2008/1/24, Paul Brook <[EMAIL PROTECTED]>: > > Is a ~10 GB SSD on PCIe costly ? It could be perfect for a system > > disk, the 3Gb/s bandwidth of the PCIe should do the difference with > > what ever sata disk. You only need to write a Linux bloc drivers (and > > the mean to boot !). > > IMHO this is a bad idea for several reasons: > > - The limiting factor for both price and performance of SSD tends to be the > flash memory, not the interface (unless you're using something really slow > like USB1).
Sur but it's far more easy to use chips in parrallel than a disk. > - You need custom drivers. This makes installation a real PITA. It also means > yuo have to write and maintain drivers for every OS you support, which is a > fair chunk of work. Sur. But for somebody in the Linux kernel dev, it did not look like a hard point. Then you could use the code for all FOSS code. > - SATA bandwidth is comparable to a single PCIe lane. If you need more > bandwidth then it's dead easy to do RAID on the host. > - Most machines have more free SATA ports than PCIe slots. > - PCIe SATA cards are cheap. That's a points. But you avoid a very complexe interface (Pata), for accessing a quite simple device, a memory chip. I think that you could also win even more latency. And cheap PCIe SATA card are rapidely saturated even by HD. > - Much of the real world speedup from SSD is due to the extremely low latency > random access (compared to disks), not increased peak bandwidth. I know. So an increased bandwith will give even more smoothness on the desktop. I don't understand why there is no chipset to connect few GB on them. > Paul > _______________________________________________ Open-graphics mailing list [email protected] http://lists.duskglow.com/mailman/listinfo/open-graphics List service provided by Duskglow Consulting, LLC (www.duskglow.com)
