The ASRock Rack EPYCD8-2T Motherboard Review: From Naples to Rome
by Gavin Bonshor on April 20, 2020 9:00 AM EST- Posted in
- Motherboards
- AMD
- Workstation
- server
- ASRock Rack
- Naples
- Rome
- EPYC 7351P
- EPYCD8-2T
System Performance
Not all motherboards are created equal. On the face of it, they should all perform the same and differ only in the functionality they provide - however, this is not the case. The obvious pointers are power consumption, but also the ability for the manufacturer to optimize USB speed, audio quality (based on audio codec), POST time and latency. This can come down to the manufacturing process and prowess, so these are tested.
For this review we are running using Windows 10 64-bit with the 1909 update as per our Ryzen Threadripper 3960X and 3970X CPU review.
Power Consumption
Power consumption was tested on the system while in a single ASUS GTX 980 GPU configuration with a wall meter connected to the Thermaltake 1200W power supply. This power supply has ~75% efficiency > 50W, and 90%+ efficiency at 250W, suitable for both idle and multi-GPU loading. This method of power reading allows us to compare the power management of the UEFI and the board to supply components with power under load, and includes typical PSU losses due to efficiency. These are the real-world values that consumers may expect from a typical system (minus the monitor) using this motherboard.
While this method for power measurement may not be ideal, and you feel these numbers are not representative due to the high wattage power supply being used (we use the same PSU to remain consistent over a series of reviews, and the fact that some boards on our testbed get tested with three or four high powered GPUs), the important point to take away is the relationship between the numbers. These boards are all under the same conditions, and thus the differences between them should be easy to spot.
Out of all the boards we tested above, the ASRock EPYCD8-2T shows impeccable power efficiency in our long idle, idle, and full load power testing. Compared directly against the GIGABYTE MZ31-AR0, it operates around 13 W lighter at full load, while it's 6 W at long idle, and 4 W lighter in an idle state.
Non-UEFI POST Time
Different motherboards have different POST sequences before an operating system is initialized. A lot of this is dependent on the board itself, and POST boot time is determined by the controllers on board (and the sequence of how those extras are organized). As part of our testing, we look at the POST Boot Time using a stopwatch. This is the time from pressing the ON button on the computer to when Windows starts loading. (We discount Windows loading as it is highly variable given Windows specific features.)
The professional level server and workstations model tend to take longer to POST into the operating system than consumer boards, and our results show this. Typically a professional model with BMC takes longer due to controller initialization, but ASRock is a good 21 seconds quicker than the GIGABYTE model at booting into Windows 10.
DPC Latency
Deferred Procedure Call latency is a way in which Windows handles interrupt servicing. In order to wait for a processor to acknowledge the request, the system will queue all interrupt requests by priority. Critical interrupts will be handled as soon as possible, whereas lesser priority requests such as audio will be further down the line. If the audio device requires data, it will have to wait until the request is processed before the buffer is filled.
If the device drivers of higher priority components in a system are poorly implemented, this can cause delays in request scheduling and process time. This can lead to an empty audio buffer and characteristic audible pauses, pops and clicks. The DPC latency checker measures how much time is taken processing DPCs from driver invocation. The lower the value will result in better audio transfer at smaller buffer sizes. Results are measured in microseconds.
We test the DPC latency at default settings out of the box, and the ASRock board performs noticeably better than any other AMD Threadripper HEDT and EPYC focused model we've tested so far.
Facebook
Twitter
RSS
34 Comments
View All Comments
SampsonJackson - Monday, April 20, 2020 - link
That is absolutely incorrect. We do it with Infiniband cards via RDMA and easily saturate multiple 100Gbps cards. Der8auer demonstrated ~28GB/s on a RAID0 using Threadripper 1st gen (~224Gbps) and was only limited by the RAID driver thread saturating a CPU core.. further scaling is possible using the inbox NVMe driver (up to endpoint/bus saturation). Are these realistic workloads? No. Is it possible? No problem.vFunct - Monday, April 20, 2020 - link
CPUs on media servers have been saturating 100G for years now. Netflix is doing that, for example. https://netflixtechblog.com/serving-100-gbps-from-...vFunct - Monday, April 20, 2020 - link
And they're delivering 200gbps now: https://wccftech.com/netflix-evaluating-replacing-...brunis.dk - Monday, April 20, 2020 - link
I think ASSRock should just rename themselves to ASRack for simplicity.kobblestown - Monday, April 20, 2020 - link
What's with the 6-pin fan connectors? Can I plug a regular 4-pin PWM fan into it?dotes12 - Monday, April 20, 2020 - link
I looked up the user manual and yes, it's keyed so that both a normal 3-pin and 4-pin fan will work with the 6-pin motherboard connector without an adapter. It appears that the extra two pins are used for a temperature sensor that's built into the fan. Per the manual, pin 5 is labeled "Sensor" and pin 6 is labeled "NC", and the custom fan speed has an option called "Smart Fan Temp Control" where you can have it increase a specific fan speed based on the temperature the fan is reporting.kobblestown - Monday, April 20, 2020 - link
Oh, that's cool. Thanks for checking it out.cygnus1 - Monday, April 20, 2020 - link
I was originally going to say "WTF are they thinking releasing such a high end AMD board in 2020 that doesn't support PCIe 4.0 when the appropriate CPU is installed. What a waste." But then I realized this board is about a year old already. As others mentioned below, the ROMED8-2T is almost the replacement for this year old board being reviewed. The biggest thing missing from that one is the x16 slots. And for whatever reason they didn't leave the x8 slots open ended to allow for x16 cards to fit.WaltC - Monday, April 20, 2020 - link
This motherboard is a cheap EPYC *server* mboard, and that is all it is...;) Keyword being "cheap"--paring down the system bus to PCIe3.x cuts the system bandwidth in half, compared with 4.0, which translates to manufacturing a lower-cost mboard relative to the layers needed to properly support the signal integrity of a PCIe4.0 system bus. A PCIe3.x system bus also requires less power than PCIex4. It's easy to forget, I suppose, that PCIe4 is *double* the bandwidth of PCIe3. But as a cheap server mboard, PCIe4 may not be a better fit than PCIe3.x.This "review" is a bit strange, imo...;) Not only does it directly compare different mboards, but it also compares those mboards running different CPUs, as well, as if to illustrate some obscure point. I would have done things a bit differently, like, for instance, restricting my choice of motherboards to those server boards capable of running this CPU--and *actually running* the EPYC CPU featured here...;) Maybe throw in a couple of system bandwidth tests and applications to illustrate advantages of the increased bandwidth PCIe4x brings to the table, along with extra costs, etc. Otherwise, what one winds up comparing are CPUs instead of motherboards, imo. As server mboards go, this one is not "high end" at all--it's actually a "budget" server mboard, imo--hence the compromises with system bus bandwidth, etc. Simply put, this mboard was not designed to "compete" with "enthusiast-class" retail mboards used for gaming--as should be obvious. People looking for budget-class server motherboards for EPYC-class cpus won't care about PCIe4, the "colors" used, RGB, multi-GPUs, etc. Those things add to cost and energy consumption, and, of course, superficial color schemes/RGB offer no power efficiency or performance enhancements of any kind.
mjz_5 - Monday, April 20, 2020 - link
I hope it’s running the enterprise version of Windows 10 because that has better performance for high core count computers.