Question Poor IOPS on 970 EVO PLUS

dippah

Junior Member
Feb 23, 2019
1
0
6
Hi all Samsung Magician is reporting the Random IOPS for my new 970 evo plus is all over the place. On multiple cold boot, cold room tests I have seen random IOPS from 200k read to 250k. write is around 190k to 205k. My read and write sequential appear to be normal ( 3544 read, 3,304 write). I have set every possible configuation I can find and yet still the sketchy IOPS whch is advertised as somwhere north of 500k. any ideas?
 

Attachments

  • test.PNG
    test.PNG
    70.8 KB · Views: 30

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,786
136
The maximum figures are shown when the drive is empty without data, and without having anything written to it before. Also, to reach its 500K IOPS, the EVO needs Queue Depth to be at 128. I doubt there's any application that can take advantage of that.

Consumer queue depth is at QD1 to QD4.

IOPS don't matter squat to the consumer. Your drive is performing fine.
 

Arkaign

Lifer
Oct 27, 2006
20,736
1,379
126
One of the relatively little known things about m.2+nvme is that these types of SSDs are more CPU dependent. In most modern configs this is not a big deal really, but can be an issue say if you have a fancy z370 Mobo and are running three of them lol. X2 drives use 2 lanes of PCIe, x4 uses 4, and Intel consumer Core series only have 16 native to work with. It's where I find Ryzen with the extra lanes help a bit, and Socket 1366/2011 with 40 lanes is just rad at that kind of thing.

This isn't to say that most users would ever be able to tell the difference, or that standard SATA 6gbps is a better way to go. Just that the optimal testbed conditions seen on box descriptions are often not achievable in real world use. It's more of a 'perfect scenario potential'. However, it does mean that it's always better to get the single biggest and fastest m.2 nvme x4 you can afford, then for all other storage a 2.5 SSD and/or bulk 3.5" storage HDD is probably the best plan for consumer class CPUs.

Speaking of hdds, many people don't realize that high capacity SATA hdds are actually pretty fast these days for the larger models. Even with spindle speeds that don't sound particularly impressive. A 2018 8TB 5400RPM 2 platter drive will generally be way way way faster than a ~2014 2TB 7200RPM 2 platter drive for example, due not just to incremental improvement to cache/controller config, but most critically to the vast increase in areal density. If the head area remains ~.8-1 square mm, the amount of data being picked up by the head at any given time is much higher when dealing with a drive with in that example : 4x the density. This applies more specifically to peak transfer rates for larger files, faster mechanisms and of course SSDs are still kings for lots of smaller files and random operations. It's just that many people are surprised when they see relatively impressive MB/sec data transfers between SSD + large HDD or a couple of large spinners. They might remember an old 7200RPM 1TB peaking out around 80MB/sec lol, while say a 10-12TB Barracuda sees north of 250MB/sec.