Kepler Nvidia GeForce GTX 780

[Andrea deluxe]

[The Jesus]

http://www.3dcenter.org/

Looks like an old graph that they just copy-pasted over, lmao. Also, that other graph is extremely deceptive in scale, lol. Its based on a 1.0 from 0.8 up to 1.5...

[flyck]

http://www.3dcenter.org/

ignoring the pink bars. The results for the 7950 and 7970 are in line with the first AMD drivers, not the ones we already have since february. Also these results are in line with 1920*resolutions. (results equivalent of techpowerup).

[Cold Fussion]

It doesn't have a scale, how can you say it's in line with anything? For all you know it could be a logarithmic scale.

[flyck]

It doesn't have a scale, how can you say it's in line with anything? For all you know it could be a logarithmic scale.

we know the performance of the 7950, 7970, 580 and 480.
At that point we know the difference between the 480 and 580, 580 and 7950, 580 and 7970 and 7950 and 7970...

So if we assume valid results, than the scale, the resolutions, the drivers are known and fixed. Using a different scale would make the results of the known invalid.

[zalbard]

An interesting observation: Shogun 2, which is probably the game that uses the most VRAM to date (without mods), also performs the worst (compared to 7970).
So if the graph is indeed legitimate then it's quite obvious that 2GB is already not enough.

[Cold Fussion]

we know the performance of the 7950, 7970, 580 and 480.
At that point we know the difference between the 480 and 580, 580 and 7950, 580 and 7970 and 7950 and 7970...

So if we assume valid results, than the scale, the resolutions, the drivers are known and fixed. Using a different scale would make the results of the known invalid.

It doesn't matter, the graph is completely worthless.

[xdan]

An interesting observation: Shogun 2, which is probably the game that uses the most VRAM to date (without mods), also performs the worst (compared to 7970).
So if the graph is indeed legitimate then it's quite obvious that 2GB is already not enough.

Do you observe that in Shogun 2 GTX 680 is worse than GTX 480/GTX 580 which have only 1.5gb VRAM.
It's not because memory. It's either bad support for Shogun 2 in current drivers, or may be some compute and rendering thinghs which are in Fermi but may be aren't anymore in Kepler.

[Kuroimaho]

An interesting observation: Shogun 2, which is probably the game that uses the most VRAM to date (without mods), also performs the worst (compared to 7970).
So if the graph is indeed legitimate then it's quite obvious that 2GB is already not enough.

I’m not questioning your powers of observation, I’m merely remarking upon the paradox of a 2GB card running out of ram while 1.5GB cards don't.

[Olivon]

[/IMG]http://www.3dcenter.org/dateien/abbildungen/nVidia-GeForce-GTX-680-Benchmarks2.jpg[/IMG]

http://www.3dcenter.org/

GTX 670 Ti scores ? Now where are GTX 680 ones ? ^^

[zalbard]

Do you observe that in Shogun 2 GTX 680 is worse than GTX 480/GTX 580 which have only 1.5gb VRAM.
It's not because memory. It's either bad support for Shogun 2 in current drivers, or may be some compute and rendering thinghs which are in Fermi but may be aren't anymore in Kepler.

That's a good point. I guess these tests are done at 1200p (or perhaps even 1080p?), so it must be something else. Still, quite strange.
The same seems to happen in Skyrim as well. Both are also CPU heavy, perhaps it could be a factor as well.

[tajoh111]

It's a amd gaming evolved title and Amd hardware has always done better with this particular game.

[El Mano]

It's a amd gaming evolved title and Amd hardware has always done better with this particular game.

But according to the graph, the 6xx performance is worse than the 580's.
This reminds me of something Charlie said, that the 680 performance would drop when playing non- nvidia optimized games.

[Andrew LB]

But according to the graph, the 6xx performance is worse than the 580's.
This reminds me of something Charlie said, that the 680 performance would drop when playing non- nvidia optimized games.

I wonder how long it took Charlie to come up with that one. The man is without question a modern day nostradamus.

Since his brain was so tired from pinching off that pathetic thought, I'll have to add in what he obviously forgot to mention on purpose. It's that the same goes for any AMD cards when you run non-AMD optimized games.

Now taking into account the fact that nVidia has far better support from game developers and an in-house driver team that actually fixes problems... the safe bet is clear.

[Rollo]

But according to the graph, the 6xx performance is worse than the 580's.
This reminds me of something Charlie said, that the 680 performance would drop when playing non- nvidia optimized games.

He really went out on a limb there!

"NVIDIA cards will not perform as well with games that are not optimized to perform well on NVIDIA cards".

What amazing predictions are next? They will be powered by electricity and cooled by a fan?

[El Mano]

He didn't say " optimized for Nvidia cards", he used another expression which I couldn't exactly recall when I wrote my previous post. I think he said Physx titles but I'm not sure about that. I invite you to have a look yourselves and tell us what you find.

[Olivon]

It seems that air-cooled limit is expected to be superior to 1.4GHz

http://bbs.expreview.com/forum.php?m...ge=1#pid392116

[informal]

1.4Ghz,wow! Impressive.
What is the top clock 7970 can achieve with air cooling, 1250ish Mhz?

PS It seems NV made their own bulldozer( high(er) clock design,a bit slower per clock) but their version actually clocks high and consumes same or less amount of power as the design it replaces. Unlike Bulldozer unfortunately.

[Olivon]

Kepler/Bulldozer comparison is really funny to read (you guy always jokin' ?). One is a winner, other is a looser.
GPU/CPU comparison is as good as car analogy IMHO.

If you want to know what nVidia thinkin' about Bulldozer, I recommend this link :

http://mediasite.colostate.edu/Media...157581c458a1d#
00:15:00

[Luka_Aveiro]

Bulldozer was supposed to be a winner...

[chispy]

" ZOTAC GTX680 2GB DDR5 256 bit 1006 6008HDCP Dual DVI HDMI DP


Bron: Factoryprices - ZOTAC GTX680 2GB DDR5 256 bit 1006 6008HDCP Dual DVI HDMI DP - ZT-60101-10P - Zotac
Url: http://www.factoryprices.nl/product_...C-ZT-60101-10P

€ 607, 59 607.00 EUR = 797.416 USD
Let the pre-orders begin !

[RPGWiZaRD]

As usual that's not how to compare EU->USD pricing, they are entirely different markets with EU having high taxes, cheapest HD 7970 for example are selling for like 480~500 EUR in Germany and in US it's $549.

Preorder prices are usually a lot higher and can jump up/down very suddenly but based on this I'd still guess it's targeted for at least $549.... which would be a shame. Totally expected though at this point.

[SKYMTL]

Do you observe that in Shogun 2 GTX 680 is worse than GTX 480/GTX 580 which have only 1.5gb VRAM.
It's not because memory. It's either bad support for Shogun 2 in current drivers, or may be some compute and rendering thinghs which are in Fermi but may be aren't anymore in Kepler.

Actually, Shogun 2 has a massive memory leak unless it is patched to the latest version. Basically, when ending one instance, changing settings and starting another saved game (a typical situation when benchmarking), the memory footprint went from ~1.1GB to 3GB+ in a matter of minutes. Luckily, CA fixed it but if you aren't using any of the last 2-3 patches, memory will prove to be a bottleneck on ALL cards to various degrees. One thing I have noticed is a near lack of sites actually updating their Shogun 2 installs to reflect the new in-game reality likely due to the hassle of retesting every card just because of a patch.

The only time when a patched Shogun 2 install uses more than 2GB is at 2560 x 1600 at the highest details with 4xMSAA+ enabled. A simple way around this is to use a compute-based AA routine like MLAA or FXAA which gives very similar image quality (due to the ultra high resolution) without any huge impact upon framerates.

[xdan]

In Europe it's parity 1€ = 1$ in prices.
So if the card it's 579$ we pay 579€ currency.
The worst it's that this applys even in coutrys that not have € currency which is very sad.
So i had to pay 220€ for my 2500K which is in $ about 270$.
A 7970 it's also about 550€ in $ about 690$

That's not nice at all....

[Lanek]

http://bbs.expreview.com/forum.php?m...ge=1#pid392116

For be honest, i will believe it when i see it lol ... Before that i will put it in the same bag of rumors who try to Hype the product. Like this graph of TDP /performance.. it is so ugly i cant imagine Nvidia will not fire the guy who have done it. Seriously, even a 12years kid will do better. ( nvidia logo blurred, the square in center with thoses circle, lol, the police used )

Kepler/Bulldozer comparison is really funny to read (you guy always jokin' ?). One is a winner, other is a looser.
GPU/CPU comparison is as good as car analogy IMHO.
..............

I think he want to do a comparison between the P4, Bulldozer who was designed for high clock... not performance or quality wise lol . ( but i dont want speak for him )

But i aggree with you, its just impossible to compare CPU and GPU...