low 4k queue depth 1 read performance across all ssd's

[AKouf]

Hello...I'm an ssd enthusiast and I would like to ask a question many people will also have. Why in every ssd and I mean from across all brands including Intel the 4k queue depth 1 read performance is so low comparing with the 4k qd1 write performance? And that is apply with hdd's too...when i crystal disk mark them the read performance is worse than the write performance! From what I read and understand is writing to a page file take longer time than reading from it isn't that right? And the thing is the 4k queue depth 1 read performance is more important factor than write as far boot times, application loading and overall user experience for light usage scenarios most consumers are dealing with (not enterprise or server usage). As far as I know until now is a typical consumer sees queue depth's from 1 to 5 (Anandtech) and the most important is qd1 for reads as for games and application loading is occurring the most from the 5 queue depths (Tom's Hardware). I own 3 Intel ssd's (2 x 320 series and one X-25m) and one samsung and I always benchmark with crystaldiskmark...4k qd1 read's are around 20mb/s or 5000iops when the writes are double that number at about 40mb/s! This is not happening when you benchmarking with qd32 as from what i see is that read performance is higher than write! Write performance from todays ssd is more that 100mb/s at qd1 when read is about 35mb/s at most and that's how I don't get it...isn't easier for the ssd controller to read at qd1 than write? When we will have ssd's with 100mb/s 4k qd1 reads which for me read and write at 4k qd1 is the most important factor for overall performance...it's not the big numbers that manufactures release in their specs sheets like the 4k queue depth 32 performance. So that's my question...I would appreciate very much if an Intel representative can answer that question...thank you

[Jose_H_Intel1]

SSD's are built using multiple NAND flash channels attached to a controller. With higher queue depths, the controller can optimize the order of the operations by executing all the operations for a particular channel all at once (even though the operations may not be in the order received by the controller).

Reads are more highly optimizable than writes so the performance numbers for reads go up faster than writes at higher queue depths.

It's become a de facto industry standard to present performance numbers using a queue depth of 32. Since all manufacturers follow that standard it allows customers to compare 'apples to apples'.