G80 De-Mystified; GeForce 8800GTX/GT

[DMA]

* 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://www.vr-zone.com/?i=4007

Yikes.. :D

[Fotis]

Lol, there is already official word(Kirk,Hara) that G80 wont be a USA and it will be close to 500mil trans.
Core clock 1.5GHz?:lol:

[walterman]

Sounds too Unreal for me :D

[retsam]

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

Yikes.. :D
at those densities, we are going to be lucky to see 400mhz ....pfftt...only an idiot would believe these specs.;)

[PeterJensen]

wtf? If by anny chance (about 1%) these specs are true.....I will have to go fab.

[SH64]

I dont believe it .. seems too high to me. & whats that 384-bit memory interface again ??!
the only thing i believe is the model name which is in order.

EDIT : & a question : what does that TCP stands for ??

[nekrosoft13]

Transmission Control Protocol ?

[Banko]

Everyone knows that G80 isn't unified so I would say since that is wrong everything else is wrong as well.

[tertsi]

When did vr-zone hired Fuad?

even snowman laughing (snowlol)

[PeterJensen]

Everyone knows that G80 isn't unified so I would say since that is wrong everything else is wrong as well.

Why do everyone know that? When did Nvidia say that? And is that 100% confirmed? I think NOT!

[PeterJensen]

Well the specs fit the rumor of 100% faster then 2x7950 in 3dmark06.

[slaWter]

Man that would be cool, especially for that price ;)

[NoWayDude]

http://www.beyond3d.com/forum/showthread.php?t=33576&page=13

Although some folks in there are giving cryptic clues this rumoured specks are way of base.

[MUYA]

I believe that just because the NV official line was ..er we likey dun da USA...doesn;t mean the whole research and dev line weren't looking into it. ie.why give ATI the PR headlines of being the first USA based card to desktops cards?? think about it? but..yeah...pessimism is the flava of the day

[PeterJensen]

I still think it will be a USA card.

[CaptNKILL]

Damn, those are some nice specs... hope they're real.

[PeterJensen]

Right now i really think it's 50/50. I mean, they could be real.

[jAkUp]

ExtremeTech: We know that DX10 has unified pixel, vertex, and geometry shaders on the API level, but it doesn't dictate the hardware implementation: It's possible to have DX10-compliant graphics cards with discrete pixel and vertex/geometry shaders. Do you think time is right to bring them together, or is there still some benefit from having separate units on a hardware level?

Nvidia: There are benefits and disadvantages to both approaches. Unified certainly is interesting in that it appears architecturally "new" and different, and has some interesting promise in terms of load balancing and extensibility.

Frankly speaking, however, the graphics industry has gotten very good at extracting a lot of performance from the current vertex/pixel shader architectures, so the competition for anything new architecturally is a highly evolved and efficient architecture. The first rule of any new GPU is to be better at the previous API. Any trade-off which might move you away from that goal has to be evaluated carefully. If you look at some of the existing GPU architectures you can see some pretty big differences in terms of architectural efficiencies, and that is with architectures that, at least at the highest level, would present themselves as "non-unified" architectures. Even within that environment, you can see huge (almost 2x) differences in terms of performance delivered per unit of area or power.

If you assume that the competitors will be bound by the same laws of physics and economics, that alone would put one competitor in a dramatically better position. Would consumers be willing, for example, to pay twice as much for a graphics card that delivered the same performance as another, just because a particular graphics card was unified, or to pay the same price, but run at ½ the performance just because it offered some new architectural block diagram? I doubt it.

There are plenty of examples, including some from recent memory, of architectures that looked new, and sounded good on paper, but in implementation suffered greatly in the first law of GPUs—which is every new GPU architecture must be better than the previous GPU architecture, as measured by the things which characterized that previous GPU architecture. Unified will come eventually, but only when being unified delivers the best performance, architectural, and power efficiency.

Don't think it will be unified.

[PeterJensen]

I say it will be unified :)

[Redeemed]

I say it will be unified :)

These are the first ever specs to be released- and you're jumping on the bandwagon that they are accurate? ;)

I do doubt the legitimacy of these specifications. Remember all the hype that surrounded R520? All the supposed specifications that many said were accurate? Turned out terribly inaccurate and the R520 got its @$$ handed to it by the 7800GTX.

I wouldn't say these specs are right. The 384-bit bus sees accurate, as this is not the first I've heard of this. The memory, then, is spot-on as well. The clock speeds definitely seem a bit high, I'm thinking that G80 will be more like 700 to 900 Mhz on the core, and 2GHz on the memory.

As far as PP, VS, and GMs- I have no idea how the card will be configured. I could only guess.

And for the card to require a hybrid water cooler/ fan cooling solution would be rather unrealistic- as not everybody wants to use water cooling.

Oh, and in regards to G80 being unified, nVidia has already openly declared that it will not be unified. Arguing otherwise is saying that you would know better than nVidia does. That's isn't too likely. :)

[SH64]

And for the card to require a hybrid water cooler/ fan cooling solution would be rather unrealistic- as not everybody wants to use water cooling.
Yep! that too gave away the other unrealstic specs.

I still think it will be a USA card.
I hope it will be a USA card.
going with the same traditional split shading architecture (i.e. more pixel pipelines & a few extra vertex pipes) dosent seem to be the right step for the future unless we are talking about a new hybrid design like the one speculated above.

[Redeemed]

Yep! that too gave away the other unrealstic specs.


I hope it will be a USA card.
going with the same traditional split shading architecture (i.e. more pixel pipelines & a few extra vertex pipes) dosent seem to be the right step for the future unless we are talking about a new hybrid design like the one speculated above.

I am rather certain that G80 wont be unified, nVidia already came out and said that it would not be unified.

What I'm banking on is something like 48 pixel pipes, 48 vertex shaders. Not sure I like that idea all that much, but it seems the most likely route nVidia would take. As far as Geometry Shaders goes- your guess is as good as mine.

One thing mentioned in that article, the new method of AA- I do think that could be accurate. Though I'm wondering the full name. Vertice AA maybe? Not sure. :)

[agentkay]

There were pretty strong rumors that the G71 would have been a 32pp card which didn´t change until maybe a month before release. I´m not sure what to think of these rumored specs, but they do sound interesting and I wonder if the G80 maybe has unified vertex and geometry units since they might be easier to unify than all 3 types of shader units (pixel, vertex, geometry).

I don´t know if 700mill. transistors are realistically possible, but I kind of doubt it, even if it will be released on .80nm.

The 384bit memory bus sounds realistic and it would fit to rumors that Samsung is selling GDDR4 memory only to ATI at this moment. With a wider bus Nvidia wouldn´t need to put GDDR4 memory on the G80 as far as I understand this correctly.

I never truely believed the old rumored specs which said 32pp + 16 vertex/geometry units, as such a card would (in theory) only need around 400mill. transistors on .80nm.

[Xion X2]

Don't see why some are giving this info any credit..G80 won't be unified. I stopped reading there. Apparently this source has been on planet Zoltar for the past four months.

[ViN86]

Lol, there is already official word(Kirk,Hara) that G80 wont be a USA and it will be close to 500mil trans.
Core clock 1.5GHz?:lol:
yea i didnt think they would be moving to a unified shader architecture cause nvidia says it isnt necessary yet or something. 1.5GHz core clock too? yeesh, if thats true, looks like the 8800GT will serve me nicely :D