Ryzen CPU Cooler vs. Performance Testing

Ryzen has been a massive hit amongst the PC enthusiast and it’s pretty clear why – cost per performance is pretty dang low!  You can get a ton of computing power for a fraction of the price of the competition, and most people will be attracted to this.  Sure they have their quirks when it comes to low resolution gaming, but they can hold their ground against the competing Intel processor in other benchmarks and high resolution gaming, where the load is less dependent on the CPU.  This got us wondering, since Ryzen will overclock the processor even further utilizing XFR (Extended Frequency Range), if a cooler cannot handle the load well, how much will the temperature of the processor affect performance?  If we jump over that magic 60*C mark, will the performance suffer greatly, or is it relatively on par with the CPU below that number?  If we’re really close to that 60*C number, will we suffer compared to a cooler that can handle the load better? We will measure this via two different methods:  Handbrake video encoding and 7-Zip.  Handbrake is an open-source, GPL-licensed, multiplatform, multithreaded video transcoder, available for MacOS X, Linux and Windows. It is popular today as it allows you to transcode multiple input video formats to h.264 output format and is highly multithreaded. In this article, we are going to take a look at three coolers, two of which we have previously benchmarked on our Intel Core i7-4770k system, and one designed specifically for Ryzen.  The coolers in question are the Noctua NH-D15, Thermalright TRUE Spirit 140 Direct, and a custom EKWB AIO Cooler (EK-XLC Predator) that was sent with our Ryzen 7 1800X.  Our friends at Noctua were generous enough to send along the brackets needed to hook up the NH-D15 to our new Ryzen setup. [gallery link="file" columns="4" ids="193669,193670,193671,193678,193679,193686"] If you’ve for some reason forgotten, the Ryzen 7 series of CPUs were designed to target high-end Intel CPUs, hoping to bring competition back to the high-end processor market.  When we had no competition in this market, Intel processors have gone up in price while yielding IPC gains of no more than 10%, which undoubtedly is frustrating for us enthusiasts.  This is where gamers were waiting for the day of AMD’s return, so we can have lower prices and better gains.  This is where we welcome Ryzen to the market!  Our $499 Ryzen 7 1800X is targeting Intel’s $1050 Core i7-6900k, both of which feature 8 cores and 16 threads. To learn more about Ryzen, you can read our full review of the 1700, 1700X, and 1800X here.  We also have our Ryzen 5 review up here. Do note that I originally started writing this article shortly after Ryzen 7 was released, but had to tend to family needs for a few weeks and got behind.  All software that was downloaded was grabbed on 03/17/17 and the Ryzen Power Plan was not available yet.  Instead I was able to utilize the Balanced Power Plan for my testing, but will spot check with the new Ryzen Power Plan and update the article if I find any difference in results. The motherboard that we’re going to use in our system is a Gigabyte GA-AB350-Gaming 3, which will run you a whopping $124.95 shipped on Amazon – Relatively inexpensive.  This board has been spotted around $110 in the past, so definitely keep your eye out for a bargain.  Since we aren’t going to be doing any gaming or storage benchmarks with this rig, a B350 board was chosen over a X370, as we don’t need the extra PCIe lanes and Crossfire support that the X370 offers.  The B350 will likely end up as the best-selling chipset for the AM4 platform, mainly due to cost and no restriction on overclocking, so it was a no brainer for us. Memory wise, we picked up a 16GB (2x 8GB) Kingston HyperX Predator 3200MHz memory kit.  Obviously we wanted the dual-channel benefit and wanted to have modules that could aide with the performance of our system.  This particular kit will currently run you $183.21 shipped on Amazon.  Again, keep an eye out for a bargain, as this memory can be had for a decent amount cheaper at times. You will find that we have chosen to do these benchmarks on an open bench, meaning no case.  There is a gap of 12-inches from the edge of the motherboard to the wall in the below picture, and the rest is open around it. Any other hardware was existing hardware that we had laying around.   Test System Summary ALL COOLERS will be using Noctua NT-H1 thermal compound, which requires zero cure time. AMD B350/AM4 Platform The AMD B350 platform that we used to test these coolers was running the Gigabyte GA-AB350 Gaming 3 motherboard with BIOS version F5 that came out on 02/24/2017.  Motherboard settings are all defaulted, however the memory had to manually be set to 2933MHz with 16-17-17-36 timings (Identical to XMP Profile 2).  Trying to achieve the full 3200MHz on the memory was not going to happen, and the Intel XMP profiles weren’t really meant for the AMD platform, though they can work. All drivers and OS version were current as of 03/17/2017 for this testing. No overclocking will be performed. All coolers will be utilizing the Noctua NH-T1 thermal paste, which requires zero cure time.

Ryzen 7 1800X Benchmarking…

In the benchmarks that will be run, we used an AMD Ryzen 1800X clocked at the factory spec of 3.6GHz.  Stock results will be utilizing auto for the multiplier and v-core, which read 1.248v in CPU-Z. Ryzen 7 1800X - Stock Frequency To record temperatures we used AIDA64 Engineer v5.80.4098 Beta. The two benchmarks that we will run to try and achieve what we want, are 7-Zip and Handbrake.  Both applications will tax the CPU at or near 100% on all cores and threads.  7-Zip is version 16.04 and Handbrake is version 1.0.3.  7-Zip's settings were left alone, with exception to changing the compression level to ultra.  The only settings that are modified in Handbrake are removing subtitles and changing the audio to Auto Passthru, otherwise it's exactly like when you open it every time. 7-Zip was used to compress 1180 image files, weighing in at 5.88 GB. Handbrake was used to compress a 43.3 GB movie file. Ambient temperature during all testing was 72 degrees Fahrenheit or 22.2 degrees Celsius.  

Click the below images for a larger view...

EKWB EK-XLC Pradator       Noctua NH-D15       Thermalright TRUE Spirit 140 Direct       Overall Results:  Comparing the air coolers to the water cooler, we found that Ryzen 7 actually stays pretty darn cool overall.  We noticed that our 1800X idled at roughly ambient and both air coolers held their ground very well with Handbrake.  7-Zip wasn't really much of a challenge, but it is a fun statistic to throw in there.  We also found that the performance difference between air and liquid cooling was negligible and yielded nearly identical results for each cooler.   Wait a minute…!!  We didn’t hit 60*C in the above charts, so we couldn’t find out everything that we wanted for this article!  We were determined to find out what exactly was needed to make this happen, so we toyed around with various configurations of the coolers and even tried overclocking. We started with overclocking and found that the system was only somewhat stable at 4.0GHz.  What happened is about 20 minutes into running the Handbrake test, the system would just power off.  It did not even hit a temperature that would cause it to fail like that, so it seemed voltage related.  As I tried to boost voltage (away from what Auto gives you), the system would power off sooner into the test.  I scratched my head for a little bit and figured I’d try backing off the voltage.  Sure enough, I got a longer running test, but it still powered off.  Frustrated with 4.0GHz, I backed down to 3.9GHz and saw the same results with auto voltage, but found that backing the voltage down let it run fully stable.  So lesson learned, was that auto threw too much voltage at the system. So with all that time fiddling around with the overclock, I found that the water cooler was just too good and handled the load very well.  As I observed, Ryzen 7 doesn’t get extremely hot, either. This turned me over to playing around with cooler configurations.  I tried running an air cooler with no fan, but it got too hot and never flat lined on its temperature – it just kept rising. I ended up finding that the best approach to this was utilizing no fans pushing air through the radiator on the water cooler, and instead have one pushing air parallel to the length of the cooler.  This allowed the cooler to not overheat and allowed it to plateau with its temperature. So, with all that said, please observe the graph below. Overall Results:  So what did we see here?  Well, while the CPU was hitting roughly 60*C, it would end up throttling itself back for a short while so it could cool off just a little bit.  In this time, the CPU dramatically dropped how much data is was capable of processing, as it was running at a whole 500MHz!  After it caught up, it went right back to its 3.7GHz run and seldom dropped to 3.6GHz for a second or two.  This test was re-run several times, some even on a different day, and the same results were observed, so it was not a fluke.  So in the end, while the CPU was hitting 60*C, we found that our Handbrake encode took another X minutes and X seconds to run.  This goes to show that you need adequate cooling if you feel compelled to overclock, otherwise you're just wasting your time in certain scenarios. Ryzen 7 1800X - Task Manager - High Temperature Throttling The above picture is what task manager was also looking like.

Ryzen 7 Air vs. Water Cooling Performance - Final Thoughts and Conclusion

When we thought about writing this little article, we ultimately hoped to find out how the CPU would perform when it was stressed to the point that XFR disabled itself.  We were uncertain if air cooling would be sufficient in keeping the CPU cool enough, so XFR would not turn off.  We knew that the liquid cooler would be able to handle the load just fine, but wanted to use it as a baseline as well.  What we ended up finding was rather interesting. To start, we found that our 1800X idled at darn near ambient temperature, even on our air coolers, which is incredible!  Typically with our Intel CPUs, we find that it idles a few degrees above ambient, regardless of how awesome the cooler is.  For the 1800X to basically not excess heat into your chassis or room is beyond great, even if we're only talking idle. [gallery link="file" ids="193678,193679,193686"] When we hit loads under 7-Zip and Handbrake, the results were still excellent for all coolers tested.  We found that there was very little difference between the three coolers tested, as they all completed their tests right around the same elapsed time. Since each cooler cooled the 1800X so efficiently, we had to find a way for the CPU to hit higher temperatures, namely that 60*C mark, to find out how the system performed.  With some fiddling around, we found a way to make this happen, and what we found was rather interesting.  We noticed that the CPU was throttling itself back for a couple seconds and then went right back at it.  In this throttle back, the CPU had a chance to cool off a little bit, but this was at the cost of performance. Ryzen 7 1800X - Task Manager - High Temperature Throttling So what did we figure out here?  We found that if your CPU is hitting 60*C, you will see a noticeable hit in performance.  This means that if you have intentions of overclocking this CPU, you will want adequate cooling - namely in the form of AIO liquid coolers.  The Handbrake test that we ran, went from 26 minutes 55 seconds to 29 minutes 26 seconds.  This is an increase of approximately 2.5 minutes, and while it isn't a lot, someone who is in a rush or doing a lot of transcoding will not be too happy. Ryzen 7 1800X - EKWB - Handbrake - No Fans Legit Bottom Line:  If you plan to overclock and use your Ryzen 7 1800X for various demanding multithreaded applications, you'll absolutely want adequate cooling or you'll find your CPU throttling itself back, costing you precious time.