The NV40 has taped-out

[micron]

yep, the higher the number.....the better the card

Frequently this is true within the same product family. I should hope that most buyers know enough not to compare the "9800" with the "5900" and automatically assume that the 9800 is better. There is much more to the comparison than that. However, it does remain pretty straightforward that "5200" is lower than "5600" is lower than "5800" and so on and so forth. There shouldn't be any confusion here, and certainly less than there used to be (with MX and whatnot in the names), though ATI's lineup is bound to be more prone to confusion than nVidia's at the moment (I'm referring to the 9000 and 9100 here...).

I was trying to make a funny in my above post Chalnoth , I hope you dont really believe that's how I chose graphics cards.
I buy whatever [H] tells me too

 

[KimB]

Yeah, I know. But that was kind of my point. I know it's a fallacy to think that higher numbers=better, but that doesn't mean it's not an okay guage for "noobs."

 

[elroy]

More info on NV40 - http://www.theinquirer.net/?article=11002. It says it'll be made up of 300M transistors! It is from the Inquirer though, so I guess it's to be taken with a grain of salt.

 

[mboeller]

found the link in the 3DCenter-forum. Link is from nvnews; original is from Darkcrow :

discuss !

 

[Joe DeFuria]

Hmmm...

No NV38?

NV35 is "it" until NV40 arrives?

 

[Ailuros]

Does it state anywhere how old that roadmap is? I wouldn't be surprised if NV38 was decided down the line to counter ATI's R360.

 

[PaulS]

NV38 was a very recent development, AFAIK.

 

[mboeller]

Upps...


here is an different roadmap from Demirug :

and

seems more in line with what we know, but here too no NV38.


And here is the complete presentation. Thanks ram

http://suif.stanford.edu/~courses/cs343/l10.pdf

 

[KimB]

Yep, according to that roadmap, it appears that the only GPU planned for this fall (other than the NV40, though that may end up closer to winter) is to be an upgrade for the NV31, the next $200-range GPU.

 

[McElvis]

Interesting stuff in that .pdf

Example -

Embedded DRAM - Tempting to embed megabytes of DRAM, But...Cannot fit the whole problem, costs are huge. Will be "just around the corner" for a long time. OK for console

 

[KimB]

My take on eDRAM was always that chip manufacturers will be forced to go in that direction when external memory just becomes too slow. I think that eventually some company will take that first step (as happened with the 256-bit interface), and others will be forced to follow.

As an aside, I think the use of better semiconductors than silicon also falls into the same boat. Somebody needs to take that first step. That first step is usually what's more expensive, not the process itself. I think this is true with eDRAM, and I think it's also true with GaAs. Once the engineering gets up to speed, engineers will find ways to get the costs down so that it isn't all that much different from the less advanced technology.