About one week ago Fudzilla reported, that the first Nvidia 40 nm chips will belong to the entry- and mainstream segment and launch at the end of the second quarter, what delimits the timespan March to April.
Meanwhile, in another report, the codename of the entry chip is mentioned: GT218.
According to the information on hand that the GT21x series will have 32 stream processors per cluster and 8 TMUs, you do not have to be a great prophet to forecast that the GT218 will have these 32 stream processors and 8 TMUs, too.
Everything else would be heavily oversized now for the smallest chip of one series. Already 32 shader units result in four times of performance than with the 8400 GS, 9300 GE and 9300 GS and after all twice as much as with the 8500 GT and 9400 GT – noting that the ALU domain clock frequency is theoretically equal all the time.
| . | G98 | GT218 |
|---|---|---|
| Technique | 65 nm | 40 nm |
| Shader units | 8 | 32 |
| TMUs | 8 | 8 |
| ROPs | 4 | 4 |
| Memory interface | 64 Bit | 64 Bit |
| TDP | < 30 W | 22 W |
| Launch | Dec. 2007 | Apr. 2009 |
The further equipment of the 40 nm chip GT218 could be a memory interface of 64 bit, that should preferably run with DDR2 memory. This would result in four ROPs, as these are in relation to the width of the memory interface.
According to Fudzilla the TDP is determined at 22 watt, while the G98 features a TDP of 25 to 30 watt.
If the development is finished correctly - a first working GT218 chip is already there (tape-out) -,
the GT218 is ought to be coming to market respectively to the OEMs in April, though a launch in March is still possible.
According to the last information there will be also launching a first Nvidia mainstream chip in 40 nm technique at the same time. The codenames here are spanned from GT216 to GT215.