Fermi production 3rd week in Feb

[Rroff]

I rather think drunkenmaster is correct on this even if its a mistake in the original article.

All indications seem to show low single digit "yield" for high end parts which wil be limited availability and sold at a loss even with the high price. Preference will be given to tesla parts with their higher price and the slightly lower spec to increase nnumber of viable parts. Most of whats left will be salvaged for 360 cores which would hint at corners being cut to get as many parts as possible... probably resulting in higher than normal failure rate in early batches.

 

[thevet]

i wish they would hurry up and release the bloody thing as im sick of waiting for it

 

[kieves1]

I rather think drunkenmaster is correct on this even if its a mistake in the original article.

All indications seem to show low single digit "yield" for high end parts which wil be limited availability and sold at a loss even with the high price. Preference will be given to tesla parts with their higher price and the slightly lower spec to increase nnumber of viable parts. Most of whats left will be salvaged for 360 cores which would hint at corners being cut to get as many parts as possible... probably resulting in higher than normal failure rate in early batches.

Bes look out for good warranties then

 

[MoRT489]

Excuse a potentially stupid question but does 4% yield mean 96% of what TSMC make is trashed?

 

[jigger]

Bes look out for good warranties then

XFX from OcUK or BFG seem the best... Oh wait.

 

[jigger]

Excuse a potentially stupid question but does 4% yield mean 96% of what TSMC make is trashed?

It looks like 96% of cores are failing as gt380. I think they would then be tested for the gt360, gt340 and so on until they pass.

It used to be the case that high end parts would be doctored to make up midrange numbers. Just go's to show how poor ( If the 4% is the true yield that is ) this 40nm plant really is.

I think TSMC should hire someone with a big foot.

 

[Meaker]

4% is just silly, you are not going to sell one of these cores as a gt340, it would just not make sense, te core would be huge for a mid range card.

 

[elpedro]

Great, so the chances are we're looking at an extortionate amount for the top dog (GTX 380) then quite a price gap between the lower end models. Sounds likes GTX 260 and GTX 280 all over again...

 

[FoxEye]

Great, so the chances are we're looking at an extortionate amount for the top dog (GTX 380) then quite a price gap between the lower end models. Sounds likes GTX 260 and GTX 280 all over again...

Sounds to me like every graphics card series ever produced

What top-end card hasn't been priced waaaaay more than a mid-range card at launch?

 

[Greebo]

4% is just silly, you are not going to sell one of these cores as a gt340, it would just not make sense, te core would be huge for a mid range card.

Except that Nvidia themselves have said there is no easy way to make the cores smaller for the mid range cards so they will be just "failed" high end cores.

Remember although the cores will be large, they will have lots of bits disabled so will generate less heat and need less power.

 

[han_solo]

Great, so the chances are we're looking at an extortionate amount for the top dog (GTX 380) then quite a price gap between the lower end models. Sounds likes GTX 260 and GTX 280 all over again...

Might not be a bad thing considering how well gtx 260's perform

 

[AMG]

oh here we go

let the price wars begin

 

[thevet]

XFX from OcUK or BFG seem the best... Oh wait.

no BFG so its EVGA for me

 

[Disjunto]

no BFG so its EVGA for me

no XFX either for nv, which was his point

 

[mmj_uk]

Excuse a potentially stupid question but does 4% yield mean 96% of what TSMC make is trashed?

They usually try to use them in lower end cards.

For example, even though 96% of yields don't make the grade for 5870, 80% of them may be fine for 5850 operation (with less processing units enabled and lower clock speeds).

 

[Rroff]

Might not be a bad thing considering how well gtx 260's perform

Heres hoping... as I'd rather put the money down on a pair of 360 than 1 or 2 380.

 

[straxusii]

Who expects it to be higher? Nvidia or TSMC? Surely with Fermi being a larger more complex core it'll be harder to manufacture and therefore yields will be lower? I'm not a fabrication expert but thats the way my common sense points.

The other thing I guess is they may have had some time to perfect the process with the 5xxx series so cards produced now may be higher yield?

That would explain all the 58xx cards that have been failing all over this forum

 

[elpedro]

Sounds to me like every graphics card series ever produced

What top-end card hasn't been priced waaaaay more than a mid-range card at launch?

There was less than £100 between the 5870 and 5850 on launch, compare that to the GTX 280 and 260...

Release prices:
GTX280 - $649
GTX260 - $399

 

[Greebo]

There was less than £100 between the 5870 and 5850 on launch, compare that to the GTX 280 and 260...

Release prices:
GTX280 - $649
GTX260 - $399

Which was why the gtx260 overclocked and then even with the added cost of a full waterblock outperformed the gtx280 and even a stock gtx285 for way less money.

 

[bru]

Is it possible that the production in the 3rd week of February is referring to the chips going off to the board makers. I know that would make it rather tight time wise, with the A3 silicon coming back early January then nvidia giving the go ahead, 6 weeks later (i think this is the time it takes to bake) and your into February, i know it would be tight but i suppose it is possible, and it would tie in with the leaked early march release.