Next up is ASRock's H81M-HDS.
Like the B85M-DGS, this is also a FlexATX board. The -HDS is simply a little wider and longer. This puts the ATX power plug slightly outside the mounting screws. With no direct support under the ATX plug, the board flexes if you're not careful when connecting the PSU.
The big story here appears right at the top, next to the four-pin EPS jack. While the other three contestants in today's story only have three-phase power designs, the -HDS benefits from a fourth phase. Four phases may seem paltry compared to premium motherboards, but they may confer a notable advantage in the limited power envelope of this budget platform.
The -HDS shares a similar layout to the B85M-DGS. The three-pin fan header next to the two DIMM slots and the four-pin header for your CPU cooler are in similar locations. Unfortunately, so is the awkward CLR_CMOS jumper. Also odd is the HD Audio header's location in the corner, between the I/O panel and PCIe slot. You either run a cable down the back of the board or across the back of the graphics card. Enthusiasts obsessed with cable management won't be happy either way.
As with the -DGS, this board's ATX plug is located just above four SATA ports. Unlike the -DGS, the H81 PCH only supports two 6Gb/s links. Those faster ports are colored grey for easy distinction. Also different than the -DGS, all four SATA ports have their latches facing forward. To disconnect the rear SATA cable, you have to detach the front one. Still, this is easier than popping a DIMM from its slot. The USB 3.0 header is also just above the PCIe slot, just like the -DGS.
Continuing down, we find a socketed BIOS ROM chip, affording easy repair to any overzealous overclocker. Next to that is the system panel header. As Paul recently found out in his SBM story, it's too high to work with long, double-slot graphics cards. Sadly, Paul ordered his SBM parts the same time I received this board for review, so I wasn't able to warn him. If your graphics card measures seven inches or shorter, you won't have an issue. Perhaps ASRock's engineers assumed only entry-level cards would be used in such an affordable motherboard. But the four-phase VRM is a major marketing feature, and surely CPU tweakers also want larger, more powerful GPUs.
Due to the H81's limitations, the aforementioned 16-lane slot is only PCIe 2.0-capable. Even if your CPU supports PCIe 3.0, it won't run at those transfer rates attached to this chipset. You're still getting the equivalent of an eight-lane PCIe 3.0 x8 slot though, and that's plenty of bandwidth for any single GPU. Just to the right is an ASMedia ASM1042 chip that powers the nearby USB 3.0 header. A single-lane PCIe 2.0 slot is spaced two rows down from the x16 slot.
Between them, there's an IR module header, a TPM header and a four-pin fan header. While the -DGS' fan header was on the edge of a double-slot graphics card, this one is clearly still underneath. Along the bottom edge of the motherboard, you'll find headers for a system speaker, a chassis intrusion alarm and two USB 2.0 ports. The -HDS also has parallel and serial port headers.
Like the -DGS, the -HDS has a modest (but adequate) I/O backplate. You get an extra PS/2 jack, adding up to two in total, and an HDMI port. Everything else is the same as the -DGS, including four USB 2.0 ports, two USB 3.0 ports and three 3.5mm audio jacks. Networking and audio are once again provided by Realtek, with the same ALC662 codec and a slightly different RTL8111G GbE chip.
The box contents carry over, too. Namely, you get two SATA cables, one with an angled connector, a manual, an I/O shield and an installation CD.
Like the -DGS, the H81M-HDS doesn't have any big feature deficiencies, given its price point. But that front-panel connector is a big layout issue for anyone using a full-size graphics card.
Like the -DGS, I started slowly with a 40x multiplier and core voltage set to Auto. One load test showed no problems at the familiar 1.201V setting. I went for a 42x multiplier at 1.215V and found that combination stable here as well. The temperatures were notably higher though, especially on the VRMs, which were edging toward 60 degrees C. Still, I didn't want to let that extra power phase go to waste.
I ran the A-Tune auto overclock feature out of curiosity, and it suggested that 43x at the same 1.215V should be good. I ran some load tests, but stability suffered. Upping the voltage to 1.24V seemed promising, only to watch the PC crash at the end of long Prime95 runs. Going for broke, I tried 1.25V and saw success! The CPU was hot at 80 degrees C, but the VRMs stayed just under 60 degrees, so I wasn't worried about blowing them.
While this configuration was technically stable on the bench, it isn't something I'd recommend for long-term use. A better cooler might improve the thermal readings. But the extra expense of aftermarket cooling destroys the premise of dirt-cheap overclocking.
The RAM was once again quick and easy to tune. After a bit of fiddling, I was able to get 1400 MT/s working. Using 8-8-8-24 timings at the stock 1.5V was fine, but I had to dial in 1.55V to get 7-7-7-21 stable because this board doesn't cheat on RAM voltage. I need to confirm with Thomas, but this may be the first motherboard in Tom's Hardware's history with no RAM voltage bias. Stock voltage read 1.5V on my meter, and my overclock of 1.55V read true as well.
My elation quickly faded once I started a full system stress test to validate the stability of my overclocks. The CPU was now peaking at 87 degrees C, averaging 82 degrees C, and Windows was crashing. Throttling the CPU back down to 4.2GHz (but keeping the RAM maxed out) solved my issues, so I licked my wounds and called it a day.