In the past we exclusively reported about the GT300 several times and inter alia revealed that the chip will be equipped with 512 shader units and a 512 Bit wide memory interface.
Now we can give first details about the complexity and size of the chip in company with Bright Side Of News.
According to our information on hand, Nvidia's GT300 will have with 2.4 billion transistors only one billion more than the GT200, despite more than twice as many shader units. Relatively this is an increment of 70 per cent. Though the performance - as much as we want to tell you - will get past this amount. So we can also determine that Nvidia worked on the transistors' efficiency.
| . | GT200 | GT300 |
|---|---|---|
| Technique | 65 nm | 40 nm |
| Transistors | 1400 Mio. | 2400 Mio. |
| Chip surface | 576 mm² | < 490 mm² |
| Shader units | 240 1D | 512 1D |
| Memory interface | 512 Bit (GDDR3) | 512 Bit (GDDR5) |
| DirectX | DirectX 10 | DirectX 11 |
| Launch | Q2/2008 | Q4/2009 |
Assuming that the transistor density will be the same as the GT200's, what is just an approximate value, and the 40 nm technique keeps its promises, Nvidia with 2.4 billions would get to about 495 square millimeters. But we have information on hand that the GT300 is even smaller than the GT200b, which is smaller than 490 square millimeters.
So it will find its position somewhere in this region and therefore be definitely bigger than the Canadian performance chips, but not take up the size of a GT200, which is the largest graphic chip by now with 576 square millimeters.