ASRock Rack X470D4U2-2T Review AMD Ryzen Server with 10GbE

16

ASRock Rack X470D4U2-2T Benchmarks

We wanted to give some sense of relative performance to more traditional server solutions on the market.

Python Linux 4.4.2 Kernel Compile Benchmark

This is one of the most requested benchmarks for STH over the past few years. The task was simple, we have a standard configuration file, the Linux 4.4.2 kernel from kernel.org, and make the standard auto-generated configuration utilizing every thread in the system. We are expressing results in terms of compiles per hour to make the results easier to read:

ASRock Rack X470D4U2 2T Linux Kernel Compile Benchmark
ASRock Rack X470D4U2 2T Linux Kernel Compile Benchmark

If you read our X470D4U review, you are going to notice these are very similar. Our Ryzen test points were effectively identical between that platform and this X470D4U2-2T. And we are using the same comparison points on the Intel side. We simply wanted to validate that we were getting the same result which is what we would expect with motherboards with the same base.

c-ray 1.1 Performance

We have been using c-ray for our performance testing for years now. It is a ray tracing benchmark that is extremely popular to show differences in processors under multi-threaded workloads. We are going to use our 8K results which work well at this end of the performance spectrum.

ASRock Rack X470D4U2 2T C Ray 8K Benchmark
ASRock Rack X470D4U2 2T C Ray 8K Benchmark

Here we can see c-ray 8K results that are solid. Performance of AMD Zen, Zen+, and Zen 2 chips tend to be great on this type of benchmark. If you wanted to build out a render farm and still have manageable nodes, then this may make a lot of sense as a solution.

7-zip Compression Performance

7-zip is a widely used compression/ decompression program that works cross-platform. We started using the program during our early days with Windows testing. It is now part of Linux-Bench.

ASRock Rack X470D4U2 2T 7zip Compression Benchmark
ASRock Rack X470D4U2 2T 7zip Compression Benchmark

We tried getting a sample set that crosses a decent spectrum of alternatives to give some anchor points. This included higher-end solutions such as the AMD EPYC 7232P and the Intel Xeon Bronze 3206R which one would find in more mainstream servers. Some of the comparisons are a bit odd. We still wanted to give some grounding as you think about Ryzen compared to other options. AMD EPYC platforms enjoy server vendor support and have more memory and PCIe I/O capacity. If you want major data center software vendor support, you likely want to use EPYC.

OpenSSL Performance

OpenSSL is widely used to secure communications between servers. This is an important protocol in many server stacks. We first look at our sign tests:

ASRock Rack X470D4U2 2T OpenSSL Sign Benchmark
ASRock Rack X470D4U2 2T OpenSSL Sign Benchmark

Here are the verify results:

ASRock Rack X470D4U2 2T OpenSSL Verify Benchmark
ASRock Rack X470D4U2 2T OpenSSL Verify Benchmark

Here we can see the AMD Ryzen 1600 AF performs just below the Intel Xeon E-2244G levels while the Ryzen 5 3600 in the platform is a bit faster than the Xeon E-2246G. AMD has significantly lower pricing at a chip level making these important comparisons in the market.

Chess Benchmarking

Chess is an interesting use case since it has almost unlimited complexity. Over the years, we have received a number of requests to bring back chess benchmarking. We have been profiling systems and now use the results in our mainstream reviews:

ASRock Rack X470D4U2 2T Chess Benchmark
ASRock Rack X470D4U2 2T Chess Benchmark

Just taking note here that the higher TDPs of these parts can outpace the AMD EPYC 3251 parts. Embedded EPYC platforms can be an alternative, especially at the lower-end where we do not get the major PCIe I/O bumps of the EPYC 3351 and AMD EPYC 3451 dual die parts.

Overall, performance is good if all you are looking to obtain is a lower core count and memory footprint server.

Power Consumption

Our test configuration is pretty lightweight and the CPU is rated at 65W, so we were expecting decently low power consumption numbers. These power consumption numbers were gathered with the Ryzen 5 1600 AF CPU, but all three should be similar.

  • Idle Power: 26W
  • STH 100% Load: 110W
  • Maximum observed power: 119W

These results were observed on 120V power using a basic Kill-A-Watt meter. The system is powered by a consumer-grade 80Plus Bronze power supply similar to what is found in many low-power short-depth server chassis.

Next, we are going to discuss some of our closing thoughts around the solution.

1
2
3
4
REVIEW OVERVIEW
Design & Aesthetics
8.5
Performance
9.3
Feature Set
9.0
Value
8.4
Previous articleSTH Turns 11 in a Very Different Year
Next articleOur SwitchNAServer or QNAP QGD-1600P Review
Will has worked in both big enterprise and small business IT since 2001. As a perpetual dabbler, he is always open to new solutions for old problems. That said, his personal IT motto has to be "if it's not broke, don't fix it" so sometimes the old ways are best

16 COMMENTS

  1. Will – thanks for this review. I have waited three years to see a uATX board with x8/x8 slots. Good points about the low transfer speeds over the BMC, and the DIMM s
    ots. Those are real problems. I have been keeping my eye on this board but it looks like it still needs some polish.

  2. Is it normal for the serial port to be male like that? I would have thought 9-pin female was standard for onboard serial.

  3. HedRat,
    DB9 serial ports are male, the cables are female.

    emerth,
    Despite the problems, they’re not insurmountable and I still think that both the X470D4U boards have merit. A taller cooler easily mitigates the DIMM issue, and while the BMC is slow that only really comes into effect during installation of an OS. For day-to-day access like rebooting the system over the IPKVM, it has no effect.

  4. I wonder if STH has actual measurements (dimensions) of the DIMMs they used that made contact with the AMD cooler? Even the exact manufactuer, model, and part number info of the DRAM used in this test would be helpful; some manufacturers might vary their product over time. Any detailed info might help buyers considering their DRAM selection when using that cooler.

  5. StiggyB,
    I hope so as well!

    Sleepy,
    For all the benchmarks and testing, I was running some low profile ECC memory. It didn’t much matter when running two sticks though, since ASRR’s recommended memory installation is in slots 2 and 4 to start with. When I first received the board, however, I plugged in some standard desktop memory (Corsair Vengeance LPX CMK32GX4MA2666C16) just for testing. This stuff does have a heatsink on it, but it’s not extraordinarily tall or anything and is barely taller than a naked DIMM. Here is a picture of one of those sticks, next to a naked stick so you can see they are essentially the same size. https://i.imgur.com/IV6pMI8.png

    With the low profile memory and the stock heatsink, you *may* be able to run a stick in the first slot, but it would be tight and you definitely would not be able to add or remove that memory without first taking off the heatsink & fan. With the Corsair memory, or anything shaped like, they are fundamentally incompatible and a cooler with more clearance is necessary.

  6. Hello,
    sorry for my english, I’m not native speaker. I’ve been using ASRR X470D4U-2T for ten months without any problem. My CPU cooler been changed to NH-C12P SE14, which have no contact with memory slot. My config is X470 + R7 3700x + 4 x 16 GB Micron 18ASF2G72AZ-2G6D1 + SAS 2008 controller and Proxmox 6.2

    Some pictures > https://imgur.com/gallery/2ihlrMM

  7. To bad that with a DIY home server setup you’re mostly forced into using UDIMMs (not counting old second hand server hardware) which are stupidly pricey.
    Really wish those prices went down, or the CPU’s support Registered ECC DIMMs.

  8. i have a threadripper for home server : main feature like double wifi rgblight control fan speed is only with windows ,iommu group is very large , no remote control like ipmi
    no more amd ryzen o TR maybe now i’m thinking to sell TR to buy a xeon or epyc entry level

  9. I like these boards other than just being the only real alternative to intel for x86 small servers.
    Having all the pcie slots coming off the cpu means is good for the add in cards and the non-2t version gives an option on the x4 for networking that I prefered and the pch contains all the onboard storage by default. It’s limited to x4 pcie so having half speeds on the double nvme drives can saturate that bus even before anything is attached to the sata array. It feels pretty well balanced in those terms considering the obvious limits of thse boards when derived from mainstream parts.

    @sleepy
    I’ve also had one of these boards using 4 corsair vengeance LPX modules and an r5 3600.
    Using all four modules had a significant impact on the fins of the stealth cooler so this board probably doesn’t obey the am4 guidelines.
    Using some destructive interferance it was possible to get it to work though. Dismantling the fan on top of the stealth was necessary and then bending the fins out of the way it was able to fit in tightly, and like Will found, there is still some contact between the ram and cpu heatsinks. (also cable tied a 120mm fan on top that got in the way of the usb3 header.

    Our long term solution was an artic alpine 64gt cooler. Dirt cheap, uses the default bracket so it’s easy to install but it’s only rated to 70w so high end cpus will need to run in eco mode. We’re running our 3600 at 45w eco mode for the extra headroom in our poor enviroment.

  10. I wonder if this motherboard will work with the NH-9Lx65, I’m going to put this in a 2U case so I need a low profile fan. Did anyone try it out?

  11. I am looking at the 1u servers based on this board as an option for game servers. A 3950x with 128gb of ram would be ideal for running my IL-2 Great battles and DCS World game servers while taking very little rack space

  12. I have the x470d4u board and use the Bequiet Dark rock 4. Since it’s my enthusiast “home server” in a FD R5 case i obtained the shorter mounting brackets to change the cooler orientation 90°. That creates clearance for all the memory slots. With original mounting brackets the first slot is blocked.
    I equipped 2xkingston 2666MHz ECC dimms in my system.

LEAVE A REPLY

Please enter your comment!
Please enter your name here