New version OCCT stability test

Thanks for the link.

I got my cpu small FFT's stable tonight in Orthos with the following settings:

Multiplier 9
FSB strap 400
FSB 450
PCIE 100
DRAM 1350
Timings 7-7-7-20-2N

Static Read disabled
Dynamic Write disabled
DRAM CLK Skew auto

AI clock twister auto
Transaction booster auto

vCore 1,46
CPU GTL 0/2 auto
CPU GTL 1/3 auto
CPU PLL 1,60
VTT 1,28
vDRAM 2,00
vNB 1,65
NB GTL 0,61
vSB auto

Load Line Cal. performance
CPU spread disabled
PCIE spread disabled
CPU Clock Skew 300ps
NB Clock Skew 200ps

All energy saving stuff is disabled.

The latest beta version is OCCT 2.1.0 beta 04 dated 02.Sep 2008

DOWNLOAD

http://www.yourpicbox.com/images/oyPHSGDVw39726.png

What are the changes in the beta version?

I see a new GPU benching module in this beta.Looks a lot like furmark

where is this option hidden? :)
never mind, it's under its own folder, OCCTGPU ;)

OCCT GPU load test gets a 9800 GTX to 80°C in less than 5min ; Furmark 1.4 still running (for more than 15min) and still at 74~75°C
(pics and details here)

anybody also up for some GPU load showdown testing?:)

Does it work with multiple GPUs?

The latest beta version is OCCT 2.1.0 beta 05 dated 10.Sep 2008
No changelog for beta version.
DOWNLOAD

OCCTGPUv0.41 dated 12.Sep 2008
HERE

The latest beta version is OCCT 2.1.0 beta 06 dated 06.Oct 2008

DOWNLOAD

http://www.yourpicbox.com/images/oOwUjGUOF41135.png

so what's new with this beta?

Temps are still wrong for wolfdale and it reads the 5v and 12v wrong on 790i ftw (12v reads under 7v). Other than that it's a great test.

For me Built-in soft Monitoring show all correct,wolfdale temp correct.
Btw,just released OCCT 2.0.1.RC1-dated 07.Oct 2008

DOWNLOAD

http://www.yourpicbox.com/images/oofLJbhDS41156.png

thanks :) good stuff here.

I thought the new OCCT was supposed to have a GPU stress test?

it's a separate exe.
in the OCCTPT folder, there is another folder called OCCTGPU

Only in beta 5 and 6,RC1 don't have.
Btw,above is link for
OCCTGPUv0.41

Quote:

---

Originally Posted by stasio

Btw,above is link for
OCCTGPUv0.41

---

All right, I tried 0.41 and I can confirm that it does not work with multi GPU configurations. Using Catalyst 8.10 RC2, it only loads one of the four GPUs on my 2x 4870X2s.

Quote:

---

Originally Posted by 003

All right, I tried 0.41 and I can confirm that it does not work with multi GPU configurations. Using Catalyst 8.10 RC2, it only loads one of the four GPUs on my 2x 4870X2s.

---

Sorry, been a while away from this post ;)

Try OCCT GPU in fullscreen mode, it should work... i have to admit i don't have any crossfire @ home to test that though ;)

As for explanations : OCCT is going through a HUGE revamp. Not as deep as the perestroika, but still, really deep.

Right now, you have 3 branches :
2.0.x : OCCT as you know it
2.1.x : OCCT with the GPU test included
3.0.x : not released yet : OCCT with the GPU test, GPU memory test done in Cuda and a new user interface (not ALL new mind you, but made to deal with those tests).

Right now, i'm working on 3.0.0. I'm releasing 2.0.1 stable in a few days to update the monitoring engine which was a bit old, and because there'll be some time until 3.0.0 is released.

Everybody there, still ?

Anyway, thanks to stasio for posting betas here :)

Quote:

---

Originally Posted by dengyong

Temps are still wrong for wolfdale and it reads the 5v and 12v wrong on 790i ftw (12v reads under 7v). Other than that it's a great test.

---

Those errors comes from HWmonitor, which isn't perfect yet. As i'm using his engine, i don't have any control on that. Feel free to contact HWmonitor's programmer to help him fix those bugs :)

Oh, and BTW, you'll always get the latests versions of OCCT by using those links :

OCCT Stable .exe file :
http://www.ocbase.com/download.php?fileext=exe

OCCT stable .zip file :
http://www.ocbase.com/download.php?fileext=zip

OCCT beta (but for 2.0.1.RC1) :
http://www.ocbase.com/download.php?fileext=beta

OCCT GPU :
http://www.ocbase.com/download.php?fileext=gpu

CUDAMemTest :
http://www.ocbase.com/download.php?fileext=cuda

PS : CUDAMemTest is reliable now, even if slow. I'm working with Nvidia on a bug i found in CUDA, happenning under vista only. Do not be surprised by OCCT's speed at testing the GPU memory : one pass executes in about 10s (on my 9800GTX), and 2 passes are almost equal to one pass of the regular Memtest. I'm still working on implementing new tests to be as close as possible to memtest86 as i can. The "Random" test is especially under way, and almost complete !

As you can imagine, implementing all those new tests without knowing either Direct3d or CUDA was quite a challenge ;)

Quote:

---

Originally Posted by Tetedeiench

Sorry, been a while away from this post ;)

Try OCCT GPU in fullscreen mode, it should work... i have to admit i don't have any crossfire @ home to test that though ;)

---

As of 0.41, you can't select stress testing and full screen at the same time. Also, in fullscreen, I would have no way to monitor if all GPUs were loaded. Maybe an FPS increase would show that they were, but then how could you be sure if it was just using 2 or 3 GPUs rather than all four?

Finally, in stress testing, the GPU fails to switch to 3D mode and remains at 2d clocks.

Quote:

---

Originally Posted by 003

As of 0.41, you can't select stress testing and full screen at the same time. Also, in fullscreen, I would have no way to monitor if all GPUs were loaded. Maybe an FPS increase would show that they were, but then how could you be sure if it was just using 2 or 3 GPUs rather than all four?

Finally, in stress testing, the GPU fails to switch to 3D mode and remains at 2d clocks.

---

Stress testing and fullscreen : true, i neevr thought of that. That is a clear limitation though, that will never be removed. I have to keep the window small enoguh so that the CPU will not be overwhelmed by pixels to check. A Picture in 1680x1050 has 1 764 000 pixels, one in 512x512 has 262 144 pixels...

For the 3d mode switch, this is driver controlled, not application controlled. The only thing i can say is that ATI's drivers are poorly made on that regard, and that they won't switch to 3d mode if you're rendering something in a window. And it's quite good for them, as some of their recent graphic cards have cooling that are really unefficient, and OCCT GPU makes them hot, hot, hot...

ATI even reduced Furmark's speed (and iu'm currently more efficient than furmark) to 20fps so that the GPU won't overheat. They did that in the drivers. Rename furmark's executable to something like "3dmark06" and things will get hot again.

I'll get on the 3d mode switch as soon as i can, in window mode, to see if i can get things running. I wish i had an ATI to test that, though :/

Thanks for your great work though! Especially the GPU test... until now, there really hasn't been a reliable GPU stress test. Even with renamed furmark, it won't make the GPUs hot enough for the fans to spin up. With your GPU test, even only stressing one of the cores, the fan spins up almost instantly after starting the test!

Quote:

---

Originally Posted by 003

Thanks for your great work though! Especially the GPU test... until now, there really hasn't been a reliable GPU stress test. Even with renamed furmark, it won't make the GPUs hot enough for the fans to spin up. With your GPU test, even only stressing one of the cores, the fan spins up almost instantly after starting the test!

---

Thanks :)

I'll try to give a shot at finding dead pixels in fullscreen mode. Maybe it'll :
1- make the driver switch to 3d mode
2- use all the GPUs

However, it'll take me some time to achieve, as i'm deep into the new user interface of 3.0.0 as of now, and i don't want to leave that until finished. Be sure i'll be posting here once i do it to see how your config behaves ;)

Thank's Tetedeiench for feedback