NVIDIA Kepler speculation thread

[AnarchX]

Not all Fermi cards were physically capable of 1/2 DP. GF104, for instance, was limited to 1/3—or 1/6, I can't quite recall.

1/12 at GeForce. One of the three SIMD16 processed DP every 4 clocks.
Did someone saw a DP result of a GF106 Quadro 2000?

 

[Kaotik]

Kepler GeForce cards are gonna be capped in DP performance just like Fermi GeForce cards. Just enough to run and debug DP code, but if you want high performance DP only Quadro and Tesla are not capped.

http://www.anandtech.com/show/2977/...tx-470-6-months-late-was-it-worth-the-wait-/6

Capped obviously, but what's the original speed it's being capped from

 

[A1xLLcqAgt0qc2RyMz0y]

Will Nvidia be releasing a 670 on the 22nd?

Previously they were released in pairs 480/470, 580/570.

22nd 680/670?

 

[trinibwoy]

Do you really think that a 1-1,1GHz / 1792 SP "Pitcairn like" (i.e. without 1/2 DP capability) and 256 bit bus chip based on GCN would have been much different in die size, power consumption and performance in respect to GK104?

Why don't you apply the "woulda-coulda-shoulda" logic to GK104 as well? You know, to balance the fairy tale equation

 

[xDxD]

Why don't you apply the "woulda-coulda-shoulda" logic to GK104 as well? You know, to balance the fairy tale equation

they say in Italy: "al cuor non si comanda"

 

[Broken Hope]

According to Gibbo.

http://forums.overclockers.co.uk/showpost.php?p=21514515&postcount=61

All I can say it looks good, its quieter than 7970 which was quiet, it clocks to similar levels and it cost less.

So if it's faster than the 7970, clocks to similar levels and costs less, then AMD is going to be forced to lower prices.

Though to be fair you can already start finding some 7970's for around £380 here in the UK down from £420-450.

 

[trinibwoy]

they say in Italy: "al cuor non si comanda"

Lol google says "the heart can't control" but I'm sure it doesn't mean that!

 

[cal_guy]

1/12 at GeForce. One of the three SIMD16 processed DP every 4 clocks.
Did someone saw a DP result of a GF106 Quadro 2000?

GF104/106/108 Quadro are 1/12 DP rate.

http://www.anandtech.com/show/3961/nvidia-launches-quadro-2000-600

 

[jimbo75]

Pretty Nvidia friendly benchmark suite there - only Metro 2033 is favouring AMD cards (and it's TWIMTBP).

How much is that WoW result affecting the overall percentage difference?

I have a feeling other sites will have the lead a bit smaller.

 

[Alexko]

Lol google says "the heart can't control" but I'm sure it doesn't mean that!

I think it means "the heart may not be commanded".

 

[Mianca]

Pretty Nvidia friendly benchmark suite there - only Metro 2033 is favouring AMD cards (and it's TWIMTBP).

HD7970 is about 30-40% faster than GTX580 with latest drivers and a reasonably balanced benchmark suite. THG's "six games average" is a joke.

Nevertheless, GTX680 looks like a very good card.

 

[Vardant]

Pretty Nvidia friendly benchmark suite there - only Metro 2033 is favouring AMD cards (and it's TWIMTBP).

So a test, that includes BF3 and DiRT 3, where AMD helped, is somehow biased in favor of NVIDIA?

 

[AnarchX]

GK107 GT 650M 3DMark Vantage including feature tests scores:
http://forum.notebookreview.com/gam...600-series-not-just-fermi-60.html#post8391440

But feature tests score seem a kind of strange.
FT1 texel-fill and FT2 color-fill are ridiculous high compared to GTS 450 GDDR5.

Mabye Nvidia manipulated it, like they manipulated GPU-Z and the hardware info panel in their driver.

 

[mao5]

a joke test

http://hardforum.com/showpost.php?p=1038514278&postcount=27

http://hardforum.com/showpost.php?p=1038514370&postcount=33

 

[entity279]

So a test, that includes BF3 and DiRT 3, where AMD helped, is somehow biased in favor of NVIDIA?

who helped is ultimately irrelevant. How the cards actually perform, on the other hand...

 

[mao5]

where's the oc test ?

 

[jimbo75]

So a test, that includes BF3 and DiRT 3, where AMD helped, is somehow biased in favor of NVIDIA?

Dirt 2 and 3 are known to favour Nvidia cards by a lot and I'm sure Nvidia "helped" more with BF3.

HAWX 2 - really?
Crysis 2 - um yeah.
Skyrim - another one heavily favouring Nvidia
WoW - massively favoured Nvidia for years

Only Metro 2033 - another Nvidia title - even remotely favours AMD cards.

Where is AvP? Total War? Mass Effect 3? Crysis Warhead? These are titles where the 7970 will probably win or have a much more realistic gap.

Fact is, only by swapping out Metro 2033 for something like Lost Planet could make this benchmark suite any more Nvidia friendly than it is.

 

[mao5]

how much bucks did toms get from NVIDIA for this joke test?

 

[jimbo75]

Funnily enough they removed Batman (another Nvidia title) when it was performing better on the 7970 - http://www.tomshardware.com/reviews/radeon-hd-7970-benchmark-tahiti-gcn,3104-12.html

And yep, they actually replaced it with HAWX 2. I mean, out of ALL the games they could have chosen, they chose HAWX 2?

 

[Mianca]

So a test, that includes BF3 and DiRT 3, where AMD helped, is somehow biased in favor of NVIDIA?

No, a review is biased when:

• it benchmarks an enthusiast card @1080p resolution
• it doesn't feature HD7870 in the perf/W metrics
• it gives a "six games average" that basically makes an HD7970 look 10-20% slower than it actually is
• HD7970 somehow loses performance compared to GTX580 in the same benchmarks at the same settings - even with newer, better drivers available; e.g. this one :

Inconsistent review is inconsistent.

As I stated earlier: GTX680 still looks like a really nice card - but those benchmarks are a joke.