Gigabyte X99-Ultra Gaming Benchmarks
Now to the main event, taking a look at the differences between this X99 board and older X99 boards. Keep in mind that before you look at the benchmarks, the differences will mostly be negligible considering everything is nearly the same. Better manufacturing techniques and higher quality components can lead to lower latency and better performance, but only just.
The test system for these benchmarks consisted of the following configuration:
- Intel Core i7-6900K
- Gigabyte X99-Ultra Gaming
- Sapphire AMD R9 Nano
- 32GB Crucial Ballistix Sport DDR4 2400 (8x4GB)
- 512GB Samsung 950 Pro
- 4TB Seagate Hybrid SSHD
- Corsair HX1000i
- Windows 10 Pro 64-bit
First we’ll see if there are any improvements in memory bandwidth and then on through the gamut of gaming benchmarks and, of course, taking a look at the interface speeds. We’ll use SiSoft Sandra, Passmark and CrystalDiskMark to acquire our data, even for the LAN score, which will later be compared against all motherboards that are tested here. For this test I used my own personal motherboard connected to the same equipment to compare against, the Gigabyte X99M Gaming 5, another particular potent, yet miniature, motherboard.
Here we see a real, though small, improvement. Any differences are going to be due to better electrical components, capacitors and even better solder quality that can amount to any difference that you might be able to discern. With relatively low-clocked RAM, we see some very expected numbers with uncompressed data.
Latency sees a similar small decrease. It isn’t much, but the average of three tests still has it lower, though by a very small amount.
The DPC latency checker measures how much time it takes to process Deferred Procedure Calls from the time that the driver invokes the call. Less latency means potentially better audio. Higher latency can sometimes be the reason for pops and hisses in streamed audio on your PC. It can be improved in a number of different ways, mainly in the way that things are, again, constructed. It’s not a large improvement, but it’s an improvement nonetheless.
Cinebech scores are nearly the same, though that’s expected. The same CPU means it’s largely going to be the same. Over a large enough sample group, it still shows a small improvement of a few points consistently.
CNTK likewise sees a small increase, but again this is expected to be small anyway. Regardless, it’s an increase, and it’s consistent across three tests.
We used SiSoft Sandra’s built in LAN test to measure just how fast each networking interface is. It measures the bandwidth on the network itself, not your Internet speed. This is how fast you can transfer file across the network. The test is interesting in that it shows that the Killer E2400 network adapter actually has a distinct, though slight, advantage compared to the usual Intel choice. That and it’s able to prioritize traffic from the Internet in a way that Intel cannot. The difference in actual gaming is said to be more significant, though such a difference wasn’t seen in our normal tests. That isn’t to say that it isn’t there, but just not easily observed. Perhaps it’s my own Internet connection that is the ultimate limiting factor.