The official GT300/Fermi Thread

[Weevil]

http://www.theinquirer.net/inquirer/...-fermi-delayed

It's official, wont see it till next year.. sometime.

[Papagaj]

I hope it wasn't posted yet

[Sn0wm@n]

LOLL at pics... it explains everything

[Boulard83]

I dont think fermi will hit in november

[Blacky]

nvidia must do something fast because if they launch the fermi about Q1 2010+ would be too late for them...since ATi probably would be launching their next-gen cards in about Q2-Q3 if I remember...

[Frag Maniac]

LOL, I'm not convinced anything Inquirer is "official", but I agree, November sounds too early. And btw, does Dwayne Johnson even know how to operate a computer? One person I know of that claims to be getting first hand info from Nvidia said he last heard we'll see something in Dec (which may have changed since then), but I'm prepared to wait thru spring if need be. I don't see what the fuss is really, just look at the 5870 status and one term comes to mind AUTO-NOTIFY. Newegg has none in stock. Now ATI and their fanboys can excuse it with shipping delays all they want, but seems to me if they had such great yields they'd have already sent loads of them to distributors and/or vendors. http://www.newegg.com/Product/Produc...5870&x=13&y=28

[LordEC911]

LOL, I'm not convinced anything Inquirer is "official", but I agree, November sounds too early. And btw, does Dwayne Johnson even know how to operate a computer? One person I know of that claims to be getting first hand info from Nvidia said he last heard we'll see something in Dec (which may have changed since then), but I'm prepared to wait thru spring if need be. I don't see what the fuss is really, just look at the 5870 status and one term comes to mind AUTO-NOTIFY. Newegg has none in stock. Now ATI and their fanboys can excuse it with shipping delays all they want, but seems to me if they had such great yields they'd have already sent loads of them to distributors and/or vendors. http://www.newegg.com/Product/Produc...5870&x=13&y=28

TSMC only has a limited capacity and the equipment they just got in recently that was suppose to increase capacity actually caused some problems, though it should be fixed by the end of the month. Evergreen cards are constantly streaming into vendors/retailers every few days and OEMs are buying up decent supplys as well.

[Andrew LB]

nvidia must do something fast because if they launch the fermi about Q1 2010+ would be too late for them...since ATi probably would be launching their next-gen cards in about Q2-Q3 if I remember...

Too late for them? You guys act as if nVidia is doomed if they go a FULL 3 months without having the performance crown.

Right now nVidia is doing excellent sales with their current line of GTX2xx cards and beat their Q3 projections by 90%! If that is failing, then I'd love to see where you hold the bar for success! lol.

Which reminds me....

TSMC only has a limited capacity and the equipment they just got in recently that was suppose to increase capacity actually caused some problems, though it should be fixed by the end of the month. Evergreen cards are constantly streaming into vendors/retailers every few days and OEMs are buying up decent supplys as well.

I don't think nVidia is overly concerned about losing a bunch of market share to ATi considering the fact that even if consumers wanted to go buy an HD5870.... NOBODY has them in stock.

This situation is in striking contrast to ATi's long history of having excellent large quantities of newly released cards available at the actual time of launch. (which I recall was about 2 months ago)


It seems to me that ATi rushed out the 5870 in order to get a jump on nVidia since Fermi was still a few months away... but they blew a great chance to grab some market share by not having very many cards to satisfy consumers.

This half-a$$ed launch by ATi doesn't do much good for consumer confidence in their products and market strategy.

[drivingon9]

Some new tidbits: http://geeksmack.net/component/conte...u-at-sc09.html

also : http://www.evga.com/forums/tm.aspx?m=21965

The latter being all speculation, but still...

[Andrew LB]

Well... it looks like nVidia and ATi are both having the same problem with TSMC. It seems ALL of their 40nm high performance products are getting a crappy 50% yield.

http://www.fudzilla.com/content/view/16421/1/

[Frag Maniac]

A yield of 50%, esp when on a radically new chip architecture with a considerable die shrinkage, is actually pretty good. It's when yields on first batch runs dip below 20% that you start getting serious concerns about profit margins. Some have rumored Fermis yields to be sub 2%, so 50% would be a huge success story by comparison.

The problem is too much talk is being generated from mere rumors. ATI may be increasing their yield, but fact remains no one has 5870s in stock, and if their yields initially were so much better than Nvidia's as implied, that would not be the case. Some seem quick to defend ATI's ability to increase yield, but not Nvidia. Product development/improvement works both ways gentlemen.

[algorhythm]

Some new tidbits: http://geeksmack.net/component/conte...u-at-sc09.html

also : http://www.evga.com/forums/tm.aspx?m=21965

The latter being all speculation, but still...

Well, people are saying December and Black Friday is November 27th. I'd say either a VERY small release of GT300s on Black Friday or more likely an announcement on Black Friday of a full launch in early December. Nvidia has been very leak-proof regarding Fermi. Just because we haven't heard anything since September doesn't mean there isn't anything going on.

[zalbard]

We've heard enough to be sure that it's not coming this year.

[ggdh]

Anybody? http://www.fudzilla.com/content/view/16470/1/ Doesn't look too real... At least GPU-Z screenshot would do...

[zalbard]

I don't see why they cannot have an engineering sample already.
AMD had samples of 5xxx months and months before the release...

[Piotrsama]

http://www.evga.com/blackfriday/
The black friday deal is...... not fermi.

[BenchZowner]

They call that a good deal ? LOL ?
Guess it was Wolfram's or Aline's decision

[zalbard]

Look at ** - looks like you can pre-order Fermi if you buy this. And get it among the first people with no hassle.

[BenchZowner]

If you could exchange the GTX 285s for 2 fermi's then yes, but if they just put you on a priority list for the first batch of cards, eff it, you can get it cheaper and maybe even faster if you know some people or just "pay up" a shop in advance

[***Deimos***]

Well... it looks like nVidia and ATi are both having the same problem with TSMC. It seems ALL of their 40nm high performance products are getting a crappy 50% yield.

http://www.fudzilla.com/content/view/16421/1/

ummm.. no.

Yield is chips without defects that make unusable. Different circuits have different defect susceptibility. Bigger chips are more likely to have defects.

If your quote was true, and AMD is getting 50% yield on HD58xx, and nVidia is getting 50% on their tiny GT220 DX10.1 40nm chips, then nVidia might as well give up. A 10x bigger complex chip would have dismal pathetic yields like under 1%.

Delay wise, AMD had A2 taped out like June, 4 months before launch. nVidia just taped out Fermi A2, and rumours A3 needed to fix bugs. So we're hoping for March.

FYI TSMC has been producing 40nm HD4770 since beginning of 09. Thats ahead of AMD which started 45nm around same time, and even ahead of Intel. Just to drive down the point, AMD wont make 40nm bulk until Q2 2010.

Finally yields do improve over time.. just ask Intel

(the faster and more steeply the line goes down teh "healthier" the process... ie 45nm was a big win for Intel)

[***Deimos***]

FYI
INTEL already spent something like $6B-$7B on 32nm. With economic crisis and banks struggling to raise capital, its difficult to borrow that kind of money for all the small fab companies.

Intel has been straining silicon since 90nm and using high-K (dielectric hafnium as gate oxide between the gate and silicon substrate) since 45nm in 2007.

Unlike AMD, Intel didn't use immersion lithography (putting liquid under lens) for 45nm. Both immersion and double-patterning will be used for 32nm and below until we run into quantum effects.

Basically, its a lot more complicated and exacting and very expensive for each die shrinks. Eazy breezy Pentium3 250nm -> 180nm days are LONG gone.

Interesting discussion of TMSC 40nm manufacturing here:
http://blog.shrinkingviolence.com/20...eld-probl.html

Liu cited two materials changes, including the use of silicon germanium to strain the silicon layers, and a move to immersion lithography contributed to problems with defects, as being responsible for being a tougher step than the shift to 65nm, which only involved a change in lithography. “We encountered major issues,” he said.

Liu explained the major problem with yield lay in the stress layers produced by adding different amounts of silicon germanium.

[skugpezz]

I hope it wasn't posted yet

november is over

[zearth]

any news on fermi?

[Sn0wm@n]

someone wants to make a picture story of the fermi is comming in december??

[Wiggy McShades]

someone wants to make a picture story of the fermi is comming in december??

what?