opencl ray tracing!

[fellix_bg]

New beta is out (1.5) -- read more here!

Users with NV cards - report in.

[freeloader]

My score...

[Pyr0]

Version 1.5 scores
(still only using 1 core of my X2 card)

RUN_SCENE_CORNELL_32SIZE : 3626.3K


RUN_SCENE_CORNELL_64SIZE : 6955.5K


RUN_SCENE_CORNELL_128SIZE : 5976.7K


RUN_SCENE_SIMPLE_64SIZE : 31614.8K

[freeloader]

V1.5 scores...

5850 1.05ghz core/1200 mem.

[fellix_bg]

ATi users should go along with 64 work group size.

[Talonman]

i'f you're going to have the scores in a list, i think it might be best to have clock speeds listed too

I agree...

But being a Nvidia guy, and knowing we need some work in this benchmark yet, think we should mess around in this thread until we have a version better suited for both GPU vendors...

Then we should start a new fancy thread, with clocks and all.

I will be glad to if we need a 'list guy'.

New beta is out (1.5) -- read more here!

Users with NV cards - report in.

http://forum.beyond3d.com/showthread.php?t=55913&page=2

Working on it!

RUN_SCENE_CORNELL_32SIZE = 2072.2k

RUN_SCENE_CORNELL_64SIZE = 2898.2K

run_scene_cornell_96size = 2662.0k

RUN_SCENE_CORNELL_128SIZE = 2898.2K

run_scene_cornell_160size = 2898.1k

run_scene_cornell_192size = 2898.7k

run_scene_cornell_224size = 2318.5k

run_scene_cornell_256size = 2526.3k

run_scene_cornell_320size = 2813.9k

run_scene_cornell_384size = 2813.9k


RUN_SCENE_SIMPLE_64SIZE = 113,564.9K

[Mechromancer]

ATi users should go along with 64 work group size.

Correct!

RUN_SCENE_CORNELL_64SIZE = 6955.5K HD4870@790/900mhz 512MB

[Talonman]

Note that us Nvidia guy's did save some face, running the RUN_SCENE_SIMPLE_64SIZE test in his v1.5 BETA directory. http://davibu.interfree.it/opencl/sm...u-v1.5beta.tgz

This was running on 1/2 of my 295@ C=621, Shaders=1512, and Memory=1152.


114,154K...

It was posted by fellix, that his 5870 was getting 70,000K in that scene.
Likewise, Sinistar posted that he was getting 71,000K on his 5870, so there is yet hope for a performance increase.

In all the other tests, still not so good.

[fellix_bg]

Not much real work to be done in this particular scene. Most of the rays are terminated right on in the void. Still, the CPU load time is worrisome.

[Chumbucket843]

nice increase in performance but cpu usage is still high.

[Talonman]

This is me on the latest version...

[SocketMan]

5770@950/1200

[BeepBeep2]

HD4850 @ 800/2250

EDIT: Sorry, here's group size 64...

[Talonman]

I believe this is what we are looking at so far...
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)
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)

If you guy's want me to start a new thread taking these scores over, linking back to here, and recording all new posts, give the word...

I can serve as an excellent 'list guy', but I would rather update the OP.

If Chumbucket843 want's the action, that's fine too. I'm easy...
Feel free to copy and paste.
I don't want to seem pushy!

I still am not sure if we want to record this app 'benchmark wise', or just throw some numbers up to get a feel of our performance based on our GPU.
It is fun to see how we progress on the new versions. I find it to be a charming app.

I do have to say that the 114,154.1K Samples/sec I can produce running the 'RUN_SCENE_SIMPLE_64SIZE', has me wondering just how wild of a ride OpenCL is going to be on performance.
FYI- You DO get your picture rendered darn fast in that one!

It looks like to me so far that ATI or Nvidia has an opportunity to shine, based on exactly what type of calculations the app is doing. Strap in!
I imagine this all could change as we get new versions of OpenCL released to us, along with diver updates.
There may currently be certain types of calculations that we are not handling the best way, but just like a switch thrown, now runs greasy fast? Hard to say.

[Calmatory]

Actually, the program is not ray tracer, but a path tracer instead. That's why the image quality is noisy at first, but improves as the number of render passes grows.

Thanks for the heads up though.

[BeepBeep2]

Actually, the program is not ray tracer, but a path tracer instead. That's why the image quality is noisy at first, but improves as the number of render passes grows.

Thanks for the heads up though.

"Path tracing is essentially a form of ray tracing whereby each ray is recursively traced along a path until it reaches a light emitting source where the light contribution along the path is calculated. This recursive tracing helps for solving the lighting equation more accurately than conventional ray tracing..."


BTW: Talonman, you can change my "2250" to 1125, it's only 2250 effective.

[Earzz]

I would like to give some results with my GFX card but i just cant get this to work, keeps saying that its missing the OpenCL.dll
Just cant figure this out, im on 9.12 and i did instal the hotfix but still without result.

[Talonman]

Thanks for the tip's boys...

I have no problem with 'path tracer', if it's a more accurate term.

2250 changed to 1125 too.

[Pyr0]

I would like to give some results with my GFX card but i just cant get this to work, keeps saying that its missing the OpenCL.dll
Just cant figure this out, im on 9.12 and i did instal the hotfix but still without result.

have you installed the ati stream sdk ?
http://developer.amd.com/gpu/ATIStre...s/default.aspx

[m0da]

doesn't look like it kicks on 3d mode on nvidia cards, or even loads the card- look at cpu usage too
8400M GS - 2808ks/s

[Talonman]

Odd...

I just checked mine, and it goes to 3D Performance speed when the app starts.


I do run with 'Prefer max performance'. in the control panel too. That might help?


High CPU utilization on systems with Nvidia GPU's installed, is still largely considered an official mystery I believe...

Thanks for the post:
m0da ------------ 8400M GS ---- Sample/sec --- 2,808.0K v1.5 (C=514, SH=1028, M=760)

[Earzz]

have you installed the ati stream sdk ?
http://developer.amd.com/gpu/ATIStre...s/default.aspx

Yep i had it already installed, this version:
ati-stream-sdk-v2.0-vista-win7-64.exe (77.4MB)

[redrumy3]

1.5
4870 875/1100
7,769K

[Talonman]

redrumy3 ---------- 4870 ------- Sample/sec --- 7,769.6K v1.5 (GPU=875, M=1100)

[Pyr0]

Yep i had it already installed, this version:
ati-stream-sdk-v2.0-vista-win7-64.exe (77.4MB)

if it installed correctly, that should have opencl.dll's included
i see 2 dll's on my sys...
one in "C:\Program Files (x86)\ATI Stream\bin\x86_64\OpenCL.dll"
the other "C:\Program Files (x86)\ATI Stream\bin\x86\OpenCL.dll"