NVIDIA GeForce GTX 580

[Rob Williams]

NVIDIA launched its first Fermi-based GPU earlier this year in the form of the GeForce GTX 480, and it was met with mixed reception. Until now, it's been the fastest single-GPU offering on the market, but certain downsides kept it from being the first-choice of many. Does NVIDIA's first proper follow-up fix all that was wrong?

Read the rest of our GTX 580 look and discuss it here!

 

[looisboo]

AMD 6990 delayed to 14th december so now wait longer if you want price war.

 

[Doomsday]

now thats what i'm talking about!

 

[Unregistered]

Unigine minimum framerates of the GTX480

Hi, referring to this page:
http://techgage.com/article/nvidia_geforce_gtx_580/11

Do you guys know why the GTX480's minimum framerates are so much higher than everything else especially at the 1680x1050 and 1920x1080 resolutions?

I'd like to know if this is real or just a measurement artifact.

 

[Rob Williams]

Hi, referring to this page:
http://techgage.com/article/nvidia_geforce_gtx_580/11

Do you guys know why the GTX480's minimum framerates are so much higher than everything else especially at the 1680x1050 and 1920x1080 resolutions?

I'd like to know if this is real or just a measurement artifact.

There should be a reason behind it, because our tests are run at least twice with the results averaged out. Unfortunately, this is something I can't immediately re-test since our GTX 480 sample died within the last month. Been trying to get a replacement in, but it's kind of tough given that NVIDIA no longer cares about it.

Good catch, though... I need to figure out an answer to this.

 

[Unregistered]

There should be a reason behind it, because our tests are run at least twice with the results averaged out. Unfortunately, this is something I can't immediately re-test since our GTX 480 sample died within the last month. Been trying to get a replacement in, but it's kind of tough given that NVIDIA no longer cares about it.

Good catch, though... I need to figure out an answer to this.

How did your GTX 480 "die"? I had one that went bad after a month. It randomly started to lock up and freeze without any software changes done to the system.

Anyway back to the question. Yeah here's hoping you can find out the gremlin behind this. There's a user on several sites who's been linking to this review and making a heck of a fuss out of this. My thought is that it's a measurement bug or something because the GTX480 is just way higher than anything else in those graphs. But I figure I should go to the source and see if we can find out for sure.

 

[Rob Williams]

How did your GTX 480 "die"?

That's a great question - another one I don't have an answer to, hehe. It happened just over a month ago I guess, when I was doing a slew of re-benchmarking to update our GPU test suite. Testing on that card was done (thankfully), and then I turned off the machine for a bit. I returned to it later to see Windows go nuts:

http://forums.techgage.com/showthread.php?t=6374

There's no physical damage on the card,and I tore it down to give it a good cleaning and replaced the thermal paste, and no cigar. It's been a while since I last touched it, so I might test it out again soon. I've had GPUs come back to life in the past, so who knows (had a Radeon card last year cease booting, then a month later, it worked again).

That aside, thanks for bringing this to my attention. It's something I noticed before but didn't think too much about, but now that the GTX 580 is out and doesn't quite match up, it's something to investigate. I'll see if I can press NVIDIA again to sending a replacement card and see if I can get it taken care of. I'll re-test the "dead" one as well again soon in case it somehow came back to life.

 

[evilives34]

rob the oven trick has worked for me on a old 8800gts i had

i think the reason it does work from what i read is that micro cracks in the solder causes the problem and putting it in the oven remelt the solder closes the cracks

 

[Rob Williams]

I am going to give the card another go tonight or tomorrow, and if it doesn't just boot up, I'll try the oven trick. If the card is indeed dead, it's not going to hurt anything to try.

It'd be kind of ironic, though. Some people would consider the GTX 480 itself an oven ;-)

 

[Doomsday]

rob the oven trick has worked for me on a old 8800gts i had

i think the reason it does work from what i read is that micro cracks in the solder causes the problem and putting it in the oven remelt the solder closes the cracks

holly crappamolly! It actually works, i'm gonna try it with my dead 9600GT!!!

 

[evilives34]

if its dead it doesnt hurt to try

 

[Unregistered]

Rob,

Any updates?

 

[Tharic-Nar]

Rob,

Any updates?

He's away for a couple days (weekend and monday), i'll bug him for an update when he gets back.

 

[Relayer]

I'll betch'a if he tells nVidia he'll try and bake his 480 back from the dead unless they'd rather replace it he'll get a new card. That's all they need is another "Bumpgate".

 

[DarkStarr]

I need an older Nvidia card, something like an 8800GT or better for folding since my card can only use ~300 shaders thanks to the client. Oh and the oven stuff does work but isn't permanent unless you can reduce temps on the card so the same thing doesn't happen again.

 

[Rob Williams]

I'll work on getting this done ASAP. Came back from the trip sick as a dog, so will do it as soon as I can bend over and not feel like falling down.

 

[Kougar]

Hi, referring to this page:
http://techgage.com/article/nvidia_geforce_gtx_580/11

Do you guys know why the GTX480's minimum framerates are so much higher than everything else especially at the 1680x1050 and 1920x1080 resolutions?

I'd like to know if this is real or just a measurement artifact.

Hey Unregistered. Rob pointed out this thread to me and asked if I could test this for him.

Using a stock Core i7 980X and my GTX 480 FTW set to stock 480 reference clocks in order to match his test configuration, I was able to get some rather definitive results:

​

Something I observed when buying my 480 is that in almost all of the DX11 tests the 480 tended to have much better min FPS rates versus the 5000 series, in addition to the edge in average FPS performance. I am more surprised that the 580 loses this edge though, I'd have to chalk it up to some peculiarity with this test given I don't know of any technical-side explanations for it.

 

[Rob Williams]

Thanks a ton for getting this test done man. I'm glad to see that the min FPS is on par with what I recorded with the reference GTX 480. I'm still not quite sure the reasoning behind the GTX 480 scoring higher than the GTX 580 in that department. I'll have to ping NVIDIA about it.

 

[Kougar]

Well one thing I also noticed... when I tested my card with a heavy overclock, say 851MHz instead of the 700MHz stock reference, the min FPS actually plummeted to ~23 while the max/average increased. I have no idea why given other programs don't reflect similar behavior that I've noticed. For that reason I don't think it's a power issue internal to the card, but that only leaves a peculiarity with the test....