The Ryzen platform is still very young. We are seeing improvements on a regular basis. As board partners and AMD iron out any existing bugs, the performance will only improve.
One of the most interesting discoveries about Ryzen was that RAM speed can have a significant impact on performance. For years with Intel’s various processors and platforms, RAM speed didn’t have much of an impact on compute performance. AMD’s new infinity fabric is directly effected by the speed of your RAM.
So I wanted to put this to the test. I was using a dual channel 32GB kit of Corsair Vengeance LPX 2400MHz RAM. I then ordered a 3000MHz kit. I wanted to control as many variables as I could. These RAM kits are practically identical aside from the speed and timings. Now let’s take a look at the test system.
For testing I used the system components featured in my latest Ryzen 7 build inside an S4 Mini case.
Test System Components
CPU | AMD Ryzen 7 1700 | |
MOTHERBOARD | Biostar X370 Mini-ITX AM4 | |
GPU | Zotac GTX 1060 Mini | |
CPU Cooler | Noctua NH-L9a | |
HDD | Samsung 960 EVO 256GB | |
RAM | 32GB Corsair Vengeance DDR4 2400MHz |
Interestingly when I began testing, the Biostar X370GTN mini-itx motherboard would not run RAM over 2400MHz. I had to download and install the most recent BIOS update which added memory support up to 3200MHz.
The BIOS update was only released a few days ago. I noticed it also added voltage stability when overclocking as well. If you are building a system with this board make sure you grab the new updated BIOS.
RAM Testing Methodology
For testing, I put the system through my typical synthetic and gaming benchmarks. I tested everything at stock settings first, then ran all tests with the Ryzen 7 1700 overclocked to 3.625GHz at 1.275v. All tests were run with the newest BIOS for my motherboard installed.
The charts will be broken down into three sections. First you will find synthetic CPU benchmarks, then Synthetic gaming benchmarks, and lastly real world gaming benchmarks.
RAM Timing & Specs
Model | Timing | Tested Speed | Voltage |
PC4-19200 2x16GB | 16-16-16-39 | 2400MHz | 1.2V |
PC4-24000 2x16GB | 15-17-17-35 | 2932MHz | 1.35V |
Full RAM Speed Test Benchmark Video
If you prefer to watch rather then read, you can see the full benchmark video here. Or you can simply scroll down and browse through the benchmarks in written form.
Ryzen RAM Test: Synthetic CPU Benchmarks
Benchmarks – (3.25GHz Stock) | 2400MHz | 3000MHz | +/- |
Cinebench CPU | 1403 | 1410 | +0.5% |
Geekbench (Single Core) | 3914 | 4297 | +9% |
Geekbench (Multi Core) | 20875 | 21643 | +3.5% |
Passmark CPU | 14170 | 15050 | +5.9% |
CPU-Z (Single Core) | 363 | 375 | +3.2% |
CPU-Z (Multi Core) | 3952 | 3957 | +0.2% |
Benchmark – (3.625GHz OC) | 2400MHz | 3000MHz | +/- |
Cinebench CPU | 1571 | 1585 | +0.9% |
Geekbench (Single Core) | 4133 | 4325 | +5% |
Geekbench (Multi Core) | 23044 | 24062 | +4.3% |
Passmark CPU | 15575 | 16104 | +3.3% |
CPU-Z (Single Core) | 412 | 415 | +0.8% |
CPU-Z (Multi Core) | 4448 | 4430 | -0.41% |
In the CPU tests I saw performance increases across the board. At stock on average I saw increases of 3.71% at stock, and 2.32% when overclocked. While these are not massive improvements they are repeatable.
Ryzen RAM Test: Synthetic GPU Benchmarks
Benchmarks – (3.25GHz Stock) | 2400MHz | 3000MHz | +/- |
3D Mark Timespy | 4272 | 4315 | +1% |
3D Mark Firestrike | 11048 | 11157 | +1% |
Valley | 3990 | 4172 | +4.4% |
Heaven | 3070 | 3133 | +2.1% |
Benchmark – (3.625GHz OC) | 2400MHz | 3000MHz | +/- |
3D Mark Timespy | 4339 | 4359 | +0.5% |
3D Mark Firestrike | 11149 | 11347 | +1.8% |
Valley | 4392 | 4605 | +4.7% |
Heaven | 3089 | 3104 | +0.5% |
Other than the Valley benchmark, you will see much lower gains in these benchmarks. The GPU benchmarks don’t rely on the CPU’s performance as much. I saw an average of +1.63% with the Ryzen 1700 at stock speeds, and +1.88% when overclocked.
Ryzen RAM Test: Gaming Benchmarks
2400MHz
Game | Avg. fps | Max | Min |
Metro Last Light | 142 | 276 | 22 |
Tomb Raider | 196 | 230 | 160 |
Batman Arkham Asylum | 101 | 130 | 61 |
3000MHz
Game | Avg. fps | Max | Min |
Metro Last Light | 150 | 282 | 27 |
Tomb Raider | 198 | 230 | 160 |
Batman Arkham Asylum | 103 | 135 | 52 |
You can expect to gain around 2-3fps in games. I would be interested to see what type of increase you could get from RAM clocked at 4000MHz. Metro Last light turned in interesting results, with +8 average fps.
Conclusion
In my testing I found that faster RAM speed does effect Ryzen CPU performance. My tests were definitive and repeatable. You should expect to see even more gains with faster RAM speeds.
Hi Jay, At first, very good post! I have a question for you: The reason why Corsair Memories (PC4-24000 2x16GB – CMK32GX4M2B3000C15) do not reach the 3000MHz and only run at 2993MHz with Ryzen processors is due to: “The AMD Ryzen™ processor does not offer memory dividers for DDR4-3000 or DDR4-3400. Users shooting for higher memory clocks should aim for 3200 or 3500 MT/s.” source: https://community.amd.com/community/gaming/blog/2017/03/14/tips-for-building-a-better-amd-ryzen-system Do you think this will be resolved in the future with a bios update? Thanks for your time, regards from Argentina!
I made a mistake, I meant “… and only run at 2933MHz (not 2993) with Ryzen processors …” regards!
Hey Adrian, I don’t think AMD is going to change their RAM frequencies in the future. Now that most Ryzen motherboards can hit 3200MHz I don’t tihnk it matters too much.