NV30 not taped out yet....

[Mephisto]

IMO, it seems quite unlikely that we see NV30 on the shelves this Xmas.

The NV20 ramp-up may be comparable to the NV30 (new process, new architecture). The NV20 had its first tape out sometime in September/October 2000, the part needed 5 respins (A5 was the production chip) and was on the shelf in May 2001. That is 8 months (!) from the initial tape-out to the _market_! If everything goes smooth, they manage to bring this down to 6 months, everything less is wishfull thinking.

 

[Vince]

I think this is being blown out of proportion in light of the recent announcements by the competition and their announed plans.

While the comment "wrapping it up" is a bit suspect, as it might not have reached GDSII tape status (been awhile, but I think thats what it's called), which means TSMC couldn't have gotten it - nVidia is a company that has historically be able to work quickly.

Your underestimating their talent and resources devoted to process and test engineering. They're cooperation and relationship with TSMC has, historically, been great and I highly doubt TSMC will give nVidia the preverbial finger when it comes to fab resources and production. IIRC, nVidia used Fabs 3, 4, 5, 6?, 8 and Wafertech (outsourced threw TSMC's investment in them).

So, aslong as the design engineers did their job well and the tools (Anyone know what nvidia uses? - Avant! Star-RCXT?) did their job, nVidia should be fine. Hopefully for nVidia, it's only a few errors that are in the upper metal layers and can be fixed quickly.

Going to be interesting to see though.

 

[Nappe1]

Memphisto: I fully agree.

 

[BoardBonobo]

With the number of people leaving and stepping down it would seem that nVidia is having a few internal problems. Have they had an internal split into two camps? One saying that they should invest soley in IQ and another wanting to retain the FPS crown. If some people want to move to a 256bit memory bus in order to keep the bandwidth up and they can't deliver the performance they want currently could that be causing a few problems? I can see a whole load of very nervous engineers burning the midnight oil.

Also is it possible that they may have dev silicon running on .15u, at a very low core speed, for driver development?

 

[nggalai]

With the number of people leaving and stepping down it would seem that nVidia is having a few internal problems. Have they had an internal split into two camps?

Hmm, I don't think so. Sometimes, people just feel like doing something new. Richard, for example, has been an "employee" for over 20 years, and obviously finally wanted to start something on his own, not the least to have fun with it. The time couldn't have been chosen better, either (regarding the market), and The Code Mafia already is in high demand for code reviews and the like. Good move.

About the whole taping-out situation: time will tell. *shrug* I expect (or suspect ) a paper launch in autumn, with production ramping up later, and the final boards being shipped in early 2003. But that's just a hunch, based on nothing, really.

ta,
-Sascha.rb

 

[Dio]

How many of nvidia's employees became millionaires after the massive rise in their stock options?

That sort of makes you lose your drive a bit, I'd guess....

 

[Evildeus]

How many of nvidia's employees became millionaires after the massive rise in their stock options?

That sort of makes you lose your drive a bit, I'd guess....

Well depends on the vision, but how many did have some big debts after the collapse of the stock?

On the topic, i think that nVidia is now clearly late in the release of NV30. I think that 2003 could be the time now leaving a great place to Ati to ramp up the production of 9700.

 

[DemoCoder]

How many of nvidia's employees became millionaires after the massive rise in their stock options?

That sort of makes you lose your drive a bit, I'd guess....

Well depends on the vision, but how many did have some big debts after the collapse of the stock?

Well, anyone who was granted options in beginning 2001 or earlier has already partially vested and could have sold before the long side down. For instance, in Jan'01, the stock was about $20, and in Jan 02, it would about $60. If you were granted 10,000 options (a standard industry size block for good engineers) on a 4 year vesting schedule, you would have vested 2,500 by Jan'02, and could have sold for a gain of $40/share or $100,000 pre-tax. If you were working at NVidia years before, say, since 1999, and you got multiple option grants (fairly standard for high achievers), you would definately be doing pretty well.

Consider that your strike price was less than $10, likely $5, and you had 3 years of vesting, possibly two separate grants, or roughly 7500-10000 shares, and you sold near the height, say $60, you would earn close to $500k.

Of course, in Silicon Vallery California, a million dollars is only upper middle class. The average house price ranges from $500,000 to $1mill for a descent 1800 square foot in the bay area.

Employees leave companies all the time for various reasons. Microsoft has a hard time preventing its employees from retiring or dropping out to start companies, and certainly Microsoft isn't a sinking boat.
R Huddy has his own personal reasons. Back in the .com heydays, people changed jobs every 6-12 mos in the bay area. The company I current work for does very well, but I still want to leave after my next round of options vest to start another company, because I'm not happy unless I have control over the direction and vision of the stuff I am working on.

 

[Guest]

The top end graphics card(Meaning the company that carries the performance crown for a prolonged time periode like we are about to see ATI do.) maker garners much attention in the OEM market this is how nvidia has managed such a large market share for some time. Now as a result of the loss of the high end market their low end market is in serious jeaprody considering how poor their low end products really are. How much will nvidia loose and how much will ATI gain is the question to ask at this point IMHO.

The only problem with your theory is that NVidia hasn't lost the performance crown yet (no R300 currently on shelves) and that even if it does lose the performance crown (I think we can safely predict this), it will only be for a few months (assuming NV30 performs as well as R300). NVidia has had the performance crown for years and they have had delayed cycles before.


I personally don't think what attracted OEMs to NVidia was the performance of their high end parts. I think what attracted OEMs was NVidia's consistent execution, timing new chips with new features to OEM release schedules. After all, the consumers technically don't buy "NVidia" GF2MX cards, but buy value cards with totally unrelated name and branding. I doubt the average GF2MX owner even knows what a GF4-4600 is!

So, NVidia could lose some design wins, not based on chipset performance, but merely because they miss an OEM cycle.

My personal opinion is that the R300 and NV30 are going to be so close in performance that it doesn't really matter performance wise. NVidia has missed Christmas before and still recovered.


I just don't see why so many people keep making this out to be some kind of war where one company has to win absolutely. The fact that we will have 2 DirectX9 parts soon with roughly equal performance and features is a HUGE BOON to us gamers. This is quickly becoming a 2-horse market like with Intel and AMD and do you honestly think we would be better off if either Intel or AMD was destroyed?

So why does it seem so many ATI fans are hoping for Nvidia to lose big time?

The problem with my theory is that it isn't mine. I got the idea from an ATI CC. Anyhow why would an OEM be interested in nvidias execution if they are still using a GF2mx series cards? No I disagree about OEMs being concerned about execution of next gen cards. They want the brand name attached to the top performing card.. Mid range/high end cards are where you see most of the "execution" occurs, yet these cards account for very few OEM deals.

 

[pxc]

The NV30 schedule isn't doomed -- yet. If NV is still making excuses in a few weeks, the fault being NV's or TSMC's doesn't matter, they are in trouble. Or at least their stock price and growth has little chance of recovery in the next several months if that happens.

It's not like they won't have at least 2 refreshed products (die horrible NV17, hello decent NV18 and the nForce2), with cheaper NV28s and NV18s selling in higher volume than high end chips. But if the NV30 does miss the whole 2002 shipping period, they will be losing marketshare form the 42% they currently have.

 

[mpegleg]

Here's the more complete data.

Project/FCS Mask Set/Tapeout to Volume

Riva TNT2 / A02 / 98 days
GeForce256 / A03 / 104 days
GeForce2/GTS / A03 / 102 days
GeForce/MX / A01 / 110 days
GeForce3 / A03 / 118 days
XGPU / A03 / 120 days
GeForce4 / A02 / 101 days

 

[Tahir2]

All are A0x based
I assume the 'x' stands for revision.

The only product that had one revision before Tapeout to volume was the Geforce MX variant - assuming that was because the chip was simply a slightly modified G256.

Only two revisions for the GF4 is very impressive by the way. Is there any data on the time between each revision in each instance?

 

[DemoCoder]

No I disagree about OEMs being concerned about execution of next gen cards. They want the brand name attached to the top performing card.. Mid range/high end cards are where you see most of the "execution" occurs, yet these cards account for very few OEM deals.

I didn't say execution of high end cards, I said "execution". If I am an OEM, I want to ship out new products before the next buying season. This means I have to have a new product ahead of time to do all the production, packaging, marketing, and injection into the channel.

Let's say OEM #1 goes with NVidia, and OEM #2 goes with ATI. NVidia misses a cycle, no new GF5MX. ATI doesn't, new value product available.

OEM#1 will be severely disadvantaged in the next season and will lose alot of sales contracts. They will have to sell existing inventory and/or try to push last years product. OEMs are not cash rich, most of them are relatively small companies who are tightly coupled with the seasonal buying patterns, and the big PC vendor's cycles. They have very thin margins on product.

Nvidia did well with OEMs because 3dfx pissed them off, and NVidia consistently delivered a new range of parts every 6 months. For most of these OEMs, it was NVidia's constantly tweaking and rebranding of the value chips that helped them the most.

In business, being able to deliver on time is one of the most important things.

 

[Mephisto]

Here's the more complete data.

Project/FCS Mask Set/Tapeout to Volume

GeForce3 / A03 / 118 days

A05 was the GeForce3 production version, not A03, so these 118 days aren't the time till NV20 was production ready.

 

[RussSchultz]

Usually the Ax is the revision number, but its not necessarily. The R300 (I think...maybe the R200) was A12, for example, and I'd be mighty suprised if they had 12 revisions.

 

[mpegleg]

Not sure about that Mephisto, I think it depended on the card vendor and your definition of production-ready. There were very early cards that shipped with A03. A05 was definitely the "real" volume rev but Malachowsky was obviously taking some artistic license...

 

[LeStoffer]

Well, it kinda explain why we never heard any of the "first NV30-silicon is back and working" rumours that I suspected would be leaked by now to counter the Radeon 9700 launch. It also explain the rumours that we might not be able to buy a NV30 in 2002.

Okay, two things:

First, I find it to be a bit shady to make any feature launch before you have the first rev silicon back and know about the problems you're might be facing before the chip can reach full production (this applies to all products and vendors).

Second, I do find it a bit fishy if we're really talking about the very first tape out of NV30 when people (read: investors) had reason to assume that the thing did in fact tape out in June. Maybe the CEO actually was talking about the final "tape-out", e.g. a production ready revision.

All in all, I think that we will probably see a papir launch in early September (when they have working silicon back). Then we'll have previews on samples very late in October (without disclosed benchmark numbers), a launch in November and the shipping product out in December (but good luck finding one in your local store...).

 

[Mephisto]

Not sure about that Mephisto, I think it depended on the card vendor and your definition of production-ready. There were very early cards that shipped with A03.

AFAIK, A03 still had some bugs (3D textures not working correctly for example) and was not intended for sale. Though it was the part that got into the Xbox developer systems and that was on the first media samples.

 

[OpenGL guy]

Here's the more complete data.

Project/FCS Mask Set/Tapeout to Volume

Riva TNT2 / A02 / 98 days
GeForce256 / A03 / 104 days
GeForce2/GTS / A03 / 102 days
GeForce/MX / A01 / 110 days
GeForce3 / A03 / 118 days
XGPU / A03 / 120 days
GeForce4 / A02 / 101 days

Correct me if I am wrong, but I thought these were the number of days from the final tapeout to volume shipping. Thus, if it took three revs to get it right, you need to add in the time for the two previous tapeouts.

 

[Unknown Soldier]

Not sure why you guys think the NV30 has been/would be tapeout.. i mean the last time i heard that the CEO said the NV30 would be released at the end of November. Meaning yes if it tapeouts from now till end Nov. it would be 120 days .. but if they tapeout in 90 days that would give them till end of August beginning September to tapeout. Saying that a 100day Tapeout would be sometime in the middle of August.

Well.. all the best of luck to Nvidia.. and even though the shares might have fallen abit .. i wish i had a few of those shares lying around.

US