NV30 - 25 to 50% faster than R9700...

[SurfMonkey]

Fresh in from Beyond3D, a new article about nVidia and the NV30. The basic gist is alongs the lines of:

a) The NV30 is late. It shoud have been a fall product.
b) There will be a few thousand high end samples around this Christmas.
c) It's likely to be 25-50% faster than the R9700. Plus the OCing.
d) Mass production will be in Q1'03.

Mostly stuff we already knew but now it's kind of confirmed. It was also interesting to see the Analyst state that if ATi had attempted to use .13u tech for the R300 it would have killed them. I guess they just aren't big enough to cope with the costs involved. But this may mean that they are screwed for bringing out the R350-400 on .13u until the process stabilises.

Read the article here: news.con (http://news.com.com/2117-1040-963195.html?tag=%20ltnc)

[Juntari]

Is ATI really screwed for using .13 process to manufacture R400? How long until the process stabilizes? I heard somewhere that they can produce in volume at the expense of performance, which means that R300 a relatively new core that can be clocked faster can be a reality with .13? NV30 is exciting, I anticipated it much. But I am also excited about R350 and I hope ATI is not screwed for living up to be a tough competitor.

[sbp]

No, ATI is not screwed. If anything Nvidia is the one who has been screwed so far.

As the article points out, ATI is gaining mindshare. And its a good bet marketshare also.

Its quite possible the R350 may come out on .15u. The R400 won't come out until 2nd half 2003.

As time goes on the .13u process will mature and yields will go up.

I don't agree with the statement ATI would have been "killed" if it went with .13 micron and couldn't afford to work out the problems with that tech. Yeah, ATI would have been in a bad position but killed? Is Nvidia killed now? And doesn't TSMC shoulder some of the cost in the switch to .13 micron?

[thcdru2k]

ouch..r400 2nd half of 03. that gives nv30 like half a year to do its damage? i don't imagine nvidia releasing anything more than a higher clocked refresh to compete with the r400.

[sbp]

thcdru2k: Don't forget ATI has the R350 cooked up. It may not be as fast and feature rich as the NV30, but it likely won't have to be. Its all about price vs. performance. Someone doesn't have to be market leader and have best thing out to make money.

In my view, ATI has the advantage now. As Carmack said a few months ago, ATI is a half step ahead of Nvidia now. I expect that continue for the forseeable future. Once NV30 comes out in full force, R350 comes out shortly afterwards. And its known ATI is working on the R400.

Its an interesting question whether Nvidia will or can release the NV35 to compete with the R400 versus a higher clocked refresh NV30. If its the NV35, the NV30 wouldn't be out there on the market that long. http://sbp777.homestead.com/files/hmm.gif

[thcdru2k]

i think this will be ati's year...nvidia can reclaim it with some massive improvements next gen...nv40.

[gemini1313]

r400 = 2nd half of 2003.

haha that makes me laugh. it will never get here that fast.

it probably wont make it until Q1 2004 if they try.

remember they still gotta get the r350 out that all the fanatics think is gonna be here next month or something stupid like that.

they gotta make their money on the r300 first and the r350 wont come out until atleast a few months after the nv30, and then the r400 wont come out until atleast 6 months after the r350.

plus the r350 is just gonna be a speed bumped r300 if they get it out sometime soon.

bottomline is the nv30 is gonna smack the r300 , so joy to all the nvidiots that held out for this beast.

oh yeah r400 has to be a beastily brand new design and didn't somebody say it was supposed to be dx10, when in fact dx9 hasn't been officially released. plus they have to migrate to .13u which will be tough for them.

anybody who thinks nvidia doens't know exactly what is happening in the market and in each others labs is crazy. they are the gawds of 3d right now and i dont think they are gonna follow in 3dfx's footsteps except for the gigapixel bandwidth the nv30 is gonna get due to its new memory controllers.

when comdex comes ur gonna see a new fsaa that is hitless if u know what im talking about so just wait another 20 days and **** ur pants then, not now over what ati isn't gonna do in time.

[thcdru2k]

:rolleyes: gee whiz

[Kev1]

I am waiting for NV30 to be released before upgrading my video card to see how the price/performance thing goes.

If it puts the smack down on the Radeon 9700 but costs $500 bucks, even $400, then I'll leave it to you bottomless pocket types to buy it :) I can't see spending more than $300 dollars for a video card no matter what the performance is. And I would like to spend $200 or less (which I always have done till now).

But this is just me. My price range for a high performance video card is $200 to $300 dollars +/- a little. Around Feb or March when I upgrade, whichever video card (ATI or Nvidia) is in that price range thats what I'm gonna get. I'm hoping the NV30 will be in that range by then, so I will keep fingers crossed. I am sure Radeon 9700 will be, as it almost is now.

[MtViewGuy]

I think nVidia better be very careful with the NV30-based boards.

Now that ATI has fully ramped up R300 chipset production, they could easily cut the price of the Radeon 9700 Pro from US$399 to US$299 and that could put a kibosh to a lot of nVidia's plans. Don't forget that ATI is working on the R350, which I believe will support the latest DDR memory technology and very likely will have improved 3-D processing to really bump up performance. We maybe talking 30% faster performance with the R350 cards compared to the Radeon 9700 Pro--probably at a lower price than what the higher-end NV30 cards will cost.

In short, nVidia's reputation is on the line with the NV30 GPU.

[Juntari]

Same here, I am trying to upgrade from GeForce 2 Ti.
I would like to do that as soon as possible, and 9500 Pro or 9700 Pro looks mightly tempting. But I am trying to hold out, to see whether NV30 can provide to be more valuable.

[Kruno]

Nvidia's main strength was in it's ability to deliver early/timely products out to the market to compete with earlier products. IE: GF3 first came out the Radeon 8500 was not even out (afaik), everything other competitors had out could not come close because they were using earlier gen chips.

Hence Nvidia developed consumer and market confidence greatly from their practices.

[Bigus Dickus]

Originally posted by SurfMonkey
Fresh in from Beyond3D, a new article about nVidia and the NV30. The basic gist is alongs the lines of:

a) The NV30 is late. It shoud have been a fall product.psst... don't say that too loud, Roscoe might be listening! :)

c) It's likely to be 25-50% faster than the R9700. Plus the OCing.Well, I suppose that guy's guess is as good as any other.

It was also interesting to see the Analyst state that if ATi had attempted to use .13u tech for the R300 it would have killed them. I guess they just aren't big enough to cope with the costs involved.I don't buy that for a second. This isn't an Intel/AMD situation where one is vastly outclassed in terms of resources by the other. ATI is as big or bigger than NVIDIA, and can cope with the costs involved just as well (or poorly, whatever the case may be). More likely the analyst was simply alluding to the fact that NVIDIA is seen as sort of faultless by consumers, regardless of the truth, while such a blunder by ATI would have been much more damaging to their image than it has been for NVIDIA. Most people here would have ridiculed ATI if they had attempted to make the transition and it caused their product to be delayed for several months. With NVIDIA doing it though, people somehow justify that "they know what they are doing, they know best, we'll all be rewarded for being patient, etc. etc. bs bs."

But this may mean that they are screwed for bringing out the R350-400 on .13u until the process stabilises.And what impeccable logic lead you to this conclusion? First off, I fully expect the R350 to be a .15u product. Second, the R400 is quite a ways off (still a year or more roughly), and that leave plenty of time for proper process validation. It might be rough, but it might not. There simply aren't any facts to presume that the R400 development is late however and that manufacturing issues associated with all process changes will be overly damaging to the product development timeline.

[Bigus Dickus]

Originally posted by gemini1313
r400 = 2nd half of 2003.

haha that makes me laugh. it will never get here that fast.Based on ATI's past three product launches, the R400 will probably be released late next year.

the r350 wont come out until atleast a few months after the nv30Why is the R350 release date dependent on the NV30 release date? Just because NV is rather late doesn't mean that ATI necessarily will be. Who knows... if things keep going horribly wrong for NV's process change, the R350 could actually show up first. I highly doubt that, but it's not like people at ATI are saying "dude, I wish they would hurry up with that NV30 launch so we can get the R350 on the shelves!"

plus the r350 is just gonna be a speed bumped r300 if they get it out sometime soon.Quite likely. I expect a clock speed increase, DDR-II memory, and perhaps some core improvements like a second TMU or support for longer shader instructions (but I doubt that).

bottomline is the nv30 is gonna smack the r300 , so joy to all the nvidiots that held out for this beast.You're fairly confident... you must be an NV employee with inside information. Can you tell us some of the specs of the NV30... please?

oh yeah r400 has to be a beastily brand new design and didn't somebody say it was supposed to be dx10, when in fact dx9 hasn't been officially released.I seriously doubt the R400 will be a DX10 part. I know that ATI has been using a new DX generation with each of it's past product launches, but DX9 is going to be out several months after the R300 was released, and I just don't think that M$ will release DX10 that soon. And, I really don't see ATI pushing back a product launch six months just to catch the next DX10 version. I could be wrong though, just doesn't look like the schedules will line up next fall for that.

anybody who thinks nvidia doens't know exactly what is happening in the market and in each others labs is crazy. they are the gawds of 3d right now... Really? Looks like ATI has the highest performance offerings in the enthusiast/high end, mainstream, value, and notebook market segments. Oh, did I mention that their offering is one DX generation ahead of NV's competing product in every market segment? Gawds indeed. :rolleyes:

when comdex comes ur gonna see a new fsaa that is hitless if u know what im talking about so just wait another 20 days and **** ur pants then, not now over what ati isn't gonna do in time. As OpenGL guy stated over at B3D... TANSTAAFL!

[PreservedSwine]

Originally posted by SurfMonkey
Fresh in from Beyond3D, a new article about nVidia and the NV30. The basic gist is alongs the lines of:

a) The NV30 is late. It shoud have been a fall product.
b) There will be a few thousand high end samples around this Christmas.
c) It's likely to be 25-50% faster than the R9700. Plus the OCing.
d) Mass production will be in Q1'03.

Mostly stuff we already knew but now it's kind of confirmed. It was also interesting to see the Analyst state that if ATi had attempted to use .13u tech for the R300 it would have killed them. I guess they just aren't big enough to cope with the costs involved. But this may mean that they are screwed for bringing out the R350-400 on .13u until the process stabilises.

Read the article here: news.con (http://news.com.com/2117-1040-963195.html?tag=%20ltnc)
Cool article, but I have a few doubts about this writers "facts"

A) I agree, Nvidia has missed a product cycly w/ the NV30
B) Sure- I agree, they will have a few samples by X-mas
C) It's likely to be 25% - 50% faster than the R300 IN SOME SITUATIONS, not all, and probably on par in other areas. Still, a far cry from the 200%-300% the R9700 crushed the Ti4600 in certain situations. 25% just isn't enough to make someone who already bought an R9700 jump ship.....
D) I agree, no mass production until 2003

Also, didn't the author state it was to be marketed as the Geforce5??? UGH! Let's hope not!

Personally, I'd like to see manufacturers sell the items as their codenames! I like the NV30!, Heck, I still have to remember it's the R9700, and not the R300!
And doesn't CLAWHAMMER sound so much better than OPTERON!!!:D

[SavagePaladin]

I love it when they 'analyze' nonexistant data...
It COULD be any of those things, but as they themself noted, NVidia does not release info on unannounced products. So how the hell would they know?

Pfeh. I'll just wait and see, as per usual.

[StealthHawk]

first off, DX9 might not even make it out this year. DX10 definitely won't be out in 2003.


Originally posted by PreservedSwine
Personally, I'd like to see manufacturers sell the items as their codenames! I like the NV30!, Heck, I still have to remember it's the R9700, and not the R300!
And doesn't CLAWHAMMER sound so much better than OPTERON!!!:D

Clawhammer != Opteron
Sledgehamer = Opteron

[Mono]

Originally posted by gemini1313
anybody who thinks nvidia doens't know exactly what is happening in the market and in each others labs is crazy. they are the gawds of 3d right now and i dont think they are gonna follow in 3dfx's footsteps except for the gigapixel bandwidth the nv30 is gonna get due to its new memory controllers.


Well i'm glad you hold so much faith in a company but I don't see your point. You think ATi dosn't know what they're doing? They're in it for the long run now, and you can expect some cut throat competition for at least the next two or three product cycles. I don't see why people would wana deny this, it just means better products for us consumers. O and btw, if they're gods how come they couldn't get nv30 out when I wanted it :(

[SurfMonkey]

Originally posted by Bigus Dickus
And what impeccable logic lead you to this conclusion?

If ATi couldn't sustain the material and image costs to develop the .13u process now then they aren't going to have much chance until the process has stabilised and yields are higher. It would seem to explain why they stuck with .15u, they knew they could get good yields and they couldn't afford to chance anything else. So maybe it wasn't a choice because they didn't have one.

And it does seem that ATi rushed the R300 out in order to beat nV to the mark which, as it turned out, they needn't have done. If the NV30 had appeared in the marketplace first then ATi probably would have been finished as a high end gfx maufacturer and would have been relegated to OEM and low end markets.

This would have been down to a rash of bad products, causing poor consumer confidence, weak advertising and developer relations. And a history of poor driver support. I think they were very lucky the NV30 was delayed, it's given them a second chance. Though nV are probably kicking thmeselves now for creating the oppurtunity for ATi to gain marketshare when in fact they could have removed them.

PreservedSwine
C) It's likely to be 25% - 50% faster than the R300 IN SOME SITUATIONS, not all, and probably on par in other areas. Still, a far cry from the 200%-300% the R9700 crushed the Ti4600 in certain situations. 25% just isn't enough to make someone who already bought an R9700 jump ship.....
[/QUOTE]

Ah, that's a bit of an unfair example, it should be along the lines of:

a) The R300 in certain highly intensive situations delivers 2x to 3x the performance of the GF4.

b) The NV30 in certain highly intensive situations delivers 2.25x to 3.5x the performance of the GF4.

And a 25% to 50% increase in performance at the top end is all we are going to see for quite awhile. And I think it will be enough for nV. We have yet to see what kind of performance gains DDRII and the .13u core will give by way of overclocking. Also if the NV30 is a deferred renderer or some form of hybrid, as looks more and more likely, then the performance hit for FSAA and Aniso filtering will be tiny in comparison to the R300.

[Lezmaka]

DISCLAIMER: The following paragraph is just something I remember, or a hallucination of something that I just think I remember, and either way, it may not have been true in the first place.

The release of DX9 and NV30 were supposedly aligned at one point. NV30 was supposed to be launched when DX9 was released. If DX9 was delayed, NV30 was gonna be delayed. But if NV30 was delayed, and DX9 wasn't, DX9 would still get released anyway.

-- End of Hallucination --

My guess is that DX9 could get released at Comdex, or if not AT it, then in the same time frame. If so, then it will have been over 2 years since DX8 was released (Nov. 9, 2000). Maybe we'll see DX9.1 in 2003, but I don't think DX10 is gonna be out til at least 2004.

[Philibob]

Originally posted by Lezmaka
DX9 could get released at Comdex, or if not AT it
So either at Comdex or at Comdex :confused: ;)

[tazdevl]

Originally posted by SurfMonkey
If ATi couldn't sustain the material and image costs to develop the .13u process now then they aren't going to have much chance until the process has stabilised and yields are higher. It would seem to explain why they stuck with .15u, they knew they could get good yields and they couldn't afford to chance anything else. So maybe it wasn't a choice because they didn't have one.


Some folks seem to be forgetting the business model ATI and nVIDIA are using. Both companies only pay for the FINISHED product. They do not bear any ongoing manufacturing costs themselves, that's borne by the fab. That's the beauty of oursourcing.

ATI played things smart, saw the tough time nVIDIA and TSMC were having... and made the right decision. They're waiting for the .13 process to be stabilized before launching a comparable product. That's called sound reasoning. nVIDIA bet all their chips on .13, it might pay off a bit in the long term, but I doubt it.

[Richthofen]

"
Some folks seem to be forgetting the business model ATI and nVIDIA are using. Both companies only pay for the FINISHED product. They do not bear any ongoing manufacturing costs themselves, that's borne by the fab. That's the beauty of oursourcing.

ATI played things smart, saw the tough time nVIDIA and TSMC were having... and made the right decision. They're waiting for the .13 process to be stabilized before launching a comparable product. That's called sound reasoning. nVIDIA bet all their chips on .13, it might pay off a bit in the long term, but I doubt it.
"

Well that is only half of the story. The yields with TSMCs 0.13 process are ok.
Problem is you need to get that process optimized for your special chip design.
And that is your task as the chip designing company. In this case Nvidia. ATI and Nvidia have squadrons of people working at TSMC to get the manufacturing processes work for their products.
ATI will have to gain that experience in nearly the pretty same hard way Nvidia had to. To use the 0.15 process pays off for ATI right now. The early process change will pay off for Nvidia in the long run.
The delay has nothing to do with TSMC. It has something to do with this certain chip @ 0.13 process.
This whole thing will pay off for Nvidia in 2003 like it payed off for Intel in 2002 compared to AMD.

[tazdevl]

I disagree.

The full picture is that yields were unsatisfactory because TSMC did not have the process down for a part as sophisticated as the NV30. Big difference between it and Transmeta's core. But the finger still points directly at TSMC, that's their business, that's their core competency.

There are always optimizations etc... that occur which is why pre-alpha, alpha, beta, pre-production etc... cards are released.

[Joe DeFuria]

If ATi couldn't sustain the material and image costs to develop the .13u process now then they aren't going to have much chance until the process has stabilised and yields are higher. It would seem to explain why they stuck with .15u, they knew they could get good yields and they couldn't afford to chance anything else. So maybe it wasn't a choice because they didn't have one.

Why is it that everyone assumes that ATI just "couldn't afford the risk of 0.13" or "couldn't afford to develop 0.13"?

Of course ATI had a choice. Same choice that nVidia had:

1) Make a hugely complex chip on 0.15
2) Or Make a hugely complex chip on 0.13

EITHER CHOICE has it's own set of challenges and hurdles. It's pros and cons.

And actually, if anyone thought they "didn't have a choice", it appears that's nVidia. The CEO is on record saying that "0.13 is required" for a chip as advanced and complex as the NV30.

Seems that ATI pretty much disproved that theory, as R300 and NV30 are of similar complexity / teransistor count.

ATI will have to gain that experience in nearly the pretty same hard way Nvidia had to.

Wrong. TSMC is gaining most of this experience themselves by fabbing the nVidia chip. TSMCs tools are becoming more mature. In a nutshell: nVidia's troubles are actually BENEFITTING ATI.

To use the 0.15 process pays off for ATI right now. The early process change will pay off for Nvidia in the long run.

Again, there's no evidence of that. IF nvidia actually managed to ship the NV30 at the same time as the R300, then I would probably agree. However, NV30 is still not out yet, and all the while ATI is now recouping R300 dev costs with the R300 core products.

By the Time NV30 is released, ATI will be able to lower R300 product prices "comfortably" because they have ALREADY made back singificant amounts of money. nVidia will either be forced to reduce their own profit margins, or keep the price high and sell fewer units.

There is no evidence to support that ATI will have the same 0.13 "problems" that nVidia had. In fact, it's likely they won't, precisely due to TSMC having to work out teething problems with nVidia.

As long as ATI is able to deliver their 0.13 part "as scheduled" (whatever that is, probably R-400 in Summer '03), I see no "long term advantage") for nVidia.