SemiAccurate.com: AMD vs. NV in 2012 - Kepler by TKO

[caveman-jim]

Famous for his 'hot, slow, late, unmanufacturable' characterization of NVIDIA's Fermi (fastest single GPU card in 2010, 2011), Charlie Demerjian has interested news about NVIDIA's 28nm products, using a new architecture codenamed Kepler:

The short story is that Nvidia (NASDAQ:NVDA) will win this round on just about every metric, some more than others.

Read the full story at Semiaccurate.com

Unless, of course, this is a trick by S|A to catch people reporting news broken by S|A without crediting, by changing key details (like 'will win' should be 'will lose').

[bittermann]

Downgrading to 256 bit memory bus....unless it's the 560 replacement then I call BS...

[caveman-jim]

IF the codename schemes continue, then yes, GK104 is the GF114 replacement - so GTX 560 replacement.

[BababooeyHTJ]

Has there been any good leaks on 7870 specs? Either way I just hope that both of these hit the market soon with some availability. I just want to see the price vs performance bar move.

[noko]

Charlie got it wrong with Fermi, maybe he figures if he reverses his viewpoint with Kepler he will get it right .

Not much in this article, maybe he lost to many Nvidia fan base so a little string here and there will bring back some attendance.

I do expect Kepler to be strong through. With some real nice surprises to boot . I can't tell ya though because some mole might get decapitated.

[acroig]

Charlie got it wrong with Fermi, maybe he figures if he reverses his viewpoint with Kepler he will get it right .

What did he get wrong with Fermi (the 400 series)? He said it would be late and hot and he was right.

[Riou]

Charlie was also right about the wood screws on the fake GTX 480 Jen-Hsun was holding.

[acroig]

Charlie was also right about the wood screws on the fake GTX 480 Jen-Hsun was holding.

, tru.dat!

[noko]

What did he get wrong with Fermi (the 400 series)? He said it would be late and hot and he was right.

A lot of stuff pure wrong: yields, unmanufacturable statements, 400 series was produced, made money, Fermi was rapidly updated with 500 series which turned out good. Mostly he was right about being late (which I predicted before he did and got the date released withing 2 weeks , pure luck on my part ) and taking more power but in the scheme of things it worked out. Plus about wood screws, that was pure hilarious .

[hazindu]

The short story is that Nvidia (NASDAQ:NVDA) will win this round on just about every metric, some more than others.

I don't see much reason to believe this, but I sure hope it's true. It would bring AMD's prices down.

[acroig]

A lot of stuff pure wrong: yields, unmanufacturable statements, 400 series was produced, made money, Fermi was rapidly updated with 500 series which turned out good. Mostly he was right about being late (which I predicted before he did and got the date released withing 2 weeks , pure luck on my part ) and taking more power but in the scheme of things it worked out. Plus about wood screws, that was pure hilarious .

Actually he got a lot right on the 400 series. A full CUDA core part did not arrive until the 500 series, it was hot, it was late and it took the 500 series to get Fermi done right.