The first information on the next generation graphics circuits, NVIDIA GT300 and AMD RV870, implied that AMD would be the first to announce its 40nm RV870 chip and actually get a big head start on NVIDIA’s GT300 circuit. Over the last few weeks the roles seems to have been reversed and we decided to wait a bit before posting anything about the ongoing rumors, especially those surrounding NVIDIA’s monster GT300. More and more stories are saying that NVIDIA will be first though.
NVIDIA has evidently cleared the first stage with a tape out of GT300. This means that it has successfully manufactured the first engineering samples of the graphics processor. NVIDIA might even have test versions ready for Computex, with DirectX 11 support and perhaps named GeForce GTX380.
Detailed information on the graphics circuit has started to leak and the 512 shader processors looks to make a stable foundation for NVIDIA.
According to German hardware site Hardware-infos the first versions of the card will operate at respectable 700MHz, while the shader processors run at 1600MHz. With seemingly exaggerated 512-bit memory bus and fast GDDR5 memory the memory bandwidth will be something really extreme. From what it seems GT300, or G300 that is the desktop model, will sport either 1GB or 2GB memory buffer.
| GTX285 | GTX385 | |
|---|---|---|
| GPU clock | 648MHz | 700MHz |
| Memory clock | 1242MHz (2484MHz GDDR3) | 1100MHz (4400MHz GDDR5) |
| Memory bus | 512-bit | 512-bit |
| Shader clock | 1476MHz | 1600 MHz |
| Theoretic perf. | 933 Gflops | 2457 Gflops |
| Memory bandwidth | 159.0 GB/s | 281.6 GB/s |
Compared to NVIDIA’s current single-GPU flagship GeForce GTX285 this is a total humiliation and the big question is when NVIDIA can get the card out in the open, and whether hardware-infos has a reliable source for all this.
XFX GeForce GTX 285
A lot depends on TSMC’s manufacturing capacity and the semiconductor manufacturer is currently having problems with its 40nm technology, which is harming both AMD and NVIDIA.
