4870/4890 design flaw?

[Wreckage]

http://www.xtremesystems.org/f...howthread.php?t=224958

Anyone experience this?

 

[thilanliyan]

Never experienced this on my reference card. Mine stays at about 77amps in Furmark with it renamed even when overclocked to 800/1000. And of course during games it doesn't get anywhere near this.

I'll test with OCCT but if overclocked furmark testing doesn't bring it above the 82A mentioned, then I doubt another test would especially at stock speeds as the poster claims.

EDIT: Another thing in my mind is why is it not crashing the 4850 cards? Considering it has the full 800 shaders and is just at a reduced frequency and I figure they would be using a "lesser" VRM and/or OCP design. If this is just OCP kicking in then it isn't a design flaw IMO.

Also, this bit is interesting from one of the French articles...I don't know what to make of it as I don't know much about coding:

"It is completely unacceptable that a graphics card (or processor) is unable to deal with a classical code (whatever it is) without crashing. Because otherwise it would curb the developers so they do not create a code optimized too which could crash the GPU."

Any thoughts?


EDIT2: I was able to replicate the findings with OCCT:GPU but this was with the card overclocked. I'd be really pissed if my games did this but they don't so I'm not really sure what to say about it.

 

[apoppin]

Originally posted by: thilan29
Never experienced this on my reference card. Mine stays at about 77amps in Furmark with it renamed even when overclocked to 800/1000. And of course during games it doesn't get anywhere near this.

I'll test with OCCT but if overclocked furmark testing doesn't bring it above the 82A mentioned, then I doubt another test would especially at stock speeds as the poster claims.

i have

Just run FurMark with the renamed .exe

You will not believe the temperature spike [well over 110C] before it all crashes
:Q

i do not see the big deal. It is technically a "design flaw" - but the benchmarks *exploit* this; you will [probably] never see this in a game

[well maybe you will; twiimtbp
- mandatory in ALL their games - from now on]

:shocked:

 

[thilanliyan]

Originally posted by: apoppin

Originally posted by: thilan29
Never experienced this on my reference card. Mine stays at about 77amps in Furmark with it renamed even when overclocked to 800/1000. And of course during games it doesn't get anywhere near this.

I'll test with OCCT but if overclocked furmark testing doesn't bring it above the 82A mentioned, then I doubt another test would especially at stock speeds as the poster claims.

i have

Just run FurMark with the renamed .exe

You will not believe the temperature spike [well over 110C] before it all crashes

Yes I said in my OP I renamed the exe and ran with it overclocked and it never broke 77A in Furmark. And the article says it's not a temperature issue (mine is watercooled so I can't test for temperature spikes).

 

[ronnn]

Question for the op, have you experienced this? Or did you finally read the thread on b3d that dried up ages ago?

 

[apoppin]

Originally posted by: thilan29

Originally posted by: apoppin

Originally posted by: thilan29
Never experienced this on my reference card. Mine stays at about 77amps in Furmark with it renamed even when overclocked to 800/1000. And of course during games it doesn't get anywhere near this.

I'll test with OCCT but if overclocked furmark testing doesn't bring it above the 82A mentioned, then I doubt another test would especially at stock speeds as the poster claims.

i have

Just run FurMark with the renamed .exe

You will not believe the temperature spike [well over 110C] before it all crashes

Yes I said in my OP I renamed the exe and ran with it overclocked and it never broke 77A in Furmark. And the article says it's not a temperature issue (mine is watercooled so I can't test for temperature spikes).

they are not the same - other than the results - crashing the card
- Furmark can actually permanently damage the card if you keep it going and it doesn't crash

but you can look at these "benchmarks" as a way to exploit a card's weaknesses

 

[thilanliyan]

Originally posted by: apoppin
they are not the same - other than the results - crashing the card
- Furmark can actually permanently damage the card if you keep it going and it doesn't crash

Yeah I think we've all heard about how much Furmark stresses a card. I figured that was what the OP was about but I definitely have not been able to replicate in Furmark what they claim about the OCP kicking in at stock.

EDIT: I was able to replicate the findings with OCCT:GPU but this was with the card overclocked. I'd be really pissed if my games did this but they don't so I'm not really sure what to say about it.

 

[sandorski]

PSA: Every GPU and every CPU has many Flaws.

 

[soccerballtux]

Originally posted by: alkemyst

Originally posted by: jandlecack

Originally posted by: toyota

Originally posted by: jandlecack
Uh, regardless of his bias, or what Nvidia has done in the past, he still has every right to point this out.

and I have every right to point out WHY he is doing it. his prior posting history clearly indicates that this is attempt to put down ATI. he doesnt care about this issue other than to flame ATI.

No, you don't really, and neither do I, because we're both off topic.

To be perfectly honest I don't give a damn who is biased for what as long as something is brought to my attention that can affect my purchase, and potentially that of others.

I would say the same thing if it was the other way around and it was to "put down Nvidia". A flaw, in case it truly is one, is a flaw, and I don't want information held back because you believe that someone has biased reasons for posting it.

Quite frankly, shut up if you can't dispute or add anything. No offense.

as a noob you may want to sit back and get a feel for the mentality of those you are defending.

All I know is flaws or not the 4890 is a superior card to most of the GTX lineup.

Yeah, this in no way affects any purchase choices of mine. No card is ever going to be stressed to the max with a game like with the fuzzy donut.

Further, I can burn out my 9800GT at stock settings with this fuzzy donut, but do I do it? No, because that's retarded, just as retarded as if I let it run for 10 hours, and the blame Nvidia for ruining my card because they didn't design it right.

 

[dguy6789]

I've run the OCCT3.1 GPU test for 12 hours full screen without a crash on my 4870. Mine is an original reference Sapphire 512MB model with a blue pcb.

 

[apoppin]

Originally posted by: thilan29

Originally posted by: apoppin
they are not the same - other than the results - crashing the card
- Furmark can actually permanently damage the card if you keep it going and it doesn't crash

Yeah I think we've all heard about how much Furmark stresses a card. I figured that was what the OP was about but I definitely have not been able to replicate in Furmark what they claim about the OCP kicking in at stock.

EDIT: I was able to replicate the findings with OCCT:GPU but this was with the card overclocked. I'd be really pissed if my games did this but they don't so I'm not really sure what to say about it.

on of my friends ran the test with his 4870-1GB and lost his installation of Win7
- i think he said the complexity was "eight"

it blacked and crashed instantly

i told him he should have had a drive image
- now he is installing all his applications again

 

[WaitingForNehalem]

This reminds me of the TLB bug. Rare, but possible to make it happen.

 

[error8]

It doesn't happen to my card and I have the stock cooler and the voltage bumped a bit. Neither OCCT GPU test or Furmark crashes my card. It might be related to my fan profiles which are a little agressive.

 

[n7]

Interesting.

That said, as since i've gotten my HD 4890, i've not had a single issue with it.
This is very unlike the constant stream of drivers issues with my GTX 280.

Therefore, i'm going to happily continue using my potentially "flawed" hardware.

 

[thilanliyan]

Originally posted by: error8
It doesn't happen to my card and I have the stock cooler and the voltage bumped a bit. Neither OCCT GPU test or Furmark crashes my card. It might be related to my fan profiles which are a little agressive.

Did you up the complexity and you're using the newest OCCT? I put it to 3 and as soon as the amps went above 83 it black screened and I had to reboot.

It barely goes above 60A during gaming though so I don't think it really matters for me. I won't be running that again though after Apoppin said someone hosed their Win7 installation trying it.

 

[error8]

Originally posted by: thilan29

Originally posted by: error8
It doesn't happen to my card and I have the stock cooler and the voltage bumped a bit. Neither OCCT GPU test or Furmark crashes my card. It might be related to my fan profiles which are a little agressive.

Did you up the complexity and you're using the newest OCCT? I put it to 3 and as soon as the amps went above 83 it black screened and I had to reboot.

I'm using 8 and nothing bad happens, no black screen, no artifacts, no overheating. Maybe if I keep it running for 3 hours, might be a different story. But I don't have all that time and doesn't look like it's going to matter at anything if I run it that long.

EDIT: I have 71 C gpu temperature, 91 C VRM and 83 Amps on the vrms, using shader complexity 8. Nothing happens. Maybe it's temperature related, even though they say it isn't. Increase the fan speed and see if it happens again.

 

[thilanliyan]

Originally posted by: error8
I'm using 8 and nothing bad happens, no black screen, no artifacts, no overheating. Maybe if I keep it running for 3 hours, might be a different story. But I don't have all that time and doesn't look like it's going to matter at anything if I run it that long.

Apparently highest complexity isn't always the most stressful. Also, what res are you running? Lol, I'm trying to get your comp to crash.

Don't listen to me...even I won't run that test anymore.

EDIT: I have 71 C gpu temperature, 91 C VRM and 83 Amps on the vrms, using shader complexity 8. Nothing happens. Maybe it's temperature related, even though they say it isn't. Increase the fan speed and see if it happens again.

The core on mine is watercooled with the VRMs staying around the same temp as yours.

 

[gdextreme]

I'd like to test this on my 4850. My core is ocd to 720mhz and RAM to 1100 mhz so its almost a 4870. Is OCCT this version : http://www.softpedia.com/progD...U-Download-123888.html

 

[thilanliyan]

Originally posted by: gdextreme
Is OCCT this version : http://www.softpedia.com/progD...U-Download-123888.html

No it's this:
http://www.ocbase.com/perestroika_en/index.php?Download

 

[nitromullet]

One of the posters in that thread http://www.xtremesystems.org/f...=3799223&postcount=259 makes some interesting points:

If the card can be made to draw current in excess of either its supply capability or its current protection threshold using nothing other than normal API commands, there's a design problem. There's nothing wrong with the chip, so maybe the title of the thread is misleading, but the reference cards' power supplies need some work. That being said, this was probably an intentional action. It makes sense to aim for above the middle as it'll provide compatibility with real life games. Making a design that is good for the upper 1% of applications that aren't widely used at an extra cost doesn't make a whole lot of business sense.

They pretty much touched on all the relevant points IMO.

 

[Qbah]

So you have to worry only if you load the card in a very specific way using this one program only? It seems no game or other application can stress the cards in a similar manner. This is not an issue at all imo.

It's like you take your Ford Fiesta (cause you can!) and do some off-road driving and then complain your suspension died way too early or some other thing was damaged

 

[BFG10K]

Guys, either address the argument or don?t post. I?ll start dishing out individual warnings if personal attacks continue.

I've removed off-topic posts too.

Video Mod BFG10K.

 

[Hauk]

Originally posted by: Qbah
So you have to worry only if you load the card in a very specific way using this one program only? It seems no game or other application can stress the cards in a similar manner. This is not an issue at all imo.

It's like you take your Ford Fiesta (cause you can!) and do some off-road driving and then complain your suspension died way too early or some other thing was damaged

My thoughts exactly. "Normal use" is what I'm concerned about.

Any chance I can borrow your Ford Fiesta though? Sounds like fun... :evil:

 

[Keysplayr]

I don't see how any of this matters if it doesn't affect any games. I mean, these cards are bought to play games, not run furmark.

 

[Candymancan21]

This is how i see this whole "design flaw" This OCCT/Furmark isnt a video game, there is no single player or multiplayer. I dont run around shooting people and building a city, so i could give a rats ass if it fails on that stupid test. It doesnt do this while gaming so why would you actually care ?

You guys are going nuts over something stupid.