• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Nvidia's Future GTX 580 Graphics Card Gets Pictured (Rumours)

Page 13 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Possible new performance slides for GTX580:

http://translate.google.com/transla...4it.it/gpu/425-gtx-580-informazioni-riservate

If you look at the game-only benches (ignore the synethic benchmarks like Vantage), then it's about 45% faster than a Cypress XT among the selected games.

Claimed perf per watt is roughly 1.25 times that of GTX480 (for certain applications; not sure why only a few were chosen), and claimed noise level for the reference cooler is a tad below GTX285.

Edit: note that if 1.25x perf/watt is accurate and wattage is slightly under the GTX480's power draw, as has been rumored, then this implies that the GTX580 will be ~20% faster than the GTX480.
 
Last edited:
Thanks for the link. The biggest thing I hate about internal benchmarks like this is that we don't know at what resolution and AA/AF these tests are being done at. 45% faster sounds amazing, but if it's only 45% better than 14fps then it's all fluff. However, in relation to this graph, it does show that AMD will indeed need a really competitive high end part if they want to take over the single GPU performance crown - especially in DX11 where Fermi performance seems to shine in relation to Cypress. Can AMD actually make a single GPU within a years time on the same node that is 35-40% faster than it's current fastest gpu? This is going to be an interesting month!
 
Blastingcap those are the same benchmark leaks that have been out for abit... only these only show from 0.8 and up to look more dramatic.

The ones that start at 0.0 are much easier to get a real idea of performance with.

These show gains in the area of 15% - 65% over a 5870 (with the exception of heavy tessellation games of Lost planet and metro2033, the new 6970 will have fixed tessellation issues so the gains comparable wont be as big).

But a 480 already has a avg of x% gains over a 5870.

_480 vs 580 (from looking at this leaked benchmark) (eyeball estimations)
140% vs 160% batman AA
130% vs 140% CoD MW2
150% vs 170% starcraft 2
_90% vs 110% Crysis warhead: (in this a 5870 beat the 480)
130% vs 145% farcry 2
110% vs 135% Hawx
110% vs 130% Just cause
105% vs 110% Resident evil5
120% vs 130% World in conclift.

gonna stop here. lets compair the % gains:

160/140 = 1.14 = 14% gain over the 480
140/130 = 1.07 = 7% gain over the 480
170/150 = 1.13 = 13% gain over the 480
110/90 = 1.22 = 22% gain over the 480
145/130 = 1.15 = 15% gain over the 480

ect
ect
ect



this might be faulty math, but to me the 580 doesnt look that impressive compaired to a 480.
The 480 is a powerfull card, and the 580 does look impressive next to a 5870.
However the 580 vs 480 doesnt look too impressive if these leaked benchmarks are what it ll end up as.
 
Last edited:
Idontcare said:
How does the math look for the 6870 vs 5870? Is it equally impressive, or more so?
Click to expand...


From the same picture "leak":

6870 vs 5870: (eyeballing that leaked picture)
85% vs 100% batman AA
90% vs 100% CoD MW2
90% vs 100% starcraft 2
85% vs 100% Crysis warhead: (in this a 5870 beat the 480)
90% vs 100% farcry 2
90% vs 100% Hawx

ect
ect
ect

the 6870 looks to lose to the 5870 in nearly everything by 10-15% or so. No the 6870 doesnt look impressive compaired to a 5870.


My point was just that they(rumor sites) claimed the 580 had over 20% performance gained over the 480, and it doesnt look like it has that (from this leaked benchmark). It looks closer to ~15% in games or so.
 
Last edited:
Thanks for answering my request, I really do appreciate you taking the time because when you use subjective adjective like "impressive" I have no reference point for how you use the term...now I do!

So the question I have for you is does 15% versus 20% come down to driver revs versus driver performance expections?

Maybe someone was given the guidance of "we expect to have 20% by the time it is released, but today the drivers are only giving us 10%" and since then they have gotten it to 15%, with another 5% expected in time?

What are your thoughts on this scenario/explanation?
 
If drivers by release push it upwards of 20% gains on a 480... thats more impressive than 15% gain is. The fact that its faster than a 480 at all, and useing less power in itself is pretty nice.

But the 15% leap just doesnt seem that big... to warrent a name generation change. I know the 6870 should just have been called the 5860 or something, but the 580 should by the same reasoning just have been a 485 or 490.


A new gen card released only haveing 15% on the last gen it replaces?
If the 6970 ends up 50%-60% faster than the 5870... thats "big" enough to warrent a new release, that is a big leap (compaired to a 15% one).
 
Arkadrel said:
If drivers by release push it upwards of 20% gains on a 480... thats more impressive than 15% gain is. The fact that its faster than a 480 at all, and useing less power in itself is pretty nice.

But the 15% leap just doesnt seem that big... to warrent a name generation change. I know the 6870 should just have been called the 5860 or something, but the 580 should by the same reasoning just have been a 485 or 490.


A new gen card released only haveing 15% on the last gen it replaces?
If the 6970 ends up 50%-60% faster than the 5870... thats "big" enough to warrent a new release, that is a big leap (compaired to a 15% one).
Click to expand...

Personally I don't care what companies call their cards. If they call it the "GTX3.14159 Tooth Fairy Edition With Military Class Components Made By Santa's Elves" and it's good perf/price and equal or better on noise/thermals/power to competing cards, and it's in my budget range, I might buy it anyway.
 
Last edited:
Arkadrel said:
A new gen card released only haveing 15% on the last gen it replaces?
Click to expand...

How much % was the increase for 6870 vs 5870 ???

We need to stop this thing about new generations related with how much more performance they have, when we use the same lithography.

Both (AMD,NV)changed the names for PR reasons
 
Last edited:
Arkadrel said:
If drivers by release push it upwards of 20% gains on a 480... thats more impressive than 15% gain is. The fact that its faster than a 480 at all, and useing less power in itself is pretty nice.

But the 15% leap just doesnt seem that big... to warrent a name generation change. I know the 6870 should just have been called the 5860 or something, but the 580 should by the same reasoning just have been a 485 or 490.


A new gen card released only haveing 15% on the last gen it replaces?
If the 6970 ends up 50%-60% faster than the 5870... thats "big" enough to warrent a new release, that is a big leap (compaired to a 15% one).
Click to expand...

Nvidia went with a new naming scheme because AMD went with a new naming scheme. At least Nvidia's naming scheme actually relates to increased performance relative to what it's brand is replacing - where as AMD decided to take the confuse all but the ultra informed approach and regroup their next gen names as to where the parts are slower than their last gen naming nomenclature.
 
tviceman said:
Nvidia went with a new naming scheme because AMD went with a new naming scheme. At least Nvidia's naming scheme actually relates to increased performance relative to what it's brand is replacing - where as AMD decided to take the confuse all but the ultra informed approach and regroup their next gen names as to where the parts are slower than their last gen naming nomenclature.
Click to expand...

*cough* 8800 9800 250 *cough*

Let's be fair here. 🙂 Doesn't change the fact that I'd still buy a Tooth Fairy card if it were good, though. 😛
 
tviceman said:
Nvidia went with a new naming scheme because AMD went with a new naming scheme. At least Nvidia's naming scheme actually relates to increased performance relative to what it's brand is replacing - where as AMD decided to take the confuse all but the ultra informed approach and regroup their next gen names as to where the parts are slower than their last gen naming nomenclature.
Click to expand...

Yea, AMD's naming structure made sense, I wish they would have called the 68xx parts a 6770 and 50, or if they really felt the performance was great maybe a 6790 and 70. But, AMD has only used the x7xx the last two generations for the mid range. Before that I think they used an x6xx for their mid range parts for a while. So I don't feel as if the x7xx name was so firmly established that they were really confusing people.


tviceman said:
AT LEAST each iteration wasn't slower than the last, amirite?
Click to expand...

The 6870 is faster than their former mid range. The 6970 will be faster than a 5870. And the 6990 will be faster than a 5970, I'm sure. 🙂
 
Last edited:
rUmX said:
Someone correct me if I'm wrong (fuzzy memory) but IIRC the 9800GTX was generally slower than the 8800GTX.
Click to expand...
you're thinking of the 9800GTX vs 8800 ultra probably. the 8800GTX 512mb was the same gpu as the 9800GTX, clocked lower (G92)
 
faxon said:
you're thinking of the 9800GTX vs 8800 ultra probably. the 8800GTX 512mb was the same gpu as the 9800GTX, clocked lower (G92)
Click to expand...

There was no 8800GTX 512, that was the 8800GTS 512. Yes, it confuses even people as informed as us. And some people think the 6870/6850 naming is confusing.

Yeah, the 8800GTS was faster than the 8800GT, which was faster than the 8800GTS. Makes sense huh.
 
Last edited:
faxon said:
you're thinking of the 9800GTX vs 8800 ultra probably. the 8800GTX 512mb was the same gpu as the 9800GTX, clocked lower (G92)
Click to expand...
the 9800gtx was g92 and was usually even or a bit faster than the g80 8800gtx. there was no 8800gtx 512mb as it was a 384bit 768mb gpu only.
 
The 9800GTX core was a more powerful GPU. G92 > G80. However, the supporting staff, buswidth, ROP's, Bandwidth, did keep the performance on an even keel for the most part. 8800GTX would often pull ahead when bandwith starved situations occured.
G92 is/was a great GPU in that it offered a more powerful GPU, using less power, ran cooler and costs less than G80 at launch. Die Shrink helped 😀
 
well you have to consider that Barts is not a replacment to any GPU it's a gap filler. the true Cypress replacment is Cayman.
 
Some pics:

88922881.jpg



74261532.jpg


77038850.jpg



16793563.jpg
 
Temperatures are good, but the performance results are not impressive. I wonder if the phenom II x6 is the culprit.
 
http://www.techspot.com/news/40876-amd-radeon-6970-benchmark-numbers-leaked.html
The site’s numbers cover both 3DMark Vantage and the Unigine Heaven benchmark, with the Radeon HD 6970 allegedly scoring 23,499 on the first one while delivering 36.6fps on the Unigine benchmark running at 1920 x 1,080 with 4x anti-aliasing and 16x anisotropic filtering. By comparison, a GTX 480 on the same platform scored 21,106 in 3DMark and ran Unigine Heaven at 29.5fps.. snip..
Click to expand...
a 480 scoreing 29.5fps in 1920x1200 4AA 16AF Extreme Tess.
a 6970 scoreing 36.6fps in 1920x1200 4AA 16AF Extreme Tess.

So the 6970 last month with early beta drivers... beat a 480 by ~25% in tessellation performance at extreme settings.

If the 6970 turns out to be better at tessellation than the 580... might happend.


I wish the guys that had leaked that 580 photo, would have tested it in the same settings as the leaked 6970 settings (1920x1200, 4AA,16AF, Extreme tess)... instead of 1920x1080, 4AA,16AF normal tess.


**54,000+ in CPU score from physx on, might be why the 580 scores in the 12,800 range... if it lost like 40,000 points there (in the CPU score), its overall score might be abit lower (when compaired to a amd card). What I mean to say is... vantage gives a egde to nvidia cards because of physx on effecting the scores a good bit.
 
Last edited:
Arkadrel said:
http://www.techspot.com/news/40876-amd-radeon-6970-benchmark-numbers-leaked.html
a 480 scoreing 29.5fps in 1920x1200 4AA 16AF Extreme Tess.
a 6970 scoreing 36.6fps in 1920x1200 4AA 16AF Extreme Tess.

So the 6970 last month with early beta drivers... beat a 480 by ~25% in tessellation performance at extreme settings.

If the 6970 turns out to be better at tessellation than the 580... might happend.

I wish the guys that had leaked that 580 photo, would have tested it in the same settings as the leaked 6970 settings (1920x1200, 4AA,16AF, Extreme tess)... instead of 1920x1080, 4AA,16AF normal tess.
Click to expand...
Great points. To add to them, isn't NVIDIA's tesselation supposed to scale better at higher loads than AMD's? So if AMD's part is (seemingly) 25% faster at higher loads, might it be even faster at "normal" tesselation? Interesting either way, I'd like to see how this develops.

Arkadrel said:
**54,000+ in CPU score from physx on, might be why the 580 scores in the 12,800 range... if it lost like 40,000 points there (in the CPU score), its overall score might be abit lower (when compaired to a amd card). What I mean to say is... vantage gives a egde to nvidia cards because of physx on effecting the scores a good bit.
Click to expand...
Another good point, only compare the GPU score, not the total. I'll google for a GPU-only Vantage extreme score for the 6970, if there is one floating around out there.

Here's another thread I was reading today with some interesting leaks: http://hardforum.com/showthread.php?p=1036345837
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top