NVIDIA Kepler speculation thread

[A1xLLcqAgt0qc2RyMz0y]

This is all fact is it??

Ever considered that 7970 is just the more popular SKU at the moment because people are gravitating to the fastest possible "new" thing?

The 7970 is currently in short supply.

Is it because "it is in great demand"?

or because it is hard to make (binning)?

Opinions will point to one of the above as FACT when it is just really an opinion.

Where can hard data (actual sales numbers) be obtained to prove which is actually correct?

 

[Dave Baumann]

Where can hard data (actual sales numbers) be obtained to prove which is actually correct?

Oh, believe me, I have them all!

 

[willardjuice]

I see a lot of posts about Charlie in this thread, we must be getting close to release.

 

[Shtal]

I expected that GTX680 should be slightly faster then HD7970, but Nvidia failed.

 

[AlphaWolf]

I expected that GTX680 should be slightly faster then HD7970, but Nvidia failed.

2% is 2%.

 

[jaredpace]

What does GT1 test measure? The Radeon is significantly trailing there.

 

[lanek]

Hard to say exactly why it is lower ...

GT1:
- Based on Deep Sea scene
- No tessellation
- Heavy lighting with several shadow casting lights

GT2:
- Based on Deep Sea scene
- Medium Tesselation
- Medium lighting with Few shadow casting lights

http://www.3dmark.com/3dmark11/


Something is sure, score was looking high, but this is mainly due to the Physic score and CPU used... stock 2600K = under 8000 physic score -> @5ghz = 11K physic score -> I7 3960@5ghz = over 16K ..
this make a big difference.

 

[kalelovil]

I expected that GTX680 should be slightly faster then HD7970, but Nvidia failed.

I may sound pedantic, but the GTX 680 IS faster than the HD7970 based on solely the screenshots you provided.

You could argue that Tahiti is faster than GK104 at the same clock speed, but retail products such as the GTX 680 and HD 7970 have a certain clock speed as one of their metrics.

 

[Broken Hope]

Oh, believe me, I have them all!

I'd love to see actual numbers, because I doubt it's selling anywhere near as well as the 5800/6900 series did considering the chatter/hype on message boards is no where near at the same level.

Doesn't help that you have cards like the 7800 series which are pretty much last years 6900 performance for the same exact pricing, sure the 7800 series is a little faster in some games but as a whole we're pretty much getting last years performance for last years pricing.

 

[Mianca]

You could argue that Tahiti is faster than GK104 at the same clock speed, but retail products such as the GTX 680 and HD 7970 have a certain clock speed as one of their metrics.

Most enthusiasts who pay $500+ for a video card tend to push their card to the limits, though.

So average overclocking headroom is a very relevant factor in that specific market segment.


EDIT:

On a side note, given the leaked benches Pitcairn @1006/1500Mhz would probably end up just about 20-25% slower than stock GTX680 in most benchmarks (~15% in 3DMark P-Score) - and that's a 212mm² chip using some 125W under load.

So much for people saying Kepler beats GCN in almost every metric.

 

[UniversalTruth]

Wait to see if all these results are based on a mature performance driver, not some beta version.

The 7970 is currently in short supply.

Is it because "it is in great demand"?

or because it is hard to make (binning)?

Opinions will point to one of the above as FACT when it is just really an opinion.

Where can hard data (actual sales numbers) be obtained to prove which is actually correct?

Neither? More like a limited production capacity, low number of wafer starts, etc.
The mistake here, if any, is that probably they take the current demand as a starting point and produce only as many. Not like to produce as many as possible, decrease prices and then enjoy higher customer interest.

 

[OlegSH]

I expected that GTX680 should be slightly faster then HD7970, but Nvidia failed.

7970 @ Catalyst 11.12 = P8111 in 3D Mark11, I don't think that results are ~20% higher with recent Catalyst

 

[jimbo75]

7970 @ Catalyst 11.12 = P8111 in 3D Mark11, I don't think that results are ~20% higher with recent Catalyst

Using SB-E @ 5GHz?

 

[Ninjaprime]

2% is 2%.

Launching 4 months later when they know the clocks of the competition... I'm guessing the rumored 950mhz was the real spec and it was bumped up to 1006mhz for just the reason of that percentage. If thats even a stock clocked model, may be a stock OC version. The power/size differences are nice though I suppose, but I'd like to see it in practice a little more, and see them memory and bandwidth limited as well.

 

[Man from Atlantis]

Gigabyte GTX 680 2GB already arrive at my shop
the prices confirm similar to hd7970.

OCN​

 

[Picao84]

OCN​

They have to adapt their boxes to the new geforce logo/font. It looks too ugly or unbalaced

 

[fellix]

The box label says "DirectX 11" -- no 11.1 support?

 

[UniversalTruth]

The box label says "DirectX 11" -- no 11.1 support?

Yup, that's really weird.
What's the MSRPrice of this?

 

[Man from Atlantis]

GTX560Ti SLi @900/4008MHz, P9127 Graphics Score
GT1 44.33
GT2 44.93
GT3 56.75
GT4 26.07

GTX680 @1006/6008MHz, P9195 Graphics Score
GT1 43.21 +71%*
GT2 43.26 +72%
GT3 57.60 +80%
GT4 27.45 +82%

EDIT:
GTX560Ti @1000/4200MHz, P5167 Graphics Score
GT1 25.32
GT2 25.14
GT3 31.93
GT4 14.82

*compared to a GTX560Ti..

 

[Love_In_Rio]

GTX560Ti SLi @900/4008MHz, P9127 Graphics Score
GT1 44.33
GT2 44.93
GT3 56.75
GT4 26.07

GTX680 @1006/6008MHz, P9195 Graphics Score
GT1 43.21
GT2 43.26
GT3 57.60
GT4 27.45

If you think about it there is no point in getting a GTX680 at 550€ when you can get two 560TI at 450€...