Shoutbox

* 0.11 micron process TSMC
* 430Mhz core / 1.2GHz 256MB GDDR3 memory
* 256-bit memory interface
* 38.4GB/s memory bandwidth
* 24 pixels per clock
* 10.32Bps Fill Rate
* 8 Vertex Pipes
* 860M vertices/second
* 400MHz RAMDACs
* NVIDIA CineFX 4.0 engine
* Intellisample 4.0 technology
* 64-bit FP texture filtering & blending
* NVIDIA SLI Ready (7800 GTX only)
* DX 9.0 / SM 3.0 & OpenGL 2.0 supported
* G70 comes with 3 models; GTX, GT and Standard
* Single Slot solution
* Single card requires min. 400W PSU with 12V rating of 26A
* SLI configuration requires min. 500W PSU with 12V rating of 34A
* Launch Date : 22nd of June

Source: VR-Zone


64 Bit FP Rendering

The smaller microprocess (0.11) results in less power need, and less heat,

the 6800 was released with CineFX 3, and Intellisampe 3.0 tech

this means that Nvidia have again rearchitected thier Antialiasing and Anisotropic filtering methods.

this new series pushes more pixels through an amazing 24 PIXEL PIPELINES and it has 2 new vertex shaders to increase from the 6 of the 6xxx range to 8 for the 7xxx series

this card looks to hit the shelves hard,

note that Clockspeeds can be modified from stock settings by nvidias partners

---

Early Samples of the G70 (7800) have pulled scores in 3dmark05 of 7800points

this is a large jump from the 6800, as drivers mature more this might increase again

note that this score of 7800 was made on the standard edition and not the ultra or GTX

this is only for the rich kids...

prices of hardware drop fairly fast tbh, it wont be amazingly expensive
for too long, i'll be wanting one maybe 6 months or so after they've
been out

No surprise that NVidia have a new card: however, the chip they've designed for PS3 wipes the floor with that one.
Then again, can't really compare them: they're designed for different types of machine.

Edit:
Call to Admins: The argument below wasn't going anywhere, and it's also started to veer offtopic. It may be an idea to split it off somewhere.

um Race..... hate to tell you this mate

but that chip kills the PS3s

the PS3 is a dumbed down version of the 7800

ie ps3 uses heavily modified GL70 core

trust me.. ain't no way the ps3 will take this on and win

heres the specs as proof

Geforce6800 - 16 Pixel Pipelines/6 Vertex Pipelines, 2 Full ALUs and 2 MiniALUs per Pixel Pipeline and 1 ALU per Vertex Pipeline - TOTAL:38 ALUs + 32 MiniALUs (Effectively a total of 54 ALUs)

G70/RSX (NV50) - 24-32 Pixel Pipelines/8-10 Vertex Pipelines, 16 Pixel Pipelines/6 Vertex Pipelines, 2 Full ALUs and 2 MiniALUs per Pixel Pipeline and 1 ALU per Vertex Pipeline - TOTAL: 56-74 ALUs + 48-64 MiniALUs (Effectively 80-106 ALUs)

Xbox360 GPU - 48 Unified Pipelines, 4 Full ALUs per pipeline - TOTAL: 192 ALUs. Also the eDRAM also has another 192 ALUs for processing FSAA/MSAA, Stencil buffers, occusion culling, and ZBuffers - Grand total between the eDRAM and shaders... 384(!!) ALUs.

Geforce 6800 Ultra:
53 ops per clock * 400MHz = 21.2 billions ops per second

Geforce 6800 Ultra Extreme Edition
53 ops per clock * 450MHz = 23.9 billion ops per second

PS3 GPU (RSX)
136 ops per clock * 550MHz = 74.8 billion ops per second

Xbox 360 GPU (R500)
192 ops per clock * 500MHz = 96 billion ops per second

Actually, I'll retract my argument until I have the FLOPS data.
Having done a course on Coputer Graphics, I can safely say the more FLOPS, the better the graphics.

Anyway, for those ops per second you calculated: you've only worked them out for one pipeline.
1.8TFLOPS/74.8GFLOPS = 24, which is the number of pipelines in RSX I think you'll find.

no

the RSX has 16 pipelines

i'll sort it out so its easier to understand

G70 - 24-32 Pixel Pipelines/8-10 Vertex Pipelines,

RSX (NV50) - 16 Pixel Pipelines/6 Vertex Pipelines,

2 Full ALUs and 2 MiniALUs per Pixel Pipeline and 1 ALU per Vertex Pipeline - TOTAL: 56-74 ALUs + 48-64 MiniALUs (Effectively 80-106 ALUs) for both


now the only thing that makes the PS3 better then the normal computer is its Cell chip running at 3.2ghz

BUT

the current Athlon64 FX55 can take a ps3 down without getting hot

First, you should know not to double-post.
Second, I worked out 24 pipelines from your figures and NVidia's own claims. I have subsequenty tried to find somewhere that confirms this, and failed so far. So while I may not have the data to back me up, neither do you for your claim of 16.
Thirdly, you can't really compare Cell and Athlon: they're totally different chips, intended for different applications.

At the end of the day, there's very few figures you can use to truly measure performance. For GPUs, the main one is FLOPS (floating-point ops per second). For CPUs, it's CPI (clocks per instruction) coupled with clock speed.

I won't carry on until I have the figures to hand. I recommend you do the same

no really.. Cell is designed as a central processor.. the same as the athlon...

and ok listen k

The 7800GT/Ultra will have the full 32pipes
The 7800nu will have 24 pipes

i have long been in the know that Nvidia didn't actually scrap the NV50, sony bought it of them

the nv50 is as it seems, just how i expected nvidia would make it

only thing is

the RSX is actually an NV50/GL70 Hybrid design, but the only thing incorporated from the GL70 is the OpenEXT extensions and CineFX 4.0

really only time will tell.. theres so much data out there right now who knows whos right,

we might just both be wrong

Cell and Athlon may both be CPUs, but that's where the similarities end. They both have different architectures (AMD64 vs PPC), and they have different internal designs.

And the RSX isn't a hybrid NV50/G70 at all. For starters, the NV50 does 32-bit FP, G70 does 64-bit FP, and RSX does 128-bit FP. Not only that, but Sony asked NVidia to make (and I quote) 'the most advanced GPU in the world'. To me, that indicates it's not some hybrid, but a whole new design.
All the info I have on the RSX came from Sony's pre-E3 press conference, available on IGN.

Edit: Y'know what, I can see this argument going nowhere. Shall we just agree to disagree and call it a day?