Today we’re looking at the different m.2 drive configurations you can use in a Hades Canyon NUC, and how they impact performance. But first, I have something I need to get off my chest. Here goes:

HADES CANYON IS NOT A NUC!

Phew, that feels better! Let me explain. Ever since NUCs first hit the stores, the concept was simple: how much computer can you get in a four inch by four inch space?  That concept has driven the NUC design ever since.  Yes, there was Thin Canyon, which was more or less a NUC sliced through the middle and folded open like a sandwich. But even that was still a NUC; an experiment to see how much a NUC would have to change to become fanless. Even then it retained much of the design of a NUC.

Then came Skull Canyon, in my opinion the first non-NUC NUC. It was huge (compared to a traditional NUC). Sure, the NUC team created it, and they called it a NUC, but it was more like two or three NUCs melted down and poured into a mould. Nothing about the design looked anything like a NUC.  Now we have Hades Canyon, an incredibly powerful tiny PC that bears no resemblance to a 4″ by 4″ NUC.

Don’t get me wrong, I love this thing. It is super fast, has every feature I could think of in a mini PC, and looks sexy as Hell.  I even like the cooling better; it’s quieter than the 4X4 i7 NUCs.  But it’s not a NUC. I mean, the power brick for it is larger than some NUCs.  The unit itself is the size of four typical NUCs. Calling Hades Canyon a NUC is akin to racing a Lamborghini against a Smart car. Yes, they’re both cars, but they’re built for different uses, and the smart car just doesn’t stand a chance.

What Hades Canyon IS is the fastest computer to come out of the NUC team, ever. It blows the doors off of every NUC that has come before it, including Skull Canyon.  By the way, if the naming trend continues, I fully expect to see Styx, Charon, and Cerberus Canyon.

Anyway, back on topic.

One of the great features in Hades Canyon is the dual m.2 slots. This opens up the option of running two storage drives in RAID 0 or 1.  So you now have the option of better securing your data from loss via RAID 1, or blowing the doors off, performance-wise, with RAID 0.

I wanted to see just what could be done with those m.2 slots, so I decided to test different configurations of drives. I’d start with a single SATA m.2 drive, then double it and try RAID 0. I’d then try a single NVMe drive, then do the same thing to see how two of them performed.  The idea was to see just how much of a performance improvement comes from each configuration.

Drives Needed!

First, I needed two sets of drives. For the NVMe tests, I decided to stick with my ADATA XPG SX6000. It was insanely fast (to my standards) and still remarkably cheap. So I just ordered another one to get myself a matched set:

For the SATA drives, I didn’t have any new drives to use. Enter Micron. Once again, the good folks at Micron were willing to send me drives in the name of science. They sent me two Crucial MX500 m.2 drives:

The folks at Micron have been super helpful in sending me memory for my tests, and this time they came through again with a nice pair of drives. Their memory has been rock-solid, and so far these drives have been too!

The Setup

To do the testing, I used the NUC8i7HVK, the most powerful of all Hades Canyons. It rocks not only an i7-8809G Processor, but Radeon RX Vega M GH graphics. It’s not just fast, it’s sooper-dooper fast. I loaded it with 32GB of Crucial Ballistix RAM.

For testing software I decided to stick with what has been working for me and use both PCMark10 by Futuremark and Performance Test by Passmark.

I only tested in RAID 0. RAID 1 is cool for protecting your data from failures, but the purpose of this test was to look at performance.

Since I would need to install Windows 10, drivers, and the testing software at each step of the process, I downloaded the latest driver pack for Hades Canyon from downloadcenter.intel.com, as well as the install packages for both benchmark tools, and put them on a flash drive. I already had a Windows installation flash drive ready.

My plan was to install Windows, the drivers, and the benchmark tools in the same way, in the same order. That would make the results at least mostly fair.  There were some drivers I skipped, like IR, which wouldn’t affect the test.

I tried to be as consistent as possible in the testing, repeating the order of everything and not installing anything unique. The only exception was the RAID driver, which as required for the NVMe drives in RAID (see below).

RAID Configuration

When using a single m.2 drive, there isn’t much to the BIOS configuration, as Hades Canyon comes configured for AHCI.  However, when using two drives, you need to configure RAID.  Here’s a quick video of the process to configure the RAID settings and create a RAID volume:

Once the RAID volume is configured, you can install your operating system of choice.

Windows RAID driver

Depending on your drives (i.e. NVMe or SATA) you will find that Windows may not see the RAID volume as a drive. If this is the case, you’ll need to load the Intel Rapid Storage RAID driver:

Once you’ve got the driver loaded and the RAID volume is listed, you can proceed with the installation as usual.

If you’re not a “visual learner” and would rather have written instructions, the NUC team has posted everything you need here.

Results

So let’s take a look at the results, to compare the different configurations.  Here’s a side-by-side comparison of the Performance Test overall results:

The results here are a bit confusing. Yes, we see a significant increase in overall performance when we move from one SATA drive to two. And we see that a single NVMe drive gives nearly the same results as two SATA drives in RAID. What’s strange, though, is that going from a single NVMe drive to two in RAID 0 yields only a marginal improvement in overall system performance. Strange.

Now, let’s look at the results from PCMark10. This should give us an idea of how real world performance would be impacted:

Notice anything weird?  The scores look almost identical. And in the Productivity test, the single Crucial drive actually won.  In the Essentials test, the NVMe drives in RAID won out, but only by a small margin.  This is confusing.

Let’s go back to Performance Test and look at JUST drive performance tests:

Alright, now we’re getting somewhere. These numbers paint a very different picture than the overall system tests.  The difference in disk performance between a single drive and a pair in RAID 0 is huge.  Interestingly, just looking at the RAID scores, the Crucial SATA drives did pretty darned well. Yes, the NVMe drives in RAID blew everything away with an overall disk score of 9318, but the Crucial drives came in at about 93% of that with a score of 8705. And if you look at random R/W, they’re neck-and-neck with the ADATA drives.

Rerunning the tests showed mostly consistent results.  There was some variation between runs in each configuration, but not enough to matter.

Conclusions

So what did we (or maybe just I) learn from these tests.  We learned that while your everyday use may not change all that much, when you’re copying large files, rendering video, sharing out data, or doing anything else that hits the disk hard, you should notice an improvement. One use that comes to mind is virtualization. When running multiple virtual machines, disk performance would be huge, and what we’re seeing here is that NVMe in RAID 0 is the clear winner.

And we learned that a high-end SATA drive can hold its own against a cheap NVMe drive in RAID 0.  Bang for your buck-wise, this is important. One of the Crucial MX500 SATA m.2 drives will set you back about $100 for 500GB. Compare that to the ADATA NVMe drive, which you can get on sale for about $40 for 128GB.  Yes, the NVMe drive is slightly faster, but for a little more than twice the cost, you can quadruple your capacity.

I realize this isn’t entirely fair, as I’m comparing entry-level NVMe to performance-level SATA. Still, when cost is a factor, you want to be sure you’re getting the best value. In this configuration (two drives in RAID 0), I would rather have 500GB of capacity than 128GB when I’m only losing a small amount of performance.