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

[haschioz]

that really suprises me. i would have figured that someone on here would at least give you the common courtisey to see if they could reproduce you results on thier machines....



i also would have thought that all the ATI ppl on here would be doing thier best to disprove what you have come up with.

i've read all comments and i must say some of you just dont know what hes talking about(i'll give you no possible reasons...)
all he wants is you try his benchmark and report your results so he can if it persists or if hes mistaken. He already begged you twice, so this should be enough.
And as one said everything will eventually reach its limits. but its nevertheless interesting that pwm is holding the core down from reaching its limits....so i dont think the core is flawed just the powermanagement is insufficient.

some should think about their reading skills, cause its important to comprehend texts.

[Hornet331]

What this shows, is that if developers coded to make 100% use of the shaders available on the 4870 / 4890 , the cards would crash.

correction, if they would make use of a specific piece of code, it would make the card crash. Thers are already GPGPU clients(F@H and MW@H comes to mind) for ati cards running and 100% with no problem at all.

[Solus Corvus]

Anyone with watercooling or the fan at 100% try this test? It might be a case of insufficient cooling rather then insufficient VRMs.

[tajoh111]

xfx offer lifetime. yay

Its something that carried over from an purely NV company those.

Also I think another problem is these things are running way to close to maximum capacity. If you look at the water cooling section, alot of chips die from vrm damage. It might be due to some overvolting, but I think alot of it has to do because with air you atleast have the guarantee air is blowing over it. With water, contact might be weak and eventually you lose the card.

These things are running way to hot and I think they should have over budget a bit rather than run just enough for these cards.

I don't even think these things should even have to be cooled. I think they should put enough or better quality so they are not so close to running at max spec.

[MsB]

Hey tetediench I have tried your test with both fan at 100% and with it stock. My card reacts quickly without fan manually set to 100% and runs the fan to 100% quicly whent he test is applied and the sapphire runs your test without problems. The test does seem to ramp up pressure on the card slowly and take it to 100 percent over about 20 seconds so thats good. It gives the card time to respond with fan . I think you are right. I think some cards are not being givin adequit on card power supplies to handle the load your test creates because this test really maxes the card out quickly. Perhaps a lot of manufacturers did see that most games cant take full advantage of the cards and in making them fudged a bit on the PS end of them to save money. It would be good if guys in here would stop bickering off topic and posturing as gpu know-it-alls and just help out by posting results of thier cards. I think having a test that can max out any hardware is a good thing. It can help seperate the wheat from the chaff and be a very nice aid when making hardware purchase decisions.

[cegras]

What 'specifications' is the card exceeding? I read through your first post but found no mention of design specs, nor links to published specs on AMD's website.

Your rather limited lack of power supplies (e.g. there's nothing special about the card crashing under both 550W and 1500W power supplies so long as the PSU itself is able to adequately supply power) as well as some rather broad sweeping generalizations you make (optimizing 3D code too much will crash the card) calls your claims into serious question.

[Nullack]

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.

[Greg83]

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.

Its not faulty , its special !

[STaRGaZeR]

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.

This. If it does, it's faulty. People saying "But but but there are no applications (games) that load the GPU like this!" is just hilarious. I'd love to see what they'll say when company X releases a game, GPGPU software or whatever that loads the GPU at this same level and the GPU turns off in consequence. Or when they're doing their favorite task with their CPU and the computer turns off, then at startup a popup appears saying "load was too high, please don't stress me so much". LOL

[LordEC911]

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.

Any hardware can fail with special software...

[STaRGaZeR]

Any hardware can fail with special software...

So now you're comparing a crash caused by an erratum with one caused by a crappy PWM?

[LordEC911]

So now you're comparing a crash caused by an erratum with one caused by a crappy PWM?

Crappy PWM?
Did you not read where some people did not have a problem with reference cards?

The fact here is this is another power virus app.

[STaRGaZeR]

Crappy PWM?
Did you not read where some people did not have a problem with reference cards?

Any special reason to why you didn't answer my question?

To your question: did you read all those that are having problems with it? Do you find acceptable that a buttload of cards fail under high load even if some don't?

[Frontl1ne]

[EMO]I'm so depressed that my CPU can't handle 365 days straight of Prime95 and i feel like killing myself because my graphics card cannot handle hours on end of OCCT and Furmark. I'm so sad, why won't anyone listen to me.[/EMO]

Show me the cards crashing on a REAL application, then I'll be vaguely interested.

[LordEC911]

Any special reason to why you didn't answer my question?

To your question: did you read all those that are having problems with it? Do you find acceptable that a buttload of cards fail under high load even if some don't?

How is that a question?
Where was I referring to an errata?

I read where the OP supposedly has this HUGE sample of cards that crash... whoops that was only 2 cards and 1 of them works.
In this thread I read where 3 people had a problem even getting the app to work and another 2 people that have a "problem card" but it worked just fine.

So right now... 4 samples:
1 reference card crashes
1 non-reference card works
2 reference cards work

also 3 cannot get app to work...

Obviously there is a problem with AMD/ATi making faulty cards, let's send all cards they make back and boycott them, that is the only solution.

Some of us like numbers and statistics to base our claim. Others like to jump to conclusions.

Edit- To the people complaining about AMD/ATi fanboys and all the supposed whining going on in this thread.
The OP tested 2 cards, jumped to a conclusion that all RV770 cards are faulty because they cannot handle his new power virus.
The title is very misleading and shows a large bias. What he could have/should have done is simply ask people here or in the GPU section to try out his new stability app and grab a consensus. After getting a good number of samples, then you might be able to post a thread like this, SHOWING the numbers.
This path he took is simply very wrong and a sad way to try and get attention for his app.

[zerazax]

Wow, this thread quickly points out who are the fanboys out there and who are sane and willing to be reasoned with.

[/b]A theory , has no facts.[/b]
This has multiple facts now.
FurMark and OCCT
Lets not turn this into a co-incidence w/o proof now.

Wow, false, FALSE. FALSE!

A theory has plenty of facts. You're reminding me of the argument around the theory of evolution and how a lot of evolution-doubters claim it doesn't have facts because it's a theory.

The truth is, the word theory and law are very misleading to the general public. Just like Newton's law's of gravity SEEM to apply universally - the truth is, they also break down and fail at certain levels in the universe - like at the atomic level, or the massive scale levels. Same for the theory of relatively (which has been tested repeatedly in particle accelerators and in observations, so there's plenty of facts that have resulted in proving that the equations and premises work)

I could argue about this all day since I did plenty of EE work in my day, but these conversations get brought up daily around the interwebs.

This. If it does, it's faulty. People saying "But but but there are no applications (games) that load the GPU like this!" is just hilarious. I'd love to see what they'll say when company X releases a game, GPGPU software or whatever that loads the GPU at this same level and the GPU turns off in consequence. Or when they're doing their favorite task with their CPU and the computer turns off, then at startup a popup appears saying "load was too high, please don't stress me so much". LOL

Until the OP is willing to release the source code, then you can't claim ththe cards are failing because the program was proper. Since the OP is unwilling to release the source code, no logical conclusion can be made that a piece of hardware is failing due to PROPER software. I quote:

Before going to sleep, i have to answer that : sorry, but the answer is no.

I don't think there are people skilled enough to analyze the effect produce on the die of a GPU by a particular shader instruction and how that will affect the crash we're encountering.

Don't you think ?

I've always kept my code for myself for the following reasons :

• I like to know where my code is used, for what purpose, by whom
• I don't want to see branches popping everywhere
• I don't want comments on my way of coding

So sorry, the answer is no

So I don't care about whether people can understand your code. There are plenty of board members here who can understand shader instructions or coding, and if not, certainly experts out there can read this and understand it.

The simple fact is, unless you can prove you're giving the GPU proper instructions and that the software cannot be to blame, then you can't conclude the hardware is at fault while the software side remains to be seen.

So to tie this back in with the science lesson, this is why research work is PEER reviewed! Until then, its just hypothesizing. And without the proper checking of work, conclusions are pure FUD.

[Nullack]

Disassemble and reverse engineer his source from the binary then.

[V2-V3]

I believe it is the OCP "Over Current Protection"

Us Overclockers on LN2 have this problem with most cards when raising the Vcore well beyond Spec and benching long hard apps, you can overcome this problem by performing a physical modification of the card that disables or increases the parameters for the Over Current Protection.

"OCP" is different from "OVP" but both are similar "OVP" is "Over Voltage Protection" a similar result would be seen if his application slowly increased the Core voltage once the "OVP" parameter is surpassed the card will fail safe and shut down or "throttle" in an attempt to protect itself.

-------OP-------
can you remove the heatsink of a reference card and the ASUS non-reference card and tell me what Vreg is being used on each?

also for the non0referance ASUS card can you tell me what amperage the GPU is pulling during your test?

and what motherboard are you using? does your BIOS support the adjustment of PCI Express power allocation? ie increasing the maximum PCI Express slot power draw.

[trinibwoy]

Hardware should never fail due to software giving it valid instructions, end of story.

Exactly. The guy said it was just DirectX and we have no reason to doubt him. So unless you're willing to make the leap and claim some combination of DX shaders constitutes a "power virus" the code isn't doing anything evil.

I don't know of any high level language or api in the x86 world for which you could make the kind of excuses popping up in the thread. But then again, GPUs don't have to be anywhere as stable as CPUs do. That's all changing now though.

[Mk]

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.

[AMDDeathstar]

How long do I have to run it for it crash?
I only let run for an hour

[LordEC911]

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.

Where are these other cards being tested?
Again... jumping to conclusions.

[STaRGaZeR]

How is that a question?

The answers to this question...

So now you're comparing a crash caused by an erratum with one caused by a crappy PWM?

...are yes or no. And why if you like. You haven't answered it yet.

I read where the OP supposedly has this HUGE sample of cards that crash... whoops that was only 2 cards.
In this thread I read where 3 people had a problem even getting the app to work and another 2 people that have a "problem card" but it worked just fine.

Some of us like numbers and statistics to base our claim. Others like to jump to conclusions.

Some of us use our little and insignificant brain to think about why ATI has capped FurMark, which is a less demanding application than this one, because it burned a few cards' PWM. Then we wonder about what will happen when we test this thing in more cards, like it happened with FurMark. The same people said back then bah, this application can't be right. It was so right that ATI capped it. What a surprise. However, right now it's just a conclusion indeed.
I've tested it in my 4870 and it FAILS. Let me ask you another question (will you answer this one?): will you admit the cards have such problem when ATI caps OCCT in a future driver release like it will probably happen?

Until the OP is willing to release the source code, then you can't claim ththe cards are failing because the program was proper. Since the OP is unwilling to release the source code, no logical conclusion can be made that a piece of hardware is failing due to PROPER software. I quote:

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?

[Mk]

How long do I have to run it for it crash?
I only let run for an hour

I think your card is fine you are safe dude.
and thank you for testing and providing the info

[leoy]

I do not think that ati says anywhere in their warranty/proper use of hardware, about occt or test programs , other than games.

But i believe that in order to cut costs, some cards may have insufficient vrm's...