ASRock submitted two models; the first on our bench is the B85M-DGS.
While listed as microATX, this is actually a FlexATX board (which admittedly is an addendum to the microATX specification). It's no more difficult to build with a Flex board than a micro board, but it is smaller, meaning less space for extras. However, this board doesn't have near the number of features of its X99 and Z97 big brothers, so real estate isn't as much of a concern.
Starting up top, we see the VRM circuitry, a four-pin EPS plugand a three-pin fan header. Conveniently to the left of the LGA slot is a four-pin fan header for the CPU cooler. Not as conveniently located is the CLR_CMOS jumper, found between the battery and CPU interface. It's very difficult to reach once your CPU cooler and GPU are installed. Lesson learned: when experimenting with overclocks on this board, remove your graphics card first if you can.
Along the front edge you'll find the 24-pin ATX power plug and four SATA ports. Those ports are mirrored, which keeps all cable latches on the outside, readily accessible to be detached—at least in theory. The two DIMM slots butt right up against the second and fourth SATA ports. If you use latched cables, they're also difficult to detach when memory is installed. At least this is a B85-based board, so all four SATA ports are 6Gb/s-capable. Your modern storage will perform identically attached to any of the connectors.
Right beneath the SATA ports is a USB 3.0 header, which is positioned just above the PCIe slot so as to not interfere with an installed graphics card. The lone 16-lane slot is PCIe 3.0-capable for maximum bandwidth. Spaced two slots down is a PCIe 2.0 x1 interface for any other add-in cards you may want. Right above that is a four-pin fan header, which leaves me scratching my head.
At first glance, it looks just far enough down that it shouldn't interfere with a dual-slot graphics card. I pulled out an old Radeon HD 6870 just to make sure, though. What I first thought was just enough clearance was actually a touch of overlap. The edge of the cooler shroud was right at the edge of the pins. I was still able to get the fan and card plugged in, but largely because this card has a small notch in the shroud at just the right place. Could I still plug in the fan without that notch? Yes. But why is the fan header located here at all? Even if it were 1/8-inch lower, I still wouldn't like my fan cables that close to an intake fan. It would be better on the front or bottom edge, or even above the PCIe slots by the CPU fan header.
The HD Audio header is in the lower-back corner. This usually irks Thomas, but because this isn't a full-size ATX board, you shouldn't have any problem with cable length in any case. Next, from left to right, you have a chassis intrusion header, serial and parallel port headers, two USB 2.0 headers, a TPM header and the front-panel pins in the bottom-front corner. A socketed BIOS chip is just above the parallel port, in case you take tweaking too far.
The I/O panel is sparse, but perfectly functional. The -DGS has a single PS/2 port, four USB 2.0 ports and two USB 3.0 ports for peripherals. A VGA and DVI-D connector are provided for anyone using integrated graphics. Realtek provides audio and networking functionality, including its ALC662 codec and RTL8111GR GbE controller.
Inside the box, you'll find an instruction manual, an installation CD, an I/O backplate shield and two SATA cables, one with an angled connector. Four SATA cables would've been nice, but given this product's price range, we can't complain.
Overall, this competent board's only notable issue is the chassis fan header. How much that bothers you will depend on how large of a graphics card you plan to use. An HDMI or DisplayPort connection on the back panel would be welcome, but not necessary. DVI covers basic home and office users interested in HD Graphics, and any power user will almost assuredly use a discrete GPU. Some test bench–friendly buttons and voltage detection points would be nice as well, but we can't expect them in this price range.
I wasn't certain what to expect from B85-based overclocking, particularly with such limited power regulation circuitry. I didn't want to go easy on the board, but of course I didn't want to blow it up either. So, I imposed some limits: 1.2V and 75 degrees Celsius on the CPU were my goals, with 1.25V and 80 degrees as absolute ceilings. The VRMs don't have dedicated heat sinks, so I wanted to keep them under 65 degrees C if at all possible. Within those bounds, I set off to find the highest stable clock rate. Paul saw 4GHz and 4.2GHz on two of his SBM machines, so I felt confident I could do the same. Igor reached 4.4GHz, but of course that was on a Z-series board with liquid cooling, so that wasn't as likely.
I started off easy, manually setting the multiplier to 40 and leaving voltage on Auto. The board set a voltage of 1.201, and a few hours of Prime95 showed no problems. Temperatures barely cracked 70 degrees C, and the VRMs hovered in the mid-40s. Auto settings tend to be generous with voltage, so I figured I could get more performance. I incremented the multiplier to 41 and left the voltage on Auto; it stayed at 1.201V as the temp climbed a little (still well within my limits). I stepped the multiplier up to 42 and the voltage didn't budge. Apparently, the -DGS has some kind of voltage limit when Auto is chosen, which honestly can be a good thing. Stability became an issue at that point, so I knew it was time to start going manual and dialing in the voltage.
With Vcore set to 1.225V, load testing proved stable with a reading of 1.226V, and temperatures were relatively mild at 75 degrees C on the CPU and 52 degrees C on the VRMs. Could 4.3GHz be a realistic setting? Selecting a 43x multiplier resulted in a quick crash under load. Upping the voltage to 1.23V and then 1.245V didn't help. Saddened, I sought to minimize voltage for the best possible temperatures. After a bit of trial and error, a 1.215V setting (1.216V actual) yielded 74 degrees C peak CPU and 50.4 degrees C on the VRM.
Perhaps I could have coaxed a little more from the board. But if you can afford to blow a $60 motherboard for the sake of experimentation, you're better off buying a Z-series board from the get-go. As it was, I got better than a 30-percent overclock using a stock cooler and a three-phase VRM.
The RAM wasn't nearly as time consuming as it was confusing. My Mushkin XMP is good for 1600 MT/s at 9-9-9-24 timings. But again, the Pentium G3258 can't support anything faster than 1400 MT/s. The default voltage of 1.5V actually measured 1.54V, which isn't too surprising. Setting 7-7-7-21 was stable at the stock voltage settings, as ASRock's cheat proved enough to maintain stability. However, switching from 1333 MT/s to 1400 required switching over to the XMP profile, which reverted back to CAS 9. Picking the XMP profile and then manually overriding the timings yielded my desired results, all while running at the "stock" voltage.