Fillrate 3240 MPixel/s

Memory Bus Properties
Bus Width 128-bit
Real Clock 2214 MHz
Effective Clock 2214 MHz
Bandwidth 34.6 GB/s

Miscellaneous
Utilization 0%
Driver Cat 9.11

[Athlonite]

@ Zebatsu yeah thanks i already tried those along with the forcegpu_ati_r600 switch

[Athlonite]

well don't need to worry about it anymore i took out the HD2600xt and replaced it with a single HD5770 from HIS and it's plodin along doin 2275ppd whether or not that good i don't know i do however know that it's better than the 545ppd the HD2600xt was doin