5900? NV4x!

[Frank]

When the 5800 failed to deliver the required preformance, the improved 5900 was there very fast to repace it.

In how far could you tack another FP unit to an existing design almost overnight? Or replace a FX unit for a dual FX/FP? In a few months?

I think it is highly likely, that that was the main improvement the NV4x would bring.

What do you think?

 

[AlphaWolf]

I think you won't be seeing any NV4x for at least 6 months.

 

[Natoma]

The only reason the 5900 seemed to appear so quickly is because the 5800 was so late and an utter failure. If the 5800 had appeared on time it still would have been an utter failure against the 9700 Pro, but the 5900 would not have appeared when it did to "save" Nvidia.

 

[Frank]

I think you won't be seeing any NV4x for at least 6 months.

Agreed.

 

[Frank]

The only reason the 5900 seemed to appear so quickly is because the 5800 was so late and an utter failure. If the 5800 had appeared on time it still would have been an utter failure against the 9700 Pro, but the 5900 would not have appeared when it did to "save" Nvidia.

Is that all there is to it?

 

[zsouthboy]

nv4x = Feb 2004, at the earliest.

 

[Frank]

nv4x = Feb 2004, at the earliest.

Yes. Will the NV4x have significant improvements over the 5900? And if so, is it really the rushed NV5x?

 

[nonamer]

nv4x = Feb 2004, at the earliest.

Yes. Will the NV4x have significant improvements over the 5900? And if so, is it really the rushed NV5x?

AFAIK, the NV40 has been in developement for a while. Before the 5800 was out in fact.

 

[Ostsol]

AFAIK, the NV40 has been in developement for a while. Before the 5800 was out in fact.

That wasn't actually all that long ago. . .

 

[mboeller]

nv4x = Feb 2004, at the earliest.

Yes. Will the NV4x have significant improvements over the 5900? And if so, is it really the rushed NV5x?

AFAIK, the NV40 has been in developement for a while. Before the 5800 was out in fact.

According to the leaked (?) I10.pdf from Nvidia the NV40 is in development since around April 2002.

 

[Arun]

Yep, nice info that is in this I10 presentation indeed...
What I find most interesting though is for how long the NV50 has been in development. I think the project began pretty much as the same time as the NV40.
The NV50 is NV's "mega project", the thing that's supposed to let them get/keep the market for the full generation time with excellent flexibility allowing them to put derivatives it in all types of devices.

If they **** up the NV50, their core market won't be able to survive it. They just can't allow it to fail. I'm sure, however, that the budget numbers for it will be significantly more impressive than the NV30's $400M one. If the NV40 fails, they won't die over it. But if the NV50 does...


The NV40 most certainly isn't a rushed NV50 AFAIK.


Uttar

 

[zsouthboy]

nv50 = TBDR? Please?

 

[Sage]

nv50 = TBDR? Please?

well if you look at the recent presentation on "The Elegance of Brute Force" Kirk pretty much says that deferred rendering totally blows and the only way into the future is brute force... of course a few years ago they were telling everyone about how much PS1.4 sucked and it would never get used and now they are trying to get devs to use PS1.4 whenever possible...

 

[Arun]

Sorry, NV50's ain't a TBDR. It is an ILDP last I heard, though!

It's amusing how he talks about Brute Force so much in that presentation - I don't quite understand how ILDP is brute force, frankly: Although it does allow to use the brute force better, kinda like if you took a really subtle approach to get to the station where you got to punch someone in the face


Uttar

 

[Brimstone]

Yep, nice info that is in this I10 presentation indeed...
What I find most interesting though is for how long the NV50 has been in development. I think the project began pretty much as the same time as the NV40.
The NV50 is NV's "mega project", the thing that's supposed to let them get/keep the market for the full generation time with excellent flexibility allowing them to put derivatives it in all types of devices.

If they **** up the NV50, their core market won't be able to survive it. They just can't allow it to fail. I'm sure, however, that the budget numbers for it will be significantly more impressive than the NV30's $400M one. If the NV40 fails, they won't die over it. But if the NV50 does...


The NV40 most certainly isn't a rushed NV50 AFAIK.


Uttar

The NV50 is Nvidia's mega project and Microsoft didn't go for the technology? Obviously more than technology has to be considered for a console. Cost being a major factor. Personally I think the NV40 will be a big step for Nvidia, not so sure about NV50.

 

[Joe DeFuria]

The NV50 is NV's "mega project", the thing that's supposed to let them get/keep the market for the full generation time with excellent flexibility allowing them to put derivatives it in all types of devices.

Incidentally...wasn't NV30 the same? (nVidia's "mega-project?").

If they **** up the NV50, their core market won't be able to survive it. They just can't allow it to fail. I'm sure, however, that the budget numbers for it will be significantly more impressive than the NV30's $400M one. If the NV40 fails, they won't die over it. But if the NV50 does...

Interesting...that's the exact same line of thought I had with 3dfx and the V3 / VSA-100 / Rampage. The V3 wasn't a great success...and then the VSA-100 problems. I thought that 3dfx would have little problem shaking off the VSA-100, and I thought that "Rampage" was it. (If 3dfx f*cked up Rampage...they'd die.)

Of course, I was wrong.

I underestimated the VSA-100 failure.

 

[RussSchultz]

3dfx didn't have a bundle of cash to fall back on, though.

Plus they made many other REALLY big mistakes, business wise.

But otherwise the similarities are striking, aren't they?

 

[CorwinB]

3dfx didn't have a bundle of cash to fall back on, though.

Money is everything.

Plus they made many other REALLY big mistakes, business wise.

Historically, their biggest mistake was to buy STB, I think. That both added loads of problems to their management (inventory, fab time and the like), as well as mightily pissed off their previous board makers, who went straight to Nvidia.

Other mistakes 3dfx did :
1) Piss off MS with Glide. Well, MS seems pretty pissed off at NV (with driver crap and CG) right now, with recent shots such as "we endorse the HL2 benchmark", the ATI XBox2 deal... I think that for Longhorn, MS needs customers to be absolutely sure that the graphical drivers and interface won't crap the computer. And, I was chatting with a guy from MS one day, and he said that on home PCs, graphical drivers (from all brands) were the biggest cause of BSODs...
2) 3dfx did piss off some developers (with late to the market 32-bits parts, specific API, no T&L...). Well, guess who is pissing off developers right now with both driver cheats and forcing them to do loads of extra code to have their hardware work ?
3) Appear arrogant. Nuff said.
4) Do some weird things with "special-purpose" drivers (like the V5 HSR "beta" drivers). Well, well...

 

[Bouncing Zabaglione Bros.]

You've got to remember the reason that 3DFX ultimately went out of business is because they failed to make money for about twelve consecutive quarters. It was a signal that we all should have seen - the consistent failure quarter after quarter to make profit. Until we start seeing the same signals from Nvidia, I doubt they are going to go under or get taken over.

 

[Joe DeFuria]

But otherwise the similarities are striking, aren't they?

No, I personally don't think an NV40 failure would spell the end of nVidia...but there are in fact, lots of humorous similarities that can be drawn between the two situations.