Performance
If you are wondering why there is no comparison to other motherboards, it’s because all results were really close to each other. I was comparing results to ASUS TUF X299 Mark 2 motherboard which is based on the same PCB as most other ASUS motherboards so results should be about the same as on ASUS Prime, Prime Deluxe or Strix.
Let’s go back to our test rig. Test specification includes the i9-7900X processor, 32GB DDR4-3733 Patriot Viper 4 and 64GB DDR4-3000 Ballistix Elite memory kits, Crucial BX300 SSD for SATA storage tests and Samsung 840 Pro for Windows 10 installation. Additional storage tests were performed on the Optane 16GB SSD in RAID/VROC. Graphics test results are based on EVGA 1080Ti graphics card. Whole rig was cooled by custom water cooling based on Watercool components reviewed couple of weeks ago.
Let’s begin with Futuremark benchmarks.
PCMark 10 is showing expected, high performance. As I mentioned results in benchmarks are within’ 0-2% difference between tested motherboards. Since the X299 chipset is designed for high-end computers then most manufacturers care to provide high quality products. Not all make everything right but most differences are in additional components which are not affecting test results much. For sure they affect stability and additional functionality of the motherboard.
The latest version of 3DMark Time Spy Extreme is also showing good results. Nothing strange since the X299 Taichi had no performance issues and the X299 XE is improved version.
Storage results
Popular ATTO Disk Benchmark is showing us really nice numbers. Typical SATA results are not much above 550MB/s. Here we see 564MB/s what is probably the highest result I’ve seen on a standard SATA3 cable. For tests has been used Crucial BX300 480GB SSD.
Also CrystalDiskMark is showing great results with maximum bandwidth above 563MB/s!
Here is also one result with two 240GB Patriot Hellfire NVMe SSD in RAID0. No issues to pass 5.2GB/s.
Even though sequential results are high then the latest generation of SSD is offering much faster random bandwidth. We were able to perform couple of tests using 4x Intel Optane SSD.
Intel VROC
Intel VROC is an interesting topic but sadly this technology isn’t perfect. Intel is changing specification and you may notice differences between paper manual and its online version. I don’t know what about other brands but ASRock has already changed description and supported drives list as what Intel provided wasn’t really true and was causing various issues ( regardless of motherboard brand ).
Below you can see our results using four 16GB Intel Optane drives. Low capacity and lower performance than higher series drives but still good enough for tests. I don’t expect that anyone will build RAID array using 16GB drives for home usage.
This is expected result but what you see is array created using Windows disk manager – stripped volume. This configuration doesn’t even require RAID controller but it still works like software RAID.
Once VROC is enabled then random 4KiB Q1T1 results are much worse – screenshot below.
I can assure you it’s not the Taichi XE fault as the same results are on other motherboards, also on other motherboard brands. Intel is not providing any additional support regading this issue. On the other hand ASRock support was really helpful and provided additional test results which are showing similar behaviour.
We’ve performed tests on a specific hardware and results can be better while in use will be single PCIE card which supports more than 1 SSD.
What is interesting is the fact that M.2_1 socket can work in VROC mode so it’s possible to enable VROC using three PCIE slots and one M.2. In this way we set our array. In use was also VROC Premium key so the highest key available on the market which is unlocking all RAID modes and offers support for all specified SSD.
I guess we have to wait on an answer from Intel and possible firmware update for VROC key, new version of RSTe drivers or motherboard microcode.
