From the Nvidia conference call (Q1 FY2011):

[Skurge]

"nV fanboy: Fermi cards are selling well.
Ati Fanboy: No they arnt.
nV fanboy: OMG you are so dumb, nV will never go under!! (posts 7 million links to financials)
Ati fanboy: Dude relax
nV fanboy: You fanboy n00b fool"

Thats what this thread looks like, the nV guys are super defensive atm.

Seriously, why the hell do people keep posting financials? They won't suddenly make a bad product good or a good product bad. It's getting annoying.

 

[Lonbjerg]

"nV fanboy: Fermi cards are selling well.
Ati Fanboy: No they arnt.
nV fanboy: OMG you are so dumb, nV will never go under!! (posts 7 million links to financials)
Ati fanboy: Dude relax
nV fanboy: You fanboy n00b fool"

Thats what this thread looks like, the nV guys are super defensive atm.

Seriously, why the hell do people keep posting financials? They won't suddenly make a bad product good or a good product bad. It's getting annoying.

I don't know what you have been smoking?
When crackpot claims get slammed by facts, facts win...get over it

 

[tweakboy]

That SF was a nice comment. What are we doing awake. lol

 

[Skurge]

I don't know what you have been smoking?
When crackpot claims get slammed by facts, facts win...get over it

Hey I was just imitating your post in a more anti-nV way, you know to balance things out

Oh and I don't believe I made any claim you could disprove by facts. "financials don't make a bad product good or a good product bad"

 

[BenSkywalker]

Seriously, why the hell do people keep posting financials?

For me the financials show how shockingly wrong Charlie was yet again. 30K lifetime shipments, unmanufacturable, 2% yields etc. When he comes out saying stuff like that it is so shockingly ignorant it is hard to explain. Financials are required by law to be accurate and remove any doubt that Charlie would have had a hard time being less accurate then he was. It is one thing to bash on Fermi for being big, hot and loud- it is quite another to believe that the part is going to lead nV into financial ruin which it seemed like a whole lot of people here wanted to believe.

 

[Soleron]

For the record (and I don't know what to believe any more with regards to Semiaccurate):

Charlie says these results are higher than he expected, but that when he said this quarter would be bad he assumed TSMC wouldn't be supply-constrained this long. He says OEMs, as it is, are buying all of AMD's stuff and then still needing thousands of GPUs after AMD's run out, so they have to buy Nvidia GT 240s no matter how overpriced they are because they wouldn't be able to sell any PCs otherwise.

So he's saying when TSMC gets another line up then Nvidia won't be able to sell.

Secondly, why did Nvidia's stock drop 15% on these results? Why did their inventory go up in the quarter when they should be able to sell everything they make per above? Why did they not see an increase in revenue despite the product launches.

He still says Nvidia shipped no more than 10,000 GTX470s and 480s, and that they are sort of lying when they said 'hundreds of thousands' because those figures include the 352-shader GTX 465s and maybe even the other 40nm parts (GT 240 etc.). It's still nowhere near the millions of DX11 parts AMD managed to ship in the same timeframe.

--

His prediction is not that Nvidia will die due to Fermi, but that Llano and Sandy Bridge will obsolete the low-end and midrange graphics where the development cost of the GPUs is paid off. Then they won't be able to afford to do R&D just for the remaining high-end and Quadro products because the GPU design cost won't have been paid off. And of course their chipset revenue should be gone by then anyway, and GPGPU revenue affected by Larrabee, which will be faster at that sort of thing.

 

[Lonbjerg]

Hey I was just imitating your post in a more anti-nV way, you know to balance things out

Oh and I don't believe I made any claim you could disprove by facts. "financials don't make a bad product good or a good product bad"

No, you where just posting FUD, because the facts don't suit you...or Char-lie...try again.

 

[DanielNenni]

40nm/yield

40nm challenges on yield and ability to supply market with adequate product. Dec-Jan was a turning point, predictable improvements. Continued throughout Q1. Initial wafer starts sucked, now margin progression good, exceeded expectations and had more die available. Shipped everything they could. Going forward expect Fermi to continue to be an important element of product mix, hence improving margin.

Yield has improved a lot. TSMC done a fabulous job of improving yield since august. Yield and number of wafers has helped a lot. "Out of the woods" with respect to 40nm manufacturability. 40nm now a world class node and very good manufacturability. Capacity/supply: supply constrained due to lots of people using 40nm. wish they had more Fermis and 40nm GPUs in general. Not enough 40nm supply for the entire world. Remain constrained for some time.

Margin at Nvidia will improve due to better yields meaning Nvidia was at fault for the yield issues not TSMC. Altera yielded fine at TSMC 40nm as did others.

Hopefully next time Nvidia follows TSMC's recommended DFM guidelines, double those vias boys.

TSMC has (2) Gigafabs making 40nm wafers and will start building a third this year. TSMC owns 40nm and will probably own 28nm. ATI will stay with TSMC at 40nm and 28nm, they will have no choice for capacity and time-to-market. GFI will be second source.

 

[DanielNenni]

I think Fermi will be fantastic with a die shrink.

It's just a massive chip.

Here is what Fermi will get from a 40nm -> 28nm shrink:

http://tinyurl.com/2bhqu4e

 

[Acanthus]

Here is what Fermi will get from a 40nm -> 28nm shrink:

http://tinyurl.com/2bhqu4e

The question for next round is "which design will end up scaling better".

 

[DanielNenni]

The question for next round is "which design will end up scaling better".

Scaling as in how many processor cores? Can the full 512 cores fit / function in 28nm?

 

[Acanthus]

Scaling as in how many processor cores? Can the full 512 cores fit / function in 28nm?

Scaling as in overall performance gained for their architecture.

I would imagine that neither Nvidia nor AMD will just do a die shrink and be happy with it. Both of them will likely add more "processor cores" or tweak the design in other ways.

What i meant was, who will benefit more from the shink + tweaks to the architecture.