NVIDIA Fermi: Architecture discussion
- Thread starter Rys
- Start date
Strict enforcement? Want to see a forum with rules and strict enforcement go post on something like doom9, they are regular forum nazis (lets hope they aren't reading this or there will probably be some rule I'm infracting). This is one of the most hand off moderated forums out there, that's the charm. Although it is of course a bit of a technocracy, reputation helps (which doesn't mean this post won't get deleted, which it will ... just learn not to care). So if you are a newbie don't call people fanboys too often, it's just not going to end well.
Coming from you well sorry doesn't mean anything
good link
no
In open gl applications the nv30 did just fine. That was good use of PR, anandtech took it as accross the board.
The fact that we've not even heard a wage statement about performance can mean only one thing: Fermi will be a disaster for nvidia ... they can't get enough performance out from a lousy, hot, expensive design - end of story.
chavvdarrr
Veteran
Oh, you were there when NV PR talked to AT?
They specifically said "opengl" and "across the board" was what AT thought, was not said from NV?
Doesn't matter how much Nvidia pays you - its not enough
Oh, you were there when NV PR talked to AT?
They specifically said "opengl" and "across the board" was what AT thought, was not said from NV?
Doesn't matter how much Nvidia pays you - its not enough
Yeah know what, I'm pretty sure I know what nV PR stated when they showed that graph
, and it wasn't across the board :smile:, it could be assumed that way, but that is the past, Fermi, isn't anything like the Fx or r600. a week and half they will be showing some more things at CES, should be fun.
No? It's just doing an uncanny impression so far I guess is the problem.GF100 is not a second nv30.
No? It's just doing an uncanny impression so far I guess is the problem.
So far we know knowthing about the performance. The uncanny impression comes only from one person.
In a way, GF100 being the "next nv30" would be good for the market. AMD seriously lags NV in GPUs and (duh) Intel in CPUs. AMD taking a 9700/5800-esque lead in GPUs for a good period time would improve the overall competitive situation - not just in GPUs but in CPUs as margin can be reinvested...
Not a technical point, but perhaps a different way to view any GF100 shortfalls.
Not a technical point, but perhaps a different way to view any GF100 shortfalls.
FrameBuffer
Banned
Yeah know what, I'm pretty sure I know what nV PR stated when they showed that graph
not sure why people keep saying that the R600 was THAT bad .. if anything, if Fermi was to follow the R600, it would probably be an almost best case scenario for nVidia. After all the same very R600 that spawned the Rv670/630/610 -> Rv770/73x/71x -> RV870/Etc/Etc and has allowed ATI to compete from top to bottom, something nvidia has had a serious problem with it's latest generation (G200 series derivatives). Sure the R600 was LATE, it also introduced a new architecture and allowed (as noted) ATI to become much more modular.. (Gee that does sound a lot like Fermi now that it's put that way.. NOT a bad thing afterall).
So I guess I would put the Fermi along the ranks of the R600, unless performance is dismal (compartively) then it would move into the realm of the FX.. that is a ways off though.
As a note, I respect Razor's opinions however I would take anything he says in relation to supposed performance/release dates with a truck load of salt.. afterall he was saying a November release, then December.. then Jan.. all being so called "internal release dates" with no means of confirmation other than hearsay. Not going to get into that whole deal (I mean I could just as easily say the RV970 has a mid '10 internal release date and would expect my word to be taken as gospil ?!) Again I respect people's opinions but when they try to pass them off as fact without any confirmation.. thats right up there with saying the G80 and the G200 series (285GTX .. 295) are all the one and same.
the r600 wasn't bad by any means, only had issues with AA performance. It did hold its own when it came to shaders.
To compare it to the Fx, thats very hard to do, the Fx series had not one problem, it had several thats what caused it to fail everything from design mishaps to Dx specs. If nV was to make those kind of mistakes again with Fermi doesn't seem likely at all, design of its shader cores (funcationality wise) isn't that far from what they had before, so shader core should be pretty solid of course ALU structure has change hence drivers are going to be a raw, missing Dx specs are they going to do something like, NO. Now do you think nV will annouce the new AA mode of x32 if performance is going to be in the shitter? Last time they had performance issues first thing to go tweak was AA and AF.
I've stated, nV stated the same publicly to some degree. Jan we have had 1 or 2 reps state them now so, if they slip, they slip.
Metal respins aren't going to increase performance greatly reguardless of certain people saying that.
So nv3x was not really bad because we got nv40, g70 and g71?
AMD lost one year to nVidia. They needed three generation of GPUs to beat them.
r600 was bad, really bad.
itsmydamnation
Veteran
3 generations???? last time i looked RV770 schooled NV, forced them to do massive price drops, refund money to customers etc, so by my count thats 1 generation............
They priced their cards very agressive, but they were not faster than GT200. And nVidia put the GT200 against the rv770 and with the g9x they don't have problems in the mainstream segment.
Look at their marketshare - they don't lost anything or 2, 3% to amd in the desktop market.
Given the massive amount of marketshare AMD lost during the 2xxx and 3xxx generations, I'd say gaining back ground is a huge turn-around. No doubt you'll continue to spin though.
itsmydamnation
Veteran
NV had to battle a 270mm GPU with a 500mm one. I fail to see how on a GPU performace basis (which is what everyone here is talking about) it didn't beat GT200. sure to consumers NV can just drop price which they had to do but here at B3D i think people look beyond that and the fact that such a small GPU could complete with a far larger one.
Not only that, but look at mobile discrete too- a market ATI had very little say in previously, and mostly only budget oriented then. And it's not as much Bumpgate as nVidia's stale low/midrange offering for so long (2y+).
