NVIDIA Kepler speculation thread

And, with that quite low clock, would that mean hot clock is still there?

Either it's still there or other stuff has moved into the hot-clock domain and the rumours misinterpreted that to mean that the entire chip ran at a higher clock.

Doubtful this is the 680. More like a 670 or even a 660.

Possibly but how are they going to meet demand? As of right now the 7970 is completely out of stock at Newegg and that's a $550 card.
 
its better compare GT2 and GT4 scores.. as source stated other tests have very low results because of driver issue or something, it lowers final score.. and graphics tests aren't effected by cpu.. unfortunately Futuremark site doesnt let me find stock clocked GTX 580 and HD7970 to compare..
 
Fake! Revealed by Photoshop gradiation curve: http://img.techpowerup.org/120116/Shot-20120116-182756.jpg

>1,1GHz 7970 with photoshoped P score.

Nice. I was about to mention that the GPU clocks were the default specs of the DDR3 GT240.

550Mhz core/1580Mhz (790x2) Memory
specsx.jpg
 
But the screenshot itself had the GT240 listed as the GPU, all this detective work wasnt really necessary :LOL:

I knew that...or are you assuming I'm blind?
I also know that programs that don't recognize a GPU will usually report a default lowest end card from their database if they can't read the actual clock speeds. It happened with the 4870 and 5870 using GPU-Z so I was covering all bases.
 
I knew that...or are you assuming I'm blind?
I also know that programs that don't recognize a GPU will usually report a default lowest end card from their database if they can't read the actual clock speeds. It happened with the 4870 and 5870 using GPU-Z so I was covering all bases.

Haha, ok Sherlock. Fair enough.
 
If it's only as fast as the 7970 then it has better be only a 670 or lower or that's pretty damn rubbish. Especially given the potential for serious performance increases from driver improvements on the 7970 and AMD's obvious ability to release a faster SKU if they need to.
 
If it's only as fast as the 7970 then it has better be only a 670 or lower or that's pretty damn rubbish. Especially given the potential for serious performance increases from driver improvements on the 7970 and AMD's obvious ability to release a faster SKU if they need to.

Not that there's anything that really makes much sense in that chiphell link, but I'm afraid they weren't pointing at the top dog of the Kepler line and no I don't believe it either. Again irrelevant when NV will launch its new family of products, until they will it's still silly season and it's expectable that we'll read and hear a whole LOT of nonsense.
 
If it's only as fast as the 7970 then it has better be only a 670 or lower or that's pretty damn rubbish. Especially given the potential for serious performance increases from driver improvements on the 7970 and AMD's obvious ability to release a faster SKU if they need to.

The 680/670 will probably fall in the 225-250w range so if nVidia's claims of improved perf/w aren't complete horseshit then both parts should be appreciably faster than a stock 7970 @ 190w. The only question is whether they arrive so late as to be irrelevant.

I'm mostly looking forward to comparisons between chips of similar size and power consumption. The architectures are converging so it'll be interesting to see who comes out on top at 28nm.
 
The 680/670 will probably fall in the 225-250w range so if nVidia's claims of improved perf/w aren't complete horseshit then both parts should be appreciably faster than a stock 7970 @ 190w. The only question is whether they arrive so late as to be irrelevant.

I'm mostly looking forward to comparisons between chips of similar size and power consumption. The architectures are converging so it'll be interesting to see who comes out on top at 28nm.

If it's releasing appreciably later than 7970, it should be (or at least better be) faster, especially if, as expected, it's also noticeably larger. Just like whatever AMD releases after that should be faster.

Right now the two companies are playing leapfrog with unaligned launch windows. IMO, that's a good thing. Competition keeps both companies on their toes.

Even if it doesn't launch in Feb./Mar. and instead launching in May/June or whenever, the leaks and rumors and everything else will do a good job of keeping some of the Nvidia faithful from jumping on the 7970.

Not saying this is true. But I can already see, rumors of GTX 680 (or whatever) in Feb. So just wait a month and see. Then rumors of March. So just wait another month and see. Then rumors of April. Etc.

Or it'll be launched in Feb. That uncertainty is great for a company with no current response. :)

Regards,
SB
 
If it's releasing appreciably later than 7970, it should be (or at least better be) faster, especially if, as expected, it's also noticeably larger. Just like whatever AMD releases after that should be faster.

I don't think there's much expectation at all of being faster simply because it launches a few months later. That gap is pretty insignificant when it comes to GPU design cycles. Also, as we saw with Fermi delays can be completely due to manufacturing issues. Kepler could very well be having similar problems.

Even if it doesn't launch in Feb./Mar. and instead launching in May/June or whenever, the leaks and rumors and everything else will do a good job of keeping some of the Nvidia faithful from jumping on the 7970.

I think the price, limited supply and nominal improvement over the 580 are doing a better job of that :) At the asking price the target market consists of ppl who are currently running 580's, 6990's, multi-GPU setups etc.

Not saying this is true. But I can already see, rumors of GTX 680 (or whatever) in Feb. So just wait a month and see. Then rumors of March. So just wait another month and see. Then rumors of April. Etc.

Or it'll be launched in Feb. That uncertainty is great for a company with no current response. :)

Heh you were expecting a more reliable launch window than "2012"? :D Honestly they don't need a response just yet because their current lineup isn't hurt by the 7970 at its current asking price. Now if Pitcairn drops soon at $249 with GTX 580 performance then they can start worrying.
 
Honestly they don't need a response just yet because their current lineup isn't hurt by the 7970 at its current asking price. Now if Pitcairn drops soon at $249 with GTX 580 performance then they can start worrying.

You think the $499-$580 GTX580s aren't hurt by the faster $549 7970s? That seems unlikely.
 
You think the $499-$580 GTX580s aren't hurt by the faster $549 7970s? That seems unlikely.

Well yeah there will be lost 580 sales. But those sales aren't going to be saved by nVidia's shenanigans anyway. If people are playing the waiting game for 7970's they certainly aren't running out to buy 580's either.
 
Wouldn't it be awfully close to expected HD 7950 performance then?
Personally I think HD 7950 should end up with performance somewhat above GTX580 level, so if top-end Pitcairn ends up slightly slower than GTX 580 that could be enough difference. Though I've no idea if it will be really priced at ~250$.
 
Back
Top