NVIDIA Fermi: Architecture discussion
- Thread starter Rys
- Start date
It is definitely late, by many months, whichever way you slice it.
Whether it will be able to join the hall of fame, manned by the legendary NV30/R600, cannot be known before it launches.
However, I do not know of a chip (cpu or gpu) that was 7-8 months late and didn't prove to be an epic fail. Maybe someone more knowledgeable can enlighten us.
They are a epic fail because the competition was to strong. That did not happen this time. Cypress performance is not very impressive. GF100 must only be 35% faster (average) than a GTX285.
Considering 5870 is probably on the order of only 30% faster than a 285...yeah, it should.
You're sure you're setting your expectations way too low? What you describe is roughly on HD5870 performance level (especially since the latter is more or less on GTX295 level). Let's be realistic and hope for at least 80% higher average performance than a GTX285 otherwise it'll be rather disappointing for a new technology generation.
***edit: oh and if you really want an example of GPU that despite being late wasn't a flop how about the R520? Yes of course can you start arguing that it was quickly replaced by R580 (which incidently was a result of the 520 delay) or nitpick on the product's shortcomings, but it was far away from anything I'd call a "flop". Au contraire it stood more than decently against G70 at the time.
I don't see anyone being overhyped considering it's a new generation. With refresh lines within the same generation it's normal to have rather mediocre expectations, but with a new generation and a hypothetical "new" architecture expectations are a lot higher than in the latter case.
Or else if NVIDIA merely wanted something that performs give or take 30-40% better on average than a GTX285, they could have just added DX11 functionalities, add a couple more SMs to it under 40nm and call it a day. Primarily the comparison in my mind is to each IHVs own predecessor. If a new generation delivers just as much as a boring refresh of their existing GPUs then ultra-yawn in the N-th degree.
Or else if NVIDIA merely wanted something that performs give or take 30-40% better on average than a GTX285, they could have just added DX11 functionalities, add a couple more SMs to it under 40nm and call it a day. Primarily the comparison in my mind is to each IHVs own predecessor. If a new generation delivers just as much as a boring refresh of their existing GPUs then ultra-yawn in the N-th degree.
If GF100 is merely neck and neck with Cypress, then it will be a financial disaster for nv in the consumer market. Dunno, how much HPC and professional markets will be able to save nv's profits.
Competition of fps makes sense for consumers.
Competition of fps/area makes sense for companies.
From a corporate perspective, competition is better defined in terms of fps/area. No idea how it is shaping up in that regard.
And what is the take on this new chip then? Is it a boring refresh wrapped up in a new GPGPU-oriented package or is it something that is going to be good at games and good at GPGPU.
I guess all this discussion about what people use for high performance computing now (home consumer style CPUs versus server Xeons) is directly comparable if NV launches the new chip and it is basically the same hardware for home consumer GPUs and Tesla. I thought NV learned their lesson with regular old GeForces just being softmodded to Quadros.
GF100 high-end needs to match 5900, not 5870. AMD could well do a price cut at the GF100 launch and spoil it, much like how Nvidia M.O. has been over the last few years.
Judging by Rys's confidence it very well might, only hope its out sooner than what is being thrown around.
Somewhere in between 5870 and 5970, according to the latest news, closer to 8 then to 9. (unless I got it completely wrong and it's 10% faster then a 5970.
Or you could take the simple route, take two times gt200b (GTX295) and use that as a reference. if it's close to GTX295, they have done a good job.
