I disagree witht eh GTX being 32pipe part, the die size is too small without some new design efficiencies. Rememeber the RSX is a 90nm part, and is not a GF7800GTX or even an GF7800GTX on 90nm, it's a variant which will have more/less features.
I don't see the GF7800GTX being 32pipe part as it stands, and it just wouldn't make sense to build it on the 110nm fab if you're planning to move to 90nm by launch time anyways. And if they're having trouble with 110nm, what are they going to have with 90nm? For those reasons I doubt any talk about 32pipe 110nm part and anything in the 'Ultra' other than clock pushes (which Asus and others are already doing, and which wouldn't match a 32pipe chip).
I'm not saying ATi wouldn' launch vaporware for the 32pipe variant, I'm just saying for their marketing purposes anything less would be dissapointment. Whatever they launch has to impress enough to make people think it can compete well against nV's part which will be scant few months away. They don't need to just match nV they need to beat them in order to then shift the balance of sales back, because right now they will be in the low minority of new high-end cards, and if they only equalize or barely beat they aren't about to make up anywhere near the ground lost in 3months. That's how the see-saw launches are supposed to work if you want to remain competitivie, otherwise ATi should have launched all the 24pipe parts they could have at the same time as nV, their decision to not do so (long before people realized that nV was shipping e-tail from day one) meant that they alwys needed to have a period where they dominated sales. If they just barely outperform the GF7800GTX by a few percent then you won't get that, and there will be a huge gap in projected and actual revenue IMO. At the very least they have to have 550+mhz core and faster memory, anything else on 24pipe wuld be a waste, and even with those spec nV coud probably launch something the next day to match performance.
This may happen, but it would be stupid IMO.
They have an overstock of their old stuff in their warehouses, they want to sell those for as much as they can before they choke those sales.
That's not what their conference call mentioned it mentioned a lack of stock. And trying to sell old stock instead of new doesn't help the situation, it would be like trying to sell XGI Duos simply because there is a glut of them in a warehouse somewhere. ATi needs to treat this as a new product line not the extension of the old one. If they have 30,000 X800PROs sitting in a warehous somewhere, well then sell them for the same as a GF6800 or even GF6600GT, get them out of there, because they are dead weight and they do not contribute to the future. MAking more R520 chips, either 32pipe, 24pipe, 16pipe or 12pipe all goes to perfecting the chip and lower the cost per unit. Worrying about inventory does nothing except for keep the accountants happy. It's the marketers and economists who need to be running the show. Accountants kill companies!
No doubt they're getting the 32-pipe version to work, but they'll have way too many 24-pipe versions to not sell those immediately as the highend for awhile.
See above.
How well did the X800Pro sell (especially if the XT and XTpe were fully available)?
That's poor pricing strategy, they were trying to sell a 12pipe card for the same price as a faster 16pipe card from nV. How many of those PROs do you think they would've sold if they sold at $100less (aka X800XL prices)? They would've outsold the GT for $100 less, but whomeever priced the cards seemed to ignore the actual market and went of some stoopid grid (Well the XT/PE is this, our next one down should be this, regardless of what the competition offers).
I don't see how everyone is so damn hopeful for a 32-pipe card from ATi,
whoa, Whoa WHOA, with the hopeful stuff. I'm not hopeful, I'm simply telling you what NEEDS to happen in order to be competitive, not hope, my hope is a 48 pipe unified WGF2.0 card from Matrox, but that's not about to happen, either cause I'm realistic about it.
I'll turn your question around on you, why are you so damn adamant that it's not a 32pipe card as it's flagship?
there's so many things that have gone wrong on the road there that they almost HAVE to have a 24-pipe card as their #1 for awhile.
Those issues don't improve with more time of NOT making chips. The best way to fix that is to move the 10% yield you can get now for a $150 premium, and then have the massive 'value' 24pipe chip/cards as well as the premium of the 32 pipe version finance your re-spin etc. With releasing a 24pipe card only, all those potentially more valuable 32pipe chips simply get shelved or crippled. That's lost revenue right there. And if they do re-spin again, then that new batch is far more attractive to the refresh so once against you lose the value of the old 32pipe chips.
ATi can launch a 24pipe card in August, but it will need to be closer 600mhz and OC from there as well in order to have any chance to make up for lost time. And how do you think increasing the speed of the chips is going to affect their bining as well? Either way they need a far better part, and a 450mhz 32pipe part can have similar performance characteristics as a 600mhz 24pipe card. Either way you're in a bind for 'volume' shipments, so it makes far more sense to sell a 32pipe expensive boutique part for the PR BS aspect and then sell your 24pipe part for your volume. However that 24pipe part need 3 things to be successful for ATi, first performance, it has to be better than the GF7800GTX (without questions or exceptions), it has to be similarly priced (which will be hard for a new part versus a part that's been selling for at least 2-3 months), and it has to be cheaper to produce than the GF7800GTX, (which once again is hard to do compared to a more mature part).
Those are the realities ATi faces, and like you allude to, making another X800Pro or even X700Pro that can't compete just makes their hole bigger and loses any moemntum they need to get to get their credibility to a level of being able to compete. Remember right now nV has a part, they and their fans/supporters are trumpeting them from the roof tops, system builders will soon be doing the trumpeting too, by the time ATi decides to launch their part, it can't just be a staid, measured and humdrum, 'look we can do it too' announcement as that will not in anyway help their situation. nVidia has the luxury of sitting back, ATi has nothing right now. And while today's meager sales don't matter, those in a months time will matter once they have more stock in retail and the prices start to fall (which they don't have to if there's nothing to compete with).
Once again I don't hope this is what they will do, this is simply what they need to do, whether they are smart enough to see that is questionable.
EDIT:
I tell you if ATi does as you suggest and if I were with an investment firm/house on conference call when next spring ATi brings out their annual report and says "we had lower than expect sales and revenue from the launch of our new product" my answer to that would be a quite vocal "Well, Fvckin' Duh!"
- You need a licence to buy a gun, but they'll sell anyone a stamp <i>(or internet account)</i> ! - <A HREF="http://www.redgreen.com/" target="_new"><font color=green>RED </font color=green> <font color=red> GREEN</font color=red></A> GA to SK
<P ID="edit"><FONT SIZE=-1><EM>Edited by TheGreatGrapeApe on 06/30/05 07:56 AM.</EM></FONT></P>