5900? NV4x!

Which is why I'd be agreeing with you.

 

Heh, my bad....I thought I had detected sarcasm in your "But otherwise the similarities are striking, aren't they?" statement.

 

You sir, wound me.

 

You guys aren't about to crank out any Voodoo dolls to stick pins in them are you?

 

I have been having back pains as of late. And stomach pains. But that might have been the week old tuna casserole I had.

 

Feel lucky you haven't had any "leaks" yet...

 

Doubt that.
The NV30 was a pretty big project for sure. But it was never really planned much in advance: The design began way before the April 2002, but a lot of changes happened when they bought 3DFX.
So, as I10.pdf says, you've got a project starting in April 2002 with an expected release date of August 2002, at SIGGRAPH, and with a lot of resources being wasted in nForce/XBox at the same time.
Plus, every single tape-out of the NV30 is a disaster both yield-wise and performance-wise, and even the last ones made them lose money on every NV30 they sold.

I'd say the huge costs for the NV30 don't really come from a particularly ambitious initial project - it was ambitious, yes, but I doubt the original plans ever thought it'd cost them $400M.

The money leaks come from the panic, created by the successive failures. They did risk productions at times, putting dozens of milions of dollars. Barely any chip came back.


The NV50, on the other hand, has been planned for an awful longer time than any other NV project. Its goals are to be a "full ILDP" - more so, even, than the original R400 design. It is questionable, however, how the NV50 will compare to the R500, technology-wise ( and timeframe-wise, of course, since we might be comparing the wrong chips there. )

Hehe, good comparaison! Rampage was on the drawing board for a lot longer than any other 3DFX chip too, the chips before it ( VSA-100 & V3 ) can approximatively be compared to the NV30 and NV40, and so on...
Only difference might be that the launch of the R420 and NV40 will be roughly synchronous. There won't be delays this time ( actually, there were, original plan for both were Comdex, but they both were delayed, so who cares )

Of course, the difference is all about money. Here's a calculation I did and that might absolutely terrify you, based on Q2 data:
Stockholders' Equity: 995,853
Operating expenses per quarter: 105,342

Total quarters NVIDIA can survive without ANY type of gross profit: 9.45
Assuming 4 quarters for 365 days: NVIDIA can survive 3450 days without any type of gross profit.

That's 28 months.
That means unless NVIDIA was bought off at a point or another because they'd be "cheap" and a company like Intel would like to get some of their IP... They can't die before the NV50 is in stores. It's not theorically possible.


Uttar

 

Hi guys,

Err... Where excatly could one find this pdf?

Cheers,

AJ

 

nVidia can't die in general, since they're big enough to refocus and make a business elsewhere if they decide that fighting to the death over standalone graphics cards isn't worth it. But no, I don't think they're going anywhere anytime soon.

Offhand, I'm curious what the next fight will be like. NV40 and R420 seem like they'll come out roughly at the same time, but if R500 is slated for Q4 2004... is NV40 going to have the same short timeframe? Will NV50 be ready by that point as well?

I think we're going to have some generational splits coming up, but I guess that's ok, since XGI and S3 are going to come in and confuse matters anyway--no one will notice. And hey, we STILL haven't heard from PowerVR.

Good times, good times...!

 

While I agree with the overall gist of your post, and that nVidia won't go away tomorrow, saying that they can't go under isn't true. Much, MUCH larger companies have both come and gone in this industry. Typically they don't simply die, they "merge" or are bought. Example: DEC, makers of successful mainframes, and then of wildly successful departmental (PDP) and somewhat larger and smaller (VAX-series) computers, designers of the Alpha processor. They hit hard times and were bought by Compaq, who themselves later "merged" with HP. At this point in time, DEC is gone. A much smaller company such as nVidia can be erased from the map way quicker than that.

It's the natural order of things when a company in the technology sector fails to compete effectively, but still sit on a thick patent portfolio and a wide business network.

Entropy

 

and nvidia had no big mistakes..

cheating, buying up companies to make them quiet, promising with tons of delaying, cg, non-support of the main opengl and dx directions, pointing with fingers onto everyone else while the whole world knew its their fault.

this all hurts nvidias business very much for sure.

 

At least now we know where the NV30's research budget went.

 

I doubt the R500 will come out in 2004, and it gets less likely the more delayed R420 gets.

 

I haven't heard anything about R420 (or NV40) being delayed past Spring '04...which has been the date for quite a while.

Having said that, I agree that R500 or NV50 shipping in the next cycle after NV40/R420 (fall '04) is practicially unthinkable.

R500/NV50 will be at least 12 months, if not 18 months, after the first R420/NV40 products ship. That would mean in one of the 2005 cycles.

 

I haven't heard of R420 being delayed either, and where I'm pulling the R500 info from is Uttar (and bandied about other sources), but of course there's just a lot of random speculation. Still, a "best case scenario" for Q4 2004 makes me wonder...

So far he hasn't come around and expounded on his commentary yet, though.

 

Hey AJ,

Ever heard of Google? Just kidding.
I actually didn't know of I10.pdf at the beggining of this thread, did a googld search, and found it. Simply type "NVIDIA PDF NV40".
Direct link: suif.stanford.edu/~courses/cs343/l10.pdf


Chtellis: It's R500 VS NV45 I think, both being H2 2004, Comdex maybe. Both chips are 0.11u.


Uttar

 

One interesting bit of information there is DDR-DRAM prices. (~3$ per 16MB)

 

That's a pretty silly thing to say considering what we saw with Enron and Worldcom. Not that I disagree with your point, just that saying it's not theoretically possible doesn't actually say much.

 

Well, I don't think Enron/Worldcom-like scenarios are likely here, considering NVIDIA had audits due to some minor fraud. That means everything they did in the last few years was audited. So either those guys are the greatest fraud geniuses EVER and managed to trick the FED in not seeing some much bigger frauds, or all their current results are very precise indeed.

And yes, as I said, it's really only theorical Plus, it seems impossible for NV to have 0 gross profit... That's simply insane, considering how diversified they are. I think the primary reason they could "poof" before the NV50 ( or even NV55 ) is that they'd be bought by a much bigger company, such as Intel ( unlikely considering their relationship with ATI, but who knows! ) - I just don't see them getting bankrupt, though.


Uttar

 

Don't they have a 1 billion dollars in the bank somewhere? Or does that go under stock holder equity?