Week in Tech: SATA Express At Last, Not Google Glass

And lo, on the first day of the fifth month during the year of Our Lord numbering two thousand and fourteen, verily did the first motherboard with SATA Express arrive. Well, it's the first I've seen sitting in front of me outside a show floor or PR event. The board in question is a new Asus Z97 beastie. Now, by some metrics, plain old SATA has been a speed bottleneck for SSDs and in turn PCs for a while. And these new interfaces will definitely release the solid-state hounds in terms of raw data throughput. But will that actually make your PC feel faster or make any difference for games? Meanwhile, I've decided I'm definitely going to buy an Oculus Rift DK2 and oddly it's Google's Glass that's convinced me to pull the trigger.

First, some basics. We've touched on this before so I won't get too granular, but SATA Express kinda combines the best bits of SATA and PCI Express to deliver faster storage performance along with backwards compatibility with existing SATA drives.

Speeds will develop over time. The complicated bit is that SATA Express uses your PC's PCI Express lanes. And these days, they're located on the CPU itself.

Now, any lanes you use for your storage obviously can't also be available for other devices. For most people, most of the time, this won't matter. But for anyone running multi-GPU systems, this could be an issue.

Anyway, SATA Express performance will be a factor of the number and type (PCIe 2 or 3, in other words) of lanes being used. Asus's first SATA Express boards will apparently support sequential read and write speeds of 745MB/s and 809MB/s which translates into single-lane configuration.

Give it up for the first SATA Express board from Asus.

According to Asus's marketing spiel, that's fast enough to copy a 10GB HD movie in 'around 10 seconds'. The maths obviously suggest something closer to 7.5GB in 10 seconds. And even that is likely theoretical. I've never seen higher than around 350MB/s for actual drive-to-drive transfers on existing SATA drives which deliver 550MB/s of benchmark throughput.

In other words, this first iteration of SATA Express a la Asus is a healthy but not earth-shattering improvement. What's more, it's arguably random access or dealing with itty-bitty slices of data rather than shunting huge lumps of sequential data that makes a PC 'feel' fast.

Now, I'm not 100% clear on this next bit. But I don't think SATA Express has any immediate implications for improving random access performance in a very strict sense. Everything about my understanding of how solid state drives operate – for whatever that is worth – tells me that random access performance is limited by the SSD controller chipset and how many I/O operations it can crunch per second.

However, with SATA Express comes a new control protocol to replace AHCI, known as NVMe. This is pretty desiccated stuff as light reading goes, but the thing to grasp is that AHCI was really designed for old-school magnetic disks. NVMe is designed for SSDs and is claimed to slash latencies by 50 per cent and generally ditch all that magnetic drive baggage. What will this mean in practice? I have no idea, but there's got to be a chance it will deliver that thing that matters most - tangibly more responsive PCs.

It's also true that sequentials do matter moderately for game level loads, so more basic bandwidth is likely to be a bit of a boon if not a gaming life-altering experience.

M.2 what? OK, this is getting confusing.

Two final points to finish this off are drive availability and the M.2 thing. For the first, I haven't seen too much noise about SATA Express drive availability, but I assume things will pick up now that boards are becoming available.

As for M.2, it's a bit confusing. It's another new drive interface that involves small plug in boards with exposed components rather than enclosed drives like SATA SSDs. In simple terms it's sort of a purer PCI Express based alternative that doesn't bother with SATA Express's backwards compatibility with plain old SATA. For me, it's more of a mobile tech perhaps than a desktop tech as there are practical limitation though boards with both standards will become fairly common. Perhaps you'll have your main super-fast boot drive as M.2 and then plug in secondary SSDs over SATA Express.

For now, the TL;DR summary goes like this. SATA Express has been a long time coming and it may not be be earth-shattering in the first iteration. But it does bring more than just more dumb bandwidth and it's essential for enabling much faster drives in the future.

Irrelevant comparison? Perhaps, but Oculus Rift DK2 for $350 looks very cheap next to $1,500 Google Glass.

As for the Oculus Rift thing, allow me a little leeway here. I can't help but notice the value delta between the upcoming DK2 kit and Google's Glass. I happened to spend a morning with Google recently at the swanky Manhattan facility where the chosen few are invited to collect their Google Glass units and are schooled in the finer points of usage and operation (this mainly involves squawking 'OK, Glass' repeatedly as the futility gradually washes over you).

To cut a long story short, I thought Glass was fundamentally crap. For sure, it's a tantalising glimpse of what will one day be possible. But in its current state, I personally decree it painful to use and pretty much pointless.

And yet somehow, Google is charging $1,500 for the privilege of buying what feels like a bit of kit in early beta at best and in many regards more like alpha. Anyway, that makes the Rift DK2 for $350 look like utterly spectacular value. For my money – and fully in the knowledge that the comparison is largely irrelevant - DK2 is a bazillion times more hardware than Google Glass and while it (DK2) is obviously not going to be the final retail experience, I have a hunch it will be much more worthwhile than Glass.

Glass is fun for about five minutes, after which you've seen enough and more importantly had enough. DK2 doubtless will have some rough edges, but I fancy it will be worth putting up with them. Here's hoping.