Originally Posted by Sean Webster
That's not necessarily more of a real-world test that you described there. It isn't a real-world workload for 90% of the people out there more than maybe a handful of times in their use. Most don't have a faster source drive, its usually slower, and most don't have workloads that involve moving 100-200-GB at a time. Not saying it doesn't have its place to show performance, but that is usually a rare workload for many. Ex. Initial clone/data migration. More often than not, people opt for a single NVMe SSD and maybe a secondary SATA HDD or SSD at best for bulk storage/games libraries.
What drive do you have? Just wondering about that 15C offset you're talking about.
A real world test, you wanted to test for max speeds. Daily usage varies of course from person to person.
I have the SP P34A80 1TB. https://www.anandtech.com/show/13955...newer-firmware
These E12 drives love to report low temperature based on reviews (comparison of reading vs thermal camera) and my own experience. 8C on start up, well the ambient wasn't lower than probably 23C, so dream on temperature sensor, dream on, while other drives SSD and HDD report 23-26C.
Why buy a fast drive? Well now these E12 based drives cost at least same as old SATA or less (at least when there isn't a shortage and you hunt for best deal for both). There are use cases where having a fast storage is a necessity, such as VMs, Photoshop cache, writing live recordings, processing large bitrate videos, ... if you don't need the speed than the access time isn't much different from SATA to notice.
On something like x99, x299 don't you have enough lanes to use a direct to CPU connection? A 3.0x4 adding card with M.2 NVMe. If you don't already have a direct M.2 port on the mobo.
You should not be getting so varying results. 500MB/s write was it? I'm getting 830MB/s on 2.0x2.
AS SSD: 761/764
CDM: 828/835 MB/s read/write
HWiNFO reports: 800+/800+ on large file read/writes, I don't copy 100GB, 200GB large folders/files directly but over 1 GB RAM cache with verification that transfer was correct. The drive burst writes max speed and that hash check isn't CPU bound so it reads at max speed.
Never saw reviewers complain that these test tools were inaccurate and inconsistent in speeds they report. They wouldn't keep using them really, plus there are also options for multiple reads/writes to average the results.
The SM2262EN don't like being close to full, they probably use SLC quite a bit to boost performance, but even latencies go up when full, go figure with those. Some places they aren't that expensive but elsewhere they can be.