Two new (conflicting) Rumors

[DemoCoder]

While these boards were down, two new rumors surfaced. In one interview with a Creative Labs Euope employee, the following statement was made

Based on what I've seen NV30 will retake the performance crown from ATI by a significant amount.

On the other side, the Inquirer posted a new rumor that says

I think many people concluded that NV30 will have a 256 bit memory interface since the Radeon 9700 PRO and Parhelia has it, but I am sorry to say my sources tell me that's incorrect.

What we learned from a chat by the heatsink a few days back, is that NV30 will have only a 128 bit memory interface but it should use much faster memory than anyone uses today. We don't have any figures yet, but I'd estimate over 800MHz

Neither really seem to make sense. The Creative Labs guy seemed to be comparing clock rates only based on other comments in the interview. The Inquirer rumor fits with former NVidia comments that "256-bit bus is overrated", but the idea of using RAM that is atleast 2x the speed of the RAM that ATI is using is ludicrous.

#1 It ain't available yet. Quad pumped memory seems to coming soon, but again, see rule #2
#2 Any RAM that NVidia can buy, ATI can buy also and put onto their card
#3 If the NV30 doesn't have a 256-bit bus, what the hell are those extra 60 million transistors doing!?

 

[Dave Baumann]

Well, Eoin from Creative is doing his job...!

 

[DemoCoder]

Spreading mis-information?

Either everything we think we know about the NV30 is wrong, or the rumors are junk.

 

[ben6]

Well, The Inquirer is no better than the National Enquirer in the US. Spreading conspiracy theories when none exist is the norm for them. Using other people's screenshots without permission or attribution is another (my Ti500 review's Final Fantasy The Demo screenshots were used without permission or attribution, hell my website's watermark is clearly in the images from the review) ,as a example. In any event I'm going to be up in Las Vegas on November 18th for Comdex. We'll see if that rumor is true

 

[BRiT]

Yup. It's the same basic addage -- downplay what you don't have and upsell what you do have. If you follow his logic, why ever make a video card beyond the Ti-4600? It can play all the games with all the features on at the highest resolutions (alright, ignore that flat out lie Creative said as its obviously dies at 16x12@32+8AF+4AA), so by the same token, why get the NV30 either? Pft. FUD...

An interesting thing about this Creative interview, is why are they touting the NV30 and not their own P10? What is Creatives roadmap? Will the P10 never hit consumer market, or only after it becomes the P11 (full DX9 support) ?

--|BRiT|

 

[DemoCoder]

The P10 is for the workstation market, not the consumer market. Kinda ironic since, I bet, the consumer chips are going to have better performance, and more over, they'll have floating point where the P10 doesn't.

The odd thing is, Creative doesn't have to downplay ATI and upsell NVidia. Creative sells chips from any vendor. They are going to be selling Radeon chipsets. There is no reason to pump up NVidia or worry about "burning bridges" with them unless NVidia really does have something up their sleeve.

If Creative took a look at the NV30 and decided it was going to suck in comparison to the R300, why pump it up, given that it will cause some consumers not to buy Creative's R300 based products, and instead waiting for NV30 based ones.

In fact, even if the NV30 is going to be better, why pump it up and cannibalize your sales? Something strange is going on.

 

[DemoCoder]

I'm going to Comdex as well. Hope the nV30 announcement occurs, otherwise, I'll be irked.

 

[ben6]

Democoder wanna meet for lunch?

 

[Nagorak]

Well, The Inquirer is no better than the National Enquirer in the US. Spreading conspiracy theories when none exist is the norm for them. Using other people's screenshots without permission or attribution is another (my Ti500 review's Final Fantasy The Demo screenshots were used without permission or attribution, hell my website's watermark is clearly in the images from the review) ,as a example. In any event I'm going to be up in Las Vegas on November 18th for Comdex. We'll see if that rumor is true

Uhh...I'd say The Inquirer is a lot more reputable than the National Enquirer. At least they only post plausible stories and turn out to be right at least 60% of the time. Meanwhile the National Enquirer runs stories like Horse Born With Human Head... It's not quite the same.

 

[Dave Baumann]

If Creative took a look at the NV30 and decided it was going to suck in comparison to the R300, why pump it up, given that it will cause some consumers not to buy Creative's R300 based products, and instead waiting for NV30 based ones.

I don't think there is a case for saying that one will suck and one won't, I'd guess at ATI and NVIDIA being farily close at this point, but thats not all ther is to it.

Creative Europe are one of NVIDIA's 'prefered' vendors, so they get the chips at a vastly knocked down rate - go with ATI and you'll likely loose that; however being a preferred vendor could mean that NV's prices are more attractive even if the performances are marginal. As Eoin says in the interview Creative are the largest Euro distributer of NV based parts, conversly ATI in Europe is looking a little crowded (Hercules, ATI, PoweColor etc) so there probably good reason for sticking with an alternative there.

Creative Asia, on the otherhand, were not on NVIDIA's 'prefered' vendor list and hence had to pay a higher price on their chips that Creative Europe did - this did not please Creative Asia very much. They tried to get pricing parity and Creative Europe tried for them but couldn't, warning NVIDIA that they would start making ATI based boards if they didn't start selling them chips at the same price. A few days after Creative Radeon 7500's started appearing there were evidently some heated phone conversations from NVIDIA.

 

[Nagorak]

It's interesting that all these Creative <Region> companies are treated as separate entities. Although in a way it make sense from Nvidia's standpoint (need an outlet in Europe and with the Taiwanese OEMs they don't really need much more coverage in Asia). Still you'd think Creative would insist on being treated the same worldwide.

 

[DemoCoder]

I still say something doesn't add up. 120M transistor chip, FP frame buffers, etc but only 128-bit bus? NVidia engineers aren't this dense. The pixel shaders would be starved trying to fetch FP textures and write to the FP frame buffer over this bus, no matter how much logic they spent making the dispatch of the instructions very parallelized. The only thing that make this work would be a huge on-chip buffer and tile rendering.

 

[Dave Baumann]

Given the workstation slant I'd have my doubts over a deferred renderer. However, we don't as yet know what overdraw reduction schemes they plan on adopting here.

 

[Jazz]

Remember he hasnt seen actual hardware yet, he has seen specs and simulations here is what he said

Now that's putting me in a really difficult position because as you know NVIDIA are pretty consistent with their "we don't talk about unreleased product" position but here is what I think I am safe in saying … (bear in mind I have seen specs and simulations not actual hardware). Based on what I've seen NV30 will retake the performance crown from ATI by a significant amount. Obviously I can't comment on clock speed etc. but its fair to say that we should see a significant jump based on the fact that NV30 is on 0.13 and 9700 Pro squeezed every last cycle out of 0.15. With regard to pricing all I can say is you should use previous pricing as a guideline for estimating SRP's. There has always been a price premium associated with having the absolute bleeding edge and in the past process changes have lead to higher development costs and thus higher street prices. Against that we have the recent trend for binning chips by speed and producing a range from a single chip and I would expect that trend to continue. Sorry I can't be more specific but being totally honest there are no final decisions yet and all I can do is use past experience as a guideline.


http://www.guru3d.com/interview/creative-092002/

i dont doubt that the nv30 will be faster but im still not sure that it will be alot faster like everyone is making it out.

Also another question id like to ask is if ATI needed a power supply to get their core clock speed upto 325MHz and has 110 million transistors would it be feasible to get an nv30 with supposedly over 120 million transistors out at 400MHz without using an external power supply ? Thanks

 

[DemoCoder]

My concern isn't if the NV30 will be faster than the R300, it's will be even be AS FAST as the R300. With a 128-bit external bus, I'd have grave concerns of them being able to pull this off.



Hey Ben,
Do you have any information as to which day of Comdex NVidia is doing their announcement? Beginning, middle, or end?

 

[Nagorak]

My concern isn't if the NV30 will be faster than the R300, it's will be even be AS FAST as the R300. With a 128-bit external bus, I'd have grave concerns of them being able to pull this off.

Seriously, if it's only got a 128 bit bus, I think Nvidia is just screwed. It doesn't matter if they have the chip running at 800 MHz, if the memory is limiting them.

 

[duffer]

A 128 bit quad-pumped bus would roughly compare to a 256 bit double-pumped bus, wouldn't it? And NVIDIA was one of the first users of DDR, so it makes sense for them to be one of the first users of QDR.

 

[Dave Baumann]

Well, they've confirmed they are using DDR-II (which isn't quad pumped).

 

[ben6]

hey DemoCoder , I'm not going to comment on a NV30 launch at all . Except that I will be at Comdex. As it's pretty local , it's a good time to visit Vegas ...

 

[DemoCoder]

If quad-pumped memory were even on the horizon of general availability, ATI could do a quick refresh of the R300's memory controller to support it, and the R350 or whatever would be 2x NVidia's bandwidth again.

And tweaking the memory interface is alot easier than doubling the number of bus lines for NVidia.

Moreover, QBM would be hideously expensive. It doesn't add up.

Either

#1 NVidia is going to be throughly beaten this round
#2 NVidia has a 256-bit external bus
#3 NVidia has some other exotic solution (deferred tiling, embedded dram, etc)

#2 seems to be the most conservative, predictable, approach if you were designing a new chip and wanted the highest probability of success.