NVIDIA HAS FINALLY taped out a Fermi GF100 derivative, but it is not what you might think. Contrary to its public bluster, Nvidia isn’t going for the high end this time.
GF108, the smallest GF100 derivative, has finally taped out. In case you are not following the Nvidia derivatives closely, they usually come out with a round number, G80, G200, GF100, then follow it with four shrinks. The shrinks are x2, usually around 90 percent of the speed of the x0 part with a much smaller die, x4, half the x2, x6, one quarter the x2, and an x8.
x8 is basically the smallest die you can get away with and still have functionality, and both Nvidia and ATI do this. Redwood, the ATI equivalent to the x6, is 400 shaders while Cedar, the x8 equivalent is 80 shaders. These parts exist simply to add features to machines that lack a port or function and are built to be cheap.
GF108 is what Nvidia taped out, so its triumphant vaporware GF100 is indeed going to be followed up by a huge wave of GPUs, all of them low end. If you are planning on getting a GF100, and you have missed the first wave, it likely is already too late to order one, so then don’t count on the derivatives to save your über-box.
SemiAccurate moles are saying that plans are in the works to tape out a GF104 and a GF106 in short order as well, but they haven’t left Dear Leader’s workers’ paradise yet. Production time at TSMC is about 10 weeks, then add two more to make boards, so you are looking at July, best case for boards, and that’s only if Nvidia buys risk wafers.
A more realistic view is late July or early August before you will be able to get bottom feeder DX11 parts from Nvidia, but that will be much sooner than real GF100 quantities will arrive. At this rate, Nvidia will have a full line of parts, in quantity, but not until Q4. Q4 2011, that is.S|A
Latest posts by Charlie Demerjian (see all)
- AMD puts massive SSDs on GPUs and calls it SSG - Jul 25, 2016
- AMD revamps workstation lines and adds three cards - Jul 25, 2016
- AMD unveils Loom and ProRender software - Jul 25, 2016
- Emulex puts NVMe over FC with NVMf - Jul 20, 2016
- ARM needed Softbank, Softbank wanted ARM - Jul 19, 2016