Review on 💻 Intel 600p M.2 80mm 256GB PCIe 3.0x4 3D1 TLC SSD Retail by Jayson Sharma

Don't buy this, there's no point. It doesn't behave any better than a SATA SSD and performance is inconsistent. When I filled the initial 16GB partition with incompressible data, I got a reasonable write throughput of 170MB/s, but after a few gigabytes the performance dropped to 20MB/s for a few seconds, then picked up again, then dropped off again. and so on until I was done recording. After reading some reviews, I realized that this is filling the external cache of the Intel SLC. I just don't know why Intel would do something so incredibly stupid. NO OTHER NVME BRAND has this type of write degradation. After completing the section, I ran my standard random reading test suite. The 600P hardware supports 8 queues, so 16 threads provides a queue depth of 4. I also ran separate tests with 64 threads (queue depth 8), but performance didn't improve, so I didn't include those numbers. I didn't bother with the single-threaded sequential read test, but that part gave such horribly bad results that I ran it as a one-time test too and got about 186 MB/s at a block size of 32 KB. Ridiculously bad too. For example, the same simple sequential read test with a Samsung 960 EVO returns 1200 MB/s. And honestly, this is probably the #1 reason I'm giving the Intel 600P a bad review. You shouldn't even be selling this junk. The last time I was this shocked was when an Intel 750 (large PCIe card) crashed completely when I/O used a block size of 64K or larger. Of course, this baby is cheap, only $110 or so. But the Samsung and Toshiba offerings are only marginally more expensive and perform MUCH better. 256 caps=00100030780100ffnvme2: Модель INTEL_SSDPEKKW256G7 BaseSerial BTPY64430Q5B256D nscount=1nvme2: Запрос 64/32 очередей, Возврат 8/8 очередей, rw-sep map (8, 8)nvme2: Interrupt Coalesce: 100uSnk2 nlkvmeSn: nsk2 12qmentries / 4 qqentries =512 lbacnt =500118192 cap=238GB serno=BTPY64430Q5B256D-1randread /dev/nvme1s1b <blksize> 100 1616 threads - INTEL 600P (16GB area, non-compressible data)blksize aggregate bw------ -------- - - - . The 600P reached its temperature limit during a longer multithread read test and throttled back automatically. To be fair, that's good. It got to 75C, marked it and muted it to prevent overheating (the 600P firmware does this, not the machine or the BIOS). Overheated flash parts can lead to premature failure. This was a read test though, and honestly these devices shouldn't overheat on a read test. I should note that in all of my testing, the only NVMe SSD that ran hotter than the Intel was the Toshiba at 80C and didn't crash or throttle (as far as I can tell). While Intel is doing the right thing, I still can't recommend it due to the myriad of other device performance issues.