Posted by Richard Solomon on December 24, 2014
When all through the house, not a creature was stirring, not even a…
Ok, so that’s probably NOT the kind of mouse Mr. Moore intended (well, sure, in 1822 it couldn’t have been – unless you buy into Erich von Däniken’s theories) but it seemed appropriate both for the date and for my finally getting around to addressing the cabling comments posted by user Porthem. S/he posted quite a lot, so I’ll exercise a little editorial license and copy in just pieces of the key questions.
“Will USB-C be viable to become the standard INTERNAL storage connector?
Currently, a big special SATA Express connector is used to connect internal disks to motherboards, and USB is used to connect consumer-level external disks to systems.
The new USB-C Alt Mode feature is going to enable using PCIe directly (via simple multiplexer, not via protocol encapsulation as in Thunderbolt) through the USB-C port as an alternative to using the USB protocol, via 4 signaling pairs each capable of 10Gbps. In other words, it will be like OCuLink, but unlike the latter, USB-C will soon be ubiquitous.“
Let me address those in pretty much reverse order…
Ouch, I can’t completely argue with the OCuLink statement as it rarely pays to bet against USB ubiquity, but I think OCuLink still offers some advantages (which I’ll try and bring out as we go).
Yes, the USB Type-C spec does offer “Alternate Modes” but with a few caveats. First, there are indeed 4 high-speed signaling pairs, but they are currently defined as two TX pairs and two RX pairs – so this is the functional equivalent of a PCIe x2 link. Even if one hypothesized an asymmetric setup, it would be at most 3 TX + 1 RX or vice-versa, which is still shy of the PCIe x4 link (4 TX + 4 RX) offered in the M.2 form-factor. Today’s NAND-flash SSDs can easily exceed the 2GB/s/direction that a PCIe 3.0 8GT/s x2 link provides, so most of them are designing their controllers to support x4 links. OCuLink as currently specified (still in draft form admittedly) supports a x4 link with some possibility of extension to x8 by the time the spec is final. OCuLink is also forward-looking to the 16GT/s “Gen4” PCIe spec, and I fear USB cables are primarily selected for their cost, so I don’t expect a lot of headroom atop that 10Gb support!
The USB Type-C spec also requires implementation of USB (at least 2.0) in anything with a Type-C connection. Entering Alternate Mode requires exchanging USB power delivery messages, so one needs at least some chunk of a USB controller to even get the Type-C into an alternate mode and really a full one to comply with the USB requirement. Porthem also said:
“Since even consumer-level external disks will soon use USB-C (with the USB protocol), it seems feasible that for a marginal increase in cost, they could add support for PCIe, since no change of connector or cabling will be required.“
To Porthem‘s points, yes, today you get the SATA or SATA Express connectors which certainly are larger and bulkier due to their legacy compatibility. The thing is though that most consumer-level disks are still round – by which I mean they’re rotating magnetic media. The magnetic disk manufacturers Scott & I talk to are pretty happy with 6G SATA today and not looking to change to PCIe (or anything else) any time soon. Furthermore, USB-SATA bridges are so cheap these days that every USB external drive I’ve ever cracked open had one connected right up to a SATA drive – even when the end-user price of the external drive was less than the internal one! I don’t think the Type-C overhead will add much cost beyond the connector (which would of course be the same adder to a native solution) so I don’t see these bridges becoming a stumbling block any more than they were in the USB 2.0 to 3.0 transition. In fact, they serve to isolate the drive manufacturer from the 3.0 to 3.0/Type-C transition.
The consumer (magnetic) disk drive market is so price-sensitive that changing connectors and adding the extra logic for USB may not make sense. Perhaps it would if the manufacturer would actually implement both their USB and SATA controllers in the same silicon and build a single PCB with both legacy and Type-C connectors on it. Anything short of that though seems like extra cost that market just doesn’t want. If the volumes of external drives don’t justify “dedicated” USB versions, it’s hard to see how they would justify these “unified” USB/SATA/PCIe versions – when change itself costs money.
I don’t claim to have the world’s best crystal ball, but in summary, to me it seems that USB Type-C doesn’t offer enough bandwidth to support “square” (solid-state) drives and adds cost for little concrete benefit to “round” (rotating magnetic) drives, therefore…
What’s that you say?
“Richard, your crystal ball is really a Magic 8 Ball!”
Yes, but I’d argue it beats Wall Street on most days 😀
In fairness, USB isn’t just for mice, so I should point you over to my colleague Eric Huang’s blog where you might find a slightly different opinion on USB Type-C… though I don’t think *HE* has a Magic 8 Ball in his entries!
As I mentioned last time, I believe Synopsys IT now has commenting fixed, so you too have nothing to dread as you read this blog post all snug in your bed with visions of sugar-PCIe dancing in your head… Ok, Ok, I’ll stop. Please do chime in with your thoughts and/or future column topic requests and make sure to click here to subscribe to ExpressYourself.
Thanks and Happy Holidays!
I’ve been involved in the development of PCI chips dating back to the NCR 53C810 and pre-1.0 versions of the PCI spec so have definitely lived the evolution of PCI Express and PCI since the very beginning! Over the years I have worked on variations of PCI, eventually moving on to architecting and leading the development of the PCI Express and PCI-X interface cores used in LSI’s line of storage RAID controller chips. For the last ten plus years I've also had the honor of serving on the PCI-SIG Board of Directors and holding a variety of officer positions. I am still involved in PCI-SIG workgroups and I present at many PCI-SIG events around the world. Back before the dawn of PCI, I received my B.S.E.E. from Rice University, and hold over two dozen U.S. Patents, many of which relate to PCI technology.