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

[Shintai]

Cause when you use it the way it's meant to be theres no problem....?

Your code produces a load behavior of the shadercores that will be never exist in the wilde (not on video encoding, not on GPGPU apps and not on games).

I would be careful claiming that. Specially after I seen it in EvE

[Greg83]

Or it can be argued
Games and applications or even drivers for the 4870 / 4970 have been programmed not to make use of the card the way it was meant to be, due to this known flaw that this tool has now uncovered

[grimREEFER]

this is pretty insane. the guy basically discovered that you cant overclock these cards, cause they arent stable at stock.

it might be planned obsolescence on the part of ati.

[Luka_Aveiro]

Cause when you use it the way it's meant to be theres no problem....?

Your code produces a load behavior of the shadercores that will be never exist in the wilde (not on video encoding, not on GPGPU apps and not on games).

So, this is ok, then?

Glad I haven't bought any of those cards...

[RaZz!]

I think we have a disconnect here. People don't use burn in programs to see how they stack against them but as a gauge of how their video card will perform in games. This is why you are not able to openly receive constructive criticism (based on how you've responded so far).

Since this is your new test method people will naturally pay close attention on the legitimacy of it based on what (and how) you decide to deliver the information.

And your car analogy is not correct. There are many cars that have a "top speed" based on the speedometer but aren't necessarily able to achieve it.

So in a nutshell we come to a question how these results effect the end user. So far, you haven't fully convinced (at least me) how we are negatively impacted in the games we play.

but isn't testing cpus with linpack or prime95 the same? you're saying the load produced by occt is "unrealistic", but isn't every benchmark that puts heavy load on a specific type of hardware "unrealistic"?

you can overclock your cpu and run windows and play games without any problems, but when running prime95 for a few minutes the test eventually aborts because of calculating errors. there are a lot of people who consider an overclock, that's not prime-stable, as a crap overclock. it didn't pass the testing methods, period.
it's the same with occt's gpu benchmark. it puts heavy load on the gpu in the same "unrealistic" way e.g. linpack does for cpus.

...and we're talking about gpus with stock clocks compared to overclocked cpus.

maybe other applications, e.g. the folding@home, for scientific calculations (especially in the future with opencl and such?) get more and more optimized and will reach these limits as well. who knows?

*edit* just tested it with my 4850 without any problems, so i can confirm that presumably only 4870/90 are affected. however, the gpu is getting really hot . even with the fanspeed of 75% it reached 80°C within a minute.

[Jamesrt2004]

That's something new

Try the very same test with a shader complexity 0 (this lower the load on the GPU by a lot). Does it launch ?

Try to monitor using RivaTuner : do you see the very same drop in the frequencies as we do record it ?

Pressing <esc> kills the App, i've always included that escape point. Maybe you could recover from that.

heya buddy sorry about late reply have to go hospital in morning etc... so was sorting that out..

having it on 0 WORKS (looks pretty cool i must say) about to try 1 then 2 then 3 see what works and what doesnt ..


-edit..-
ok
0 = works...
1 = reproduces the thing in OT
2 = same ..
3 = what i said before seems like it doesn't show up just black screen with mouse working and can exit into windows pressing esc.

[Manicdan]

It can't be a driver failure, because the test runs @ lower than stock clocks...

The chance that it is really a design flaw is very likely, you might want to think about considering it as a fact.

drivers failing have nothing to do with card stability. look into how amd chips can do a soft reset then tell me thats not what happened

[Eastcoasthandle]

but isn't testing cpus with linpack or prime95 the same? you're saying the load produced by occt is "unrealistic", but isn't every benchmark that puts heavy load on a specific type of hardware "unrealistic"?

you can overclock your cpu and run windows and play games without any problems, but when running prime95 for a few minutes the test eventually aborts because of calculating errors. there are a lot of people who consider an overclock, that's not prime-stable, as a crap overclock. it didn't pass the testing methods, period.
it's the same with occt's gpu benchmark. it puts heavy load on the gpu in the same "unrealistic" way e.g. linpack does for cpus.

...and we're talking about gpus with stock clocks compared to overclocked cpus.

maybe other applications, e.g. the folding@home, for scientific calculations (especially in the future with opencl and such?) get more and more optimized and will reach these limits as well. who knows?

No, because we know that many have used burn-in programs and achieved 0 problems to only discover stability problems in the games they play. Lets not forget that this is an overclocking venture that many enthusiast use to determine how well they are able to play the games/(non burn in) programs they use most. Not a self serving need to determine if this (or that) program works for them alone. Then based their decision on that program alone there something is wrong (or not) with the hardware.

[Greg83]

Don't games stress more then just the cpu though.
If your test only stresses the cpu and not the north bridge or ram, your games could easily have issues, even though your cpu test couldn't find them after an overclock.

Games don't make good stability tests either though.
Just the fact that FurMark was being throttled by drivers, to prevent something from being noticed shows there was something to hide and now its been found what was trying to be hidden.

[zanzabar]

any chance of a win7 compatible occt, i know that its kinda off topic

[Tetedeiench]

any chance of a win7 compatible occt, i know that its kinda off topic

Latest version is

[Tetedeiench]

heya buddy sorry about late reply have to go hospital in morning etc... so was sorting that out..

having it on 0 WORKS (looks pretty cool i must say) about to try 1 then 2 then 3 see what works and what doesnt ..


-edit..-
ok
0 = works...
1 = reproduces the thing in OT
2 = same ..
3 = what i said before seems like it doesn't show up just black screen with mouse working and can exit into windows pressing esc.

Ok, so you have the same problem as other people

If you underclock your card drastically, let's say 500/500 i'm sure Shader Complexity 3 will run just fine

And really, don't worry about the reply delay

[Eastcoasthandle]

Don't games stress more then just the cpu though.
If your test only stresses the cpu and not the north bridge or ram, your games could easily have issues, even though your cpu test couldn't find them after an overclock.

Games don't make good stability tests either though.
Just the fact that FurMark was being throttled by drivers, to prevent something from being noticed shows there was something to hide and now its been found what was trying to be hidden.

Greg83 you are on the right track. Hardware specific burn-in application can't and more often don't tell the who story of how stable your PC really is. Which is why we have seen those use them still having game related problems even though their burn-in program tells them otherwise.

However, I do disagree that games do make for good stability tests. Although it's not the best way to determine stability nor is it convenient having to reboot your PC. However, if you really want to know if your overclock can handle that game or application you will have to eventually test it using that specific non burn-in program/game.

So the question become more simplified IMO. If the hardware tested fails the burn-in program test yet works fine in the games and applications you play; what tangible effect did using that program have? IMO, none...

[zanzabar]

Latest version is

i cant get the gpu test to work

[Tetedeiench]

i cant get the gpu test to work

Don't forget to install DirectX9 from Microsoft's website. 7 is shipped with DirectX11, and DirectX9 is installed separatly (and not from Windows Update).

It's in OCCT's FAQ

[skycrane]

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....

whats the worst that could happen??? they fry the gpu, and RMA it... its not like ist never been done before.... expicaly wht the benchers on here

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. the only reason im not checking this right now, is that Milkyway, has not got off their high horse and 1. not got the new servers up and running, 2. not given the new code the CP so he can optimize it for FREE for the ati cards. i guess nvida is giving them to much money so they dont support thier competor... kinda like intel does ... thats right i said it

ive got a great idea then to all the ppl who dont believe what he is saying.... quit complaining on here with your words, and spend 10 min and see if you can get your card to do the same thing. then come back on here with something intelligent to either disprove, or approve of his methods .

[Tetedeiench]

Greg83 you are on the right track. Hardware specific burn-in application can't and more often don't tell the who story of how stable your PC really is. Which is why we have seen those use them still having game related problems even though their burn-in program tells them otherwise.

However, I do disagree that games do make for good stability tests. Although it's not the best way to determine stability nor is it convenient having to reboot your PC. However, if you really want to know if your overclock can handle that game or application you will have to eventually test it using that specific non burn-in program/game.

So the question become more simplified IMO. If the hardware tested fails the burn-in program test yet works fine in the games and applications you play; what tangible effect did using that program have? IMO, none...

Yet, if it takes 4 hours for the game to crash, and if the stability test will tell you in 4 minutes, which one will you choose ?

I mean, i'd rather be stable in the stability test, and be sure all my games will run fine, than test for 1h my favorite game, and only know for that my game lasted for that long without crashing.

Stability tests are great for telling you quickly how stable your system is. And i'd rather learn that quickly, instead of learning than in front of Sephiroth after a 2 hour long battle... aaaaaaaaaaand when the final cinematic launch... cccccccrashhh.

Yes, i was lvl 99 when i faced sephiroth. That's a silly example.

[Luka_Aveiro]

ive got a great idea then to all the ppl who dont believe what he is saying.... quit complaining on here with your words, and spend 10 min and see if you can get your card to do the same thing. then come back on here with something intelligent to either disprove, or approve of his methods .

enough said

further silence can be interpreted as positive answer, too, you know

[Mk]

ive got a great idea then to all the ppl who dont believe what he is saying.... quit complaining on here with your words, and spend 10 min and see if you can get your card to do the same thing. then come back on here with something intelligent to either disprove, or approve of his methods .

yeah
+1

[MsB]

I have a sapphire 4870 1 gig with the nicer than reference design heatsink design shown here , sorry its easier to just use newegg linky, http://www.newegg.com/Product/ShowIm...d%20-%20Retail although your test makes it hot as holy H it doesnt make it black screen. btw I used the settings you described. Test says gpu is at around 110 c but when I click it off ati sensor shows 85c btw, I am running my fan at full speed cause I dont want to damage card.

[Eastcoasthandle]

Yet, if it takes 4 hours for the game to crash, and if the stability test will tell you in 4 minutes, which one will you choose ?

I mean, i'd rather be stable in the stability test, and be sure all my games will run fine, than test for 1h my favorite game, and only know for that my game lasted for that long without crashing.

Stability tests are great for telling you quickly how stable your system is. And i'd rather learn that quickly, instead of learning than in front of Sephiroth after a 2 hour long battle... aaaaaaaaaaand when the final cinematic launch... cccccccrashhh.

Yes, i was lvl 99 when i faced sephiroth. That's a silly example.

The end result remains the same; you still will need to run that program or game. That's the focal point. Regardless if its 4 minutes or 4 hours

[Praxis1452]

The end result remains the same; you still will need to run that program or game. That's the focal point. Regardless if its 4 minutes or 4 hours

Regardless, he has found a very interesting bug or aspect of the design. That in fact, a program that can take full effective use of the gpu, will cause it to shutdown.

While as solely a stability testing utility, level 3 is not needed, it is an interesting thing to explore and should not be simply dismissed.

Personally I think this is quite an awesome discovery, not that it would stop me from purchasing a 48xx.

[RaZz!]

No, because we know that many have used burn-in programs and achieved 0 problems to only discover stability problems in the games they play. Lets not forget that this is an overclocking venture that many enthusiast use to determine how well they are able to play the games/(non burn in) programs they use most. Not a self serving need to determine if this (or that) program works for them alone. Then based their decision on that program alone there something is wrong (or not) with the hardware.

i really don't get why people here disregard Tetedeiench efforts so much. it's not like he wanted to produce that gpu failure, he stumpled upon it while creating a gpu stability test.
instead of investigating the issue everyone claims to know how unrealistic and far away "from real world" situations the benchmark is.

he indeed discovered a flaw in the design of the 4870/90 that leads to such behavior.

however, these are just my 2cents.

[Greg83]

This is why I run 48hour prime95 25.* 64bit blend test.

and atleast 12 hour of memtest86+ followed by hci memtest for another 12 hours and won't mention the other memtest I use, but it works much quicker

When overclocking anyway.

I'm happy with my q9650 @ 1.2Ghz and my 4870x2 at 100/200 for 24/7 and 400/200 when I play team fortress 2 maxed out though.

[BrowncoatGR]

Looks to me that this test is a power virus for ati cards and causes them to exceed specifications