OCCT Perestroika - 3.0.0.b04 released ! Stable : 2.0.1

Printable View

Show 100 post(s) from this thread on one page

12-13-2008, 07:02 AM
Tetedeiench
Quote:

Originally Posted by 003

I can confirm this issue, and add to it. In b10, neither the linpack test or the normal OCCT test loads the CPU at all. Have not tried the earlier betas.

It should work with this one :
OCCT PT 3.0.0.b12 Out !

Changelog :
- Several graphical glitches were fixed
- the Linpack test is now working again
- The OCCT test has been updated - check it out ;)
- the GPU test now correctly works on Crossfire configurations and SLI config too (SLI needs a profile)
- A few warnings were changed (just made more explicit, really)
Enjoy !

http://www.ocbase.com/download.php?fileext=beta
12-13-2008, 07:05 AM
Tetedeiench
Quote:

Originally Posted by BababooeyHTJ

In b10 specs in sig the gpu test does not work at all. Just does not start.

mmmh, Quad-GPU crossfire, interesting.

I made some heavy changes to the GPU test recently. Could you test it and tell me if it works :
- In windowed mode or fullscreen ?
- In error check mode ?
Keep in mind that it takes 1 min for the test to start in Custom and Auto modes (the test always starts after 1 min).
12-13-2008, 10:36 AM
davidk21770

Quote:

Originally Posted by Tetedeiench

Keep in mind that it takes 1 min for the test to start in Custom and Auto modes (the test always starts after 1 min).

d'oh -- you should remind those of use that usually use Infinite. And, why?

In error checking mode here, I get 15fps and no movement of the pattern -- but GPU temp goes up like it's under load.
Without error checking, I get 75 fps and the pattern moves nicely.

Same thing in full screen except it's 12.5 & 84 fps. And, why do you use a low res screen? 640x480!

(Video card in sig. -- eVGA 7900 GT KO RoHS, Zalman VP900CU Cooling, mild OC to 645 MHz)

Edit: Also -- a feature request. I usually run LinPak for a certain number of tests, not for a fixed amount of time. Can you include that option?
And, if I run 90% of memory it won't run (no cpu load). I'm assuming that you're not setting memory correctly for 32 bit XP with 4G of memory (IMU, less is usable because of reserved space for memory mapped PCI cards).
12-13-2008, 05:16 PM
BababooeyHTJ
Quote:
Originally Posted by Tetedeiench

mmmh, Quad-GPU crossfire, interesting.

I made some heavy changes to the GPU test recently. Could you test it and tell me if it works :

In windowed mode or fullscreen ?
In error check mode ?

Keep in mind that it takes 1 min for the test to start in Custom and Auto modes (the test always starts after 1 min).
No, it was just 2 refrence HD4870 512mb on Vista x64. Really wish it worked. That would be the only crossfire stress testing tool with error checking.
12-13-2008, 05:40 PM
Ket

I have to ask, are you sure the temp monitoring is accurate? With OCCT 2b at least it seemed to not be.. said my E4400 was well into the 90s under load :eek: So I put a temp probe in and ran OCCT again, sure enough the CPU was nowhere near those temps (was about 73c)
12-14-2008, 12:27 AM
Tetedeiench

Quote:

Originally Posted by BababooeyHTJ

No, it was just 2 refrence HD4870 512mb on Vista x64. Really wish it worked. That would be the only crossfire stress testing tool with error checking.

If you launch directly OCCTGPU.exe by clicking on it in the folder <OCCT>\bin\OCCTGPU ? It should report an error.

Your DirectX9 version is probably not up to date ;)
12-14-2008, 12:28 AM
Tetedeiench

Quote:

Originally Posted by Ket

I have to ask, are you sure the temp monitoring is accurate? With OCCT 2b at least it seemed to not be.. said my E4400 was well into the 90s under load :eek: So I put a temp probe in and ran OCCT again, sure enough the CPU was nowhere near those temps (was about 73c)

The temp monitoring from OCCT is as good as HWmonitor, which is integrated into OCCT ;)
12-14-2008, 12:41 AM
Tetedeiench

Quote:

Originally Posted by powellandy

Hi

What does the error 'debordement de pile' mean? Got it after 18h of OCCT RAM (Larger) on 3 beta 10 on system below.

Thanks
Andy

That's an internal error of OCCT. it can mean your system isn't stable, as well as being a bug of OCCT.

I'll try to reproduce it, however i don't have a quad-core system :/

Oh, and i apologize for the logn reply delay, i knew i forgot something.
12-14-2008, 02:56 AM
powellandy

Quote:

Originally Posted by Tetedeiench

That's an internal error of OCCT. it can mean your system isn't stable, as well as being a bug of OCCT.

I'll try to reproduce it, however i don't have a quad-core system :/

Oh, and i apologize for the logn reply delay, i knew i forgot something.

Anything I can do to help?

Andy
12-14-2008, 05:05 AM
stasio

Quote:

Originally Posted by Tetedeiench

The temp monitoring from OCCT is as good as HWmonitor, which is integrated into OCCT ;)

Voltage +12V(-12;-5) wrong and GPU temp missing.
HWmonitor 1.12 has several issues to be fix :(
Btw,
When loading OCCT PT 3.0.0.b13 in WINDOWS 7,receive notes:
-Your DirectX9 isn't up to date.OCCT GPU and Power supply tests will be disabled.
-Your graphic card doesn't support CUDA.You need Geforce 8 or higher.GPU memtest will be disabled.
12-14-2008, 07:42 AM
powellandy

Possible bug reports

- How is OCCT 3b12 supposed to behave on clean vista x64 sp1 install. Says dx9 not installed and no GPU tests - is this normal - should I install dx9?

- How is linpack OCCT 3b12 supposed to behave on Hyperthreading Intel P4 630 (Prescott). System info says 1 core and running linpack test shows 100% CPU usage in OCCT but only 50% in Process Explorer. Using CPU OCCT test with hyperthreading box ticked gives 100% usage in both. Should linpack stress both 'cores'?

Andy
12-14-2008, 10:48 AM
Tetedeiench

Quote:

Originally Posted by powellandy

Possible bug reports

- How is OCCT 3b12 supposed to behave on clean vista x64 sp1 install. Says dx9 not installed and no GPU tests - is this normal - should I install dx9?

- How is linpack OCCT 3b12 supposed to behave on Hyperthreading Intel P4 630 (Prescott). System info says 1 core and running linpack test shows 100% CPU usage in OCCT but only 50% in Process Explorer. Using CPU OCCT test with hyperthreading box ticked gives 100% usage in both. Should linpack stress both 'cores'?

Andy

Those are not issues, but they need explanations ;)

* You have to update DirectX9 even on vista (it is DirectX10, but it does have everything for DirectX9 too). That's why OCCT is complaining ;)

* That's hyperthreading ;) I'll try to keep the explanation short & understandable. Hyperthreading causes your OS to see twice the real amount of real cores on your system (for instance, if you've got 2 cores, Windows will see 4). However, you can view those 4 cores as "2 are real, 2 are "benefits only" from HT". So if you load the 2 real cores, and the 2 others are IDLE, you'll see, In the task manager, 50% workload. In reality, you'll have both of your physical cores loaded, and no benefit from HT taken. Those 50% means "95% of the real horse power of your CPU", to keep it short.

OCCT CPU test has 2 different behaviours :
* Without HT as DC, it will launch one test per physical core (i.e. "real" core). Thus you'll see 50% load, even if 95 or so% of your CPU is busy.
* With HT as DC checked (by default, as it proved more efficient on HT CPUs), OCCT will launch one test per core, physical or logical (i.e. real and "fake" cores). You'll see 100% workload in the task manager.

The linpack test has only one mode : 1 test per real core. Thus the 50% in the task manager.

The hard part is understanding that only half of your cores in the task manager are "real", and that the others represent the benefits from HT.
12-14-2008, 10:52 AM
Tetedeiench
OCCT PT 3.0.0.b13 OUT !

Changelog :
- OCCT CPU Test fixed : Now the mode "large data set" works on core i7s
- A message now warns you to use fullscreen mode when you're using SLI/Crossfire. I'm currently working on the profile part in Nvidia's SLI (you do need one). Crossfire should be already supported flawlessly.
- Fullscreen mode is checked by default if you're using SLI/Crossfire, even in Auto mode
- Going back to "Auto" mode doesn't make you switch back to 500x500 anymore. 1024x768 is instead used, as intended.
- The Linpack test is now working again with 32 bits OS with more than 2GB of memory installed.
Enjoy !

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

Can someone with a crossfire confirm me that everything is well in Fullscreen mode with the GPU 3D test, even in error check mode ? Thanks !
12-14-2008, 11:09 AM
kiwi

Tetedeiench,

What are ci7-1, ci7-2, ... versions? Are they specifically made for core i7?
12-14-2008, 12:20 PM
Brama

I see that italian translation is incomplete.
If you need an help, I am available. ;)
12-14-2008, 12:22 PM
Tetedeiench

Quote:

Originally Posted by kiwi

Tetedeiench,

What are ci7-1, ci7-2, ... versions? Are they specifically made for core i7?

They were debugging versions for the CPU OCCT problem. I'm currently deleting them :D
12-14-2008, 12:23 PM
Tetedeiench

Quote:

Originally Posted by Brama

I see that italian translation is incomplete.
If you need an help, I am available. ;)

That would be great ;) I'm not done yet, but there will be alot of translation changes soon ;)
12-14-2008, 12:24 PM
Tetedeiench

There was an error in b13, i forgot to remove a debug message. It has been fixed and reuploaded ;)
12-14-2008, 01:40 PM
Tetedeiench
OCCT PT 3.0.0.b14 OUT !

Changelog :
- The linpack test now works with 32 bits OS with more than 2GB of Memory
- OCCT's way of remembering window position as been changed for some people for which it didn't worked
Enjoy !

http://www.ocbase.com/download.php?fileext=beta
12-14-2008, 03:07 PM
powellandy

Quote:

Originally Posted by Tetedeiench

Those are not issues, but they need explanations ;)

* You have to update DirectX9 even on vista (it is DirectX10, but it does have everything for DirectX9 too). That's why OCCT is complaining ;)

* That's hyperthreading ;) I'll try to keep the explanation short & understandable. Hyperthreading causes your OS to see twice the real amount of real cores on your system (for instance, if you've got 2 cores, Windows will see 4). However, you can view those 4 cores as "2 are real, 2 are "benefits only" from HT". So if you load the 2 real cores, and the 2 others are IDLE, you'll see, In the task manager, 50% workload. In reality, you'll have both of your physical cores loaded, and no benefit from HT taken. Those 50% means "95% of the real horse power of your CPU", to keep it short.

OCCT CPU test has 2 different behaviours :
* Without HT as DC, it will launch one test per physical core (i.e. "real" core). Thus you'll see 50% load, even if 95 or so% of your CPU is busy.
* With HT as DC checked (by default, as it proved more efficient on HT CPUs), OCCT will launch one test per core, physical or logical (i.e. real and "fake" cores). You'll see 100% workload in the task manager.

The linpack test has only one mode : 1 test per real core. Thus the 50% in the task manager.

The hard part is understanding that only half of your cores in the task manager are "real", and that the others represent the benefits from HT.

Thanks for this. I know what hyperthreading is... Just wanted to ensure it was working the way it should.... which it is! Thanks :D

Cheers
Andy
12-14-2008, 05:08 PM
Ket
Quote:
Originally Posted by Tetedeiench

OCCT PT 3.0.0.b13 OUT !

Changelog :

OCCT CPU Test fixed : Now the mode "large data set" works on core i7s
A message now warns you to use fullscreen mode when you're using SLI/Crossfire. I'm currently working on the profile part in Nvidia's SLI (you do need one). Crossfire should be already supported flawlessly.
Fullscreen mode is checked by default if you're using SLI/Crossfire, even in Auto mode
Going back to "Auto" mode doesn't make you switch back to 500x500 anymore. 1024x768 is instead used, as intended.
The Linpack test is now working again with 32 bits OS with more than 2GB of memory installed.

Enjoy !

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

Can someone with a crossfire confirm me that everything is well in Fullscreen mode with the GPU 3D test, even in error check mode ? Thanks !
I can probably check that tomorrow at some point for you, stability testing a different system atm.
12-15-2008, 07:32 AM
Brama

Quote:

Originally Posted by Tetedeiench

That would be great ;) I'm not done yet, but there will be alot of translation changes soon ;)

I confirm I am available for translation of OCCT 3 from english to italian.
If you need my help, let me know here or by a PM. :)
12-16-2008, 05:01 AM
Tetedeiench

Quote:

Originally Posted by davidk21770

d'oh -- you should remind those of use that usually use Infinite. And, why?

In error checking mode here, I get 15fps and no movement of the pattern -- but GPU temp goes up like it's under load.
Without error checking, I get 75 fps and the pattern moves nicely.

Same thing in full screen except it's 12.5 & 84 fps. And, why do you use a low res screen? 640x480!

(Video card in sig. -- eVGA 7900 GT KO RoHS, Zalman VP900CU Cooling, mild OC to 645 MHz)

Edit: Also -- a feature request. I usually run LinPak for a certain number of tests, not for a fixed amount of time. Can you include that option?
And, if I run 90% of memory it won't run (no cpu load). I'm assuming that you're not setting memory correctly for 32 bit XP with 4G of memory (IMU, less is usable because of reserved space for memory mapped PCI cards).

I've got so much to read and answer, sorry about the long reply delay ;)

Error checking mode : this behaviour is normal. OCCT GPU checks the 3d output by comparing to a reference frame that was recorded a while ago. Thus, i disable all movement to be sure i've got the same rendering for each frame.

More, CPUs aren't capable of checking alot of frame per seconds. To prevent the CPU from dropping frames because the pile is full, i limit the fps to 15, which is already not bad. And for that, i augment the complexity of the scene. That's what you see at the beginning. It may happen if the pile is full : OCCT will then augment the complexity to avoid dropping frames (and thus, not cheking the results).

What do you mean by low-res screen ? You can configure it in OCCT's UI ;)

As for the linpack "run a fixed # of loops", yes, it is planned. I don't think it'll be done for 3.0.0 (i want to stop implementing features to be able to release a stable version), but in 3.0.1, i'll definitly think about it, as well as, in this case, outputting the GFlops value calculated by linpack.
12-16-2008, 05:03 AM
Tetedeiench

Quote:

Originally Posted by Brama

I confirm I am available for translation of OCCT 3 from english to italian.
If you need my help, let me know here or by a PM. :)

Sorry to ask you, but can you send me your email by PM or to iench 'attttt' ocbase 'dot' com ? I'm usually communicating by email with people doing translations ;)

Thanks ;)
12-20-2008, 01:21 AM
Tetedeiench
OCCT PT 3.0.0.b15 OUT !

Changelog :
- News version of Linpack test : can someone checdk with XP 32 bits with more than 2GB of memory ?
- The Linpack test isn't limited to 2GB of memory on 64 bits systems. It was successfully tested on a 16GB memory computer - it takes about 13s to allocate all that memory :D
- The tab order is now correct in the option window
- SLI detection fixed : it should now work and enable fullscreen by default in GPU and Power Supply tests
- The warnings about SLI and Powersupply test now shows correctly (i.e. only once).
- Fixed several errors (invisible for you all) preparing for the translation process
Enjoy !

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

Show 100 post(s) from this thread on one page