I wonder what 4 X SSD in RAID0 could do against that thing. I'm already doing 950MB/s sequential read with my 2 old Crucial M4 128GB...
This should slay the 4x SSD array pretty handily, especially at low QD and in steady state. The problem with RAID is that it is constrained by the PCH, which communicates via DMI interface. There is only a PCIe Gen2 x4 connection (ouch) back to the proc, which limits speeds to 1.8 GBps maximum. Keep in mind, this is the entire bandwidth available for everything on the DMI, so other devices will also feed from this same pool. Question for Intel: Why are we using PCIe 2.0 x4 DMI in 2015?
Also, is there a way to RAID0 2 of those SSD PCIe cards? The speed would be crazy, unless it would choke the PCIe bus at some point, let's say with 2 X GTX 980 in SLI on top of everything else?
You can raid them into a dynamic volume, but the system will not be able to boot from the volume. I would *theorize* there will not be an appreciable performance loss when used in SLI, there are plenty of lanes (40) on high-end chipsets.
Hopefully, we'll see this as an M.2 card with IMFT 3D-NAND soon.
This is inevitable, of course, but not at this speed...at least not based upon this particular architecture. In the briefing with Intel they indicated that attempting to roll a new m.2 would result in unacceptable tradeoffs, there are incredibly tight tolerances for power/heat with m.2. Even SATA based devices are running into heat issues with m.2 designs, thermal throttling, etc.
Of course, just the size of the 18-channel chip precludes its use in an m.2....