[Carl B]

Originally Posted by neliz

True, DX10 architecture.
But we assume NV50 = G80 ... NV50 was explicitly denied as being unified while Jen Hsun said that G80 would (possibly) be..

Very true. Back in the day when people were talking in terms of NV47 and NV50, rumors were that NV47 and NV50 got canceled; of course we saw NV47 go to G70 and herald the beginning of a 'half-step' generation for NVidia founded on NV40 tweaks. There's at least some reason to believe that G80 may not represent NV50 at all, but rather NV60.

[Geo]

http://www.theinquirer.net/?article=29314

This seems to also point at March for G71. Actually, if he's got his facts right, it sounds like they are on second rev, which is often the sweet spot for these things.

[Joe DeFuria]

Unfortunately, he can't possibly have his facts straight...if it taped out in early January, I don't believe there is any way they could actually have samples up and running right now. (It just takes physically longer from tape-out to delivery.) I assume that they do have samples up and running now, and it actually taped out earlier. I'm guessing that nVidia received the first samples earlier this month, and now they are "up and running" in some fashion.

So, it's hard to tell exactly where they are in the test/debug process, but I would guess that an early March "hard launch" at this point would be a best case scenario.

[Geo]

He seems to use "tape out" to mean getting first samples back. So when he starts early January vs early March, I start wondering if they went back for another spin.

[Joe DeFuria]

Originally Posted by geo

He seems to use "tape out" to mean getting first samples back. So when he starts early January vs early March, I start wondering if they went back for another spin.

If January is "when they got the first samples back that will not require a respin," then I'd say Early march (one month from now) is a best case scenario for a hard launch. If a (another?) respin is required, I think you'll have to add on another month at least.

The key question in my mind is: are the January samples the very first ones for G71? Or are they the samples of expected "final silicon"? It's possible that the answer to both questions can be "yes", but unlikely.

[Ailuros]

I'd rather ask first what Fuad specifically understands under "tape-out" before even start speculating over it.....*runs*

[Razor1]

when was the g71 first in drivers? Wasn't it well before the release of the GTX 512?

[Sunrise]

I wouldn´t read that much into it. Everything that can be said about INQs new rant is that Joe actually just proved that Fuad´s "knowledge" is more than a little bit flawed here. There´s a lot of guesswork in there, even to the point where the whole post looks rather silly. So NV can actually ramp it up in about one month´s time, have enough chips to please AIBs and release it at the end of February / beginning of March ? Are we even still living on earth ?

He even contradicts himself by stating "It is a long time from tape out till final chip and we know that Nvidia taped out the chip in the first half of January."

Please gimme a break.

[kemosabe]

First vague rumour of a possible G71 yield issue. Release the hounds!

[CJ]

Originally Posted by geo

Since VR-Zone tends to be a little more accurate when Big Green is in the crosshairs, I lean towards taking this seriously: http://www.vr-zone.com/index.php?i=3168

7600 is now after CeBIT (which ends on March 15, btw)? Have we reached a place yet where we can officially call it "late"? And I'm still not seeing the Signs and Portents we would expect heralding G71 launch as imminent (say within a month).

Well this kinda explains the"G73 -> 8-24-12-16 CATZ".

G73 would be 8 ROPs (or Color Ops), 24 ALUs (12 fragment pipes), 12 TMUs and 16 Z/Stencil Ops. Looks like a worthy challenger to the X1600XT which will probably get beaten by a RV560. And the G73GS should be 4-16-8-8.

I'll try to double check if the G70 figures weren't a typo with the 24-48-24-48 instead of a 16-48-24-32. But if it was indeed a typo, that would make G71 one weird card.

[Jawed]

Originally Posted by kemosabe

First vague rumour of a possible G71 yield issue. Release the hounds!

I released mine long ago...

Jawed

[Sxotty]

When has there not been a yield issue on a top end SKU lately? Hmm never...

[Arun]

Originally Posted by Sxotty

When has there not been a yield issue on a top end SKU lately? Hmm never...

G70 had some insanely good yields and margins.
And I'm not trusting those "yield issues on G71" stuff. They might have required more respins than they had hoped, but that's all.

Uttar

[Matasar]

Has there only been 1 tape out ?

[neliz]

Originally Posted by Matasar

Has there only been 1 tape out ?

If you read the article, you'd know they're at rev. A2.


But G70, wasn't it late then? Ati was showing the r520 in march and was hoping for a may launch at the latest. at least the G70 was "later" with a end of june launch.. while all "we" got was a 512mb x800

[Hellbinder]

Originally Posted by Sxotty

When has there not been a yield issue on a top end SKU lately? Hmm never...

R580 has no yield issues. Nor did it have yield issues. The issues were worked out on the R520 at roughly the same time.

[Chalnoth]

Originally Posted by Hellbinder

R580 has no yield issues. Nor did it have yield issues. The issues were worked out on the R520 at roughly the same time.

How do you know for sure? January is, after all, a strange time to launch a product.

[Hellbinder]

Originally Posted by Chalnoth

How do you know for sure? January is, after all, a strange time to launch a product.

muahahahahahaha, i have ways to know things,,, yes i have waaaaaaays.

[IbaneZ]

Sweet. ATI as back on top and HB is suddenly an ATI fan again.
Where are the other turncoats?

[Sxotty]

Uttar, It doesn't count on g70 b/c their was no competition. When they came out with the 512MB edition it seems things weren't so peachy.

Same thing with x1900, wow they can beat a year old piece of hardware, kudos. Anyway it seems parts that come out in competition to a higher part have yield issues b/c they aim for too aggressive of clocks to beat the competition.

[Chalnoth]

Originally Posted by Sxotty

Same thing with x1900, wow they can beat a year old piece of hardware, kudos.

Well, the G70 isn't a year old. It was released sometime around July-August, if I remember correctly. But what's impressive is that it still holds up pretty well, even on a larger process but with the same die size, while being a nearly two year old architecture.

[neliz]

Originally Posted by Chalnoth

How do you know for sure? January is, after all, a strange time to launch a product.

They delayed the R580 launch for a month.. it was planned to be a december thing from the start.. only to save x1800 sales I guess..

[neliz]

Originally Posted by Chalnoth

Well, the G70 isn't a year old. It was released sometime around July-August, if I remember correctly. But what's impressive is that it still holds up pretty well, even on a larger process but with the same die size, while being a nearly two year old architecture.

June 22 was the launch..

[digitalwanderer]

I just don't see how they'll have availability for a March 9th launch.

[CJ]

Originally Posted by Chalnoth

How do you know for sure? January is, after all, a strange time to launch a product.

The fact that there aren't any gimped parts out there with 1 quad disabled could indicate that ATi has good yields... unless they're stockpiling those cores for the X1800/X1900GT(O).