Since so much former chipset functionality now resides on the CPU die, and since there are only a handful of third-party peripheral controllers out there, we rarely see meaningful performance differences between motherboards these days. That said, we still test performance, if for no other reason than to ensure everything is functioning correctly.
When it comes to testing motherboard performance, we've usually gathered benchmark results using the CPU's peak stock memory multipliers. Since DDR4 is so new, however, and Haswell-E's conservative 2133MHz maximum stock speed, we've taken to testing X99 boards with the memory clocked at the highest speed we can attain while keeping the CPU at its stock clocks.
We tested Gigabyte's X99-Gaming 5P against another Haswell-E board, Asus' X99 Deluxe, which Geoff reviewed last year. Asus' board was able to clock our DDR4 DIMMs at 2800MHz while maintaining stock CPU clocks. The X99-Gaming 5P, on the other hand, had to boost the base clock to 127.3MHz in order to run our DDR4 memory at 2800MHz. Since we had to settle for a 2666MHz top speed on the Gaming 5P to play by these rules, these results aren't strictly an apples-to-apples comparison. As you'll see below, however, this makes very little difference to real-world performance. Using targeted memory benchmarks like Stream makes the difference easy to spot, though:
Since the X99 Deluxe is able to run our memory 5% faster than the X99-Gaming 5P, we should expect a synthetic memory benchmark to show a difference in performance. We're not let down: Stream shows just more than a 6% higher sustained memory bandwidth for the copy test.
As expected, a 5% deficit in memory clock speed doesn't put the X99-Gaming 5P at a disadvantage in real-world tasks. In fact, the board pulls out a decent win in our x264 encoding test. We're not exactly sure what's behind the higher x264 result for the Gigabyte board, but repeated testing shows a consistent 7% advantage over the Asus board. For the rest of the tests, the Gigabyte board is happily trading blows with its rival.
Boot times were remarkably close between the two boards, with both taking around 28 seconds to go from power-on to the Windows 8.1 desktop. With Z97-based boards boasting boot times in the low-to-mid teens, the above times may look long, but they're not unexpectedly so. After all, Intel's high-end desktop platforms borrow heavily from their two-socket server/workstation kin, which play in a market where boot time performance is rarely considered, much less discussed.
Unlike performance results, one's choice of motherboard can have a notable impact on power consumption. We measured total system power draw (sans monitor and speakers) at the wall socket with our test system idling for a period of five minutes in the Windows desktop, and then under a full load combining Cinebench rendering with the Unigine Valley demo.
While the X99-Gaming 5P consumes only 3W less than the X99 Deluxe at idle, this difference grows to 25W under load. Asus' LGA2011 boards have a history of higher power draw, but even so, that's still a sizable difference. It seems unrelated to the difference in the boards' respective memory speeds—when we run the memory at 2666MHz in the X99 Deluxe, the load and idle power results remain unchanged.
Perhaps we're seeing the advantage of Gigabyte's purportedly more efficient power delivery system here. That said, we do have to remember that the Asus board is powering more third party controllers, like extra SATA, USB, GigE, and Wi-Fi chips.
As part of our investigation into the above differences in power consumption under load, we took readings with both boards running the memory at the stock speed of 2133MHz. Interestingly, this config dropped load power consumption by 12.7W on the X99-Gaming 5P and by 13.3W on the X99 Deluxe. The corresponding idle power consumption dropped by 11W on both boards. Perhaps requesting an increase to the memory multiplier results in the firmware boosting some uncore voltage to ensure stability when clocking the memory this fast.
The following page is loaded with detailed motherboard specifications, system configurations, and test procedures. For those who take the path less traveled by not skipping straight to the conclusion, thank you. Those tables don't populate themselves. You'll also get to see pictures of the test components inside and out of the Antec P380 case we use to house our test rigs.