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

[Tetedeiench]

Wow, lots of crash reports all of a sudden !

We had a period orf scepticism "your talking nonsense", but all of a sudden, we have a rush of reports !

Thanks guys. Your reports helps me in several ways :

• The more cards are affected, the less likely it is a software problem
• The more cards affected, the more likely i will finally get a professionnal attention, and by that i mean someone with the means and knowledge to discover what's really going on (wether he be a user of this board, or a website, or somebody else, i don't care, i want to know what's going on !)

Maybe by going on we'll pin down the problem. I have to admit i don't what to do to go further though. Your thoughts ?

[Hornet331]

Why you don't contact ati directly, im sure they can answer your questions, since they designed the card (would be bad if they can't ).

If they are acting stupid you still present your findings to "news" sites like fud/inq and then other sites will scoop it up and ati needs to respond.

As for testing, i can't really participate since i own a powercolor 4870 PCS+ and they have the beefed up VRM (that thing already sucks 93Amps peak in Furmark, i don't want to see how your test performs ).

[Tetedeiench]

Why you don't contact ati directly, im sure they can answer your questions, since they designed the card (would be bad if they can't ).

If they are acting stupid you still present your findings to "news" sites like fud/inq and then other sites will scoop it up and ati needs to respond.

As for testing, i can't really participate since i own a powercolor 4870 PCS+ and they have the beefed up VRM (that thing already sucks 90Amps in Furmark, i don't want to see how your test performs ).

Previously, i tried to contact them directly. I never got an answer. I have to admit i gave up completly trying to contact them directly :/

Remember i'm just a regular user just as you are. Think of me as yourself, that, when you come back from work, instead of playing World of warcraft, launches his VMWare/Visual Studio, and dwelves into programming. that's how OCCT have been done, for 6 years. So if i had to write a mail, i'd mail... support@ati.com or something like that . I could get my hands, after research, on some direct emails, but they just ignored them, so well...

[Tetedeiench]

Hehe,
OCCT GPU test draws 360W from the wall whereas Furmark stays at 310W...

Skipped that post.

I truly didn't know the power consumption difference was that huge between furmark and me. Impressive. That may explain why some cards are OK with Furmark, and not with OCCT GPU.

[Final8ty]

I stopped reading after the 4th page.

All i see here is that a program can push a gfx card to overheat.

The coolers are not good enough and a compromise may have been made to keep the noise down.

[Tetedeiench]

I stopped reading after the 4th page.

All i see here is that a program can push a gfx card to overheat.

The coolers are not good enough and a compromise may have been made to keep the noise down.

Well, the crash type doesn't look like an overheating crash. If it was an overheating crash, you would have artefacts, etc. More : the crash wouldn't be THAT immediate : it would take at least a few seconds. Here, it's instantaneous. That's the thing that's "weird". You don't even see the donut appearing.

[Pyr0]

just tried this on my 4870X2.
it's cooled with an xspc waterblock but still just shows a blank, black screen as soon as the occt gpu test starts. no gfx, just blank screen

to get the test to run, i need to underclock my 4870X2's memory to around 800 or so (i tried 850 and still got a black screen when starting occt)

[edit]
790core and 820mem, i see the graphic for approx 2 secs before the screen blanks.
750core and 820mem seems to work, i ran it for couple of mins before pressing esc.

[Tetedeiench]

For all those whose card fails, can you do the same and post a rivatuner screenshot showing the VRMA value when the test runs, and see at which value the test seems to fail ?

I'd like to see if, for you too, it fails at 82A

YOu can find the screenshots i'm talking about on the very first post on this thread.

That might confirm/infirm the 82A hypothesis, since your cards display the same behaviour as those affected before by the black screen plague

[Tetedeiench]

Somebody just sent this thread to AMD's website, and here is their official position.

Notice that the thread has been closed, not allowing further questions on the matter :
http://forums.amd.com/game/messagevi...&enterthread=y

I would have liked to exchange with AMD about such a matter, i cannot

Your limited testing is flawed. If the VDDC's overheat they will shut down the card to protect it. If you dont want to have a crash when testing artificial means such as this stability benchmark then increase the fan speed to cool the VDDC's. Obviously underclocking acomplishes a similar goal as they dont get as hot. No games are affected just artificial stability tests.

Before paranoic people start to wonder : i DO NOT have any relation with the guy that started the thread in the AMD board. None whatsoever. i discovered it with a link sent to me by one of my french betatester.

[W1zzard]

good job on the stresstest. dont listen to teh naysayers here. expect your app to be castrated just like furmark in the next catalyst releases

1) The test uses close to 100% of the transistor in the die.

i doubt it is anywhere near 100% (uvd, tesselator, rops, mc, tmus), dont spill out random claims

If it was an overheating crash, you would have artefacts, etc.

no

More : the crash wouldn't be THAT immediate : it would take at least a few seconds. Here, it's instantaneous.

yes. my guess is vrm ocp as well. nvidia delayed a launch because of a similar issue

then increase the fan speed to cool the VDDC's. Obviously underclocking acomplishes a similar goal as they dont get as hot.

fail. clearly has no idea what he's talking about

[theystolemyname]

Your limited testing is flawed. If the VDDC's overheat they will shut down the card to protect it. If you dont want to have a crash when testing artificial means such as this stability benchmark then increase the fan speed to cool the VDDC's. Obviously underclocking acomplishes a similar goal as they dont get as hot. No games are affected just artificial stability tests.

That is a wonderful response. I was considering a 4890, but now I'm not so sure. Does XFX make one with the 4 phase VDDCs?

[Pyr0]

the 82A theory seems like it may hold true for me/my card
i was unable to capture the last RT screenie before the screen blanked.
but at 770 core and 830 mem, the test ran for almost a minute before blank screening

the last run here was 760/820

[Hornet331]

Isn't the 4890 4 phase by default?

Afaik the asus cards (top) have 5 pahses + 2 for memory. ( 4890)

edit: seems like all 4890 have 5 phases + 2 memory (at least the one i have checked: xfx, asus, powercolor, HIS, Club3d)

which brings up another point, why is the 4890 affected by this, it even has a stronger vrm circut then the 4vrm 4870.

[W1zzard]

does anyone have a card with VR modded OCP ? would be an ideal test subject. test with normal ocp, hope that it crashes, increase ovp limit, if it doesnt crash it's ocp.

[Humminn55]

Apparently I'm a tad confused in all this, esp. with the "82A" hypothesis being bandied about.

From what I've been able to gleen, Tetedeiench is saying, and please correct me if I'm wrong....PLEASE!....is that the 4870's that were being used in Crossfire in his original post by one of his testers using OCCT 3.1.0 and the gpu test was pulling 82A+ into each card? Apparently that's what is being said and displayed with the hardware monitoring software graphs presented on the first page...up to 86A per card, which comes out to, since the cards run on +12V DC, 1032W for the card pulling 86A and 980W for the card pulling just below 82A.

And Tetedeiench also states, "....we had a 4870 PCS+ (Asus) overclocked that could pull 106A, flawlessly....", which turns into 1272W of load being processed by the VRM's, correct?

So, what power supplies are being used that can supply 2kw of power, for the 4870's in Crossfire as first mentioned since the two were pulling a combined 168A, or for the 4870 PCS+ card that was pulling 106A? Just curious.....or am I missing something and this is system draw, but that's not what's being presented, at least from what I've been able to read. And I'm also wondering what gauge wiring you're using to sustain 86A of draw since typical 18ga wire won't support such amperages without beginning to melt its insulation.

[perkam]

the 82A theory seems like it may hold true for me/my card
i was unable to capture the last RT screenie before the screen blanked.
but at 770 core and 820 mem, the test ran for almost a minute before blank screening

the last run here was 760/820

A very well designed test, but it has no bearing on practical and everyday applications that the card is used for.

If it was that big a problem we would have seen a recall.

Perkam

[W1zzard]

Apparently that's what is being said and displayed with the hardware monitoring software graphs presented on the first page...up to 86A per card, which comes out to, since the cards run on +12V DC, 1032W for the card pulling 86A and 980W for the card pulling just below 82A.

its not 12v current that is measured but vgpu current

[Humminn55]

its not 12v current that is measured but vgpu current

So what is vgpu current, then? It's expressed in amperage which can be converted to wattage since amperage x volts = wattage.

So what is the voltage being used?

Or, essentially, how is vgpu derived and what does it represent in draw?

Please explain to an unknowing one so she may understand. Thx.

[Pyr0]

stock 3d voltage for a 4870x2 is 1.25-1.26v
if you look at the rivatuner graphs, they show the gpu voltage

[Humminn55]

OK.....I see. So the vgpu amperage, at like 82A at the 1.26V is only like 103W of power. DUH! My stupidity.

[Tetedeiench]

the 82A theory seems like it may hold true for me/my card
i was unable to capture the last RT screenie before the screen blanked.
but at 770 core and 830 mem, the test ran for almost a minute before blank screening

the last run here was 760/820

Thanks for your test This seems to confirm we're hitting a limit there, wether this might be OCP or OVP, which i didn't know about before today. Remember, i'm NOT a hardware guru, i'm just a regular user, a developer, an enthousiast, nothing else. So don't flame me

Now, how can we pin that down

[cegras]

I wanted to bring these topics back to your intention, in case you either missed or intentionally ignored me.

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.

Also, it doesn't really matter if the card crashes on a 1500W or 550W.

If the OCP is, as you hypothesize, set to 82A maximum, then are you not exceeding design specifications by running this test? In this case, is it not the same as overclocking - do so at your own risk?

OK I downloaded it, set the settings, and ran it. Nice fuzzy red donut thingy waving around on my screen. Got bored of it after a few minutes and shut it off. How long is it supposed to be before it was to black out?

My 4870 is overclocked to 790 core too by the way (highest stupid CCC will allow me).

No black Screen here. Gecube 4870, and Sapphire 4870 1gb in Xfire. I didn't like how high the temps went up, in such a short time though!! That is an intense test.

To address the real world application of it... Any game that would put this kind of stress on hardware, would be out of business pretty quick, considering most of us here, have decent hardware, versus the masses... I dont think many Dell's or E-machines would last haha.
As for not being able to use the "full" potential of the card, it seems ATI designed the card for the 99.9% of people who will never hit this wall, they could have made the card impervious to this, by adding more pwm/vrm to the card..but at a cost. I don't see why the cost benefit analysis isn't surprising.. this condition will not surface in the wild, and for those that it does, its cheaper to RMA those few cards, than it is, to design change it to meet such strenuous rare occurances. That said, my cards worked fine, i just dont want to fry them for no reason :P

[lockee]

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'?

If the processor was under liquid helium, out of spec and caused a black screen I don't think anyone would declare the processor had a design flaw.

[W1zzard]

wether this might be OCP or OVP, which i didn't know about before today

those are very simple protections that have been added to recent voltage regulator designs. if an output stage mosfet blows (overheat) it is possible that it pulls the whole output stage to 12V = bad(tm). ovp protects against that

ocp is for cases where a short circuit occurs somewhere in the consumer circuit. short circuit = lots of current = bad(tm).

in a proper configuration ocp is set in a way that it is never triggered during "normal" use. apparently ati didnt do enough testing and set ocp too low.

in addition to those two, volterra has added overtemperature protection as well, which is triggered when an output stage gets too hot. there are some more fault triggers as well which are not important for us

[Hornet331]

If the processor was under liquid helium, out of spec and caused a black screen I don't think anyone would declare the processor had a design flaw.

indeed, but somehow people claim it will happen under normal conditions...