G80 Specs!

EternityZX9

Limp Gawd
Joined
Sep 15, 2004
Messages
236
* Unified Shader Architecture
* Support FP16 HDR+MSAA
* Support GDDR4 memories
* Close to 700M transistors (G71 - 278M / G70 - 302M)
* New AA mode : VCAA
* Core clock scalable up to 1.5GHz
* Shader Peformance : 2x Pixel / 12x Vertex over G71
* 8 TCPs & 128 stream processors
* Much more efficient than traditional architecture
* 384-bit memory interface (256-bit+128-bit)
* 768MB memory size (512MB+256MB)
* Two models at launch : GeForce 8800GTX and GeForce 8800GT
* GeForce 8800GTX : 7 TCPs chip, 384-bit memory interface, hybrid water/fan cooler, water cooling for overclocking. US$649
* GeForce 8800GT : 6 TCPs chip, 320-bit memory interface, fan cooler. US$449-499


http://vr-zone.com/?i=4007

Also there is a dicussion going on at:

http://www.xtremesystems.org/forums/showthread.php?t=116474

Enjoy! :D
 
Anyone know the reasoning behind the goofy amount of ram?


768MB memory size (512MB+256MB)

I would think theyd just double it or something. but what do i know. :cool:
 
EternityZX9 said:
I believe it has been suggested for the new AA mode, although that could be wrong.

The whole thing could be wrong.

It is all rumor at this point.
 
seing as how this info comes from a place other than the INQ it may hold some merrit as to its validity, but I still stand behind waiting for the actual chip makers to release info before jumping to any conclusion on any up coming new hardware.
 
768MB? Maybe a dedicated AA buffer then? *me wants* NO, dual 7900GTXs are NOT fast enough :)
 
I would honestly like to see expandable ram like a motherboard has, and an interchangable gpu. ^_^
 
w1retap said:
I would honestly like to see expandable ram like a motherboard has, and an interchangable gpu. ^_^
Wouldn't you need a bigger PCB for that? I mean I like the idea but I don't know about the space/heat/bandwidth issues that would arise....I did hear something about external PCI-E though as if the GPU is almost a seperate computer. If that happened (and latency wasn't an problem) I could see swappable GPUs and memory.
 
Intersting, but I dont believe it. I can see the G100 series having "ELVENTY BILLION" transistors.
 
I'm skeptical because I thought it was pretty much confirmed nvidia weren't going the unified architecture route with the G80
 
tornadotsunamilife said:
I'm skeptical because I thought it was pretty much confirmed nvidia weren't going the unified architecture route with the G80

just a thought, but maybe the VS/GS are unified, but not the VS/GS with the PS only the instruction sets between them



/note that I'm just posting an opinion here, I in no way am confirming or denying specifications
 
ThreeDee said:
I sure hope they make an AGP version ...





...umm ...yeah

:)



all you old geezers on agp! slackers! update your interfaces! :p


Or think of it as an excuse to upgrade your entire rig to be pci-e compliant ;P
 
EternityZX9 said:
* Unified Shader Architecture
* Support FP16 HDR+MSAA
* Support GDDR4 memories
* Close to 700M transistors (G71 - 278M / G70 - 302M)
* New AA mode : VCAA
* Core clock scalable up to 1.5GHz
* Shader Peformance : 2x Pixel / 12x Vertex over G71
* 8 TCPs & 128 stream processors
* Much more efficient than traditional architecture
* 384-bit memory interface (256-bit+128-bit)
* 768MB memory size (512MB+256MB)
* Two models at launch : GeForce 8800GTX and GeForce 8800GT
* GeForce 8800GTX : 7 TCPs chip, 384-bit memory interface, hybrid water/fan cooler, water cooling for overclocking. US$649
* GeForce 8800GT : 6 TCPs chip, 320-bit memory interface, fan cooler. US$449-499


http://vr-zone.com/?i=4007

Also there is a dicussion going on at:

http://www.xtremesystems.org/forums/showthread.php?t=116474

Enjoy! :D

I also will comment this with a rather big grain of salt, but if this holds true, this is indeed a beast and even the external power supply rumor might be true.

In what games are concerned, if these specs are true, you probably just double-click the game's icon and the first thing that appears on your screen (after entering 3D mode) is: GAME OVER...the card is just too fast...

I am joking of course :)
 
Dunno why but my feeling is that those specs are bogus. nVidia never announced official specs and others told it has been postponed so we must take those specs with a fat grain of salt ;)
 
tskiller said:
Anyone know the reasoning behind the goofy amount of ram?

Supposedly it is the side effect of desiring a wider memory bus. Obviously they are not doubling the memory bus like we normally expect in the byte world (256 to 512). You can either make a wider bus by adding chips with a wider path or just add additional chips and their corresponding traces of course.
 
tskiller said:
Anyone know the reasoning behind the goofy amount of ram?

768MB memory size (512MB+256MB)

I would think theyd just double it or something. but what do i know. :cool:
Each memory controller likely supports a 64-bit width (2 x32 chips). A 384-bit interface like that would have 6 controllers and use 12 x32 RAM chips. It's not really 512MB + 256MB, it's more like 6 128MB channels.

If that part of the spec is true, it seems like that was a trade-off. A 384-bit memory interface uses fewer pins than a 512-bit memory interface so cards will be cheaper to manufacture. And moving from a 256-bit interface with GDDR3 memory to a 384-bit interface and GDDR4 speed memory gives a huge increase in memory bandwidth anyways. A 512-bit memory bus might not be required for the G80.
----

VRZone has been very off about other leaked specs in the past. I'd take most of the "specs" listed above as being wrong until other confirmation. ;)
 
Even though these are just rumors, it's fun to discuss :)

Besides, with my (hopeful) hiring for an on-campus job, I can upgrade to an 8800GTX and keep my 7900GTX for physics, then sell the other.... :) :) :)
 
Sovereign said:
768MB? Maybe a dedicated AA buffer then? *me wants* NO, dual 7900GTXs are NOT fast enough :)

No they aren't.

I must say that if the pricing has gone up to $649.99 as suggested in the first post, I am a little pissed about that. I am getting tired of the price of these things going up every six months.
 
LOL at the 700M transistors. That's insane.

That means it has 2x as many transistors as the 7900 series, the GT is also 2x the price as the 7900gt. That means it should use about 2x the energy, and had damn well be twice as fast to merit all this!
 
EternityZX9 said:
True, but like Blauman said, at least it's not from the INQ! :)


Though much of what it says was already said a couple days ago at the Inq. It would be so easy to take what they said and build specs around it (like the weird dual memory interface crap).

I don't believe a word of it. Though the specs are at least rather entertaining. Cheers to the OP.
 
I just don't see 700M transistors happening, especially at up to 1.5GHz unless they're now borrowing technologies from the CPU market to fab these chips. The proposed process is what, 65nm? Even then, this one's mighty tough, and we haven't even talked about heat yet. At $650, I think I'd be adding a second radiator, another pump and a good block, so the upgrade would cost me upwards of $800. Yikes indeed.

An interesting rumour otherwise.
 
Not likely to be used for physics.. I dont think there are enough transistors to have a dedicated physics chip onboard. besides, that would have been rather fast to create a new gpu AND a ppu, and get them onto a single PCB. I rather like the dedicated AA memory though.
 
there is a severe lack of Fab size, so power consumption could still be anywhere from as bad as the 7950, to as good as the 7300 dangit.

seriously, grain of salt with this one.
 
couldnt possibly be as bad as the x1900's power consumption if its on a smaller process... its not like most people here dont have the overhead to handle it anyways...
 
yeah, but who the hell would really want to deal with it in a new age where power consumption is as critical as actual computing speed?

(still, wouldnt stop that 1% of PC gamers that just need the best)
 
Hmmm, Brent speculated about the interpretation of the specs, but he didn't cast any doubt on the accuracy of the specs. Maybe they are pretty close to the real deal. :cool:

Same deal with Dan the motherboard editor--unhappy about the price, not skeptical about the specs. Do these boys know things we don't? I sure hope so!
 
None of this makes sense.

for one why would nVidia allow a complete U turn on power consumption and heat.

The 7950GX2 draws less power and makes less noise than a stock X1900.

As the chewbacca defence goes.

It does not make sense.
 
Fawkes said:
It does not make sense.

Sure it does. Nvidia has seen the performance of their gpu's against ATI's and probably know the performance of the R600 and have built up their offerings to be competitive and hopefully to be the sampe jump in performance was the case with the 6 series vs the 5's, 7's vs 6's.

Hope the new card is twice the performance of the 7900GTX.

Maybe its also a sandwich card like the 7950 GX2.
 
I know these might all just be rumors but if you look back from history:

http://www.rojakpot.com/showarticle.aspx?artno=88&pgno=1

Going from Geforce 2 to 4 to 6....Look at the transitor count and correlated other aspects those cards. Notice how many of the aspects double in retrospec of the previous two generations. This is just something I notice, it doesn't make things fact, but it makes me believe that this next card from nVidia could just as well have specs close to what is said in the article (Exameple: ~700M transitor count). However, imho I'm certain nonetheless the G80 will be a beast whether or not these specs are accurate (Same goes for the R600).

I mean it's completely possible to have another era of 9800 performance or shattering FX to 6800GT performance are we not?

Now, we will just have to wait and see how it all plays out.
 
Commander Suzdal said:
Maybe this will be the generation that allows us to play Everquest 2 at a quality setting above Balanced?

haha, I will DEFINITELY test it! thanks for reminding me
 
Back
Top