GTX580 reviews thread

[bryanW1995]

No, Online retailers are gouging. MSRP for the 580 is 499.00.
Newegg is mega gouging.

hopefully they're gouging b/c they think they can rather than b/c the cards are limited availability.

 

[MentalIlness]

http://www.hardwarecanucks.com/foru...ws/37789-nvidia-geforce-gtx-580-review-8.html

1150MHz core speed with voltage tweak? Can't wait for that review...

Repost. See above

 

[exar333]

@MentalIlness
the 580 either scales horribly in SLI (with current drivers) or its CPU bottlenecked.
That might change in the future with new cpus comeing out, and with nvidia haveing time to work on SLI drivers.

Definitely. Let's make this decision based on one resolution run, from one game, on a brand-new card. Good research method....

 

[Obsoleet]

On 11/9/10, you could go as far to say the 580 is a good GPU.

As soon as the 6970 is here, I believe everyone will look back and see how marginal of an improvement the card really was. It's hard for reviewers to deny the state of affairs today, but I'm sure they are all sitting there with a 6970 with a smirk.

It's a small step for Nvidia.. but they are playing catchup on thermals, still not there yet. It had to be released now, so it got some glory days and sales to at least the NV diehards. There's no point in trying to release it after the 6970s..

 

[notty22]

Here is Linus taking apart a 580 down to the Vapor Chamber.
Very cool. http://www.youtube.com/watch?v=ncEU1mrN2oc&feature=channel

 

[bryanW1995]

Really you think every rumor springs from truth ? You sure were fast to jump on the rumor that nv's price cuts were not permanent really fast. You set that record straight right quick. Guess it's only the rumor you prefer to be true that have that basis of some truth.

Can you field a question to nv for us ? Ask them why they've named a GTX 480 refresh as a brand new series ? The 580 is just the 512SP 480 we've been hearing of for months now. Why the weird naming ? Have you got your free pair of GTX 580s from nvidia yet ? Or did they not bother with the focus group process for this launch as the card is really just a refresh not a new series as it's name denotes.

Perhaps in your capacity as a mod you can lock the umpteen 580 threads and create the stickied 580 review thread for us all as we saw with the 6870/6850 launch. Thanks.


Yes the 6970 sure is looking to be an exciting launch when it comes out this month. Brand new architecture. Fastest single-gpu. Big improvements in crossfire scaling that will make it the multiple-card champ as well. Then there is the 6990 in December to look forward to.

no need to get hostile on him groove. also, you know he's not a video mod. I already pm'd idc and bfg to ask them to sticky the review thread, I'm sure I'm not the only one.

 

[Paratus]

So did anyone read the [ H ] review?

What struck me was how well the 5870 2GB held up against the 480. It also basically tied the 580 in F1 2010 (lost by 1.2 FPS avg at 25x16 8X TR MSAA).

The 580 is a very nice card, but I've come to the conclusion that due to the architectural differences, (scalar vs vec5 or 4) no similar generation card will win every bench.

 

[Daedalus685]

Give me a break. This isn't AMD Zone here.

If 20% performance improvement, lower power usage, cooler temps, and quieter operation is just a "refresh" then I want EVERY product to get refreshes all the time.

The fanboy comments on the forum here are getting rediculous. The 580 will only get better with driver updates, and it's already within spitting distance of the 5970. I seem to remember when we said a new card at the speed of the last gen's 2xGPU was a GOOD thing. Sure it's not perfect, but its a definite improvement over the 480.

Its naive and foolish to compare this to a 280->285 refresh.

It is an amazing product.. but how is it not a refresh akin to the 280 > 285?

The only difference was that the 480 had many more issues than the 280 that needed refreshing.

What we have is a chip that took the 480, enabled all of the cores and increased the clocks while correcting power issues. It seems to fit the definition of refresh to me, and is very similar to the jumps we used to see in the 7800 to 7900 days.

I suppose this is both a refresh and a repair at the same time, as it corrected faults while increasing clocks and whatnot. But the feeling that this is "more than a refresh" is merely reflective of how many issues bothered us with the 480, and how mundane refreshes have been the last couple of generations (generally only targeting a shrink to reduce power and increase clocks and not increase the shading power like this).

 

[Stoneburner]

So.... next question, is FUD right about a dual gpu card version of these? It looks like a downclocked, lower voltage dual 580 gpu would actually work unlike dual 480's.

The reviews i've checked, I haven't seen TRI SLI either.

 

[bryanW1995]

On 11/9/10, you could go as far to say the 580 is a good GPU.

As soon as the 6970 is here, I believe everyone will look back and see how marginal of an improvement the card really was. It's hard for reviewers to deny the state of affairs today, but I'm sure they are all sitting there with a 6970 with a smirk.

It's a small step for Nvidia.. but they are playing catchup on thermals, still not there yet. It had to be released now, so it got some glory days and sales to at least the NV diehards. There's no point in trying to release it after the 6970s..

I dunno, 6970 info has been awfully scarce here lately, and amd doesn't appear to have anything in channel yet so that they can pull up the launch a few weeks like they did with 48x0.

 

[Daedalus685]

So.... next question, is FUD right about a dual gpu card version of these? It looks like a downclocked, lower voltage dual 580 gpu would actually work unlike dual 480's.

The reviews i've checked, I haven't seen TRI SLI either.

I really really doubt we would see two 580s on a board unless it were something rare from ASUS.

While the power/performance is great in these it still consumes more power than a 5970, which at the time many wondered if it would fit in the power envelope itself.

If there is a dual GPU card I'd expect it to be full gf104s (aka 560) for something akin to 470 SLI performance just squeezing under 300W. Blind guess of course.

 

[Obsoleet]

So did anyone read the [ H ] review?

What struck me was how well the 5870 2GB held up against the 480. It also basically tied the 580 in F1 2010 (lost by 1.2 FPS avg at 25x16 8X TR MSAA).

The 580 is a very nice card, but I've come to the conclusion that due to the architectural differences, (scalar vs vec5 or 4) no similar generation card will win every bench.

I noticed that. I pretty much only read H's reviews and a random selection of other sites on launch. H is the best, most unbiased site out there IMO.

 

[Grooveriding]

no need to get hostile on him groove. also, you know he's not a video mod. I already pm'd idc and bfg to ask them to sticky the review thread, I'm sure I'm not the only one.

I know my posting style can seem hostile at times. New launches are usually the only time to point out bias that has nothing to do with tangibles and is just brand loyalty. Was just pointing out some inconsistencies.

580 is nice taken as what it is, GTX 480 as it was meant to be.

 

[notty22]

I'm not saying they can or will do it. But dual 580's with this cooling approach and the engineering polish they are using on the electrical circuitry, I think they could do it.
They have mentioned explicitly the monitoring of voltage draw on the 6 - 8 pin and the pull through the m/b. Its able to level off and throttle where necessary. This is to address those melted m/b 24 pin cables in SLI that popped up in the extreme useres.
This gives me a little inkling that 600mhz 580's might be the ultimate trump card
Just some conjecture.

 

[RussianSensation]

@Russian, i was aware AMD used to driver hack and throttle in Furmark (and coped a "shellacking", well deserved). I wasn't aware they still did it. Do they?? Cos my CF 5850s heats up nicely in Furmark well above any gaming scenario.

Yes, it will still heat up much more than in any game. However, without the throttling AMD implemented, you very well may have ended up with blown VRM/MOSFETs that countless GTX275s/HD4870s had experienced. As a result AMD decided to protect its customers by lowering the power levels so that the newer HD58xx cards can never reach 100% maximum levels in Furmark. That's a good thing.

 

[BathroomFeeling]

Silly/obvious question, but do those Furmark throttles affect framerates?

 

[RussianSensation]

Silly/obvious question, but do those Furmark throttles affect framerates?

Yes, definitely; the Furmark benchmark score falls when the card starts throttling. How would you measure it though since there is no way to disable hardware PowerPlay in HD58xx series? So you wouldn't be able to arrive at the non-throlled score. :awe:

I think reviews providing 2-3 scores (i.e., Left 4 Dead 2, Crysis, Furmark, GPGPU like Folding@home) allows readers who like Furmark scores to get the information they want, while giving others real world application power consumption figures. I just think the videocard's power consumption shouldn't be unilaterally measured by FurMark. A lot of times, you'll see members comparing videocard's power consumption with FurMark's numbers only. I don't think placing more emphasis on a theoretical power virus over power consumption in games or even GPGPU computing is the right approach.

If Cayman isnt at least 30% faster it will be a failure

Cayman on the other hand should be a full refresh, not just a "tweaked" 5870, so personally I am expecting ~50% more performance, specially in DX11

50% faster than HD5870 or 50% faster than GTX580? Considering GTX480 is ~ 20% faster than Hd5870 and GTX580 is another 15-20% faster, GTX580 is about 35-40% faster than the HD5870. So you think HD6970 should be between 85-90% faster than HD5870 on the same 40nm process? That's not realistic. 50% faster than HD5870 is possible though.

You embarress yourself by posting that. Not that most folks don't write off your posts immediately anyways. You made endless threads ranting about 460 SLI being the second coming. Not that any forum goers here will be surprised by this considering it is nv.

It's true that happy medium did recommend 460 SLI; and so did many of us. Of course unlike HD5970 with 10.10c drivers, GTX460 SLI or HD6870 CF actually work the way they were intended to. Even with 10.10c drivers, HD5970 CF scaling/performance is sub-par compared to either of HD6870 CF or GTX460 SLI in majority of modern games. HD5970 is a fast card, but when you can get HD6870 CF for the same amount of $, unless you only have 1x 16x PCIe slot, then HD5970 doesn't look so good.

I noticed that. I pretty much only read H's reviews and a random selection of other sites on launch. H is the best, most unbiased site out there IMO.

1) The only review on the Internet where HD5870 is only 2.2 fps slower than GTX480 in Civilization 5.
2) The only review on the Internet where HD5870 is only 1.4 fps slower than GTX480 in Metro 2033.
3) The only review website on the Internet that thinks MLAA > MSAA by deriving their conclusion after testing it in 1 game.
4) The only review on the Internet that chooses to use Medal of Honor console port where GTX285 is only 2 fps slower than HD5870 and F1 2010 instead of more demanding games like AvP, STALKER:CoP, Lost Planet 2, Arma 2, etc when testing $300+ modern graphics cards.

Yes definitely, the best review website.

 

[Idontcare]

can we get a sticky?

Done.

Moderator Idontcare

 

[BathroomFeeling]

Thanks!

 

[nitromullet]

AT's is up too...

http://www.anandtech.com/show/4008/nvidias-geforce-gtx-580

 

[ShadowOfMyself]

50% faster than HD5870 or 50% faster than GTX580? Considering GTX480 is ~ 20% faster than Hd5870 and GTX580 is another 15-20% faster, GTX580 is about 35-40% faster than the HD5870. So you think HD6970 should be between 85-90% faster than HD5870 on the same 40nm process? That's not realistic. 50% faster than HD5870 is possible though.

I was comparing it to the 5870 of course... 50% faster than GTX480 is something only a fanboy would dream of

And btw, looking at the computerbase ratings (I love that they do this, spares us the trouble), the 580 is actually 25 to 30% faster, except at 2560x1600 where it takes off because of 1.5GB memory, obviously

Also cant help but think Nvidia rushed the 580 out for a reason... Anyone who buys one now instead of waiting is either rich, or a fanboy

 

[RussianSensation]

And btw, looking at the computerbase ratings (I love that they do this, spares us the trouble), the 580 is actually 25 to 30% faster, except at 2560x1600 where it takes off because of 1.5GB memory, obviously

This is what happens when a review actually includes demanding games instead of Medal of Honor and F1 (HardOCP) or Call of Juarez 2, Dawn of War 2, Unreal Tournament 3 (TPU). Computerbase arrives at GTX480 and GTX580 as 36% and 58% faster, respectively, than the HD5870 at 2560x1600 8AA. I am sure Kyle Benett doesn't agree with that review. Computerbase shows DX9 vs. DX10 vs. DX11 performance averages too.

 

[Obsoleet]

1) The only review on the Internet where HD5870 is only 2.2 fps slower than GTX480 in Civilization 5.
2) The only review on the Internet where HD5870 is only 1.4 fps slower than GTX480 in Metro 2033.
3) The only review website on the Internet that thinks MLAA > MSAA by deriving their conclusion after testing it in 1 game.
4) The only review on the Internet that chooses to use Medal of Honor console port where GTX285 is only 2 fps slower than HD5870 and F1 2010 instead of more demanding games like AvP, STALKER:CoP, Lost Planet 2, Arma 2, etc when testing $300+ modern graphics cards.

Yes definitely, the best review website.

It is. HardOCP / Brent Justice are the real deal. I would honestly bet it's the other sites with malevolent intentions, rather than [H]. There are certainly obviously biased sites out there, both ways, than HardOCP.

 

[Skurge]

This is what happens when a review actually includes demanding games instead of Medal of Honor and F1 (HardOCP) or Call of Juarez 2, Dawn of War 2, Unreal Tournament 3 (TPU). Computerbase arrives at GTX480 and GTX580 as 36% and 58% faster, respectively, than the HD5870 at 2560x1600 8AA. I am sure Kyle Benett doesn't agree with that review. Computerbase shows DX9 vs. DX10 vs. DX11 performance averages too.

I agree if those results were done with a 1GB 5870.

 

[exar333]

It is an amazing product.. but how is it not a refresh akin to the 280 > 285?

The only difference was that the 480 had many more issues than the 280 that needed refreshing.

What we have is a chip that took the 480, enabled all of the cores and increased the clocks while correcting power issues. It seems to fit the definition of refresh to me, and is very similar to the jumps we used to see in the 7800 to 7900 days.

I suppose this is both a refresh and a repair at the same time, as it corrected faults while increasing clocks and whatnot. But the feeling that this is "more than a refresh" is merely reflective of how many issues bothered us with the 480, and how mundane refreshes have been the last couple of generations (generally only targeting a shrink to reduce power and increase clocks and not increase the shading power like this).

GTX 480 to 580 is not very much like 280 to 285 at all. The GTX 285 had all of the same internals of the 280, and the only differences were the clocks bumped up 10% and 55nm instead of 65nm. It was more of a die-shrink with a small bump in clocks.

For those who had 280s and 285s (myself included - I had 2x280s and 2x285s due to free step-up at the time) they were exactly the same. The 280 was already very overclockable, and I actually regretted "upgrading" because I lost over-volting ability due to NV using a cheaper voltage tech on the 285. Also, it wasn't really any more power efficient either (same with 55nm 260s if you recall).

The 580 not only is speed-bumped, but has more architecture enabled. The amazing thing is this WASN'T a die-shrink at all, AND it manages to be more power efficient. I know it's not hard to get worse than GF100, but you get my drift....

I will say again that a 20% speed bump is not trvial. We get cooler, quieter, more overclockable on non-exotic methods, plus slightly more efficient. Thats a big win for NV compared to where they have been. If Fermi was this good from the get-go, I probably could have gotten my 5870 for $50-75 cheaper.