There seems to be a pre-COMPUTEX lull in terms of hardware announcements, but there's nothing like a bit of high-end GPU conjecture to kick start a dull and overcast British morning.
If you believe everything you read on the web, you may recently have heard that NVIDIA's G300 architecture - successor to the current G200 - has been making steady progress. According to various reports, the forthcoming 40nm GPU has already been taped out and engineering samples should be just around the corner.
Today, the rumours are cranking up a notch as German site hardware-infos.com claims to have the details of the G300 GPU's core, shader and memory frequencies. NVIDIA's keeping quiet, of course, and we can't comment on the authenticity of these figures, but should the numbers be accurate, here's what we're looking at:
| Graphics cards | NVIDIA GeForce GTX 3** (G300) | NVIDIA GeForce GTX 295 | NVIDIA
GeForce GTX 285 1,024MB |
NVIDIA GeForce GTX 280 1,024MB | NVIDIA GeForce GTX 275 896MB | NVIDIA GeForce GTX 260 896MB |
|---|---|---|---|---|---|---|
| PCIe | PCIe 2.0 | |||||
| GPU(s) clock | 700MHz | 576MHz | 648MHz | 602MHz | 633MHz | 576MHz |
| Shader clock | 1,600MHz | 1,242MHz | 1,476MHz | 1,296MHz | 1,404MHz | 1,242MHz |
| Memory clock (effective) | 2,200MHz | 1,998MHz | 2,484MHz | 2,214MHz | 2,268MHz | 1,998MHz |
| Memory interface and size | 512-bit, 1,024MB, GDDR5 | 448-bit (per GPU), 1,792MB, GDDR3 | 512-bit, 1,024MB, GDDR3 | 512-bit, 1,024MB, GDDR3 | 448-bit, 896MB, GDDR3 | 448-bit, 896MB, GDDR3 |
| Manufacturing process | TSMC, 40nm | TSMC, 55nm | TSMC, 55nm | TSMC, 65nm | TSMC, 55nm | TSMC, 65nm |
| DirectX/ Shader Model | DX11, 5.0 | DX10, 4.0 | DX10, 4.0 | DX10, 4.0 | DX10, 4.0 | DX10, 4.0 |
| Vertex, fragment, geometry shading (shared) | 512 FP32 scalar ALUs, MADD dual-issue + MUL (unified) | 480 FP32 scalar ALUs, MADD dual-issue + MUL (unified) | 240 FP32 scalar ALUs, MADD dual-issue + MUL (unified) | 240 FP32 scalar ALUs, MADD dual-issue + MUL (unified) | 240 FP32 scalar ALUs, MADD dual-issue + MUL (unified) | 216 FP32 scalar ALUs, MADD dual-issue + MUL (unified) |
| Peak GFLOPS | 2,457 | 1,788 | 1,063 | 933 | 1,011 | 805 |
*unannounced GPU, rumoured specification |
||||||
Sounds bigger and badder - as always - but we must reiterate that the numbers are anything but confirmed. Interesting nonetheless, and more so following last month's rumoured-reveal of ATI's Radeon HD 5870.
Will we be seeing samples of either GPU at next month's COMPUTEX? Seems a little too soon, we reckon, but the Taiwanese show is as good a place as any for the green or read teams to pull out a surprise.
