NV30 Sampling in October

[GraphixViolence]

This is pretty significant information, if it is in fact reliable. If NV30 isn't shipping until January at the earliest, that puts it 5+ months after R300, basically an entire product cycle. That means it will have to compete against ATI's next product refresh (R350?), about which we know as little as NV30.

More importantly though, this is distinctly at odds with the information NVIDIA has been giving out... I don't think anyone could really mistake January for "fall". On the other hand, given today's business climate, it's hard to imagine a CEO taking the risk of appearing dishonest. Something doesn't quite fit.

 

[KimB]

Traditionally, tape out refers to first silicon. Had nVidia had any silicon, he could have answered yes. If he could have answered yes, he would have.

He didn't have hours to think of the most "PR-appropriate" answer available. He only had a few moments to think, and so the definition of tapeout that he thought of was the one that first came into his head.

Since we pretty much know that nVidia currently has beta NV30 boards, it's obvious that he wasn't talking about first silicon.

 

[Guest]

This is pretty significant information, if it is in fact reliable. If NV30 isn't shipping until January at the earliest, that puts it 5+ months after R300, basically an entire product cycle. That means it will have to compete against ATI's next product refresh (R350?), about which we know as little as NV30.

More importantly though, this is distinctly at odds with the information NVIDIA has been giving out... I don't think anyone could really mistake January for "fall". On the other hand, given today's business climate, it's hard to imagine a CEO taking the risk of appearing dishonest. Something doesn't quite fit.

Paper launch?

 

[Ichneumon]

Yes, but Nvidia can say they got it out by "fall" If they manage to have even just a few boards on shelves or available for purchase. Then they can say how they can't supply chips fast enough for the demand etc.etc...

Oh, there's plenty of ways to step around the comments that have been made so far without it being "lying" from a legal standpoint.

I still don't think we'll see volume availability until the beginning of next year, but that doesn't mean we won't see some few boards available before the holiday season.

 

[GraphixViolence]

I don't think it's that simple. To have boards ready and on shelves for the holiday season, they would have to be in production by November at the latest. Lets say their production date shipped into December, maybe they could still get a few boards on shelves just before Xmas. But January is just way too late, unless they wanted to ship with beta boards or incomplete drivers, which seems highly unlikely.

They couldn't get away with a paper launch either, because they didn't just commit to launching NV30 in the fall, they publically stated they would have product on shelves for the holiday season.

Are there any obscure holidays in January they might be referring to?

 

[RussSchultz]

A small obscure holiday known as Chinese New Years, only celebrated by about 20% of the worlds population.

Actually, I think it happens in February this year.

Amazing what silly little facts one learns working in the consumer electronics business.

 

[GraphixViolence]

Hmm, I can see the press release now...

"NV30 Now Shipping Just In Time for the Chinese New Year Holiday Season"

 

[Hellbinder]

I didn't say that the CEO misunderstood the question, I said that the people who read that statement have different assumptions about what he said.

I really wish you guys at B3D would put a stop to this. EVERYONE I have asked from inside the industry, and everyone of you here who are inside the industry ALL say the same thing about what tape-out means. Why are you letting 3-4 Product loyalists from Nvnews conduct a smear campaign on the term? The ONLY place the word tape-out is up for debate is the Nvnews forums...

Anyway, If they have samples available for October they surley they will release some boards to the press for *previews* ala 9700 style. It could be very interesting to see how that all pans out. Although they might not do that if the Nv30 does not outperform the 9700 pro by a significant margin. It may cause people to not way for it.... Looks like we could be in for an interesting chess game over the next few months. Probably why the ATI CEO mentioned a 9700 MAXX product This is gearing up to be a good ol' fassion Japanimation style battle.

 

[John Reynolds]

]I really wish you guys at B3D would put a stop to this. EVERYONE I have asked from inside the industry, and everyone of you here who are inside the industry ALL say the same thing about what tape-out means. Why are you letting 3-4 Product loyalists from Nvnews conduct a smear campaign on the term? The ONLY place the word tape-out is up for debate is the Nvnews forums...

Because people are entitled to their opinions. Yes, I think certain Nvidia supporters are being wildly optimistic over the NV30's market arrival (in volume), but they're entitled to that, IMO. Besides, it's not really all that hard to ignore such statements, is it?

 

[jvd]

ince we pretty much know that nVidia currently has beta NV30 boards, it's obvious that he wasn't talking about first silicon.

um we do ? how exactly do we know this ? sorry if i missed where this was public knowledge i'm just asking incase i missed it in a post or something

 

[multigl2]

Traditionally, tape out refers to first silicon. Had nVidia had any silicon, he could have answered yes. If he could have answered yes, he would have.

He didn't have hours to think of the most "PR-appropriate" answer available. He only had a few moments to think, and so the definition of tapeout that he thought of was the one that first came into his head.

Since we pretty much know that nVidia currently has beta NV30 boards, it's obvious that he wasn't talking about first silicon.

i've been out of the "loop" so to speak for a little while, please direct me to where there is conclusive evidence that Nvidia has NV30 beta boards.

 

[Oats]

He didn't have hours to think of the most "PR-appropriate" answer available. He only had a few moments to think, and so the definition of tapeout that he thought of was the one that first came into his head.

First, there is one defintion of tape-out. There is no room for confusion. Second, nVidia's CEO is a smart guy, understatement of the year. He is more than familiar with chip design. He knows what tape-out means. If someone was confused, it wasn't Huang. The CEO of nVidia doesn't get confused by such a question. A few moments is plenty of time. Finally, he was prepared for that question. In a PR conscious company, conferences called are prepared. He knew that question was coming. He knew what he could answer to it, reason why he tried to avoid it.

Since we pretty much know that nVidia currently has beta NV30 boards, it's obvious that he wasn't talking about first silicon.

Sure we do...

 

[RussSchultz]

]

I didn't say that the CEO misunderstood the question, I said that the people who read that statement have different assumptions about what he said.

I really wish you guys at B3D would put a stop to this. EVERYONE I have asked from inside the industry, and everyone of you here who are inside the industry ALL say the same thing about what tape-out means. Why are you letting 3-4 Product loyalists from Nvnews conduct a smear campaign on the term? The ONLY place the word tape-out is up for debate is the Nvnews forums...

As strange as it sounds, the term tapeout is very context sensitive. Chips get a tapeout, revisions get a tape out. Generally, if you ask if the design has taped out you'd mean initially. If you were talking for accounting purposes, you might mean "how many total tapeouts, including revisions" (since each one of them has a cost and a schedule), if you were both talking revisions, you still might say "has the NV30 taped out" both knowing full and well you meant this revision of the NV30.

There's been too much countering information for me to say "absolutely, positively" it was the initial tapeout they were talking about, and not some revision.

 

[TheMightyPuck]

I don't know jack about this biz so apologies if these questions annoy: I wonder what an IHVs budget per board is in order to sell in the retail channel for $400 (classic best buy high end), and what portion of that is the cost of the chip. Based on a particular die size there is going to be a required yeild (which i guess can be acheived by clocking the chips down). Anyone know what the NV30 yield needs to be for an IHV to buy it? I'm guessing companies like Leadtek are going to wait until they get a fully functioning ref board from NVIDIA and firm chip delivery dates before they start buying memory and PCBs and whatever else you need to manufacture the things. Any idea how long it takes for an IHV to get something into the channel after they get the ref. specs. and volume chips? Just curious. I know these modern manufacturers work fast and absolutely hate inventory. (Do IHVs like Leadtek etc even build without orders?). I don't think Best Buy is going to want a bunch of stuff on the shelves in January so I doubt they would make a major order without some guaranteed delivery date. Then again see caveat above Thanks for any insight.

 

[Guest]

As strange as it sounds, the term tapeout is very context sensitive. Chips get a tapeout, revisions get a tape out. Generally, if you ask if the design has taped out you'd mean initially. If you were talking for accounting purposes, you might mean "how many total tapeouts, including revisions" (since each one of them has a cost and a schedule), if you were both talking revisions, you still might say "has the NV30 taped out" both knowing full and well you meant this revision of the NV30.

There's been too much countering information for me to say "absolutely, positively" it was the initial tapeout they were talking about, and not some revision.

Hrm IIRC a tape out is normally only done the once. With revisions called "metal revs". Now it seems the NV30 is not in fact taped out you make arguments to say otherwise. Where as before when everyone thought it was taped out you made arguments for one type of tape out. I quess technically speaking you are right though. Funny that??

 

[McElvis]

I'm not sure where now, but I do also remember nVidia's CEO stating earlier in the year that they would have the FIRST DX9 capable card on the market...

Not everything a CEO says comes true...

 

[Ty]

Hrm IIRC a tape out is normally only done the once. With revisions called "metal revs". Now it seems the NV30 is not in fact taped out you make arguments to say otherwise. Where as before when everyone thought it was taped out you made arguments for one type of tape out. I quess technically speaking you are right though. Funny that??

If that's true then what was the earlier statement from nVidia about a product with several tapeouts? Can anyone refresh us as to what exactly was mentioned then? My memory isn't so hot nowadays.

 

[Joe DeFuria]

If that's true then what was the earlier statement from nVidia about a product with several tapeouts?

nVidia never said that. They said in a conference call that they anticipated a "record number of tapeouts" this fall. NOT for any one single product though.

 

[RussSchultz]

Yes, generally you'd ask "how many revisions did that part have" and not "how many times did it tapeout", if what you were getting at was how many revisions the part had.

Yes, generally, if somebody asks has the chip taped out, they'd mean initially.

But, as I stated, if I knew and he knew we had already had first silicon, I'd understand the question if posed to me to mean "the latest revision", especially if we were knew there was a revision being worked on currently. I'd probably ask the question differently though.

I think its funny as hell watching people KNOW for an ABSOLUTE FACT all sorts of stuff, simply by reading about it on a forum. Its also entertaining to see how it these ABSOLUTE FACTS usually fit some forgone conclusion, rather than the conclusion fitting the facts.

I personally don't care if it had or hadn't taped out initially or finally. I'm only posing one situation based on personal experience where these ABSOLUTE FACTS aren't so absolute or factual.

You can repeat your mantras all you want, but the facts are the facts and when the reality is revealed, it won't be any different based on anybody's prayers or wishes.

So can't we PLEASE quit beating the drum about has/hasn't taped out? We'll all know soon enough either way.

 

[Guest]

Hrm IIRC a tape out is normally only done the once. With revisions called "metal revs". Now it seems the NV30 is not in fact taped out you make arguments to say otherwise. Where as before when everyone thought it was taped out you made arguments for one type of tape out. I quess technically speaking you are right though. Funny that??

If that's true then what was the earlier statement from nVidia about a product with several tapeouts? Can anyone refresh us as to what exactly was mentioned then? My memory isn't so hot nowadays.

There were a few here that said that it had in fact been taped out. Be carefull how far you go with this some don't like to talk about it but here is the link.... Of course there is no evidance from nvidia to confirm what was said in this particular thread.

http://www.beyond3d.com/forum/viewtopic.php?t=1587&postdays=0&postorder=asc&start=0

This is an explaination of how a tape out is worked out I read it once a while back.

http://www.beyond3d.com/forum/viewtopic.php?t=1742&postdays=0&postorder=asc&start=20

Generally, the fabless company will start 5-10 wafers, but only allow one to go the full run and leave the rest untouched once the transistors are placed (apparently this takes 3-4 weeks, whereas the metal layers take less time 2-3 weeks). This allows them test the product, but still have a stock of wafers that are ready to go for whatever modifications are required.

The term "tape out" is generally only used for the initial time the database of netlists, etc is sent to the fab to make the mask sets. Every other time after that is generally called a "metal rev", because the only thing that changes is one or more of the metal layers (the connecting wires between the transistors). Very rarely is there a full layer change(i.e. throw out the wafers and start new), though even then, I don't think its commonly referred to as a 'tape out'.

Had you listened to the conference call, you would have heard for yourself that the plain english of the statements were not "NV30 had taped out a record number of times", but several statements in a row that was explaining why the engineering costs were so high for the quarter: record number of tape outs, including the NV30 which was in .13. He then went on and whatever statements following made it clear that it was multiple products and re-iterated that .13 is expensive and NV30 was in .13, and it was their fall product.