opencl ray tracing!

[Earzz]

Thanks guys i finally got it to work, seems like something went wrong with my previous install

Well here are my results with my XFX 5750 512MB @ 870/1338

[Andrea deluxe]

strange...

on my system the bench refuse to run...

tell me: OpenCl.dll is not aviable....

[Talonman]

Thanks guys i finally got it to work, seems like something went wrong with my previous install
Well here are my results with my XFX 5750 512MB @ 870/1338

Thanks...

I believe this is what we are looking at so far...
Lightman ----------- 5870 ------ Sample/sec -- 19,088.2K v1.6 (GPU=1056, M=1247)
freeloader ---------- 5850 ------ Sample/sec -- 17,298.6K v1.5 (GPU=1007, M=1152)

freeloader ---------- 5850 ------ Sample/sec -- 13,719.6K v1.4 (GPU=1007, M=1152)
Toysoldier ---------- 5870 ------ Sample/sec -- 13,719.6K v1.4 (GPU=875, M=1300)
fellix bg ------------ 5870 ------ Sample/sec -- 13,719.6K v1.4 (GPU=900, M=1250)

safan80 ------------ 5970 ------ Sample/sec -- 11,012.8K v1.4 (Unknown)

redrumy3 ---------- 4870 ------- Sample/sec --- 7,769.6K v1.5 (GPU=875, M=1100)
SocketMan --------- 5770 ------ Sample/sec --- 7,535.1K v1.4 (GPU=950, M=1200)
Earzz -------------- 5750 ------ Sample/sec --- 7,535.1K v1.5 (GPU=870, M=1338)
mattkosem --------- 4890 ------ Sample/sec --- 7,520.9K v1.4 (GPU=1056, M=1000)
BeepBeep2 --------- 4850 ------ Sample/sec --- 7,172.0K v1.5 (GPU=800, M=1125)

Mechromancer ------ 4870 ------ Sample/sec --- 6,955.5K v1.5 (GPU=790, M=900)
PyrO --------- 1/2 of a 4870X2 - Sample/sec --- 6,955.5K v1.5 (GPU=790, M=915)
redrumy3 ----------- 4870 ------ Sample/sec --- 6,375.8K v1.4 (GPU=875, M=1100)

PyrO --------- 1/2 of a 4870X2 - Sample/sec --- 5,796.2K v1.4 (GPU=790, M=915)
NovoRei ------------ 4870 ------ Sample/sec --- 5,616.1K v1.4 (512mb, 790mhz)

Talonman ----- 1/2 of a 295 ---- Sample/sec --- 2,898.1K v1.5 (C=621, SH=1512, M=1152)
m0da ----------- 8400M GS ---- Sample/sec --- 2,808.0K v1.5 (C=514, SH=1028, M=760)
Chumbucket843 -- GTX 260 ----- Sample/sec --- 2,068.7K v1.5 (C=602, SH=1369, M=1159)

Talonman ------ 1/2 of a 295 --- Sample/sec --- 1,159.2K v1.4 (C=621, SH=1512, M=1152)
Chumbucket843 -- GTX 260 ----- Sample/sec --- 1,123.2K v1.4 (C=602, SH=1369, M=1159)
DosDuoNo ------- GTX 260 ----- Sample/sec --- 1,093.2K v1.4 (C=655, SH=1125, M=1125)

[Talonman]

Just for trivia boys, we are up to v2.0 Alpha, with multi-GPU support for this app.

If you have any interesting results, I would be glad to point David to this thread.

http://forum.beyond3d.com/showthread.php?t=55913&page=5

[Lightman]

My precious:



Soon I'm going to try V2.0

[Talonman]

Nice... What was your GPU set to?

Lightman ----------- 5870 ------ Sample/sec -- 19,088.2K v1.6 (GPU=???, M=????)

[Pyr0]

the new version doesn't seem to be working very well here

[Mechromancer]

I love the workload load balancing between my CPU and GPU. This new version works quite well for a single GPU config. Pyr0, did you try a larger work group size? Such a small size is probably not giving the second GPU much to do. Try 160 and 256 (160 = 800/5 clusters that our GPUs have).

[Talonman]

Thanks for the posts guys, going to put the word out now to David, to look this way too.

Update: Done, and I'm sure he is looking at your results now!

Note that Dave posted this: "This version should run on all GPUs (and CPUs on Apple/ATI) available. It is highly untested and I'm experiencing a problem under Windows: my PC resets after few seconds like if my power supply couldn't sustain the load Everything works fine under Linux ... quite strange. Well, it may crash but you can give it a try."

If any Windows ATI guy has the system power off, please put the word out. (He is on Windows XP)

[redrumy3]

2.0 does seem better on single card at least for my 4870
875/1100 need better cooling for my 4870 if i want to push it more but this is my 24/7 clock on stock cooler.

no problems here running on windows 7

8,396.1K

[Pyr0]

i've tried different sizes, but they don't make any difference...
i notice when i first start any of the tests, it looks like this and has good cpu utilization too:



but after approx 5 or 6 seconds, the display window changes to look like my previous post and cpu usage drops to around 20% or less:

[Mechromancer]

i've tried different sizes, but they don't make any difference...
i notice when i first start any of the tests, it looks like this and has good cpu utilization too:



but after approx 5 or 6 seconds, the display window changes to look like my previous post and cpu usage drops to around 20% or less:

It takes a few seconds for the program to load balance all the different OpenCL devices. Because the CPU is so much slower at these calculations, it gets a very small percentage of work to do. The GPUs appropriately handle the heavy stuff . They'll probably fix how the GPU workload is split between multiple GPUs in the Beta and final release (crosses fingers).

[Lightman]

Nice... What was your GPU set to?

Lightman ----------- 5870 ------ Sample/sec -- 19,088.2K v1.6 (GPU=???, M=????)

Everything is in the pic

GPU - 1056MHz
MEM - 1247MHz

[TJ TRICHEESE]

for some reason my 285's dont throttle up?

[Talonman]

That is odd, both sides of my 295, and 280 all go into 3D Performance Mode correctly.

[Dade]

i've tried different sizes, but they don't make any difference...
i notice when i first start any of the tests, it looks like this and has good cpu utilization too:

Pyr0, my sounds strange but the problem is that some of the hardware here is just "too" fast for a so simple scene. It looks like the automatic profiling isn't able to correctly evaluate the performances of your GPUs/CPUs.

As suggested by Lightman, I'm going to add hand tunable workload % so, even the profiling doesn't work well, you can hand tune the % of work assigned to each device.

[Talonman]

Thanks for joining David!! Make no mistake, we are glad to have you as a member here...

[r1rhyder]

Same as Pyr0, cpu usage drops out. But all 4 GPU's are clocking up and working great.

[Talonman]

A manual GPU workload distribution feature has been added, to insure optimum performance by hand tuning.

http://forum.beyond3d.com/showthread.php?t=55913&page=6
(Post 137 to download.)




Might want to give it a run if you have more than 1 GPU to task!

[Pyr0]

i think my problem might be driver related or something

with the new build, it starts just as before, only showing half of the scene, then a few seconds later it changes to show approx 1/10th (just like the pics in my previous post)

even using the manual load balancing can't fix it
i cannot get it to display a full scene. it's like one half of my X2 card is not working
samples/sec fluctuates wildly anywhere from 4800 to 8000, and again only one half of my X2 card is using 3d clocks

[r1rhyder]

As you can see in my post above, with auto load balancing, only a small sliver of the scene at the bottom is visible. With v2.0 Alpha2, the more I increase the load on the cpu, more of the scene becomes visible.
The same happens when I increase the load on the fourth OpenCL device "GPU'. So, the more I increase the loads on devices 4 and 5, more of the scene becomes visible. Have no idea why, just what I noticed.

[Lightman]

Your problems seem to be coming from broken OpenCL install or you are simply still using Beta4 version of AMD Steam SDK.

If that's the case just uninstall old beta and install final version.

On my rig SmallPT works correctly and displays whole image.

[r1rhyder]

Previous versions work fine. It's only v2.0 Alpha's 1 and 2 that this happens with. With multi GPU's.

[Talonman]

Lightman...

http://www.luxrender.net/wiki/index....der_and_OpenCL

FYI - "SmallptGPU is a porting of original Kevin Beason's [1] to OpenCL. The performances were again quite impressive: SmallptGPU was about 10 time faster than the SmallptCPU. A video of SmallptGPU can be find here or here. It has attracted a noticeable amount of interest and has been published on the front page of Khronos group too. It has been used as benchmark at Beyond3D, XtremeSystems and in other websites. Some really impressive results have been achieved by running SmallptGPU on an overclocked ATI HD5870 by Lightman: "

Looks like it paid off to be 'King of the List'!

[Lightman]

Oh nooo!

Now I'm sure someone will beat my score today!



ps: You're mentioned as well!