Surely you've read our 3rd generation Ryzen review, including our look at the budget model Ryzen 5 3600 for $ 200. While we were testing that we asked how well do these processors work on a really affordable B350 motherboard? The test person for this experiment is the Asrock AB350M Pro4. Why this micro ATX motherboard? It's easy, this was one of our top picks for the best "ultra cheap" B350 motherboard that was available for only $ 75 … it was a bargain in 2017.
It is also a great candidate for our 3rd generation Ryzen tests on a B350 board as it is as bald as they come. It packs a 3-phase vCore VRM using the 4-phase ISL95712 controller, but we get a component doubling so each phase contains two high-side MOSFETs and two low-side MOSFETs. In that sense, it's pretty decent for an AM4 motherboard under $ 100.
In this article, we would like to examine two main things: Can an inexpensive B350 motherboard safely support the Ryzen 9 3900X? How hot does the VRM get during a torture test, for example? We're going to strap on a couple of k thermocouples to find out.
Suppose we don't cook the board. How's the performance? Are there thermal problems, performance limitations, or other factors that could affect performance? We will test the 3900X, 3700X and the Vanilla 3600 on the Asrock board and then compare the results to those collected on the MSI X570 Creation.
Using the BIOS version 5.90, the Asrock AB350M supports Pro4 Zen 2 processors with AGESA version 22.214.171.124. The MSI X570 build was tested with the latest BIOS using AGESA version 126.96.36.199. This version was used for all previous tests and works as expected.
First we have VRM temperatures and we will first look at a worst case scenario – at least until the Ryzen 9 3950X is released. We don't expect many to try to pair the 3900X with an entry-level B350 board, but it's useful information for a few years later at which point you'll likely be able to get the 12 core Part to snap peanuts.
All of these tests were carried out on an open test bench without additional cooling. However, this is not so unrealistic as we use the box cooler that directs air over the VRM heat sink, at least one of them. This setup has the advantage of cool air circulation because it is not clogged in a poorly ventilated case, with the disadvantage that there is no direct airflow from the case fans.
The Ryzen 9 3900X extracted 145 watts in stock and this pushed the VRM temperatures to a maximum of 65 degrees for the bottom of the circuit board, where most of the heat is dissipated while the top of the MOSFET driver only reached 51 degrees. Both are very safe operating temperatures for these components and with 8 sensors we covered the board very well. What you see here are the hot spots.
Activating PBO with the box cooler is pretty pointless. It only increases power consumption and leads to almost no additional power. Nevertheless, the VRM is more stressful, so we are interested in doing this for this test. This increased the package consumption to 185 watts, which corresponds to a performance increase of 28%. The peak temperature at the bottom of the circuit board increased by 22% and we can see that the circuit board rises by 80 degrees.
Although these temperatures are not dangerous, they get a little too hot to sustain for long periods of time. The board is now also on the edge and since PBO + AutoOC is activated in the Ryzen Master software, the Asrock AB350M Pro 4 has come off.
The board did not experience a major failure or crash, but Blender's 1-hour stress test closed after about 20 minutes for the desktop. This happened three times in a row. We took the hint and gave the small B350 board a rest.
In short, it looks like the Asrock AB350M Pro4 can comfortably deliver 150 watts, but once you start increasing over 180 watts it will want to wrap it up.
When switching to the 65-watt TDP Ryzen 7 3700X, the B350 board did not break a sweat. Unfortunately, we didn't measure the package's power consumption with this processor, we just didn't think about it at the time, and that was before we tested the 3900X. But even with PBO + AutoOC activated, the 3700X ran flawlessly on the Asrock B350 board and the VRM temperatures only reached a maximum of 46 ° C, which is very safe. In this case, we didn't waste time testing the Ryzen 5 3600, as this won't burden the VRM.
In fact, there was a problem with the R5 3600: memory support was somewhat unreliable. While the Ryzen 5 processor on the X570 cards works with DDR4-3600 memory, it only worked on the Asrock B350 card up to DDR4-3000. This is interesting because both the 3700X and the 3900X with DDR4-3200 worked perfectly on the Asrock board, but DDR4-3600 did not.
In other words, memory support is nowhere near as good as the X570, but that would almost certainly be the case. Future BIOS revisions may improve memory support, but it is equally likely that the AB350 Pro4's signal quality is not good enough to go beyond DDR4-3200, and even lower 3200 CPUs are out of the question.
This means that for the benchmarks you will see, the 3900X and 3700X were tested with the DDR4-3200 CL14 memory, while the 3600 was tested with the DDR4-3000 CL14 memory.
Despite the change in memory frequency, the R5 3600 has the smallest performance difference between the two cards tested. The X570 boards were almost 1% faster, so chalk up to the error rate.
The 3700X was 3% faster on the X570 board, not a big margin, but it was consistently 2-3% faster. Then we see that the 3900X on the X570 board was 4% faster … which also costs about 7 times as much.
When testing with Blender, we found that the typical all-core clock frequency increased a bit, but after half an hour it adjusted to the clock rates specified here. Basically, the 3900X on the X570 board was 2.5% higher, the 3700X was 2.7% higher and the R5 3600 was 0.1% higher.
In short, the performance of the R5 3600 was the same, that of the 3700X was 2% longer to complete the test and that of the 3900X was 4% longer, which is in line with the margins observed in Cinebench.
This time the 3900X was 5% faster on the X570 board when tested with V-Ray, while the 3700X was only 3% faster and we can see that the R5 3600 does 1% better on the B350 board. Anything 3% or less is within the margin of error.
In terms of gaming performance, the R5 3600 was again able to deliver the same level of performance on both boards, the same was true for the 3700X, while the 3900X lost a few frames.
Assassin's Creed Odyssey is a much more demanding game for the CPU, and here the 3900X and 3700X models were 4% faster on the X570 board, while the R5 3600 saw no real difference.
After completing the benchmarks, we have the total power consumption of the system while running the Blender test. Total system consumption was reduced by 10% for the 3900X, 16% for the 3700X and 11% for the 3600.
There are probably a few reasons for this. The high-end X570 cards may waste more power due to their extreme VRM implementations. We know that the X570 chipset uses a lot more energy and these cards are generally littered with RGB LEDs, although the MSI Creation is good in this regard. In both cases, AMD's Zen 2 CPUs appear to be more energy efficient on 300 and 400 series boards.
The good news for those who own entry-level AM4 motherboards is that they can easily handle the new 3rd generation Ryzen processors from the R5 3600 to the Ryzen 9 3900X. We haven't found any issues other than questionable memory support, although this wasn't new to Ryzen 1st and 2nd generation parts.
The Ryzen 9 3900X works fine in its standard configuration, but don't expect to be able to overclock the 12-core CPU. Here too, memory support on this special B350 card was limited to DDR4-3200. However, this is all you need to get maximum performance, provided you are using low latency storage. We don't think 3900X support is relevant right now, but in a few years, if you can buy these CPUs second-hand, we will expect them to be between $ 100 and $ 200. Then get one and stick it on a cheap B350 board will be a nice option.
With high-end B350 boards, we still have to test them all, but those with a more powerful VRM should be even better, provided there is BIOS support. We are also in the early stages of supporting older motherboards so the situation could improve. There should be no performance difference between AGESA 188.8.131.52 and 184.108.40.206, but memory compatibility and stability in general could be improved.
We are aware of the challenges AMD and its board partners face due to the promise of backward compatibility. For example, the memory capacity of the SPI flash EEPROM chip that stores the motherboard's UEFI firmware is not large enough to repair the latest AGESA microcode. This is a problem for a number of entry-level and mid-range boards that only use a 16 MB EEPROM.
Asrock has created a lite version of the BIOS that contains no colors and has a very simple layout. The X570 motherboards have been upgraded to a 32MB EEPROM, no doubt to help tackle this problem in the future, although they are discontinuing support for first-generation parts anyway.
The bottom line is that AMD's AM4 upgrade path can be used by your choice sooner or later.