Back when Intel launched their SSD 750 Series product line, their hope was that the motherboard industry would adopt the U.2 PCIe connector and add those ports to all motherboards. Well, it's over a year later, and we've only seen U.2 appear on a very small fraction of currently shipping motherboards. It's far more likely to see motherboard manufacturers simply tossing in a U.2 to M.2 adapter than to incorporate both onto the board itself. Since things have panned out the way that they have, Intel has recently let us know they will be introducing a new SKU of the SSD 750 products:
Instead of the bundled U.2 cable (seen above next to an ASUS Hyper Kit), the alternate will include a U.2 to M.2 cable, eliminating the need for an adapter for consumers who have no native U.2 port on their systems.
For now, the only available products on the market have the U.2 cable, so don't worry about getting the wrong one. Once the new SKU hits the to market, we should see product descriptions indicating which cable is included. Those purchasing starting this summer should be aware that in the future there will be an additional product with the alternate cable, and be careful to purchase the desired product/cable.
Ugh, U.2 is weird. It
Ugh, U.2 is weird. It specifies the
I don’t like the push to U.2
I don’t like the push to U.2 in favor over M.2 on the motherboard. A 2242 M.2 to U.2 adapter will allow for the greatest flexibility.
I like M.2 drives. No cables make for a very clean build, or at least a much cleaner build if you are crap at cable management.
Any word on if these adapters
Any word on if these adapters will be available separately?
Allyn,
Can you explain why
Allyn,
Can you explain why there are two connectors
on the new Intel SKU?
Thanks!
MRFS
Answered
Answered here:
http://techreport.com/news/30251/intel-bridges-the-u-2-gap-with-an-m-2-cable-for-its-750-series-ssd
Builders will still need to hook up the SATA power connector toward the business end of the drive to get it fully up and running.
p.s.
“… their hope was
p.s.
“… their hope was that the motherboard industry would adopt the U.2 PCIe connector and add those ports to all motherboards.”
I may be in a small minority, but
here are my 2 cents on Allyn’s sage observation:
The upstream bandwidth of Intel’s DMI 3.0 link
is EXACTLY the same as a single NVMe M.2 SSD:
x4 PCIe 3.0 lanes @ 8G per lane, 128b/130b jumbo frame
(130 bits / 16 bytes = 8.125 bits per byte)
And, if one wishes to build a bootable RAID array
with multiple NVMe SSDs, one must connect to
ports that are controlled by RST — which are
downstream of that DMI 3.0 link.
I’ve been advocating 4 x U.2 ports for a long time,
but of course there won’t be any performance gains
until the bandwidth of the DMI link is increased.
That should happen at PCIe 4.0, which uses a 16G clock.
(Software RAIDs don’t count here, because they are
presently NOT bootable, and we want an OS to run
on a fast solid-state partition e.g. we host our
workstation OS on a RAID-0 w/ 4 x SanDisk Extreme Pro SSDs
and a Highpoint RocketRAID 2720SGL.)
Alternatively, a bootable NVMe RAID controller with x16 edge
connector would be another viable solution e.g.:
http://supremelaw.org/systems/nvme/want.ad.htm
And, a parallel approach would expand the
number of lanes in the DMI link to at least x16.
In that way, there is an elegant symmetry when
four x4 NVMe SSDs are using an x16 DMI link:
4 NVMe SSDs @ x4 = x16
Thus, upon arrival of PCIe 4.0 chipsets,
upstream bandwidth of a single x16 expansion slot is:
x16 lanes @ 16GHz / 8.125 bits per byte = 31.50 GB/sec
(or, roughly speaking, 2GB per lane x 16 lanes)
31.5 GB/sec upstream bandwidth should suffice for a while,
particularly if Intel’s Optane claims turn out to be true.
Final dream here: over-clock-able storage,
just like overclocked CPUs and DRAM became standard fare —
possibly with pre-sets like 6G, 8G, 12G and 16G.
RETURN OF THE JUMPERS! Yea Team 🙂