Iops?

[BRENNANX]

Hello to everyone who may see this probably uninteresting question,but I must ask it still.When running Passmark,the final test (as everyone knows) is the disk benchmark and the two final tests for the disk are IOPS 32... and IOPS 4K may i ask what are they/And why would my 970evo do so miserably in them (compared to other 970 evos?Thankyou for your time and hi Dave(I presume Dave will answer as usual)....BRENNANX (RSA)

[David (PassMark)]

IOPS means Input / Output Operations per Second.

From the Help file,

IOPS 32KQD20 A large test file is created on the disk under test. The file is read randomly; a seek is performed to move the file pointer to a random position in the file, a 32KB block is read or written then another seek is performed using a queue depth of 20. The amount of data actually transferred is highly dependent on the disk seek time (previously called Random Seek). IOPS 4KQD1 Similar to the IOPS 32KQD20 test but uses 4KB blocks and a queue depth of 1.

Queue Depth (QD) refers to how many operations are "queued" up by the application accessing the disk. Most applications (and their programmers) take the easy approach and always used QD1. Some sophisticated apps, like databases and web servers will be doing a queuing however. Queuing allows the disk & it's device driver to reorder & merge the open IO requests to optimise the access speeds.

As to why your drive is so bad, there is no one answer. Start with this page
https://www.passmark.com/support/per...erformance.php

[BRENNANX]

THANKYOU Dave...

[TehCyberJunkie]

Semi-on topic, I'm starting to see some baselines (#1553969 & #1561345) get posted up with IOPS 4KQD1 scores of over 1,000MBytes/Sec! These numbers are blowing my mind because I'm in the final stages of designing our next database server where maximizing this specific stat is absolutely critical - and I have no idea how it's possible for those numbers to even exist. Can anyone shed some light on how it was done?

[Richard (PassMark)]

Semi-on topic, I'm starting to see some baselines (#1553969 & #1561345) get posted up with IOPS 4KQD1 scores of over 1,000MBytes/Sec! These numbers are blowing my mind because I'm in the final stages of designing our next database server where maximizing this specific stat is absolutely critical - and I have no idea how it's possible for those numbers to even exist. Can anyone shed some light on how it was done?

The old saying likely applies, "If it is looks too good to be true, it's probably not." We'll take a look at those baselines you mentioned. Likely, some disk cache (e.g. https://forums.passmark.com/performa...msung-evo-ssds) is being utilized to obtain those numbers. There is some detection for disk caching but a few still sneak though. If a small set or single baseline is straying too far from the average then it probably the case.