After multiple attempts, the fastest result was a little over 4.5 GB/s, falling just short of 4.0 GB/s after about the 128 GB mark (total file transfers were 173 GB, which is a large 7 was the total size of the folder). zip archive I was copying). Even though I set it to Gen5 and used HWiNFO to verify that it was indeed running at the correct link speed, the problem was with his T700 in his M.2 slot on this Z690 board think. The top M.2 slot is supposed to be PCIe 5.0, but I’ve had better results using a PCI-E 5.0 to M.2 adapter card in the top slot (without processor graphics). (the drive is in the bottom M.2 slot).
Sounds verbose, but we had a lot of problems testing these Gen5 drives. Presumably they were all related to a specific test platform. For example, I usually present different graphs with sequential and random 4K results compared at different queue depths, but my test platform under custom workloads with CrystalDiskMark fails to produce consistent results on the T700 was. It happened on both AMD X670E and AMD X670E boards. Intel Z690 board. Usually benchmarks he puts on QD1, then QD2 and so on, but these results are very inconsistent from run to run, especially with QD2 compared to QD1 with an inexplicable setback in performance. was seen
In order to find the best way to test this Gen5 SSD, we went from AMD AM5 platform to 13th Gen Intel platform, from Gen5 M.2 slot to dedicated PCIe 5.0 to M.2 adapter card, and many others. effort was made. I also tried the Micron storage driver instead of StorNVMe by plugging it into a PCI-E 5.0 x16 slot (which helped). Ultimately, trying to run custom benchmarks with varying queue depths and threads yielded inconsistent results and failed to generate meaningful numbers needed for a detailed review. So this is just an “overview”. (I was also unable to retest his two T700 drives on the AM5 using a fork of dual M.2-Gen5 PCIe adapters – not supported on the Z690 board.)
In short, storage is complicated and we miss Allyn.
