NV35 correction

[Johnny Rotten]

exellent speculative roadmap Joe. I like it. probably will be something along those lines. although I'd prefer a more aggressive roadmap from ATI and Nvidia.

Agreed. ATI will have another 'clean kill' opportunity come in Q1/early Q2 of 2003 (and by clean kill I mean indisputable superiority like we have today). But in my mind that only happens with a die shrink of the existing 9700. Anything less than that (ie speed binned .15u 9700's) leaves nvidia more wiggle room than ATI *could* have left them with.

 

[Evildeus]

Not sure you are right

 

[demalion]

Hmm...I have some disagreements based on other discussions. Over on Rage3D someone pointed to a comment on warp2search that mentions the 9100 (that thing mentioned in the leaked driver set) will be announced in December. I think your March '03 for the 9000 "AGP 8x" variant is way off based on that.

I think the nv30/nv35 timetable is likely if the nv30 can beat the R300 in a fair amount of benchmarks. If it can't, that would imply that IF the R350 is out by then nVidia would have very strong reason to not delay the nv35. I don't think the .13 issues would necessarily impact the possible release data of the nv35, only the marketability/profitability of the nv30...if they have to lower the price of the nv30 anyways, then they would have no reason not to release the nv35 around the time it was scheduled (should we call that "summer"?) if it could prevent lost "speed crown" mind share.

 

[Xmas]

2) Other NV30 variants (different speed grades to compete more with the 9700 non pro and 9500 Pro) start to ramp up a month or two later (March/April). About the same time, NV31 (NV30 MX?) starts to ship to compete with the 9000.

I don't think so. I believe NV31 to be to NV30 what R9500 is to R9700, a chip with identical functionality but reduced speed.
NVidia doesn't need a new chip to compete with R9000.

And I don't think the delay of NV30 will affect the launch of NV35 that much. If they recognize that ATI is quite a bit ahead, NV30 won't last very long.

 

[Nagorak]

As I believe Bigus is suggesting though, recent history does not support nVidia being able to ship a first gen product, (NV20) followed by a "tweaked" version 6 months later (NV25), followed by a new core 6 months after that. (NV30). Which is what is being suggested nVidia execute here. It's been more like first gen product (NV20) ...speed bin same product 6 months later (NV20 ti) ... twekaked core 6 months later (NV25)....tweak of the tweaked core 6 months after that (NV25 AGP 8x).

So even if nVidia's has been planning on NV35 in the spring, and NV40 in the fall, I have high doubts that they will be able to execute that plan. I expect that we'll have either the NV35 in fall '03, or possibly a NV35 "refresh" at that time. (Speed binned NV35 ti, or NV35 with 3GIO?).

Recent history does seem to support that conclusion, but if it's true Nvidia may be in trouble since ATi has been releasing next gens approx. every 13 months.

 

[Joe DeFuria]

Just defending my speculation a bit.

Over on Rage3D someone pointed to a comment on warp2search that mentions the 9100 (that thing mentioned in the leaked driver set) will be announced in December. I think your March '03 for the 9000 "AGP 8x" variant is way off based on that.

Well, there were "rumors" of the RV250 (Radeon 9000) launch as early as April last year (every new trade show was a new supposed launch date), and it just kept on getting pushed further back. So I would take the Dec 1 launch with a grain of salt.

Also, I was intending to mean March as the shipping date of the AGP 8X RV250 (Radeon 9100?), so even if ATI "launches" it in December, March shipping can still be close.

I don't think so. I believe NV31 to be to NV30 what R9500 is to R9700, a chip with identical functionality but reduced speed.

Possibly...but that would be a first for nVidia. nVidia has never introduced two different chips for the $150-$250 dollar segment and the $250+ segement. They have always either just speed binned the parts to cover both markets or used "the last generation" parts for the mid-range.

Traditionally nVidia has used one chip for cards from the <$100 to about $125 range (MX), and a second chip for $150+. (Ti). Now, nVidia tried to use the GeForce4 MX in the $150-$200 market with the GeForce4, but competition forced nVidia to go with the Ti 4200 instead.

In any case, I'm willing to bet that the NV31 is designed to be the GeForce4 MX replacement...and whether or not nVidia tries to sell a part based on it to compete with the 9500 Pro, or limit competition to the 9000, will depend on its performance and feature characteristics.

NVidia doesn't need a new chip to compete with R9000.

I disagree. nVidia's only "plus" for their GeForce4 MX line relative to the R9000, is that the MX has AGP 8X support. Performance and features wise, ATI is the clear winner in that segment. nVidia does need new competition in that segment...especially if ATI gets out an AGP 8X version of the 9000 as we are hearing rumored.

 

[Randell]

I disagree. nVidia's only "plus" for their GeForce4 MX line relative to the R9000, is that the MX has AGP 8X support. Performance and features wise, ATI is the clear winner in that segment. nVidia does need new competition in that segment...especially if ATI gets out an AGP 8X version of the 9000 as we are hearing rumored.

but wait Joe, the Geforce Mania day slides clearly show the Gf4MX trouncing the Gf3 in AA scores

You need to consider that after all

 

[Xmas]

They have always either just speed binned the parts to cover both markets

Now read again what I wrote, Joe

edit: Oh, you only mean reduced clock speed? Well, what about GF2 GTS - GF2 MX? Same functionality, reduced speed.

NVidia doesn't need a new chip to compete with R9000.

I disagree. nVidia's only "plus" for their GeForce4 MX line relative to the R9000, is that the MX has AGP 8X support. Performance and features wise, ATI is the clear winner in that segment. nVidia does need new competition in that segment...especially if ATI gets out an AGP 8X version of the 9000 as we are hearing rumored.

Again, NVidia doesn't need a new chip. They only need to reduce Ti4200 prices. NV31 will be above NV25/28 in performance.

 

[Fuz]

NVidia doesn't need a new chip. They only need to reduce Ti4200 prices. NV31 will be above NV25/28 in performance.

I disagree. Unless nVidia are in the business to lose money, we will never see the 4200 compete with the 9000 in pricing. You see, if the 4200 drops in price dramatically, so will the 9000.

I don't know the exact figures, but I believe the margins on the 4200 are already low. I could be wrong though.

 

[Xmas]

NVidia doesn't need a new chip. They only need to reduce Ti4200 prices. NV31 will be above NV25/28 in performance.

I disagree. Unless nVidia are in the business to lose money, we will never see the 4200 compete with the 9000 in pricing. You see, if the 4200 drops in price dramatically, so will the 9000.

I don't know the exact figures, but I believe the margins on the 4200 are already low. I could be wrong though.

They have GF4MX below R9000, and GF4Ti above it. Why would they need a card at exactly the same price as a R9000? And I surely don't believe NVidia will produce a new chip that comes between GF4MX and Ti.

 

[Fuz]

At the moment, nVidia does not have any DX8 hardware for the budget gamer, and no the 4200 is not a budget card.

ATI are winning alot of OEM deals with the 9000, simply for the fact that its a DX8 class hardware. Its one of those 'check list' items that will be a must this holiday buying season.

This christmas, we will also start seeing alot of games using DX8 features, and OEM's know this.

My point is, nVidia does not have any hardware for the budget gamer that is DX8 compliant. There is too much of a performance gap between GF4MX and the Ti4200. The gap is huge.

Edit: I have both cards, so I am speaking from first hand experience. Believe me, the gap is monsterous.

 

[hkultala]

They have always either just speed binned the parts to cover both markets

Now read again what I wrote, Joe

edit: Oh, you only mean reduced clock speed? Well, what about GF2 GTS - GF2 MX? Same functionality, reduced speed.

NVidia doesn't need a new chip to compete with R9000.

I disagree. nVidia's only "plus" for their GeForce4 MX line relative to the R9000, is that the MX has AGP 8X support. Performance and features wise, ATI is the clear winner in that segment. nVidia does need new competition in that segment...especially if ATI gets out an AGP 8X version of the 9000 as we are hearing rumored.

Again, NVidia doesn't need a new chip. They only need to reduce Ti4200 prices. NV31 will be above NV25/28 in performance.

reduce prices of >60M transistor chip to the level of <40M transistor chip?
Think they could make profits with that?

 

[DadUM]

reduce prices of >60M transistor chip to the level of <40M transistor chip?
Think they could make profits with that?

The actual cost to produce the chips probably isn't all that different. Another set of costs come in for the "return on investment" of chip development and driver support. For a "mature" chip like the NV28, my guess is that most of those types of cost have already been incurred. Therefore, they can cut the profit margin, and sell them at a lower price and still make money. Now, whether or not I am wrong or if nVidia just wants to keep higher profit margins doesn't mean they will sell the cards for less.

 

[Joe DeFuria]

edit: Oh, you only mean reduced clock speed? Well, what about GF2 GTS - GF2 MX? Same functionality, reduced speed.

No, they are different chips. GF2 MX is a dual pipe board, and GF2GTS is a quad pipe board. Same functionality, yes. Not just the same chip downclocked.

I'm not saying that NV31 won't be a "effectively same functionality as NV30, just reduced performance and cost via elimination of pipes." It very well may be.

I'm saying that whenever nVidia has made such a chip, it was targeted at for sale in the VALUE segment, not the "mainstream performance" one. For the "mainstream performance" market, nVidia has always used the exact same chip, just different clock speeds and memory clocks, as that used for the enthusiast market.

Again, NVidia doesn't need a new chip. They only need to reduce Ti4200 prices. NV31 will be above NV25/28 in performance.

You can't just reduce prices to the point where you are not profitable. NV25/28 chips are not only more expensive in terms of transistor count / die size...but they use more expensive memory to get their performance. THis makes for a more expensive board.

Everyone pretty much admits that the Geforce3 Ti is a better part than the GeForce4 MX. That's an "older" part too. Why doesn't nVidia just sell that part in the value segment?

Most people would also say that the Radeon 8500 is as a whole, a better part than the Radeon 9000. Why doesn't ATi just lower prices of the 8500, instead of designing a new chip?

 

[Xmas]

No, they are different chips. GF2 MX is a dual pipe board, and GF2GTS is a quad pipe board. Same functionality, yes. Not just the same chip downclocked.

I'm not saying that NV31 won't be a "effectively same functionality as NV30, just reduced performance and cost via elimination of pipes." It very well may be.

That was my original quote:

I believe NV31 to be to NV30 what R9500 is to R9700, a chip with identical functionality but reduced speed.

R9500 is a quad pipeline board, R9700 an 8 pipeline board. That's exactly what I meant. It's identical functionality, but reduced speed.

I'm saying that whenever nVidia has made such a chip, it was targeted at for sale in the VALUE segment, not the "mainstream performance" one. For the "mainstream performance" market, nVidia has always used the exact same chip, just different clock speeds and memory clocks, as that used for the enthusiast market.

When GeForce2 came out, GF2GTS was high-end, GF2MX was the mainstream board and TNT2 boards (and some GF256SDR) filled the value segment. Of course that shifted over time, with GF2 Ultra boards coming out.

You can't just reduce prices to the point where you are not profitable. NV25/28 chips are not only more expensive in terms of transistor count / die size...but they use more expensive memory to get their performance. THis makes for a more expensive board.

Everyone pretty much admits that the Geforce3 Ti is a better part than the GeForce4 MX. That's an "older" part too. Why doesn't nVidia just sell that part in the value segment?

Most people would also say that the Radeon 8500 is as a whole, a better part than the Radeon 9000. Why doesn't ATi just lower prices of the 8500, instead of designing a new chip?

So what do you expect NVidia to do? Design yet another chip that takes place between 4MX and 4Ti? Maybe a NV3x with 2x2 or 4x1 pipelines?

 

[Joe DeFuria]

When GeForce2 came out, GF2GTS was high-end, GF2MX was the mainstream board and TNT2 boards (and some GF256SDR) filled the value segment.

Well, I consider the "value" segemnet $125 and less. (This is where the cards like the MX's and 9000s generally debut at.)

http://www.anandtech.com/showdoc.html?i=1266&p=21

At $119 the GeForce2 MX should make everyone that paid $300 for a GeForce 256 (SDR) feel a bit cheated right about now. The performance numbers speak for themselves, the GeForce2 MX is about as fast as a GeForce 256 and sometimes a bit faster but still slower than the GeForce 256 DDR.

So what do you expect NVidia to do? Design yet another chip that takes place between 4MX and 4Ti? Maybe a NV3x with 2x2 or 4x1 pipelines?

No, design another chip that REPLACES the 4Ti. A chip that pushes down the GeForce4 MX to the segment where the GeForce2 MX is selling (sub-value?), and create a chip that can be made CHEAPER than the 4Ti (so it can be sold for a profit in the current GeForce4 MX market), but retains as close to the same performance and features as possible as the Ti, if not exceeding it. This is something that the move to 0.13 can make possible.

Much like the Geforce3 / 3 Ti is phased out, and the Radeon 8500 is phased out. They were phased out because they are too expensive to make.

It's time to phase out the GeForce4 Ti, and replace it with a part that is about the same in performance / features, but costs less to make.

 

[demalion]

There was actually a rumor of the sort concerning what nVidia would be introducing in the near future when we were discussing the NV18/28 etc. Personally, that makes sense to me, but what part number would it have if such a thing is in the works? The only number being thrown around is the NV31...could be to NV30 what RV200 was to R200, I suppose.

No, I'm not saying I have any info to support this guess, just that it is what makes sense to me for nVidia's profits in my knowledge vacuum of their chipset line up.

 

[Sharkfood]

The way I look at it, NVIDIA needs two (2) new chips to compete against ATI, to capture both the middle and high-end.

Obviously the NV30 will either compete with, or far surpass the R300. If it merely "competes"- then this will warrant some serious efforts for NVIDIA for Q3/Q4 2003 to rush out something that "far" surpasses for the priceline we'll be looking at (i.e. the $350+ US range).

For the middle end, you really cant over-market some new, spit shined AGP 8x Geforce4 to compete against the 9500 Pro, especially given DX9 and possible Doom3 texture/pass issues that may occur at the same time.

Perhaps what is partially responsible for the delay with the NV30 involves an attempt to rearchitecture this chip for failed yield recovery for sub-product (i.e. reduced pipelines from failed yields captures new products) *or* retool/rearchitecture in such a way that it can be specifically produced with ease in a demasculated form, yet similar architecture. Cost could be troubling with such a design.

All speculation of course, but I simply see all existing NVIDIA products being very challenging to market against the 9500/Pro... especially if this middle market product with exceptional featurelist enjoys inevitable reductions in cost over the next 3-4 months.

 

[antlers]

Aren't the GeForce 4 Ti and the ATI 9500+ relatively expensive cards to build (big, with lots of power regulation hardware)? This may set a floor on their price even if nVidia or ATI gives away the chips to OEMs.

 

[Fuz]

All speculation of course, but I simply see all existing NVIDIA products being very challenging to market against the 9500/Pro... especially if this middle market product with exceptional featurelist enjoys inevitable reductions in cost over the next 3-4 months.

You could also say that nVidia will have a tough time marketing against the 9000 this holiday season. By that I mean they have no bugdet DX8 card at the moment. Although the G4MX and the 9000 are similar in price, they really are a generation apart.

Aren't the GeForce 4 Ti and the ATI 9500+ relatively expensive cards to build (big, with lots of power regulation hardware)? This may set a floor on their price even if nVidia or ATI gives away the chips to OEMs.

They are expensive to build, that is why we will never see the 4200 compete against the 9000. nVidia are not like Sony or MS, they can't afford to sell the hardware below cost.