Developer of OCCT 3.1 "HD4870/4890 Can't handle GPU Stress tests."

[Squall-Leonhart]

Who are you ?
I'm the guy behind the well-known program OCCT, which has recently become an all-rounder stability check program (CPU, GPU, Power supply). English is not my native language, so please excuse the mistakes

What's the point ?
Recently, i've been working alot on the RC1 version, and especially on the GPU:3D test. It has been improved dramatically, and this new test revealed a hardware design flaw in the new Radeon HD4870/4890 cards who followed the ATI/AMD reference design. Cards like PowerColor 4870PCS+ are NOT affected, as they are custom design.

They basically crash because they can't handle the load. Early testing shows the VRM cannot supply enough power to the card. That's the only thing i could dig until now, with limited testing means. They seem limited to 82/83A. Please help us dig more !

For a 3d developer, that means : "Do not optimize your code too much for ATI cards, or you could reach the limits, and crash them".

What's this test doing ?
The new test is still a furry donut, but it sports, among new features, a shader complexity parameter, which is, as its names states, how complex the shader will be, i.e. the amount how work the graphic card will have to do in one pass. The highest value is not always the best. For HD4XXX cards, 3 is the best value.

Let me stress the following points :

• This test uses DirectX9, which is updated SEPARATLY from DirectX 10 and 11. Install it from Microsoft Website
• I do NOT use other functions than DirectX's basic function. Really, nothing fancy. Shader Model 3 shaders, alot of Alpha blending... and that's it !

Is there a specific test configuration to reproduce it ?
Yes. First download OCCT by going on the official website and grabbing the RC1 : ocbase.com • View topic - OCCT PT 3.1.0.RC1 Available !

Next, the goal is simple : maximise the GPU load. Here is the way to do so on those cards. Be sure to use these settings !

• Enable Fullscreen Mode
• Disable Errorcheck Mode (comparing images is NOT effective)
• Use a High resolution. Preferably the native resolution of your screen (i.e. 1680x1050 for a 22" LCD, etc)
• Shader Complexity 3 for HD4XXX cards

Click "Go" and watch your screen go black. Your card has gone into protection mode. Frequencies dropped to 200Mhz. Reboot is needed.

How many cards are affected ?
Right now, we've successfully crashed about 10 different cards using this test, using alot of different power supplies (ranging from 550W Antec to 1500W ToughPower (!!!). We had Seasonic, Corsairs... etc).

OCCT 3.1.0 shows HD4870/4890 design flaw - they can't handle the new GPU test ! - XtremeSystems Forums

[ElijahTW]

Wow, that sucks for HD4870/90 owners :S

[@ruantec]

Ati cards just as AMD CPUs has fallen behind since long time now... sad because i was a very big fan of that combination in the past.

__________________

Current development tools:

Visual C++.net, Visual C#.net
Visual VB.net, Visual Webdeveloper.net
Bloodshed Dev C++, Borland C++
Visual Basic 6

[Kurbster]

this is stupid....

just because a stress test that's meant to push above and beyond what the GPU normally handles causes some problems; you call the card "flawed?"

come get me when the problem shows up in regular games instead of 1-man stress programs.

__________________

[Strike105X]

I agree with Kirby, only when something (a declaration) comes from AMD in these regards i may see this as something serious, until then its all just random bashing, maybe if you look enough you can also find a bug in NVIDIA cards to freeze them.

__________________

DeviantArt|AnimeList|PC Config

---

Last edited by Strike105X; May 31st, 2009 at 18:02..

[bkwegoharder]

Of course when you push a card or any electronic device to a certain point in usage, it will break down. I bet that you could do the same with any graphic card no matter what company if you find a good method.

[Squall-Leonhart]

Originally Posted by Kirby

this is stupid....

just because a stress test that's meant to push above and beyond what the GPU normally handles causes some problems; you call the card "flawed?"

come get me when the problem shows up in regular games instead of 1-man stress programs.

the card is flawed. the test is showing that the 3 phase VRM can't feed the card at FULL load.

It also means that ATI's hardware is capable of higher performance, but they are using driver profiles to limit that performance to prevent the VRM issue from showing up.

The problem does not occur on nvidia hardware, as it has been using atleast 6 phase VRM's since the 8800GT.

MSI also just released a card with 10phase VRM which completely eradicates coil squeal at full load.

Smells of design flaw to me.

Originally Posted by strike105x

I agree with Kirby, only when something (a declaration) comes from AMD in these regards i may see this as something serious, until then its all just random bashing, maybe if you look enough you can also find a bug in NVIDIA cards to freeze them.

Nvidia cards have no VRM drop when full load is placed on them.

---

Last edited by Squall-Leonhart; May 31st, 2009 at 19:10.. Reason: Automerged Doublepost

[bkwegoharder]

You act like Nvidia cards don't have problems

[Squall-Leonhart]

Posting images from toms hardware is going to get you flamed here.

The site is biased as **** towards ATI, and the fact is the single 295 actually pwns the 4870x2 in crysis

amazing that the 295 can't do 2560x1600 in toms benches, but can in HardOCP's

and 4x AA at that resolution is overkill, you barely need 2x.
Besides, that issue is a lack of framebuffer space,

the issue here is cheaping out on the VRM/PWM implementation to cut costs, while severely impeding the full power of the product.
Rename furmark.exe and it will also crash out the card. (Clear case of DAAMIT using driver optimisations to impede a products true operation.)

4 Phase designs don't have this problem. only reference 3 phase.

[bkwegoharder]

If anything you're biased against ATI. Also, it's not that just the GTX 295 can't do it. It's because of a driver issue.


Also I bet that a program can be made to do the same thing to Nvidia cards, or a program can be made to crash AMD processors or Intel processor by overworking them. Of course when you overwork a piece of hardware it's going to shutdown. It's like expecting to have a car work at full performance after driving 200 Miles an hour down a dirt road.

[Squall-Leonhart]

Furmark and OCCT 3.1 do the same thing as Prime95 and Lynx. If the hardware can't handle 100% load then ITS FAULTY BY DESIGN. PCSX2 emulation is as intensive as prime95 in some aspects. I don't see properly configured computers, with quality hardware crashing due to the cpu fail.

and sod off with your biased ****. I don't go spreading fud about ATI that doesn't exist (unlike how Charlie does about nvidia at TheInq). Everything said and stated is researched, factual and reproduced by multiple people

Furmark blackscreen
HD4870 Furmark black screen crash. - techPowerUp! Forums

adding a 4th VRM would lower the heat for one thing.....

Without a doubt the most breath taking game I have seen yet, Crysis is from the developers of Far Cry, yet another title that was ahead of its time. This game will bring any system to its knees, but look good doing it.

Crysis Test Settings

The only new game in our test suite over the previous GTX 295 preview I wrote, Crysis does in fact see a large performance advantage over the AMD 4870 X2 cards - by as much as 15% at 1920x1200. Quad SLI also sees some good scaling over the single GTX 295, around 30%, and we have FINALLY reached the 60 FPS mark in this title at a full HD resolution!!

As i said. Not a driver bug. just a lack of VRAM.

---

Last edited by Squall-Leonhart; June 1st, 2009 at 00:21..

[Squall-Leonhart]

>.>

Enabling 2x AA reduced the GeForce GTX 295's frame rate by 15.4% to about 42 fps. That's actually very playable. Even 4x AA is quite playable at about 38 fps - a performance loss of just 9.6%. This is amazing performance. For the first time in history, we have a single NVIDIA graphics card that can actually deliver playable frame rates at 1920 x 1200 with 4x anti-aliasing enabled.

[MrPink]

Looks like computer sex to me

I wasn't aware of this potential design flaw in the ATI cards; I was considering getting a 4890 at one point, and have even recommended several 4870s to friends in their computers. While they haven't yet complained of getting black screens, still....

I also wasn't aware of Toms being biased ass hell towards ATI . I'll have to broaden my gamut of regularly checked hardware sites.

Those charts are making a GTX 285/295 seem like a very appealing upgrading..

[Squall-Leonhart]

lol, not all the 4870's and xx90's are bad. just make sure if you do get one, it is a custom design with 4 phase VRM, and not 3.

[stpdrgstr]

I wonder is the MSI R4870 512 MB has this problem. And if it does, should I rage about it?

Because it would be yet another reason to rage.

__________________
My rig: ASUS P5QC - c2d wolfdale e8400 @3.60ghz - ASUS Silent Knight - 2x1 GB RAM DDR2 800 Geil - Ati Radeon MSI HD4870 512 MBs 780:1000 - Enermax Modu82+ 620w - 2x160GBs SATA Western Digital - Antec P182

[Squall-Leonhart]

try renaming furmark and running it for abit, if the screen turns black, then yes

This could potentially be why the capabilities of the card are far less then they really should be. the driver uses profiles to create idle states in game processing.

[Strike105X]

This could be the reason why ati doesnt let access to there drivers game profiles like nvidia does

__________________

DeviantArt|AnimeList|PC Config

[Squall-Leonhart]

I was thinking the same thing.
nvidia makes some fixes at the code level, rather then profiles

Still, nvidia's cards are designed to regulate the power the gpu needs using the reference 6 phase vrm.

[Gossamer]

I sure won't be using ATI in my next project. I've had too many driver problems from them not to mention it's virtually impossible to use wine with the current ATI driver setups. I wouldn't mind forking over the extra $$ for stability and re-assurance.

__________________

[bkwegoharder]

Originally Posted by Squall-Leonhart

>.>

Enabling 2x AA reduced the GeForce GTX 295's frame rate by 15.4% to about 42 fps. That's actually very playable. Even 4x AA is quite playable at about 38 fps - a performance loss of just 9.6%. This is amazing performance. For the first time in history, we have a single NVIDIA graphics card that can actually deliver playable frame rates at 1920 x 1200 with 4x anti-aliasing enabled.

If VRAM were the issue, why would an SLI GTX 295 perform less than a single GTX 295?