GTX 580 performance a FRAUD?

V

vjcsmoke

Supreme [H]ardness
Joined
Dec 5, 2006
Messages
4,511
According to the legitreviews article, when the GTX 580 was downclocked to match stock GTX 480 speeds the performance difference was only 3.5%. So the extra shaders and such don't really give as much a boost as we thought, it was mostly the increased clock speed??
http://www.legitreviews.com/news/9429/
 
haha nice, it's one game but I'd imagine it's around that across the board.
 
Wow, that's a lot of pre-rendered high definition video they must have stored on the video card, then. My question is, how did they figure out in advance what inputs Brent would be giving to the games he chose? Plus the timing of the pre-rendered video had to be extremely precise so Brent wouldn't notice that it wasn't actually interactive.

Basically, if nVidia really did figure out how to perpetuate a fraud so complete that Brent didn't notice he was looking at pre-rendered video rather than video rendered by the GTX 580, I'd think nVidia would be really rich by now.
 
Well, Metro is an extremely demanding game.....I'd need to see more comprehensive testing.
 
Rumor is nvidia has a driver ready but are holding back until after Cayman launches. I've said this before, they may be holding performance back with the drivers. Rumors are already leaking that they are going to unleash a driver after Cayman launches to boost about 10% in games with the GTX 580. Now this has history as nvidia always supposedly has a magical driver and I'm sure people are sick of that tactic but in the past they have delivered that magical driver. We'll have to wait about 30 days and see.

BTW anandtech does a even better test

http://www.anandtech.com/show/4012/nvidias-geforce-gtx-580-the-sli-update
 
vjcsmoke said:
According to the legitreviews article, when the GTX 580 was downclocked to match stock GTX 480 speeds the performance difference was only 3.5%. So the extra shaders and such don't really give as much a boost as we thought, it was mostly the increased clock speed??
http://www.legitreviews.com/news/9429/
Click to expand...

Well that isn't suprising considering the difference between 480 and 512 is only 6%

the difference is that the 480 couldn't be clocked any higher and was a harvested chip to begin with.

the 580 just brought efficiency to their manufacturing process and another side benefit of that is the clock frequency could be higher (probably what they were targetting to begin with).

So i wouldn't call it a fraud but more of a fix.

Either way you look at it the 580 is faster than the 480 - doesn't matter how.


As far as holding a driver back, if they are then i'd be pretty upset as an Nvidia customer..
 
A driver that boost a card's performance by 10%? I'm dying to see that.
 
Why? If you buy a 580 right now, you are buying the fastest single GPU card. Hypothetically, when Caymen launches and they release this magical driver, its like a free upgrade. That would be sweet.
 
H-street said:
As far as holding a driver back, if they are then i'd be pretty upset as an Nvidia customer..
Click to expand...

By that comment, I believe that you bought a GTX 580. I'm not sure why that would bother you as it's already performing about 20% faster than GTX 480 and it's only 30 something days away but at the same time I understand your perspective of it too if that is the case. Not sure how long you've been an nvidia customer but they have done this before and there were also times they said it was coming and it falls short of expectations and only delivers a very tiny boost. We'll have to wait and see I guess.
 
JacobC1983 said:
Why? If you buy a 580 right now, you are buying the fastest single GPU card. Hypothetically, when Caymen launches and they release this magical driver, its like a free upgrade. That would be sweet.
Click to expand...

LOL. That driver would be magical indeed.
 
Lord_Exodia said:
Rumor is nvidia has a driver ready but are holding back until after Cayman launches. I've said this before, they may be holding performance back with the drivers. Rumors are already leaking that they are going to unleash a driver after Cayman launches to boost about 10% in games with the GTX 580. Now this has history as nvidia always supposedly has a magical driver and I'm sure people are sick of that tactic but in the past they have delivered that magical driver. We'll have to wait about 30 days and see.

BTW anandtech does a even better test

http://www.anandtech.com/show/4012/nvidias-geforce-gtx-580-the-sli-update
Click to expand...
If Nvidia has a driver that gives a 10 % boost on tap, why would they withhold it until Cayman launches? That makes no sense..
 
heflys20 said:
A driver that boost a card's performance by 10%? I'm dying to see that.
Click to expand...

Now lets be clear here. I DOUBT LOL that it'll be 10% in all games, it could be anywhere from 6-12% boost across the board and there may be 2 or 3 games here or there that see absolutely no boost or no boost worth mentioning e.g. 1% etc..

I've been here done this many time with nvidia already.
 
heflys20 said:
A driver that boost a card's performance by 10%? I'm dying to see that.
Click to expand...

its like 256 driver..

When they say its 10%, its probably 2-3% at most in general..

driver update never and ever magically give 10% raw performance out...

If GXT 580 does ever get the performance boost, same goes to all GTX 400 series since they are pretty much the same chip beside couple tweaks..
 
x3sphere said:
If Nvidia has a driver that gives a 10 % boost on tap, why would they withhold it until Cayman launches? That makes no sense..
Click to expand...

Kind of like the sea creature mentioned by Joaquin Phoenix in Gladiator. He talks about the sea urchin that lets his enemies biet at it and it holds still so that they dont know how dangerous it is, then more come and keep biting yet it still holds still etc.. Then it eats them all.

I believe that is the psychology in this. They've done this before. Sometimes it works sometimes it doesn't. In this case if Cayman launches with the price they are saying, the performance may be less significant as I'd gladly save $100 to sacrifice 5% of performance IF and that's a big IF the are outperformed.
 
I disagree.

GTX 480 3x SLI
Heaven%20Benchmark%20v2.0.%203x%20SLI%20GTX%20480.jpg


GTX 580 3x SLI
Heaven%20Benchmark%20v2.0.%203x%20SLI%20GTX%20580.jpg
 
Also it could be that the new games aren't taking advantage of the GTX 580 fully from the driver yet. Maybe there is no magical driver as some of the rumors say but I'm sure nvidia can and will squeeze out more performance from Metro 2033 with the GTX 580. They're pretty well known for doing wonders with their drivers in games.
 
If you downclock a 5870, it will run like a 5850 i.e. 5970. What's the point?

It has always been known that a fully working Fermi 512 chip was only 5% faster than the 480 chip. Increasing clock speeds is normal in a "refresh". The GTX 580 is not a new generation of cards.
The 6800 on the other hand is a hybrid of last gen and new parts. The performance increase at the same clock will be higher than Nvidia's, but AMD will also increase clocks to 900MHz on the 6970, up from 850MHz last gen.
Is AMD also not telling the truth by increasing their 6000 series core clocks?
 
Last edited:
vjcsmoke said:
According to the legitreviews article, when the GTX 580 was downclocked to match stock GTX 480 speeds the performance difference was only 3.5%. So the extra shaders and such don't really give as much a boost as we thought, it was mostly the increased clock speed??
http://www.legitreviews.com/news/9429/
Click to expand...

And, that's a surprise because? The clock speed is increased %10, and the core width is increased only %6.6. You tell me which one is going to have the greater effect.

Also, the test is ambiguous because they clock the memory down as well. If you starve those new shader units, there's no telling how much you're affecting their performance. A better test would be if they included a GTX 580 with stock memory clocks and the 701 MHz core clock.

But I wouldn't expect anything less than this from Legit Reviews, the home of brain-dead, pointless reviews without insight. Anyone who actually knows how GPUs work and has a calculator could have told you about the tiny increase in performance from adding 32 shaders.
 
defaultluser said:
Also, the test is ambiguous because they clock the memory down as well. If you starve those new shader units, there's no telling how much you're affecting heir performance. A better test would be if they included a GTX 580 with stock memory clocks and GTX 480 GPU clocks.
Click to expand...

Aww good point!
 
heatlesssun said:
Yes though the 580s are Superclocked editions, like 20Mhz over stock cards.
Click to expand...

That's what I figured. The argument isn't whether the 580 is good at stock, but how good it would be if it possessed the same clocks as the 480. I wonder how much difference it would make?
 
heflys20 said:
That's what I figured. The argument isn't whether the 580 is good at stock, but how good it would be if it possessed the same clocks as the 480. I wonder how much difference it would make?
Click to expand...

But lets not forget Heatlessun's point that by down-clocking the ram to match GTX 480 speed you are also effectively bottle-necking the extra SP's so it isn't a good way to find out what the extra SP's are netting you in performance gains.
 
Last edited:
Must be a slow news day. A card with 6% more units gets 3.5% higher performance at the same clocks. Fraud!!!
 
heflys20 said:
That's what I figured. The argument isn't whether the 580 is good at stock, but how good it would be if it possessed the same clocks as the 480. I wonder how much difference it would make?
Click to expand...

Why not just look at the Anandtech article?
http://www.anandtech.com/show/4012/nvidias-geforce-gtx-580-the-sli-update/3

The point about memory isn't correct
- if they didn't down clock the memory, then they question would be 'how much of that is due to the memory speed being higher?'
- so, anyway, these figures show you what you get when the clocks are the same
- which shows, therefore, the effect of the architectural changes.
- some of which (up to 6%) is due to the additional SM / 32 SPs (which isn't an architectural change)
 
samurai1999 said:
So, looks conclusive ...
- 24.8% more FPS on the same system
- with a 13% difference in clocks (701 vs 797)

- so, MythBusted!
:D
Click to expand...

I'm never afraid to admit it but I'm lost at this point. What's your take on this and why did you come to that conclusion?
 
Looking again at the Anandtech article

http://www.anandtech.com/show/4012/nvidias-geforce-gtx-580-the-sli-update/3

- something that is CPU bound like Civ V gives almost no improvement

- something that is shader bound, like Crysis, Folding@home, Wolfenstein, SmallLuxGPU, gives 7% at equal clocks, and an additional 10% when the shader clocks are boosted by 10%

- with DiRT & HAWX2, there's 16-17% on the architectural changes, and a further 10% with clocks...
(maybe the Z-thingy?)

- Metro2033 also seems to do 10% better at equal clocks according to this test ...
 
samurai1999 said:
Looking again at the Anandtech article

http://www.anandtech.com/show/4012/nvidias-geforce-gtx-580-the-sli-update/3

- something that is CPU bound like Civ V gives almost no improvement

- something that is shader bound, like Crysis, Folding@home, Wolfenstein, SmallLuxGPU, gives 7% at equal clocks, and an additional 10% when the shader clocks are boosted by 10%

- with DiRT & HAWX2, there's 16-17% on the architectural changes, and a further 10% with clocks...
(maybe the Z-thingy?)

- Metro2033 also seems to do 10% better at equal clocks according to this test ...
Click to expand...

Ok, thanks for clearing that up. Even without the higher clocks most apps get a boost of 5-10% that was expected by only adding the extra 32 SP. The real great thing about the GTX 580 is the improved TPD and power draw and cooler. I can imagine how that same GTX 580's temps/noise/heat would compare to the GTX 480 with the same exact clocks etc.. I'd imagine people last year would have done a even trade in seconds as it would be a no brainer. But with the higher clocks they added a bit more. overall 20% boost over a very powerful card isn't a bad boost as it's a refresh and not a new architecture.
 
If I downclocked my Phenom 965 to 3 GHz, I wouldn't be surprised that it ran exactly like a 945. I don't see what the big deal is here.

On the flip side of that coin, this kinda shows that this card isn't much more just an OC'd 480. Maybe it's capable of a moderately higher OC on its shaders, so I guess thats an improvement.
 
So these sharder performance figures are fake even though they perform just a bit better and at a much lower power draw? Ok.
 
what is a new geforce 580 good for then? you can run any game with a 470 gtx. what's the point, maybe it's better to wait for the 680??
 
arcr said:
what is a new geforce 580 good for then? you can run any game with a 470 gtx. what's the point, maybe it's better to wait for the 680??
Click to expand...

Why wait for the 680 beacause at that time somebody will say what you just said.

I'm kidding but I think you get my point. I know where you are coming from. The PC is supposed to be the platform of choice where games are supposed to be developed. The graphics cards are supposed to be more powerful than the developers need so that sky is the limit for them. Unfortunately we live in a world where money rules everything.

People dont feel they need to pay money for games so they torrent them

Game developers dont feel they should give games away without money

so what do they do, they develop their games on consoles which limit and slow down progression. (there are positives to consoles but irrelevant here)

What we get is a bunch of console ports that make people like you ask the question you just asked.

However it's all a vicious circle as many people refuse to spend money on console ports. I know quite a few people that actually BUY every game they can that helps advance progression and torrent any console port.

Not sure how this will ever sort itself out, but that's my 2c on it.
 
Considering its already cooler and less power hungry at stock clocks I wonder how the GTX 580 would compare to the GTX 480 when downclocked to its levels?
 
Apply the vapor chamber technology equally to the GTX 480 then do the benchmarks at GTX 480 clocks.
 
The downclocked 580 looks a lot less impressive for sure. So let's go by the theory that it would be 'bandwidth starved' when downclocked. The next step would be to OC a 480 to the exact same speeds as the stock 580 and measure how much of a difference there really is. From the numbers I'm seeing, it looks like the difference would likely be only about 10%.
 
AceGoober said:
Apply the vapor chamber technology equally to the GTX 480 then do the benchmarks at GTX 480 clocks.
Click to expand...

That's not going to affect the gaming performance.. not sure what your getting at. And no the temps will still be higher on the GTX 480. It wasn't just the vapor chamber tech that dropped the temps. You do know this right?
 
You must log in or register to reply here.
Back
Top