Originally Posted by psyside
Let me explain this.
For 4K QD1 reads, from 40 on launch, to 36 now.
4K QD1 writes, 130 on lauch (had 128 on mine) to 116 now.
Acces time, from ~ 30/40 ms to 74.
I know this is not noticeable, but i wonder how come instead of improvements, they downgrade the performance with firmware releases?
Maybe they decided to favor reliability instead of peak performance? i don't know.....but i'm generally satisfied with my drive, never had a single issue, this is not really an serious issue, so yea. I have OCD for peak performance/tweaks so that's why i make big thing out of this stuff, same like when we oc, and chase the very last mhz
In general read speed performance is not affected by the amount of free space available on a SSD unless it is performing GC to accommodate pending writes. So I'm reluctant to blame the free capacity of your SSD for the change in Read Access Time. Also, a single access time number for read or write is a simplification of reality, and I don't know if that is a best, worst, or average value. But it's all we get from AS SSD.
Frankly, looking at your AS SSD benchmarks, I noticed the differences that BradleyW
mentioned. The offset is very different, and in the first one the SSD is identified as "Samsung 840 PRO Seri SCSI Disk De...", and in the second, "Samsung 840 PRO Series", with the same version of AS SSD. While both show "iastorA" as the driver, the driver may be a different version, given the identification as "SCSI" in one and not in the other.
Were those benchmarks done on the same PC, same OS, same IRST driver version? The offset looks like GPT formatted for the first one, and UEFI booting OS drive (541696K) for the second?
No big deal, just asking. BUT, driver differences, mother board differences, empty drive, OS drive, etc, all can and do influence benchmark results. IF the only difference between the first and second benchmark is the firmware (which is assuming a lot) then the firmware might be the cause of the difference. The decrease of ~15MB/s for 4K Write is within the margin of consistency for AS SSD, and also could be caused by an OS being installed and used on the second benchmark.
I would also say that the 840 Pro never quite reached 40MB/s for 4K read, the EVO is the first SSD to do that in AS SSD. Over time all SSDs that are used and not secure erased will be in their steady state performance, which is slower than a new, empty drive.
So IMO it's a bit early to declare the new firmware has increased read latency, but time and other benchmarks will tell.