Megadrive1988
Veteran
I'd be surprised if G92 (or whatever Nvidia calls their true G80 refresh, the near 1TFLOP part coming Q4 2007) doesn't have a 512-bit bus. otherwise, AMD will have an opportunity to pull ahead with the R700.
NVIDIA: Beyond G80...
- Thread starter kyetech
- Start date
DegustatoR
Legend
Dual-chip is a niche product. It can be their only top-end in case if ATI's next top-end would too be a dual-chip product which we already know isn't the case (R650). So i'm still very doubtful of G92x2 being their next top-end -- i mean this is certainly nice for them from the business pov but it leaves them too unprotected for the potential ATI retaliation.
Why? What about the 7950GX2? That wasn't a niche product but the fastest single PCI-e solution last gen with the X1950XTX coming in second.
The G92 may well be a more refined G80, cut down in some places specwise to produce targeted low power consumption and heat so that these can be fitted in a GX2 style card. (while there could be other improvements of course)
However i would like a single chip high end solution rather than dual if that was the case.
Oh yeah, because the 512-bit bus helped AMD out so much this generation. Or are you implying that they are goign to go to a 768 or 1024bit bus next gen? Because the chances of that are less than 0. It's an imaginary number.
But will G9x be one chip?
R650 is apparently a fictional product. ATI was concerned about leaks and deliberately planted stories about R650 and even R670 to find out who was leaking.
There is supposed to be an RV670 product (better mid-range) but probably not till January.
But R650 won't happen and never was going to happen. (Supposedly).
Not sure where the January is coming from, but that's a little too far out as far as my estimations go. October's where my money is.
Hehe, I was right after all (as always?
).http://forum.beyond3d.com/showpost.php?p=1005859&postcount=12
vertex_shader
Banned
Ops wrong forum sorry
Well, technically G9x will be a family of chips. And there is always the possibility of a multi-chip board like the 7950 GX2. But for a primary product? I think we can all be sure that the primary product will be single-chip in nature. There may be a niche high-end multi-chip part, but it will still be a niche part.
Well, what I mean specifically is the following:Well, if you listen to people like Entropy, you could be led to believe that anything above $299 (let alone $399!) is niche. In that context, I'm not sure how to consider Chalnoth's comment...
1. I expect there to be a single-chip high-end flagship product released among the first batch of G9x products.
2. There may or may not be a higher-performing multi-chip solution released shortly afterward (~2-6 months) that outperforms the flagship product.
Well, if you listen to people like Entropy, you could be led to believe that anything above $299 (let alone $399!) is niche. In that context, I'm not sure how to consider Chalnoth's comment...
It's an interesting point, and really one of the major drivers behind SLI in the first place. Cater to the high-end niche market that *is* willing to pay outrageous amounts for ultimate graphics goodness, while not greatly adding to your own costs with a special asic that you can't possibly make a profit on for that volume.
If you go back a few years, you could make a case for $399 as a more natural high-end point to leverage from, and above that you're really deally with the lunatic fringe (god bless 'em, every one, said the fellow wingnut
).