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

[STaRGaZeR]

Really when did you test it?
Did you read the thread?
Why didn't you post your results before?

LOL again, you didn't answer any of my questions, yet you write more. I'm putting you in the list of XS people that love to argue a lot, think they're always right but when you ask direct questions that compromise their argument they just STFU or act like you, with evasives and answering a question with another question not related to it (Speederlander, Shintai and the like). You know what list is it, don't you?

Learn how to do things properly:

- Some hours ago.
- Yes.
- Because I didn't want to. Last time I checked I decide when and what I write in the Internet. But who knows, maybe there's a conspiracy and I was hiding the result? Maybe I have invented it? Ahhh...

[Macadamia]

Hardware should never fail due to software giving it valid instructions, end of story. Period. You whiners cant cope with the idea of having faulty hardware.

I guess you never heard of Intel/AMD's in-house power viruses.

Although it might be an insane case of the ALUs not being TEX limited at all, plus heavy ROP utilization we're seeing here.
No, a GPGPU application will NOT get to this level of utilization.

To every one saying that's the application flute...then why don't we see any other NV cards fail? why don't we see any other OLD ATI cards fail? why its only HD 4870 / HD 4890 ? why not GTX 260 or HD 4650 for example? its logic every other card works BUT the HD4870/HD4890 and some of them whit a good quality PCB works so in logic again it the application flute? of course not if every other card works then there is something wrong with this one..end of story.
so please less blabla chit chat more testing..
thank you.

Try it with 4770/4650. I bet you'd see an INSANE power draw increase over the general load.

[Tetedeiench]

Really?
So gathering the actual results and revealing the glaring bias of the OP is "stubborn and biased interventions?"
Thanks for your insight.

So like it was mentioned before, if this is a huge RV770 design flaw why is the 4850/4830 unaffected with their even "cheaper" VRMs?

The bug is happening only on 4870/4890 so far. We found 4850's to be unaffected (we even had a Crossfire of 4850 ro be "bug-free").

And we reproduced this bug on about 10 different cards, i stated that clearly in the first post, using AMD and Nvidia Processors, alot of different northbridges, and power supplies brands and such. Alot of hypothesys came live during this, and all were cut out, until the VRM one remained. Not 2. Please be constructive.

I mean, we even had a reference design underclocked running fine under 82A, bumped only the vGPU, and... well... black screen.

The 82A limit on the VRM right now is just a hypothesis. Really, it's the sole thing we can think of that could cause a crash that quick. I can't see temperature rising that quickly that will cause such a crash. If temperature was at stake, the crash wouldn't be that frank, you would see the donut with artefacts first, as usual.

Anyway, my goal is to understand what is going on. Again :

• Only 4870/4980 are affected (that's what our testing shows) at stock frequencies
• Only reference design (i.e. cards that use the reference design 3-stage VRM) are affected. That is a huge number of cards, but indeed, that is

I have to admit i'm surprised of the amount of "hate" i'm getting there. I mean, on any other board i'm going, OCCT is already pretty well known, and people know i'm not doing this for fame, or anything else (hell, OCCT has been around for 6 years now, why would i seek that just NOW ?).

Right now, i've seen RC1 news popping around everywhere (for my own discouragement - i hate people spreading betas), and people complaining that OCCT isn't compatible with ATI cards. Websites refuse to help me understand what is going on... they ignore my emails. I finally could get my hands on a french website - i do hope it'll start the chain.

Posting in the boards is the sole thing that is left for me to understand what is going on. Because i truly think my test is not at stake there, as underclocking make the cards support it.

[Nullack]

I guess you never heard of Intel/AMD's in-house power viruses.

How can a power virus be considered valid software instructions?

[Solus Corvus]

Hardware should never fail due to software giving it valid instructions, end of story. Period. You whiners cant cope with the idea of having faulty hardware.

How do we know that the software is giving it valid instructions without looking at the source code?

[570091D]

I have to admit i'm surprised of the amount of "hate" i'm getting there. I mean, on any other board i'm going, OCCT is already pretty well known, and people know i'm not doing this for fame, or anything else (hell, OCCT has been around for 6 years now, why would i seek that just NOW ?).

Right now, i've seen RC1 news popping around everywhere (for my own discouragement - i hate people spreading betas), and people complaining that OCCT isn't compatible with ATI cards. Websites refuse to help me understand what is going on... they ignore my emails. I finally could get my hands on a french website - i do hope it'll start the chain.

Posting in the boards is the sole thing that is left for me to understand what is going on. Because i truly think my test is not at stake there, as underclocking make the cards support it.

well, i'll say 'thanks' for being open and honest with us.

i'm guessing that there was no way for ATI to know that there would be such a powerful graphics application available for these cards. i'm sure the pwm's are just fine for everything but this, and if that's the case, just leave the program at 2. there is no reason to ensure your card's stability under a load that no game can reproduce.

[HiJon89]

I don't think any optimisation of a game would make use of all 800 shader cores, the rops, the TMU, etc all at the same time. Aka, unrealistic load.

The point of a stability test is to put an unrealistic load on a component Why don't people complain about running Memtest or Linpack for hours? These tests also stress specific components beyond what is to be expected under normal use. These tests are designed to put as much load on the components as possible, if you don't like that then you're rejecting the idea of a stress test as a whole so don't direct your anger at just this one person.

Another thing to think about, what would you say if a new Intel CPU couldn't Prime95 at stock speeds, that it is an unrealistic load?

How do we know that the software is giving it valid instructions without looking at the source code?

As the OP has stated multiple times already, the same cards that fail will pass the test when sufficiently underclocked, suggesting a hardware rather than a software problem.

[LordEC911]

LOL again, you didn't answer any of my questions, yet you write more. I'm putting you in the list of XS people that love to argue a lot, think they're always right but when you ask direct questions that compromise their argument they just STFU or act like you, with evasives and answering a question with another question not related to it (Speederlander, Shintai and the like). You know what list is it, don't you?

Learn how to do things properly:

- Some hours ago.
- Yes.
- Because I didn't want to. Last time I checked I decide when and what I write in the Internet. But who knows, maybe there's a conspiracy and I was hiding the result? Maybe I have invented it? Ahhh...

And I find it amusing that you keep pushing me to answer a question in regards to something I didn't say.
You are intentionally misunderstanding my statement, twisting my words and then trying to provoke me into answering a question comparing things I didn't mention.
Also when you say "questions" you mean question right? There was only one.

Well I guess I will explain my thought to your question.
I was stating any special made software can cause hardware to fail, which was my original statement. You then ask me to compare an errata, which I didn't mention, to a VRM failure, also wasn't in my statement.
Why am I answering a question about myself comparing an errata to a VRM failure when I never did so?

This "supposed" VRM failure makes no sense, seeing as how lesser models are not affected, 4850/4830, and there is simply no proof to backup the claim that it is a VRM failure.

Also, when I asked you about why you didn't post your results right away, yes I was implying you didn't do the test.
Why would you wait until your fourth post in the thread to post the results, which were in a response to my start of gathering data, when you and the OP are trying to help the community with this thread?

Also with some updated results...
#Samples- 9

4 Crashes- Stargazer's 4870, cowie's 3x4890

5 Noncrashes- SparkyJJO's 4870@790mhz, Telperion's Asus DK, AMDDeathstar's VisionTek 4870's in CF@780mhz, MsB's Sapphire 4870 1Gb

Plus two people with a 4850 tested and didn't have a problem.

Then some are having problems with the App-
Zanzabar, Jamesrt2004's 4870,
Plastok's 4890

Again, there is no conclusive data here.

And we reproduced this bug on about 10 different cards, i stated that clearly in the first post, using AMD and Nvidia Processors, alot of different northbridges, and power supplies brands and such. Alot of hypothesys came live during this, and all were cut out, until the VRM one remained. Not 2. Please be constructive.

I am trying to, I guess I am not allowed to forget everything covered in the first post ~ 9 hours after I read it.
I guess what I was remembering was the test which you had the screenshots of.

[Toysoldier]

@Tetedeiench

Have you presented your findings to AMD/ATI ?
If yes, what did they say ?

[Solus Corvus]

As the OP has stated multiple times already, the same cards that fail will pass the test when sufficiently underclocked, suggesting a hardware rather than a software problem.

Suggesting, but by no means proving.

Assuming that there aren't any bugs in the software, testing if it is a PWM problem or a cooling problem should be fairly easy. Take a card that fails and increase it's cooling capacity.

[Toysoldier]

Suggesting, but by no means proving.

Assuming that there aren't any bugs in the software, testing if it is a PWM problem or a cooling problem should be fairly easy. Take a card that fails and increase it's cooling capacity.

I don't think that would help since water cooled cards fail too and cooling has been ruled out as the cause of the failure.

[LordEC911]

I don't think that would help since water cooled cards fail too and cooling has been ruled out as the cause of the failure.

How so?
There are members here with reference cards that are not crashing.

[iddqd]

There are also certain stability tests that can permanently damage some Intel CPUs, and nothing really stresses them that much in real-world use.

But one could argue that doing everything you possibly could at once is not a meaningful mode of operation.

[zerazax]

Anybody can. Why? Because the same cards that fail the test at default clocks pass it with 0 problems if they're underclocked. This has been proved too in this thread several times by the author you don't want to believe. Or did you miss it?

Oh really? He proved two cards? How many other members on this board just posted they encountered no such problem?

All sorts of cards will have varying tolerances, and unless the OP has more than a SMALL SAMPLE SIZE, his "conclusion" is irrelevant

Unless you've got a hundred cards that can all REPRODUCE this same issue, you're playing with small sample sizes

This is like basketball when someone shoots 12/15 for a game. Can you conclude this guy is a .800 shooter? And right now, from our own thread's results, the guy isn't even at 50% for the small sample size, so how can you conclude either way?

[lm358]

It's silly to compare GPU tests like this with CPU 'stress tests'.

CPU is very complex with a number of different units and execution paths and a bunch of cache.. it is very difficult to really stress the CPU simultaneously.
GPU on the other hand (especially one like 4xxx series) is a whole bunch of math units, which can easily be synthetically activated to stress the entire GPU.

Obviously ATi know that in 'real world use' (even GPGPU) applications can't utilise their chip in this way.. hence they make a design decision to use lower-cost power circuitry to meet 'real world' power designs. OK oh well, that's their choice and if you're the 0.0001% of people who care about 'synthetic stress testing' then you should avoid ATi products!

You can kind of think of it like a car engine. It can run from 0-7000rpm. If you run your car engine at 7000rpm 24/7 it won't be very happy Car markers will tune the engine according to durability vs cost and make design decisions based on real world usage. But sure, go stress your car engine :P

[zanzabar]

i fixed the app, i had to reinstall the dx redistributable but games were working fine and i had installed it previously so im just confused.

i get an extra 3c on my 3870 (i know its not the problem card) but it is doing more than before


im thinking that it is the ram

[Tuvok-LuR-]

I tried to get more professionnal testing done. So far no good. People are just plain ignoring my emails. I thought that having more than 1.500.000 downloads would be enough to get, at least, some listening, or a professionnal test when you uncover something in your own field (which is, stability testing). Seems like not. I have enough proof to get, at least, a "duh, let's check, he may be right". I mean, we had this thing happening on a PC with a ToughPower 1500W. I doubt it was at stake.

what about asking in a game developers forum like Gamedev?

[Solus Corvus]

I don't think that would help since water cooled cards fail too and cooling has been ruled out as the cause of the failure.

I searched the thread for watercooled cards and cards with the fans at 100% but didn't find anything. I haven't finished reading the thread though, want to point out some specific posts?

[awdrifter]

For me, I still uses games as stability tester for video cards. I think it would be more helpful for OCCT to simulate realistic gaming loads rather than an unrealistic load that will crash the card.

[Tetedeiench]

Ok. Stop the flame war, please.

First of all, i'll happy to Show my source code to any expert. But i'll not give it to anybody. That's the first point. No copies.

Here are the tests that lead to the conclusions shown there. I'll sum them up for you.

And please understand that i'm not a hardware guru. I'm a developer. i've got a good understanding of Hardware design, yes. But when it comes down to knowing how a VRM works, well... that's well beyond my knowledge. I've always been honest about that. That's why i'm here. Help. That's what i need.

Cards affected

• HD4870
• HD4890
• All in all, G-DDR5 based ATI cards @stock frequencies.

All of them ?

• No. The only card that could withstand the test was of specific Powercolor design, with a 4-stage numerical VRM built-in. All others failed
• All others claims, before this board, that "it runs fine", were because they did not use the test configuration for the maximum load. Investigation on the cases on this board is needed. Can you help me there ?

PC and components used

• Watercooled cards.
• Temporarily over-cooled cards (added 120mm fans on the cards)
• Cards were plugged on AMD and Intel PCs, i7, C2D, and other CPUs and northbridges.
• PowerSupplies were all trusted brands (Corsair, Antec, Seasonic), Min 550W, Max 1500W (yes, we had the bug happening on a 1500W ToughPower).

Test 1 : Debugging the app

• I spent about 3 days doing that. The only debugging tests that worked were those that castrated the load. Made no sense whatsoever.
• Conclusion : non conclusive. or i suck at debugging.

Test 2 : OSes

• Problems Occurs on the following OS-es : XP32, Vista32&64, 7 32&64
• Conclusion : This rules out the OS as the cause

Test 3 : Driver version

• 4 driver versions were used : from most recent, to older version. Same behaviour. Official or betas.
• Conclusion : problem is not related to a Specific driver version. It may be common to all driver versions though, but unlikely, as HD4850 supports the test, and they use the same chip. Memory related problem ? Seems unlikely to me, but hey... who knows.

Test 4 : Lowering the load on the card

• This was done using the following : lowering the resolution of the test, lower the shader complexity value (even if this is less relevant as this changes the shader that is used), and running into windowed mode. Worked.
• Conclusion : Lowering the load makes the test works, even with the same algorithm applied (in the case of Windowed mode / less resolution used). The shader program is not at stake.

Test 5 : Lowering the frequencies

• Lowering the frequencies made the test stable. it ran perfectly fine.
• Conclusion : the program is not at stake. The shader, or program, doesn't depend on the itnernal clocks of the card of any sorts. If lowering the frequency of the cards makes the test stable, it strongly indicates a hardware failure, wether this be temperature, or a power supply stage failure

Test 6 : Increasing the vGPU

• We had a card running the test fine, at lower frequencies. Then we increased the vGPU, and only the vGPU. This only increased the load on the VRM, and the temperature on the card. It did not have any impact on the frequencies and such. The test was made unstable by changing only the vGPU
• Conclusion : This shows that increasing the temps & VRM load makes the test unstable. We do not touch the frequencies, yet the problem is reproduced. We seem to have isolated the problem. The card IS the cause, as it is the only thing that changed between the tests.

Test 7 : Switching cards

• We switched the non-affected, because of specific design, powercolor HD4870 card, with a reference card. Problem was not happening with the powercolor card. problem showed up with the second card. Of course, same OS, same app, same driver version.
• Conclusion : i don't think software is at stake anymore

Lets's combine the conclusions :
Test 1 : non conclusive.
Test 2 : OS is not the cause.
Test 3 : A specific driver version i not the cause, or it is shared by all driver versions. Later conclusions shows that this is VERY unlikely.
Test 4 : Lowering the load make the test stable. With the very same algorithm behind. So the algorithm is supported by the card. Why only in a castrated version ?
Test 5 : Lowering the card frequencies instead of the castrating the test make the test stable again. The test ran at full power, which was unstable before. So the test is supported by the driver, and OS. This indicated a hardware failure.
Test 6 : making the test stable and increase the vGPU, and only this value, makes the test unstable. This changes only the teps and load on the VRM and GPU. This narrows down our research even more.
Test 7 : switching an unaffected, specific design card (the powercolor one) with an affected card makes the bug appear, where it never existed before. As everything else is the same in the PC but the graphic card, the card IS the cause.

Now, we came to the "82A" hypothesis (which is STILL a hypothesis, i remind you) because on ALL cards the crash was trigger when we reached this value, reported by rivatuner. Temps were rivatuner, but this was the common value shared by all cards.

And you got it. All we've done. All the research, with all our limited means.

My final point is : What can you do to :

• help me know what's going on ?
• Narrow why the thing is happening to some cards, and not on others ?

Remember, only G-DDR5 based cards (i.e. HD4870 and HD4890) seems affected by the problem, and only those with a 3-stage VRM power supply, which is the vast majority of the cards.

I am at work at the moment, so i can't do alot of research. Could you dig, for the card that works, what kind of VRM stages they use ? Reference ones, custom ones ? Thanks for your help

[kl0012]

Obviously ATi know that in 'real world use' (even GPGPU) applications can't utilise their chip in this way.. hence they make a design decision to use lower-cost power circuitry to meet 'real world' power designs. OK oh well, that's their choice and if you're the 0.0001% of people who care about 'synthetic stress testing' then you should avoid ATi products!

How they can know that? What is the point to release monsters with 800 shader processors if you can't utilize all of them?
We can't know it for sure but it is possible that some games would look mutch better (with the same FPS) if game designers did better optimization during development. Or may be they did it and then ran into the same instability issue? So may be we need less shader processors but better game optimization?

[p2501]

Okay, one thing I'd be interested in would be if those cards that failed were ever taken apart to see if there's proper contact between HS and VRM's? I can imagine that if there isn't, temps with this test would skyrocket so fast that they'll just shut down very fast. Was this checked? Does it even apply? Seeing that we have vanilla cards that do not fail this test I think the question is valid.

(i.e. maybe it's manufacturers fault. See mainboard VRM cooling sinks, which don't always have good contact.)

[MpG]

How so?
There are members here with reference cards that are not crashing.

Which begs the next question - why? Ideally, we could pop the sinks on all the cards tested, and see if there are differences in the components that correspond to pass/fail. Obviously, that's not feasible, but that doesn't mean it might not be interesting nonetheless.

That aside, there have been cards coming out lately with boosted power circuits - if the manufacturer is going to make a big deal over their non-reference boosted power circuits, aren't we allowed to at least question why there's room for improvement over the base design? It's not as though this industry doesn't have its share of incidents where a manufacturer tried to intelligently cut corners, but flubbed it.

Seriously - somebody puts a processor under liquid Helium, and we all cheer and throw in our armchair opinions. We'll discuss/argue whether a motherboard having "only" 8-phase CPU power is a "problem". But somebody writes some software that can overpower certain stock cards, and suddenly a whole bunch of people are stonewalling questions and shouting that it's not a 'realistic scenario'?

[p2501]

Seriously - somebody puts a processor under liquid Helium, and we all cheer and throw in our armchair opinions. We'll discuss/argue whether a motherboard having "only" 8-phase CPU power is a "problem". But somebody writes some software that can overpower certain stock cards, and suddenly a whole bunch of people are stonewalling questions and shouting that it's not a 'realistic scenario'?

QFT!

Many people in this thread sadly fail to see this.

[Solus Corvus]

Ok. Stop the flame war, please.

First of all, i'll happy to Show my source code to any expert. But i'll not give it to anybody. That's the first point. No copies.
[...]
Test 1 : Debugging the app

• I spent about 3 days doing that. The only debugging tests that worked were those that castrated the load. Made no sense whatsoever.
• Conclusion : non conclusive. or i suck at debugging.

I hope you don't think I was flaming you with my source code comment. I am just saying that we can't necessarily rule it out either. Even the best programmers make mistakes and it helps to have extra eyes looking at it.

If it only happens on GDDR5 cards, what about the 4870x2 or 4770? I would test it on my x2 except that I sold it. I'll be sure to test it on the 4890 when I get it - it better pass at 1.45v and 1100mhz core .

What kind of water blocks were used? Full cover?