NVIDIA: 2005 Updates to be Shader Model 3.0 Based

I'm sure it's more a question of resources and allocating them to get the best return on your investment. In this case ATI had more resources allocated to early PCIe adoption and it payed off big for them in OEM design wins. Unfortunately with so many OEM orders to fulfill they had problems getting the right volumes to retail, and with RIALTO's delay they had problems getting additional AGP models out for retail upgrades.

 

But as someone else already stated, at the time of the 9700 ATI had very few OEM wins. It was at this time that ATI started to gain favor with OEM's because of the heat, and sound of the NV30/NV35, not to mention sub par performance compared to the 9700. So at that time ATI had nearly nothing going to the OEMs, now the situation is reveresed.. Everything is going to the OEM's and next to nothing was going out for discreate buys. I think they have solved their yeild issue now.. Better late than never I guess.

The thing is ATI could of kicked Nivida while they were down, but because of the part shortages they were not able too. I think this is a good thing because in all honesty we need two players in the market.. I don't care for NV, but that doesn't mean I wouldn't buy one of their cards if it performed better or offered a better price/performance ratio. Right now I don't think they do.

 

From what i'm reading it looks like NVs gonna launch NV48 in the spring (110nm, "a little bit faster versions") and NV47 in the autumn ("we’ll have one or two products this year going from 90 in the second half").

NV47 just doesn't bode well for "a little bit faster version" and it appeared in the drivers just around the same time as we heard rumours of NV50 cancellation. NV50 was presumed to be NV's fall'05 product. Let's suppose that they've put it back to the drawing boards (maybe b/c of LH coming out sooner than was originally expected?) and decided to make NV47 instead of it.

How's that for a version? Guess we'll see if it's true soon enough already...

 

Well, i read a few days ago that NV48 is NV45 with 512MB RAM with the same clock and memory clockspeeds-isn`t it???

Maybe NV47 will be much improved NV40 with 6 quad and many other improvements+some features from (cancelled????)NV50 and it will be made in 90nm.
About NV50 cancellation....i think Nv haven`t cancelled NV50 at all....
IMo NV47 will be what NV50 was said to be and ``new`` NV50s architecture will be release and known as NV60.

i only wonder why NV is likely to release only ``a little bit``faster GPU in Spring when ATI will release powerful r520????
Maybe explanation is that NV know something about performance of r520 and maybe it isn`t as fast as is was said to be and maybe they have something faster than r520 but at this moment it isn`t necessary to release it????

 

Err neither NV or ATI are going anywhere anytime soon. It would take a monumental collapse for either to go bankrupt in the forseeable future. I mean it would take Microsoft or Sony going bankrupt AND Intel/AMD as well. There really isn't anyone to take up the slack at the moment (XGI? S3? (err supposed to get a couple of cards from them next couple weeks), Matrox? Intel's GMA950? HAHAHA

To that matter, NV could have sold a lot more cards and chipsets as well. It really didn't matter which company sent cards last quarter, manufacturers sold what they got.

Interesting fact, ATI said at RS480 launch in November that they sold 15 million chipsets since their first one IGP320, NV sold ~1.35 million in December alone.

 

I was referring to Radeon 9800 series, not Radeon 9700 series in the post you quoted.

OEM's are fairly evenly split from what I understand, maybe slightly in favour of ATI right now. I know that Dell are using several NVIDIA parts at the moment. NVIDIA will not disappear, they own too much of the chipset business, along with the fact that they own over 40% of the video card market too. They are likely to take even more of the chipset market, now that they are ready to release NForce 4 for Intel.

 

They could have bought bridge chips from nVidia if that's all that was stopping them

 

I doubt it. 6800LE with 512MB memory? Something's fishy here I believe that NV48 line will have a new top card with 512MB and higher clocks.

I'm not sure we'll see NV5x or NV6x at all. Looks like NV's changing it's codenames to Gxx.

We'll have to see how 'powerfull' this R520 will be. Maybe NV's waiting for this too, letting ATI to make the first step in this year and then decide what kind of chip with what clocks they'll release.

That's another possibility, however i find it hard to believe that NVs todays chips could compete in any form with 90nm R520.

 

The other explaination is underpromise and overdeliver. Perhaps the phrase "a little bit faster" is deliberately cryptic. You don't want to overpromise and underdeliver, because if you lose or fail to deliver, the surprise and shock will hit your stock price hard. On the other hand, if you don't promise anything spectacular, but in fact, deliver something better than expected, you are likely to get a boost.

Managing expectations is the name of the game. What does "a little bit" mean? 10%? 30%? I think the probability of ATI delivering something in Spring which is 2x-3x the R420 is extremely low, and I would not expect them to make such a claim in a CC even if they could do it.

 

Well. not exactly GF6800LE must have 512 MB but GF6800GT and GF6800U could have...

i thought about architecture when i was writing about NV50 an NV60...
If they have changed their codenames what G70 will be-NV47, NV50???

Yes -i think NV is waiting what ATI will do with r520 and how fast it will be and then will decide what GPU is necessary to release but they must have some ``secret weapon`` if r520 will be very fast and i suppose that they have it...They must know something about performance r520 and i`m sure that they have in their labs chip which could beat r520...

 

Not even close really. ATi was THE dominant OEM player before nVidia even released their Riva128. Intel has now made it that ATi will likely never seen the kind of OEM dominance they had back then(although the market itself was a lot smaller).

 

... Using words like "could be" and "hinted" and "might be," I wonder what it is you are trying to convince yourself of....

 

Well if Xbox 2 developpers kits with R500 are just starting to go to devs, and if R500 was used to iron out the 09 process, I guess it could mean that R520 should be a bit later than may. September maybe ?

 

A lot of that has to do with the emerging PCIe graphics bus standards, which requires additional marketing and product description, at least according to the way they all seem to be approaching it.

nVidia was implying changed architectures for its PCIe products by changing the numbers when in fact such changes didn't exist. ATi did the same with its model descriptions, etc. Personally, I'd like to have seen them simply go nV40A and nV40E for AGP and PCIe versions, respectively; and likewise for ATi to simply have an "A" and an "E" affixed to its model names, and so on. But, of course, such was not the case...

The PCIe introduction has confused a lot of things, but then so did the PCI introduction into the ISA era, the AGP introduction into the PCI era, etc. There's a period in which the claims made by marketing departments have to be analyzed for veracity by way of holding the products under a microscope before things begin to stabilize. I think we were all somewhat spoiled by the relative predictability and uniformity of the AGPx4-x8 era, when the basic question was only who got it right and who didn't. At the moment the pressing question from the enthusiast-consumer end is only whether PCIe is sufficiently better than AGP to justify the expenditures necessary to make the change. Certainly, looking at PCIe primarily as an expansion bus (as opposed to PCI), the lack of PCIe peripherals (other than graphics cards) indicates the answer is currently "no," with 3d-cards and graphics products being the only possible exceptions.

 

That was a long time ago - before the 3D accelerator revolution when ATI deliberately left the market to Nvidia and 3DFX. Intel didn't have anything in graphics back then. While Nvidia and 3DFX grew to dominance, ATI spent the best part of a decade in the doldrums producing low power, low cost, high volume parts.

ATI were nowhere, and until R300 arrived and Nvidia/3DFX stumbled, they were nowhere with today's OEMs and modern 3D accelerated systems. The last couple of years have seen ATI make massive inroads into a market owned by Nvidia. Nvidia took their eye off the ball, and ATI capitalised.

That's how competition works. Years back, ATI took their eye off the ball and Nvidia and 3DFX made their inroads into the market and became the incumbents. Now ATI are doing the same.

 

It did confuse things, but that wasn't what I was referring to

My issue is with the sheer number of products that they have.

X300 LE
X300 SE
X300 SE HyperMemory
X300
X600
X600 Pro

Do they really need all of those products at entry-level?

The high end is pretty much the same too IMHO.

Having some parts that are exclusive to AGP, and some that are exclusive to PCI-Express (which only recently changed) was another issue I had. That goes down to the issues they had with getting RIALTO out at the time when people wanted it - before they could purchase NForce 4 motherboards. I say NForce 4, because NVIDIA are AMD's biggest chipset manufacturer; I'm not forgetting VIA's K8T890, SIS's 756FX and ATI's Xpress 200 series chipsets here - the companies just don't hold the same market share that NVIDIA does.

I was speaking with several board partners over the last week, and they all said that they are shipping more PCI-Express parts than AGP parts in the high-end now.

 

Why is this an "issue"? They have products that are meeting different segements and they are presumably selling them otherwise they wouldn't do it. Its not like the x300's are anything more than an oem product anyway. Take a look at the number of processors intel are pushing, all because oem's want different configurations and different prices - ati aren't allowed to do similar things for some reason?

http://www.intel.com/products/processor_number/info.htm

 

NVIDIA SLI solution by having two 6800 Ultra 512MB cards together is going to be faster than single r520.

 

Woah, someone who can see the future!!!

 

I'm not going to rehash the history in full, but ATI was all over the 3D revolution. Just not the cutting edge of it, nor supporting add-in 3D. What do you think was the sellingist 3D accelerator of all time, for a Very long time. Hell, you can STILL buy them retail from stores right now. Goto compusa and look at what is on the shelves. Thats right you can still buy a Rage Pro based card (actually based on the mobile version, but it's functionally the same core)... That same basic chip is still being sold in the low end today, and going onto server motherboards all the time.

I'd not put it past them that that new server chip that was mentioned the other day from ATI is probably some bastardized/pcie version of the rage pro core. (note: that is total speculation, I know nothing right now on the specs of that server chip that was mentioned.)

Anyway, that was a long way of saying ATI didn't concede anything to anyone. They just let themselves be lulled into completed OEM domination. When the competition proved they were not only faster and more advanced, but that they could deliver more, regularly, ATI was abandoned. It did indeed take them years to recover from that. The Rage128 series was a series of blunders by ATI. The Radeon was the beginning of the recovery, and the 9700 was really the fruit that was born of the seeds planted when the original Radeon DDR was laid.