NVIDIA GF100 & Friends speculation

I thought it was the 14th

 

Hm, the 14th is the launch of ATIs 5670.
might fit for fermi, too. ^^

 

I thought it was the 14th as well. As a matter of fact I thought Rys split off the architecture discussion for exactly that reason. You know, make room for the discussion of the article he's gonna post at midnight tonight.

 

Thanks. Since I'm not a regular there and am on a lazy day (thus not wanting to read the entire thread) what were/are IceColds GT200 complaints and what exactly does Chris mean with:

What is "this" problem ?

 

They have all one setup-engine. Means they are able of doing one triancle/clock. nVidia's problem is, that the base clock is lower than the clock of AMD cards. Cypress with 850MHz vs. (maybe GF100) 650MHz. And you need this for tessellation.

 

So were IceCold's complaints about GT200, related with the lack of a tessellation engine ?
Because Chris seems to be talking about the problems that GT200 had (of which IceCold complained about) and that GF100 addresses i.e. fixes them.
One of them being this "impressive tessellation engine" which if indeed is "crippled" by the lower clock frequency, when compared to Cypress, won't be that impressive...so I'm still in doubt that that's what Chris meant.

Care to help out Chris ?

 

Geometry performance on Fermi is going to be extremely good.

 

Icecold's main complaint was that the GTX280 didnt wipe the floor with his GTX8800. NoT powerfull enough for him to upgrade to I guess. So Im assumming Chris means that Fermi wont dissapoint him in this regard, but in comparison to his 8800 or to the 280 ??

 

The sticker on Fermi's fan is going to be extremely good.

 

Telling the truth, if you want to me to be more specific, cards out there now aren't going to get close to it when it comes to geometry processing.

 

On CES presentation, there was (only) one performance graph that compared NVDA card with ATI one.

That graph made people laugh, somebody was disapointed, others were amazed.

My first post, and first questions is: can anybody tell us what was the title of the graph, what number was on the green side, and what number was on the red side?

 

"Some people laughed, some people cried. Most people were silent. I remembered the line from the Hindu scripture the Bahgavad ghita: Vishnu, trying to persuade the prince to do his duty, takes on his multiarmed form, and says "Now I have become death, Destroyer of Worlds, I knew we all felt that, one way or another."

 

I split it off to keep GF100 product speculation away from serious architecture discussion. I've just moved the stuff that hung off Chris's post to this new thread too. And if there's anything coming out tomorrow, it won't come from me

 

Can you tell us why?

 

that I don't know the reason why its faster, just faster , I'm sure you do though!

 

Having a faster setup rate is nice, but it's not the be all and end all ... geometry processing is not necessarily ALU light.

 

Faster geometry processing does not necessarily mean higher triangle setup rate, although that would be nice.

 

I thought that supposed NV engineer said the NDA expires on the 12th? Wheres the info?

Edit: Rahja, that was his name.

 

Maybe

Naw thats not what I am talking about. I'm not gonna say anymore beyond this point. Since revealing anymore would be a blatant break of NDA. Knowing that there are some big GPU level improvements and an impressive hardware tessellation engine isn't that big of a deal.

Shouldn't be so quick to believe everything you read on the web...

 

Yes.