Nvidia GTX 590 Reviews

If that 1,2 volts is true, then that's one hefty OV (~+28%) for default cooling and all. The VRMs shouldn't burn up though.

 

Had to do a recap, to see if I didn't lost count. It's actualy 8 reviewers so far, one of them reported also a dead card in a retailers' demo system.

sweclockers.com,TPU,hardware.fr,pclab.pl,lab501.ro (reports two),tbreak.com,tweaktown.com,itfiles.ro

Then there are some unconfirmed or undetailed reports, like here: http://forums.electronicarts.co.uk/battlefield-3-pc/1396168-gtx-590-has-arrived-3.html , or Fellix' pic above.

 

Posters on the above forum alledge some 590s have blown even at stock volts, something I've not heard anything about elsewhere so far. Has any of you heard of any such confirmed cases? I'd rate a B3D poster of somewhat higher credibility than some random battlefield kiddie on an EA board...

 

Hmm yes.. seems like this is list atm:
http://www.techpowerup.com/reviews/ASUS/GeForce_GTX_590/26.html
http://tbreak.com/tech/2011/03/zotac-gtx-590-review/3/
http://www.sweclockers.com/recension/13695-nvidia-geforce-gtx-590/16 (2 cards)
http://www.hardware.fr/articles/825-2/dossier-nvidia-repond-amd-avec-geforce-gtx-590.html
http://www.itfiles.ro/2011/03/nvidia-geforce-gtx-590-noul-rege-al-placilor-video/13/

And then those not mentioning it in their reviews, but some are praising the great OC potential Wonder how many more we have of those..
http://www.tweaktown.com/news/19192...590_why_some_have_gone_up_in_smoke/index.html
http://pclab.pl/news45334.html
http://www.xtremesystems.org/forums/showpost.php?p=4791289&postcount=37 (lab501.ro - 2 cards, no OV)

Some of them on the newer, recommended drivers, not 267.52. One could also say that those lucky ones who survived shows better OC results (due to lack of OCP) than retail drivers..

 

Wonder how long before Amd PR makes some negative PR about it
Figure they made fun of Sandy Bridge, why not this?

 

Yeah, those two reported by lab501.ro were not overvolted. itfiles.ro reports their blew at stock clock, so probably while being at higher voltage, like TPU's card. pclab.pl's voltage remains unknown, but they got 700mhz at stock I think, so they probably wanted to push higher with voltage.

Hardware.fr reports their blew during a stress test, no word about clocks or voltagest. You can ask Damien if you are interested.

 

hmh? All but sweclockers happened on 267.71's

 

I'd say the GF FX 5800 Ultra is still worse, but not by much.

 

NV apologists seem quick to point out most if not all cards that have blown did so while overvolted, however this is a hardcore market product, if you gotta drive the darned thing like a little old lady on her way to church for fear of the DC regulators blowing up then something's very wrong.

NV should never have released a cutting-edge product that's as fragile as this. It's not as if reviewers haven't ever overvolted hardware in the past without smoke and flames shooting out of it...

 

I'm of the opinion overvolting anything is friggin' dumb. It's asking for trouble. So, I'm not surprised cards are blowing up because of that.

What does surprise me is that cards are blowing up when not overvolted, or even overclocked. That is what says to me that this card should not exist.

 

The board wasn't overclocked. I was measuring power draw of a couple of boards and due to a bug the GTX 550 Ti review driver (267.59) ended up running with the GTX 590...

 

So the OCP was not working with the driver - that would explain it, thanks.

 

I agree, but my post further up above highlighted the only case I knew of with the exact voltage. Now, I am not against careful tweaking of the cards including more juice. It was just the amount or more, that I think in that particular case was a bit optimistic. It probably was for a good reason Nvidia used screened chips with very low operating voltage - something which AMD apparently did not have to resort to on HD 6990. While 1.2 volt is only a bit on the high side for my personal taste on a regular 570 or 580, it is, as I've said, around 28% above default for GTX 590. For that amount of OV, I would not say the card was fragile by design. Again, this is only about this one case at TPU and no, it should have shut down in time, given a sensible board design and not reliying on drivers to do that.

That further highlights the superiority of a hardware-based solution to excessive powerdraw. AMD apparently is 1.5 generations ahead, albeit not flawless either.

 

I actually think that 1.2v is quite high, but a card like this should be built like a rock with a lot of head room. 150-200$ card is another story. I feel like the entire 5x0 series, atleast the bigger ones have suffered from some quality issues. Lots of cards being broken, including mine. My card broke with 1.06v. Warranty covered it though.

Maybe they have taken the cost cuttings too far.

 

Just to get some perspective, a realitevly similar increase for HD6970 for example would result in about 1,5 Volt. That's clearly out of the range I would endorse to even try.

 

Carsten, both cayman and fermi are manufactured at the same process by the same company, likely in the very same fab and engineered to the same tolerances. So don't get overly worked up by percentages - they don't really tell an accurate story while absolute numbers do.

Fermi is run at really low default volts due to its monstrous power consumption, not because the chip would neccessarily be damaged by higher volts. I'm wagering it could manage just fine with the same volts as cayman, except it would mean the chip's power and heat dissipation envelopes burst through the roof. Hence why NV didn't do that.

 

Not to mention that only one reviewer as far as I know went over ASUS's reviewers guides limits, in fact I think it was the only one that even went over 1V, and nVidia told about the new suggested limit after everyone had already blown their cards.

 

Grall,
To what extent AMD and Nvidia followed TMSCs design rules, we cannot possibly know.

Absolute numbers don't tell anything in this case for exactly one reason - at least as far as I am aware of: The GPU didn't blow but the power circuitry. And that is engineered towards the default level plus a certain amount of tolerance.

 

My point was, since the GPU is fine with a certain amount of volts that will encourage people (particulary those buying a bleeding edge product like the 590) to want to run their GPUs at that volt.

And since power follows volts with a certain relationship, the power circuitry should be properly dimensioned to handle that situation. If not, the VRM should be physically incapable of being set at a volt that would draw more amps than the unit is capable of delivering.

Either way one twists and turns this one particular issue it all comes back to "Nvidia fucked up".

 

Newegg isn't even listing the 590 anymore. I could be wrong, but don't they usually just have something as out of stock with a auto-notify-button?

Could Nvidia be re-engineering a second batch?