DegustatoR
Veteran
Who says it'll have 9900 moniker? =)
I think that 9900 will be G92b based.
GT200/G100/D10U will use the GF10 naming.
Rumours are mixed up it seems.
NVIDIA GT200 Rumours & Speculation Thread
- Thread starter Arun
- Start date
- Status
I think that 9900 will be G92b based.
GT200/G100/D10U will use the GF10 naming.
Rumours are mixed up it seems.
How unlikely is it anyway that some folks have confused 99xx-whatever with GT200? Again look at the newsblurbs in the initial post of this thread. Am I the only that sees a supposed "9800GTX" on 55nm and a 512bit bus in one of those?
Yup, all three are correct and proper codenames, which are used for different purposes but all relate in one way or another to the same chip. I wouldn't be surprised if G100 originally referred to something else but that's not the point.
As for the 9900GTX, that's probably just G92b... While GT200 is a GF10, as the D10U codename implies. As for GT2xx, I'd expect a whole bunch of 55nm derivatives to be there in time for the Winter OEM cycle. Clearly won't be the same for Back-to-School though so it'll be interesting to see if G9x manages to be competitive against RV620/RV635/RV670.
Are you're expecting high-end G10 to debut on 55nm as well?
The U standing for "Ub0r-Enthusiast" or plainly an Ultra?
Stop reminding me there might be at least one more codename internally and possibly even more. Sigh - as if three codenames used by different people wasn't ridiculous enough already! Some time ago I saw The Tech Report saying Intel generally had an insane abundance of codenames, more than everyone else. Clearly this is all about NV being jealous and wanting to beat Intel in this most important metric... :| Larrabee only seems to have one codename, that does make it sound rather boring.
Berek
Regular
The bolded part sounds more like a step into the right direction LOL
I live in China and can safely say the Chinese NEVER know what is going on, except for perhaps video cards... and THAT says something.
Sorry, I just had to post that in here... I've been living here for a year and I need a break, hehe.
In order to avoid misunderstandings I am against the ton of leaks before any release and not anything else. And yes of course most of us here and anywhere else are dying to know more, but if there's an intentional "confusion campaign" from the side of IHVs I can fully understand it too.
Freak'n Big Panda
Regular
those specs contradict pretty much all the rumors we've herd so far. BS IMO. EDIT: I'm pretty positive at this point that GT200 is 2 G92s sandwiched together on one die with performance comparable to G92s in SLI, although probably a little bit faster (+10-20%).
DegustatoR
Veteran
What?
Freak'n Big Panda
Regular
well if it's not that then I think it's some sort of dual GPU design ala R680.
Freak'n Big Panda
Regular
I'll explain myself a little bit better. Maybe this isn't even technically possible (someone please tell me if it isn't) but really do think GT200 is 2 G92s on one die @ 55nm. With some sort of internal bus connecting the two together and then working in a fashion similar to SLI. I don't think it'll be a worthy upgrade for 9800GX2 users as it should offer very similar performance. Reasoning? Simply put monolithic single chip designs are on the way out. Both manufacturers are having a very hard time topping G80 performance 1.5 years after release. I don't think either manufacturer will have a true single chip solution that will significantly outpace G80 until at least this fall.
Sorry, but this absolutely nonsense.
G9x is a scalable design, where your solution would be uneconomic and inefficient and NV would never do such a solution.
NV could do [strike]192[/strike]128+SP/384-Bit@65nm SKU in 2007, but there was no need, since ATi offered nothing competitive.
But GT200 will more than just beat G80.
Last edited by a moderator:
Jawed
Legend
If we talk about 24SPs per cluster I see no problem, even if it would be 16 SP clusters, it would be around G80-size.Nonsense
- Status
