The World's First DX10 GPUs
On November 9th, 2006, NVIDIA launched their long-awaited G80 GPUs, now christened as the GeForce 8800 GPUs. Not only does this new series boast an entirely new massively-parallel unified architecture, it also marks the beginning of a new era in 3D graphics - DirectX 10.
The new GeForce 8800 GPUs have an amazing 681 million transistors and are fabricated at TSMC using the 90nm process technology. NVIDIA will be delivering two different GPUs based on the same core, with only differences in the number of stream processors and memory bus width.
Leading the new GeForce 8-series is the NVIDIA GeForce 8800 GTX. Armed with 128 stream processors and a 384-bit memory interface, this GPU boasts a fill rate of 36.8 GigaTexels per second and a memory bandwidth of 86.4 GB/s.
Following right behind is the lower-end NVIDIA GeForce 8800 GTS. It comes with 96 stream processors and a 320-bit memory interface, allowing it to achieve a fill rate of 24 GigaTexels per second and a memory bandwidth of 64 GB/s.
Model |
Price |
GeForce 8800 GTX | $ 599 |
GeForce 8800 GTS | $ 449 |
GeForce 7950 GT | $ 299 |
GeForce 7900 GS | $ 199 |
GeForce 7600 GT | $ 159 |
GeForce 7600 GS | $ 129 |
GeForce 7300 series | < $ 99 |
Graphics cards based on the new GeForce 8800 GTX will be priced around $599 with 768MB of GDDR3 memory, depending on the board manufacturer. GeForce 8800 GTS cards, on the other hand, will be priced at only $449 with 640MB of GDDR3 memory.
As you can tell, this immediately pushes down the prices of older NVIDIA graphics cards. According to NVIDIA, this will be the new product line-up (see table on the right).
There's no mention of the GeForce 7900 GTX or the GeForce 7950 GX2 in the price line-up, but they should be priced between $400 and $449. Meanwhile, we expect to see the GeForce 7900 GT around $250.
Specifications
Architecture |
GeForce 8800 GTX |
GeForce 8800 GTS |
Architecture |
G80 |
G80 |
Process Technology |
90 nm |
90 nm |
Number of Transistors |
681 Million |
681 Million |
Die Size |
Est. 484 mm2 |
Est. 484 mm2 |
Stream Processors |
128 |
96 |
ROPs |
24 ROPs |
20 ROPs |
Core Clock |
575 MHz |
500 MHz |
Shader Clock |
1350 MHz |
1200 MHz |
Fillrate |
36.8 Gigapixels / sec. |
24.0 Gigapixels / sec. |
Memory Interface |
384-bits memory interface |
320-bits memory interface |
Memory Buffer |
768 MB GDDR3 |
640 MB GDDR3 |
Memory Clock |
900 MHz |
800 MHz |
Memory Bandwidth |
86.4 GB/s |
64.0 GB/s |
Power Consumption |
145.5 W |
NA |
Graphics Bus |
PCI Express |
PCI Express |
Shader Model Version |
SM 4.0 |
SM 4.0 |
NVIDIA IntelliSample Technology |
4.0 |
4.0 |
NVIDIA SLI Technology |
Yes (Two SLI connectors) |
Yes |
NVIDIA Quad SLI Technology |
NA |
NA |
NVIDIA TurboCache Technology |
No |
No |
64-bit Texture Filtering and Blending |
Yes |
Yes |
DVI Support |
Two dual-linked DVI, both integrated into the GPU |
Two dual-linked DVI, both integrated into the GPU |
Now, let's take a look at the new unified architecture and see what's new and interesting in the two GPUs based on this architecture.