nvidia "D8E" High End solution, what can we expect in 2008?

well if there's the addition of FP64 and DX 10.1 support, that's already much more than the difference between G8x and G9x (at least the two G9x we really know, G92 and G98)

 

http://bbs.chiphell.com/viewthread.php?tid=14253&extra=page=1

 

To late for wish-lists, christmas is over...

 

Is that guy known for "leaks"?
Just wondering due this in the end of the post:

Bolding by me

 

I do not know Chinese language . I had problems with the translation of the page

 

Neither do I, that's Google's translation

 

Yes, but the Google's translation - and Babelfish - was not working from my computer this morning in this Chinese page . For this motive I had put this it here, here some people are polyglots

 

thank you. that makes sense.

Nvidia's next new architecture won't arrive until 2009 at the soonest. maybe not until 2010.

NV60 / G100 / GeForce 10 doesn't have to be a Direct3D11
part, but it could be. it probably will be.

 

They still have to deliver D3D10.1 GPUs in the meantime.

I've no idea what they name internally things anymore, nor does it really matter what the final products will be named. We'll probably see a series of high end D3D10.1 GPUs until the D3D11 thingy finally arrives and the first may as well be name GF10 in the meantime. In the past I had heard about a supposed "G200" intermitting "G90" and "G100".

If those codenames really existed and it wasn't just a fool's errand, then I'd guess what was supposed to be "G200" could still arrive in next year but definitely not as early as Fudo wants everyone to believe. If s.o. from NV is in a humorous mood tomorrow and tells him that "G100" comes in May with 640SPs he'll plaster it on his page w/o even double checking if it's even possible heh...*shrug*

 

Memory at 3200Mhz .. wow .. ok it is GDDR5 .. but that's huge bandwidth. How much bandwidth can we be expect if it is 780Mhz core and (2GB)3200Mhz mem. with a 512Bit bus? I know we missing ROP's etc. but what is the minimum/maximum bandwidth(ballpark figure is fine)?

US

 

Sounds like Nvidia have pretty good feeling that they are well ahead of ATI R700.

 

I thought R700 is something for 2009. From AMD slides it seems that 2008 products are R6xx based. I think now is the appropriate moment to use the roll eyes smiley, no?

 

Bandwidth = memory bus width/8bits per byte * effective memory clock speed

That gives 204.8 GiB/s.

 

There seems to be at least a whispering campaign going on from the AMD camp that this isn't so and R700 will show up in 2008.

 

The question is, which "R700" is meant.

I think it is 3.2GHz real clock, since GDDR5 starts at 2-2.5GHz.

 

Any idea time frame apart - different between R680 and R700?

Edit: Maybe 4 months??

 

There is a rumor that 9800 GTX could be G92 based, so if this is the case, nothing more to expect than an heavy overclocked 8800 GTS

http://www.hwupgrade.it/news/skvideo/geforce-9800-gx2-i-primi-dettagli_23750.html

it's in Italian , I will report the phrase in the news:

"A seguire NVIDIA presenterà dei die shrink delle proprie soluzioni GeForce 8800 GTX: parliamo della scheda che dovrebbe prendere il nome di GeForce 9800 GTX, costruita utilizzando chip G92 con tecnologia produttiva a 65 nanometri mantenendo presumibilmente le stesse specifiche delle schede GeForce 8800 GTX in termini di stream processors as the 8800 GTX.

Una scheda GeForce 9800 GT dovrebbe venir presentata da NVIDIA tra i mesi di Marzo e Aprile: mancano informazioni sull'architettura utilizzata da questa GPU ma presupponiamo possa trattarsi sempre della GPU G92, con frequenze di clock incrementate e presumibilmente sempre con bus memoria da 256bit di ampiezza."

Translated, is (in Italic my comments):

"Following (the 9800X2) Nvidia will announce the die shrink of their GeForce 8800 GTX solutions: we are talking about the board that should take the name Geforce 9800 GTX, built using G92 chips with 65 nm process technology, presumably keeping the same specification for the number of stream processors.

A 9800 GT board should be also announced by Nvidia between March and April: informations about the architecture used by this GPU but we suppose it could be G92 based, too, with higher clock (than the 8800 GT) and the same 256 bit bus"


Obviously this is not G100, but (again , if this is true) it could mean that G100 will not arrive before 5-6 months at least.

 

Kewl thx AlStrong

US

 

We all know how well heaps of bandwidth helped the R600 chip...

 

I think it's more because they used it un-effeciently. The R600 just couldn't do AA properely .. it wasn't because of the heaps of bandwidth available.

US