Anand says - now that NV30 is taped out - we can reveal...

[Joe DeFuria]

My only comment - this is the first time I have seen such an article from Anand. Wonder why....

Ok, I'll bite too.

Obviously, it's because nVidia is no longer the "technology leader" in terms of shipping products. (ATI 9700...close enough to shipping.)

So, nVidia must combat a competitor's shipping product with "hype" about a new product. Standard PR move from any company. This is the first time since the TNT era that nVidia has been in this position, hence, the first time we see such articles from Anand (and others probably soon.)

As with all PR "hype", you can expect PR to give Anand the "best possible scenario" with regard to the specs and release of the future product. So, 90-120 days is a best case scenario, and the "paper specs" (whatever they are, Mhz wise and bandwidth wise) are also "best case."

Typically, but not always, the best case never comes to fruition.

 

[Oats]

Reverend wrote:

My only comment - this is the first time I have seen such an article from Anand. Wonder why....

nVidia's desperation is Anand's desperation to.

 

[Hellbinder]

Ok, ok, ok.....

I just have a coupple things...

First. Will at least a majority of you decide wether Anand was saying they *got* first Silicon back.. Or they just sent the final designs to Fab.... There is still confusion on This. I am thinking it has to be *just sent out*... but..????? Or is it just that Nvidia told him it was *taped out* and he yet again is showing the world that he is clueless???

Next. In his closing points he says this...

Despite very low yields on the parts, NVIDIA will make sure that enough NV30 parts are available to meet demands for this year.

Exactly *how* are they going to do that???? Ask Santa to bring them some??? This is the defenition of Cheerleader reporting. I wonder if we will get a rant from *Ed* about this... I seriously doubt it

I hope we see the same thing with the Nv30 launch! I want to see R350 comparrisons to it. It will only be 3-4 months behind the Nv30 after all! Of course we wont. The door of the internet only swings one way. It would be different if the Nv30 had even been ANNOUNCED yet. But it is not even official. Not even a press release nothing.

 

[Guest]

They will meet demand by pricing the board high enough to limit demand

Even if they get very low yields, pricing well above the R300 (maybe they'll throw in 256 M of memory) will limit demand to only the hardiest of die-hards . They wanted to sell the GF3 for $499, remember.

And when people complain--"Our OEM partners set the final prices for retail boards. This is a board aimed at the enthusiast market. It's not expensive considering the unique features it provides."

(Sorry I'm guest. I can't seem to get the registration e-mail.)

 

[RussSchultz]

]

Despite very low yields on the parts, NVIDIA will make sure that enough NV30 parts are available to meet demands for this year.

Exactly *how* are they going to do that???? Ask Santa to bring them some??? This is the defenition of Cheerleader reporting. I wonder if we will get a rant from *Ed* about this... I seriously doubt it

Either the fab will suck up the cost , or NVIDIA will and file it under 'marketing costs'.

Its not impossible, just costly.

 

[Joe DeFuria]

Well, if the fab "sucks up the cost", then we can be pretty sure that they won't exactly be prioritizing the production of those chips....

 

[msan_ewu]

Couple of interesting points:

1)IF the problem is not the .13 fab, and instead was the chip design, then maybe the rumors about a changeover from a 128 bus to a 256 bus due to R300 is right.

Just adding this because I hadn't heard of a NV31

NVIDIA is expected to announce new mainstream NV30 solution in Q1 2003, aka NV31. This chip is supposed to be based on the NV30 architecture, though it will be cheaper because of fewer rendering pipelines (their number will be reduced from 8 to 4) and narrower memory bus (128bit instead of 256bit). NV31 is also likely to undergo some other changes compared to NV30. By the way, NVIDIA is going to base its new notebook solution exactly on NV31.

http://www.xbitlabs.com/news/story.html?id=1029123535

 

[alexsok]

DELETE

 

[martrox]

Jesus, alexsok, you are as bad as Doom & Hell.....hmmm....they do kinda fit together. I think your whole problem with this thread is the source of it refutes your flamewar postings.

You know, all of this stuff is BS.....ATI's BS & nVidia's BS....because no one has even seen either of these cards. Maybe, soon, ATI's BS will come into focus & the BS will be seen for what it is. And, maybe, at some point down the road, nVidia's BS will be seen for what it is. Then we can have a talk about whats REAL!

You know what I really love about fanboi's? The ability to be true believers.....even if the belief is just BS! Now, if we could just teach them that silence is golden........

And, it's a sad fact that most of the features that everyone is talking about won't even be used during the life of the card! Unless playing tech demo's is your thing... Just how many features of DX8 (or DX8.1, would hate to leave the ATI'ers out!) are now being used, in TODAYS games? Hell, I'm still waiting for someone to get a really good idea....... Like really high quality FSAA!

 

[alexsok]

Jesus, alexsok, you are as bad as Doom & Hell.....hmmm....they do kinda fit together. I think your whole problem with this thread is the source of it refutes your flamewar postings.

You know, all of this stuff is BS.....ATI's BS & nVidia's BS....because no one has even seen either of these cards. Maybe, soon, ATI's BS will come into focus & the BS will be seen for what it is. And, maybe, at some point down the road, nVidia's BS will be seen for what it is. Then we can have a talk about whats REAL!

You know what I really love about fanboi's? The ability to be true believers.....even if the belief is just BS! Now, if we could just teach them that silence is golden........

And, it's a sad fact that most of the features that everyone is talking about won't even be used during the life of the card! Unless playing tech demo's is your thing... Just how many features of DX8 (or DX8.1, would hate to leave the ATI'ers out!) are now being used, in TODAYS games? Hell, I'm still waiting for someone to get a really good idea....... Like really high quality FSAA!

You're right, I was wrong in my previous post, sorry

 

[Tahir2]

aleksok

if you have certain priviliged information then maybe you should refrain from posting in these threads as we dont want another case where a whole thread is deleted because you said something you shouldnt have.

you do realise though, what you might know is misinformation anyway i hope.

 

[woolfe99]

The way that Anand keeps stating that NV30, on paper, is faster than R300 makes me think it's a 8x2 pipe architecture, compared to R300's 8x1. It's gonna' need massive bandwidth to keep those 16 TCUs fed, unless it has greatly enlarged caches.

The question is whether Anand is enough of a dolt to make the statement "NV30 will be faster on paper" *without* a 256 bit bus, knowing that it won't have enough bandwidth to feed that many pipes. I tend to think the answer is "no, he isn't that much of a dolt." But I also think he doesn't know one way or another about the bus size (a piece of NDA info he will probably get with his review sample), and he IS enough of a dolt to say it will be faster on paper on the basis of a presumed faster core and memory, without considering that the bus size is a hugely important (and unkown) variable. I guess what I am saying is that the best policy is to ignore Anand's speculations about performance, realizing that he a) probably doesn't have all the specs, b) is being fed with a lot of Nvidia PR, and c) his perceptions are colored by past experience with Nvidia having expected or better than expected performance in past releases.

For that matter, I would ignore his speculations about a release date as well. The only thing we can take from this article is a piece of factual information - that NV30 has "taped out." But unfortunately, we do not necessarily know exactly what even *that* means.

 

[Typedef Enum]

If anything, that article raised more questions than it solved...which is kinda' funny, because Anand's post from a few days ago basically stated that he was going to put this thing together in order to dispell rumour's, etc.

Oh well...I'm sure it's only a matter of time before the 'real' low-down is put out there.

 

[Nagorak]

But Russ - in the collorabative design and die shrink process - who would own the IP of how to do it optimally - is it joint or 95% exclusive to TSMC - to give to any competitor of NVidia who paid them to gain this IP - time and time and time again?

Cause if I were NVidia I would have liked to say the IP is mutual and we get a royalty - or 100 days etc before you can reveal the insights needed to make this work to our biggest competitor.

I presume your can't patent this knowledge - but is it protected in other ways?

Ummm...I really doubt tooling the fab to work properly is considered "IP" in any sense. But if it is TSMC would obviously own 100% of it. What you're suggesting is basically akin to hiring a construction crew to build your house and then trying to get them to pay royalties on how they cut their wood since they "perfected their technique" at your house (as if they hadn't been working on their technique for a long time, and would continue to do so in the future).

Not to mention even a company like Nvidia wouldn't want to own IP on fab refinements, because that meant other companies would hold IP on it, and in the end it would just come around to screw everyone over.

I honestly don't see why you'd even think there'd be any IP involved in this to begin with though.

 

[Nagorak]

They will meet demand by pricing the board high enough to limit demand

Even if they get very low yields, pricing well above the R300 (maybe they'll throw in 256 M of memory) will limit demand to only the hardiest of die-hards . They wanted to sell the GF3 for $499, remember.

And when people complain--"Our OEM partners set the final prices for retail boards. This is a board aimed at the enthusiast market. It's not expensive considering the unique features it provides."

(Sorry I'm guest. I can't seem to get the registration e-mail.)

Won't work...they have competition now (R300). They could price the GF2:Ultra and GF3 insanely because they had none. Err, well I guess they can "limit demand" but they won't be able to charge nearly as much. Frankly if the NV30 is in the same ballpark as the R300 and getting yields of 15%, I think they are screwed. We may see a paper launch without ever seeing more than a handful of products on the shelf. Does anyone have any idea what yields ATi is getting? It has to be more than 15%, and they're trying to maximize them by releasing gimped chips as R9500s.

 

[ram]

The NV30 can be faster on paper without being 8x2. If it just has a higher clock speed than the 9700 (which seems likely), it can theoretically put out more pixels than the 9700 if it is still only 8x1. Does it have that many more transistors than the 9700 that it can support twice as many texel units and longer shaders as well as a higher-precision pipeline?

I guess the NV30 can do more PS instructions per clock than a R300 and therefore should be faster clock per clock for complex shading code.

 

[David G.]

The way that Anand keeps stating that NV30, on paper, is faster than R300 makes me think it's a 8x2 pipe architecture, compared to R300's 8x1. It's gonna' need massive bandwidth to keep those 16 TCUs fed, unless it has greatly enlarged caches.

Greatly increase caches .... Well it can't have so great caches as they only have 8 million tranzistors in plus which would form maybe less than 128 Kb of cache ( somebody who has experience with this should givea better aproximation ) .

Those 16 TMU's won't help that much .... Look @ Radeon 9000 vs Radeon 8500 : the difference is less than 15% ... and remember that TMUs units are extremely bandwidth dependant and that DDR-II offers more speed indeed but not the same real bandwidth as DDR . It's more like ... a 900 Mhz DDR-II chip is able to offer a bandwidth equal with that offered by a 750 Mhz DDR chip array .

So it would be relatively easy to fight the NV30 who's only real and evident advantage is the 400 Mhz working speed . ATi would just have to switch to 800 Mhz DDR memory or even 750 Mhz would be good and raise to woking speed to 350 Mhz .

Also , bare in mind the other advantages ATi will have : mature drivers , established market position for the product along with a mature marketing campaign and , most of all , the fact that 0.15 micron technology is mature and cheaper than the 0.13 tech .

Also price will be a good advantage as when the NV30 will be released it should cost more than 400$ as the memory is more espensive , the PCB for the DDR-II memory will be more epensive , the chip is bigger ( more tranzistors ) , the 0.13 , the iyelds are so small .... and of course ATi will lower the price to around 299$ .

 

[David G.]

Intresting on the whole R300<->NV30 story for me is who made the first below $300 DirectX 9 card.
ATI will anounce a Radeon 9500 card this month and it should be ready for the holiday sales too.
If this one is nearly as fast as the current Geforce 4 and the price is between $200-250 that card would be the "real" seller in the holiday sales.

You're being to modes my dear friend .... The 9500 , a DX9 card with R300 technology will surely be , at least 30% faster than the Radeon 8500 whic is less than 15% slower than GF4 Ti 4600 .

As a matter of fact they even have similar architectures : 4 pipelines .

 

[kid_crisis]

aleksok
if you have certain priviliged information then maybe you should refrain from posting in these threads as we dont want another case where a whole thread is deleted because you said something you shouldnt have.

you do realise though, what you might know is misinformation anyway i hope.

Heh, if it was misinformation why did he get in trouble with his source for posting it?

Don't let them chase you away Alexsok.

 

[multigl2]

aleksok
if you have certain priviliged information then maybe you should refrain from posting in these threads as we dont want another case where a whole thread is deleted because you said something you shouldnt have.

you do realise though, what you might know is misinformation anyway i hope.

Heh, if it was misinformation why did he get in trouble with his source for posting it?

Don't let them chase you away Alexsok.

thats most certainly definitive.

i got a good one for you: the R300 will feature all the functionality of the NV30 with driver tweaks, says so my source....

see, just because someone says something online doesnt make it true, okay?

edit: nobody is chasing anyone away. The readers tend to be slightly more skeptical than other forums; if someone posts largely unsubstantiated information then it will get questioned.