NVIDIA Kepler speculation thread

I hope thoses number are fake:

The figures presented below were tested on an Intel Core i7 3960X 3.3GHz CPU on an MSI X79A-GD65 motherboard with 16GB of DDR3-1600 memory under Windows 7. Nvidia's 320.18 WHQL and AMD's Catalyst 13.4 WHQL drivers were used to drive the Geforce GTX 770 and Radeon HD 7970 cards, respectively, with both cards running at reference speeds. All games were run at 2560x1600 resolution and all numbers are frames per second. More details on the benchmarks are listed in the appendix.


http://www.theinquirer.net/inquirer...comes-close-to-the-radeon-hd-7970-ghz-edition
 
I grabbed this:
E8a5px8.png
 
The inquirer? Yeah hahah ok. The gtx770 being about 10% faster than a 680 will not be slower than a 7970GE in BF3, Crysis 3, , Skyrim, or Batman AC. So yeah that graph looks like crap.

BF3, 680 is already a smidge faster than the 7970GE
54880.png


Crysis 3, 680 is essentially tied to 7970GE
54890.png


And who knows why Skyrim is sooo slow with the Inquirer's graphs.
47482.png


That is just 3 examples. Official reviews will be out in 2 days. Of course AMD will be faster in titles like Sleeping Dogs, Dirt Showdown, and Hitman. But Nvidia should have the edge in Crysis 3, Far Cry 3, Tomb Raider, Bioshock, Max Payne, and BF3.
 
BF3, 680 is already a smidge faster than the 7970GE
FYI - AnandTech happens to use one of the few maps that is a little more unfavorable to Radeon than the other maps. If you check other reviews I believe most will paint a different picture, certainly at the higher resolutions.
 
FYI - AnandTech happens to use one of the few maps that is a little more unfavorable to Radeon than the other maps. If you check other reviews I believe most will paint a different picture, certainly at the higher resolutions.
Wait, are you telling me that Anand only bench one map on BF3 and call it a day?
 
If Ryan is around I'm sure he can explain their procedure. IIRC our performance labs standard sweep uses around 4 or 5 maps in BF3 and we can replicate Anand's positioning with one of them, the rest shows positioning more akin to what you'll see at many other sites.
 
If Ryan is around I'm sure he can explain their procedure. IIRC our performance labs standard sweep uses around 4 or 5 maps in BF3 and we can replicate Anand's positioning with one of them, the rest shows positioning more akin to what you'll see at many other sites.

I never trust on internal tests :LOL:
 
I never trust on internal tests :LOL:

Weird how you've been linking Nvidia ones ever since I can remember.

Regardless of that you don't have to trust internal tests you just have to check what most of the reviews are showing, and it's a 3-4 fps win for AMD.

For as long as I can remember the main sites showing a win for Nvidia are...

Anandtech
Hardware Canucks
Bit-Tech

No need to comment on SKYMTL, Olin Coles at Bit-Tech was once threatened with being sued (allegedly) by AMD for an article he wrote and Anandtech has long been known to be Nvidia friendly in benchmarks hence the constant outlier results.

The chances of all 3 of them choosing the same map by luck is pretty low.
 
Weird how you've been linking Nvidia ones ever since I can remember.

Regardless of that you don't have to trust internal tests you just have to check what most of the reviews are showing, and it's a 3-4 fps win for AMD.

For as long as I can remember the main sites showing a win for Nvidia are...

Anandtech
Hardware Canucks
Bit-Tech

No need to comment on SKYMTL, Olin Coles at Bit-Tech was once threatened with being sued (allegedly) by AMD for an article he wrote and Anandtech has long been known to be Nvidia friendly in benchmarks hence the constant outlier results.

The chances of all 3 of them choosing the same map by luck is pretty low.

Everybody says "my products are the best", so i look at every reviews as i can, to have an idea, and then i wait the "real tests": mine or the users' ones in forums (yeah, god bless internet LOL!)
 
The inquirer? Yeah hahah ok. The gtx770 being about 10% faster than a 680 will not be slower than a 7970GE in BF3, Crysis 3, , Skyrim, or Batman AC. So yeah that graph looks like crap.

BF3, 680 is already a smidge faster than the 7970GE

Crysis 3, 680 is essentially tied to 7970GE

And who knows why Skyrim is sooo slow with the Inquirer's graphs.

Crysis 3

geforce-gtx-780-performance-review,3516-9.html
crysis-avg.png


BF3

geforce-gtx-780-performance-review,3516-6.html
bf3-avg.png


Skyrim

skyrim-avg.png


geforce-gtx-780-performance-review,3516-12.html
 
Oh btw just incase you thought that was a fluke tviceman -

arkhamcity_2560_1600.gif


bf3_2560_1600.gif


crysis3_2560_1600.gif


skyrim_2560_1600.gif


Yup the 7970 GHz wins them all again. It's Anandtech who is the outlier here.
 
The inquirer? Yeah hahah ok. The gtx770 being about 10% faster than a 680 will not be slower than a 7970GE in BF3, Crysis 3, , Skyrim, or Batman AC. So yeah that graph looks like crap.

That is just 3 examples. Official reviews will be out in 2 days. Of course AMD will be faster in titles like Sleeping Dogs, Dirt Showdown, and Hitman. But Nvidia should have the edge in Crysis 3, Far Cry 3, Tomb Raider, Bioshock, Max Payne, and BF3.

Games are not made up of only one scene ;)
Different benchmark, different result.
 
Oh btw just incase you thought that was a fluke tviceman -
Yup the 7970 GHz wins them all again. It's Anandtech who is the outlier here.
As boxleitnerb notes, a lot will depend on the benchmark. I'd prefer to look at results that give a detailed breakdown of game i.q. levels. As such:
Crysis 3 at Hardware France looks pretty even (BF3 favours the 7970GE)- as does PCGH
Skyrim at ComputerBase would tend to go against your TPU results (BF3 still favours the 7970GE)
I generally steer clear of Tom's Hardware. They don't do themselves any favours by doing supposed in-depth testing of the Titan's CUDA abilities using an outdated 2.66 version of Blender for instance when there has been a Compute 3.5 supported 2.67 available for the last 2+ months
 
Some features of DX11.1 are hardware and software ( only 4 require hardware ), so any DX11.0 cards will be compatible but only by software. ( allready discussed with GTX600 series, as AMD support hardware DX11.1 and Nvidia just by software ).

And like allways the words of Nvidia is " thoses features are useless anyway ".. ( who is not true but thats an other story )

No they're not. NVIDIA supports (on the GK1xx-chips anyway) DirectX 11.1 Feature level 11_0, in other words, they support DX11 only.
Every card with modern drivers "supports" DX11.1 in same way they do, feature level being dictated by the card (so GTX 280 would be "DirectX 11.1 Feature Level 10_0" for example)
Only GCN (and now possibly the GK208 or whichever it was) actually supports DX11.1, aka DirectX 11.1 Feature Level 11_1.
 
Those specs on the Nvidia site are clearly in need of an overhaul. They give a dedicated 2x processor clock for a supposedly Kepler-based card for example.

WRT DX11.1: The site talks explicitly about "DirectX 11.1 API", that three extra letters should make it clear enought that it's not HW-level 11.1.

WRT Kepler-GT630: There is already an OEM-version in the wild. It's GK107-based and has only 192 ALUs, thus one SMX fused off.
 
BF3, 680 is already a smidge faster than the 7970GE


Crysis 3, 680 is essentially tied to 7970GE

A lot of factors will determine whether a card performs 1 or 2 fps better than another, but for all intents and purposes performance is basically the same. So your 10% prediction may not be far off. Below are the results for BF 3 and Crysis from Guru3D ... ;)

index.php



index.php
 
The chances of all 3 of them choosing the same map by luck is pretty low.
Actually it's pretty high. We're using Thunder Run, the only significant ground based on-rails map in the game. I don't know what anyone else is running, but if they followed the same train of thought they'd end up on the same map.

And why on-rails? It's very easy to replicate (I'm not the only one that needs to be able to do this) and very consistent run-to-run, which makes it easy for other editors to repeat it while not having our results bounce all over the place for no good reason. Also, it's not CPU bottlenecked at the high end, which gives us room to grow for more exotic configurations like tri-Titan. I'd much rather have a proper recorded benchmark (and I'm annoyed to this day that DICE took that out of the game; it was in the betas) but in a pinch on-rails will have to do.
 
Actually it's pretty high. We're using Thunder Run, the only significant ground based on-rails map in the game. I don't know what anyone else is running, but if they followed the same train of thought they'd end up on the same map.

And why on-rails? It's very easy to replicate (I'm not the only one that needs to be able to do this) and very consistent run-to-run, which makes it easy for other editors to repeat it while not having our results bounce all over the place for no good reason. Also, it's not CPU bottlenecked at the high end, which gives us room to grow for more exotic configurations like tri-Titan. I'd much rather have a proper recorded benchmark (and I'm annoyed to this day that DICE took that out of the game; it was in the betas) but in a pinch on-rails will have to do.

I stand corrected on the luck part, let's just say it's fortunate for Nvidia that the sole map in the game where they win instead of lose by 10% happens to be the easiest one to benchmark.

It is throwing out an unrealistic benchmark however and we're talking about a game that sold millions of graphics cards. For last years premier game and premier tech site to use it must be quite frustrating for anyone at AMD.
 
Back
Top