Build a robust testbench PC

Zak Storey introduces one of the machines we’ll be using to test hardware over the coming year.

THE CONCEPT

Every issue, we dissect a plethora of hardware for your reading pleasure. It could be anything from headsets to processors, graphics cards to motherboards. And ideally, much of that hardware needs to be put through its paces on a quantifiable, standardised test bench. In short, as much as is feasable, we try to keep the systems we use for each product as close to identical as possible. This helps ensure that our results across multiple motherboards, GPUs, CPUs or memory kits aren’t affected by any spurious system irregularities or instabilities.

Getting these things set up correctly imperative for providing solid, reliable results, and something many a hardware reviewer will pride themselves upon, ourselves included.

But when it comes to building a solid test bench, there’s a whole variety of factors to consider. First of which is the choice of chassis. Generally speaking, you have two routes. You can go with a fully enclosed case or you can opt for an open-air test bench. The latter, however, tends to be less conducive to accurate cooling benchmarks (after all, it doesn’t replicate a typical, everyday fully enclosed system).

But case choice, accessible airflow and ambient temperatures vary so widely among our audience, that it’s often less about what specific temperature the coolers and fans can maintain, but more what variances in temperature and noise are noted between two different cooler designs, something we then balance with strict ambient temperature controls and thermal load testing.

On top of that, opting for an open-air case makes swapping out hardware far easier and quicker. Which means that, when it comes to supertests and error testing, it can be achieved far faster than when we’re using a closed chassis.

As you’ve probably guessed, we’re about to show you one of our main test benches for 2017 — which hardware we’ve chosen, which niggles we’ve found and how we’ve overcome them.

HARDWARE VARIATIONS

So where did we start with this new test bench? With the fixed components. We knew we wanted a new test bed, and Raijintek’s Paean Aluminium ATX Bench Table caught our eye. Squeaky clean and providing far better mounting options than our previous one, it was the obvious choice. Finding it in Australia wasn't an option, however; we had to source ours from the UK. Next up was the cooler. We stuck with our trusty NZXT Kraken X61, coupled with two Noctua NF-A14 iPPC fans. It’s not the fanciest all-in-one water cooler, but its vast surface area and fantastically easy mounting system make it a snap to set up.

We also decided to play around with our storage this year. Moving the OS from the 500GB 850 Evo to the 250GB 960 Evo M.2 and using the Crucial MX300 1TB as our games and file storage drive, as opposed to the 2TB Samsung 850 Pro, made sense. Finally, Be Quiet!’s Dark Power Pro 11 1,200W PSU would be providing the juice, as it has done for the last 18 months.

As we test a variety of products, the test bench has to rotate to compensate. A new X99 board? Then it’s four sticks of DDR4 Dominators, alongside an Intel Core i7-6950X. AM4+ and Ryzen? A new motherboard and high-end processor should do the trick. GPUs launched? We’ll stick with the Z270 setup, and swap the GTX 1080 out instead. A test bench constantly changes, so while it has to adhere to solid principles on what hardware is included, it must also be flexible, depending on the gear being tested.

1 FRONT I/O PROBLEMS

Test beds take the most abuse of any system we put together here in the APC office. They have all sorts of hardware installed and pulled out again several times in one day, often three or four times a week. And because of that, they have to be built from the right stuff. Unfortunately, just one week in, we accidentally broke the front I/O on Raijintek’s Paean, which meant we had to repair it using good old packing tape, just to keep it in one piece. It doesn’t look particularly glamorous, but most test benches we’ve used prior to this lacked a front I/O altogether, with us either booting via an onboard switch or shorting the two power pins on the motherboard itself most of the time.

2 GHETTO RADIATOR MOUNTING

Next up were cooling issues. By design, the Paean is meant to be a fully liquid-cooled test bench solution. The theory is, you place a combi res in here, with a pump to the right of the motherboard, then a triple thick radiator to the right of that. Although that’s certainly a possibility, we’re not keen on deviating from the more common AIO yet, because it’s more relevant to consumers, at least in our eyes. That said, mounting it the traditional way means our tubes aren’t long enough — so, to circumvent this, we attached the radiator sideways on the radiator panel, by carefully threading cable ties through the corners of each fan, using the radiator screw as our anchor point. This keeps it secure, and means we can still reach the processor with ease.

We try to keep the systems we use for each product as close to identical as possible. This helps ensure that our results across multiple motherboards, GPUs, CPUs or memory kits aren’t affected by any spurious system irregularities or instabilities.

3 THE NEW STANDARD

We’ve been a big advocate of M.2 PCIe drives for quite some time now. In fact, if you’re building a modern system, using any of the new chipsets from either Intel or AMD, your best bet is to go for an M.2 PCIe SSD for your OS. The sequential read and write speeds alone are mind-bending. It was a tricky decision to make, though, because they’re still not widespread just yet, and on top of that, booting from them requires Windows 10, and you also need that chipset support or an add-on card to make it a reality. However, the 960 Evo won us over, purely due to its exceptional value for money performance-wise .

4 MULTI GPUS

One of the neat features of the Paean is the PCIe bracket. Now, we know what you’re thinking — why on earth is this such a big feature? The thing is, the vast majority of test bench manufacturers seem to fail every single time they attempt this, with many using either pins or disproportionate supports for the GPUs, leading to frantic and troublesome mounting solutions. Fortunately, with its tempered glass, the Paean is just as much a showcase as a test bench, so the mounting solution here is exceptionally solid, and enables us to test up to four-way CrossFire or two-way SLI if we so desire.

5 STORAGE CHANGES

Not only did we decide to swap out the operating system to an M.2 PCIe SSD, but we also transferred all of our benchmarking files across from the Samsung 850 Pro 2TB on to this Crucial MX300. Now, let’s be honest, it’s not going to blow anyone away when it comes to performance, but as a workhorse drive for mass storage, it’s exactly what we need. The Sammy 850 was being underused acting just as our games and files storage drive, and it has been put to better purpose elsewhere because of this change. On top of that, this neat mounting mechanism means we can pull the Crucial MX300 out whenever we need it, to hook it up to any system we choose with a USB 3.0 to SATA cable with relative ease.

6 CABLE MANAGEMENT? WHAT’S THAT?

OK, so this isn’t the tidiest build we’ve ever done (actually, we’ve always been pretty lax when it comes to cable management), but this is entirely for good reason. Because we swap out hardware at an alarming rate, we need slack cables to cater for the constant shuffling around. The last thing we want is to struggle to unplug a CPU or PCI power on the fly. You’ll also notice we’re still running our trusty 1,200W Be Quiet! Dark Power Pro 11 PSU. It’s super, uh, quiet, and has enough headroom to plug in multiple GPUs without much issue, making it one sweet piece of kit. We’ve also thrown in a few CableMod sleeved cables for good measure — although, due to the fact that this $370 PSU is semi-modular, we refused to plug in the extension lead for the 24-pin ATX.