I put this in the "AMD" section because this applies to multiple different motherboards and NVMe drives I have tested! Also, please use this thread for similar interesting results!
https://www.reddit.com/r/Amd/comments/5yua7z/what_are_your_ryzen_nvme_pcie_benchmark_results/
The first post shows with the Samsung NVMe driver he got around 1900 MB/s, and without he got much higher. I have the opposite occurring! Seems to me like Ryzen chipset drivers and how they handle bandwidth from the m.2 slot is kind of messed up with its communication between the OS and the hardware.
I also noticed someone there posted the same drive as me which is a Mydigital BPX 480GB NVMe drive also with 1900 MB/s read. This drive uses a Phison controller, and I also believe so does the Corsair MP500, but I've seen benchmarks online with that drive running as advertised!
If you look at the above benchmark, I am also getting slightly above 1900 MB/s read and much lower write, like on the linked reddit page. Yet, the compression benchmark shows something different for us! I also noticed similar results with ATTO showing around what we would like to see, but crystal disk info shows much lower speeds than what we would expect! Seems to be the case that with people who use a Samsung NVMe drive without the driver you do get the results as desired. I can't attest to this because I don't have a Samsung drive handy right now. I did see however other sites like TweakTown say they did use the driver and everything is fine!
I also noticed that you must check both options under device manager under properties for your drive. You must check write cache / and Flush, otherwise, the write performance will tank to abysmal numbers. I noticed this mostly on AS SSD tests!
Has anyone else here seen similar issues across Ryzen. The tests above were on a Crosshair Hero x370 board. As far as I can tell this seems to be some driver / OS / Program related issues. I say this because I have everything set in the BIOS as it should make sure all bandwidth is available to that only m.2 slot. I have no other drives plugged in, and I've done my research and know that in the manual if you populate some SATA slots, or add additional cards it can take away from the slow and cause Gen 2 speeds. I'm only using my GPU and have nothing else causing the slot to run slower! I have the latest BIOS as well, so I'm a bit at a loss here as to why the drive is running slower.
Additionally, I tested the drive on my Z270 test bench and had ZERO issues with the drive running at the rated speeds, so the drive isn't faulty. I also tried multiple drives on the HERO board with similar results where the driver ran a little slower than expected. The only thing I haven't been able to test has been a Samsung drive on this board. I wonder if anything other than approved drives cause bottlenecks or driver related issues through the default windows NVMe driver?
What is very strange here is the lack of consistency across benchmarks, I understand they all do different tests, but we shouldn't be getting this level of variance across tests, even within the same damn program. I am baffled that crystal disk and AS SSD don't show proper results, yet ATTO and the AS SSD compression test show what we need it to! I think these programs need to be updated somehow to reflect this!
https://www.reddit.com/r/Amd/comments/5yua7z/what_are_your_ryzen_nvme_pcie_benchmark_results/
The first post shows with the Samsung NVMe driver he got around 1900 MB/s, and without he got much higher. I have the opposite occurring! Seems to me like Ryzen chipset drivers and how they handle bandwidth from the m.2 slot is kind of messed up with its communication between the OS and the hardware.
I also noticed someone there posted the same drive as me which is a Mydigital BPX 480GB NVMe drive also with 1900 MB/s read. This drive uses a Phison controller, and I also believe so does the Corsair MP500, but I've seen benchmarks online with that drive running as advertised!
If you look at the above benchmark, I am also getting slightly above 1900 MB/s read and much lower write, like on the linked reddit page. Yet, the compression benchmark shows something different for us! I also noticed similar results with ATTO showing around what we would like to see, but crystal disk info shows much lower speeds than what we would expect! Seems to be the case that with people who use a Samsung NVMe drive without the driver you do get the results as desired. I can't attest to this because I don't have a Samsung drive handy right now. I did see however other sites like TweakTown say they did use the driver and everything is fine!
I also noticed that you must check both options under device manager under properties for your drive. You must check write cache / and Flush, otherwise, the write performance will tank to abysmal numbers. I noticed this mostly on AS SSD tests!
Has anyone else here seen similar issues across Ryzen. The tests above were on a Crosshair Hero x370 board. As far as I can tell this seems to be some driver / OS / Program related issues. I say this because I have everything set in the BIOS as it should make sure all bandwidth is available to that only m.2 slot. I have no other drives plugged in, and I've done my research and know that in the manual if you populate some SATA slots, or add additional cards it can take away from the slow and cause Gen 2 speeds. I'm only using my GPU and have nothing else causing the slot to run slower! I have the latest BIOS as well, so I'm a bit at a loss here as to why the drive is running slower.
Additionally, I tested the drive on my Z270 test bench and had ZERO issues with the drive running at the rated speeds, so the drive isn't faulty. I also tried multiple drives on the HERO board with similar results where the driver ran a little slower than expected. The only thing I haven't been able to test has been a Samsung drive on this board. I wonder if anything other than approved drives cause bottlenecks or driver related issues through the default windows NVMe driver?
What is very strange here is the lack of consistency across benchmarks, I understand they all do different tests, but we shouldn't be getting this level of variance across tests, even within the same damn program. I am baffled that crystal disk and AS SSD don't show proper results, yet ATTO and the AS SSD compression test show what we need it to! I think these programs need to be updated somehow to reflect this!
