My take on the disaster that is the GFFX. (long post)

[Arun]

The NV35 most certainly looks VERY impressive based on all rumors specs.

But the R400 rumored specs are even more impressive IMHO.

However, some of it gotta be BS and false sources. So unless we get more reliable info from good sources ( *cough* CMKRNL, please? *cough* ) , we won't know for sure. The R400, based on current rumors, would be godly. But it would surprise me if ATI could pull it off with 0.13, it would sound a lot more like something to do with 0.09


Uttar

 

[Dave Baumann]

Very few people considered what's been achieved with r300 would be possible on .15u, especially at the clock rates they achieved them at. Personally I'd be inclined to think that ATI are more likely to be able to do it than anyone else!

 

[Unknown Soldier]

I think most people are disappointed becuase this wasn't just another product launch from just another company. It was to be "the biggest contribution to the 3D graphics industry" from the biggest company in the industry. People just expected more.

I think what most of us are disappointed about is the product that Nvidia released to the public. 4-6 months ago we had the Nvidia CEO and main PR man both saying that the NV30 would thrash the R9700 Pro by alot and would even be double over the 4600. Now with benchmarks out and the lot and no thrashing of the ATI card or even doubling of the 4600, I believe people think they have been sidelined by Nvidia.

The NV35 sounds more like what the NV30 should have been, without the dustbuster on the card.

Nvidia, you disappoint me.

The good thing about the wait though is that if I do deceide to buy a graphics card now, the 9700 PRO is alot cheaper than when it came out.

Next time Nvidia, if you gonna say the cards gonna do double, make sure it does double.

US

 

[LeStoffer]

Russ,
It's a disaster because nothing on this scale has happened since pre-RIVA128 eras. From NVIDIA's internal point of view, this is the most uncomfortable launch yet. Parhelia is more than a disaster, it's a terminal error.

So very true. But ack, now you have reminded me of the Riva128 when neither Unreal nor Quake II worked. By far the annoying card I ever had...

 

[martrox]

While there are excellent posts here disagreeing about my conclusion of the dire strates that nVidia is in - How do you guys feel about my statements on the causes of this. Did nVidia get blindsided by the R300? Did they push the GFFX beyond it's intended speeds to make up for the lack of performance as compared to the R300?

 

[RussSchultz]

.13u is behind the problem, there's no doubts about it. That is (generally) what screwed them.

We fully expected .13u to be ready for our next generation chip (we're a conservative company when it comes to process), but we're having our doubts that it will be mature enough for us when the time rolls around.

Mature enough for us means that it will work, period and we won't have problems with it, period. (Small companies like ours would have been devastated by what happened to NVIDIA)

 

[Dave Baumann]

I'd also put too wide a diversification too fast in there as well.

 

[KimB]

I'd also put too wide a diversification too fast in there as well.

I don't think so. I think that that diversification will help to significantly reduce the impact of this mistake.

 

[nggalai]

Hi Russ,

but nevertheless, "disasterous" describes the current NV30 launch situation pretty well.

I'll have to disagree with that statement. If we want to talk 'disasterous', Parhelia was 'disasterous' because it essentially was the last hope for that company, and came out being much slower than the other competition out--essentially a non-starter.

While NV30 is certainly not as astounding as everybody was expecting, it would have been a solid release 6 months ago, whereas now it looks like it'll get leapfrogged within several months rather than coming of age along with the R300.

But a disaster? No. NVIDIA will weather the storm and it will take at least one or two more serious missteps before they're in danger of having their business go under.

Agreed, in this perspective. I myself felt that the NV30 launch itself was "disasterous," i.e. that the launch was very bad in itself, not so much for NVIDIA. The launch was prepared badly, reviewers got very little time to do their work, important information wasn't delivered by NV in time (such as possible screenshot issues), etc.

ta,
-Sascha.rb

 

[tamattack]

OMG... Chalnoth called NV30 a mistake!?!

(Just raggin' you...)

 

[antlers]

Shader benches are pretty disastrous in this thread

It means they have no technological foundation for their CineFX claims--long shaders will be just too slow to run. NVidia, in fact, will be holding back the advancement of shader technology if they get a big installed base. Even for off-line rendering ATI's solution is starting to look more attractive.

 

[Hellbinder]

I am shocked that so many of you think that Nvidia is going to be able to kick out Nv35 tomorrow.. It is simply TOTALLY unrealistic.

Nv30 is not even on the shelf yet. And some of you are all convinced that Nvidia is going to tape-out Nv35 and kick it out by summer. It toally strikes of wishful thinking. They are still working day and night on getting retail levels of Nv30's to market. If xbox supposedly distracted them fom getting Nv30 on time.. then how do you suppose the now 6 month long fiasco of the Nv30 has not distracted them at all??? i say Nv30 is on its slightly revamped planned schedule.. which is full retail by November.

And as far as cmkrnl's comments on R350's impact on the Nv30. I wonder if you are really clearly thinking this thing through? I Know what the R350 is bringing to the table, and I know near down to the fractional % of what kind of lead it has based on the Nv30 previews. As i am sure you are as well. What do you call a product that gets released with various rather distracting negatives that evereyone is talking about, that is about to lose in every single benchmark currently used. By a oft significant factor with the settngs people want, with No distracting negatives??? In my Crystal ball.. thats a big deal. No R350 is not an all encompassing DX9 wonder card.. That comes later.. But it is about to open a can of whoup ass.

Yes, we all know that Nvidia still gained market share. Even with Ati's sudden takeover of the pole-position. However, the tide is slowly building. It wont take very many more cycles if things stay the current course, couppled with continually posotive reviews for Errosion to set in.

 

[duncan36]

Actually the Mercury numbers showed that Nvidia's desktop % remained stable and their integrated #'s declined.

ATi had much higher revenues on stable market share, so basically what happened is that they traded 8500 sales for 9700 sales which isnt a bad thing.
ATi's biggest mistake in the recent past was screwing up 8500 sales by mismanaging the launch of the 9000, and i'm not sure renaming the 8500 the 9100 solves much. The 8500 was attractive in that it performed close to the ti4200 but was much cheaper, launching the 9000 confused buyers and so 8500 sales suffered. They should have called the 9000 the radeon 8000.
Anyway going forward things look better for ATi than Nvidia despite these numbers the 9500/9500pro are excellent products that have been well received, so inevitably these will start to eat in on Nvidia's share.

 

[demalion]

Shader benches are pretty disastrous in this thread

It means they have no technological foundation for their CineFX claims--long shaders will be just too slow to run. NVidia, in fact, will be holding back the advancement of shader technology if they get a big installed base. Even for off-line rendering ATI's solution is starting to look more attractive.

I think the most reasonable interpretation of those results is driver issues regarding shaders. I don't automatically consider the GF FX faster than the R300 due to its clock speed, but those results seem to me to be outside the bounds of reasonable expectation.

 

[speng]

I think Nvidia noticed they'd be late and not much of a competitor and shifted more resources towards getting the NV35 out earlier, in case of a failure on the part of the NV30. 6-7 months later the low-k problems and others that faced the original design might be fixed.

Speng.

 

[duncan36]

Well the six+ month delay of the nv30 is going to impact Nvidia at some point, one of their design teams was basically chasing their own tail for the better part of a year.
Also another consequence of the nv30 debacle is that whatever plans for a downclocked/efficent mobile part based on the nv30 is now down the toilet.

 

[MuFu]

]If xbox supposedly distracted them from getting Nv30 on time.. then how do you suppose the now 6 month long fiasco of the Nv30 has not distracted them at all???

Well obviously it will have slightly, but most of the hard work has been done already. NV30 was a leap forward in terms of technology - NV35 is just a small evolutionary step. You mentioned that NV30 was initially fabbed with low-K dielectrics - they obviously will have learned from that experience and a future transition should be fairly straightforward.

NV31/NV34 will be production qualified and ready to roll soon. Teams probably begun the ~6 month run (layout, actual delay simulation, mask verification etc) into NV35/NV36 ramp back in December, maybe earlier. I know that the NV36 layout was drafted back then for sure. It's not as if the same group of guys works on each project from start to finish.

They can probably get the thing qualified by June/July.

MuFu.

 

[MuFu]

Also another consequence of the nv30 debacle is that whatever plans for a downclocked/efficent mobile part based on the nv30 is now down the toilet.

Not they're not. That's NV36.

MuFu.

 

[Tahir2]

How many engineers are working for NVIDIA devoted to just 3D Core design? Must be at least a 100 since they gained that many people from 3dfx... I reckon right now they are all twiddling their thumbs and rolling their eyes.. NOT!

If NV35 is based on NV30 (and history tells us that it will be with maybe a tweak here and there) then NVIDIA are in trouble. Even with low k dialectrics in the future running at 500 MHz now was a mistake. Add a mature .13 process with low k dialectrics which can gain you about 20% performance and you dont actually get much back. Now what am I missing? I think NV35, much like NV25, will have to add more than just clockspeed and memory bandwidth to compete.

 

[jvd]

I dunno with almost a 200mhz clock advantage over the r300 the nv30 still looses in some areas . So if ati can get another 150mhz on the core of the r300 and name it the r350 nvidia would need to get another 150mhz to once again just match ati. THe problem is that ati can release the r350 a month after the fx and no one will complain cause those who spent 400 on the pro have had a good 6+ months to play with the card. Nvidia has to wait a few months (I'd say at least 4) before they release another higher end card or else who ever bought the nv30 will be pretty pissed off. Also if the r350 comes out and is say 10% faster( I say at least 20& faster) and its sold for 400 or 350 nvidia will have to drop the price to sell the geforce fx which should cut way into profits cause that card has to be one expensive thing to produce.