Tistory View

The CrystalDiskMark is powerful testing tool to measure SSD, USB-Stick, Sd-card and so on. Many users already use this good tool. by growing the size of these devices, 1GiB option is not report actual speed

 

Default 1GiB

Chnage default value

Default value is 1GiB. actual spped, specially for test benchmark, This value must be almost to drive size. but, This testing will spend more times and not good at Flash-memory device such as SSD, so, must be tested with proper size. To measure actual or reasonable speed, I suggest 16GiB for testing. This value eliminate the distortion caused by cache memory.

Test count 3 is suitable for most case.

 

Why?

To improve performance, Devices has cache memory(Don't confuse cache of OS). This cache is very faster than Flash-memory. In case SATA3 SSD, This speed is even higher than bus-speed. When drive receive data, The controller of device save these data to CACHE, and refetch it, and then actually save on flash memory. This behavior make that small size data will be handled maximum speed. so, Test actual speed of flash-memory, Must be use more than cache on testing. This testing can be "benchmark"

 

Actual Test

Testing with 1GiB option

 

Testing with 16GiB option

These two figure report differences. specially  seq. write. In real world, User can be expierenced 1GiB Testing speed, But For someone to transfer Large size file, 16GiB test is reasonable.

16GiB option also has a little distortion. you can see by next figure.

 

On This testing, By USB-Stick is not clean, 1GiB Test result slower than clean device.

 

Behavior of CACHE

tested file size : 4.64GiB

Cache VS Cache Full

 

At pos. 3.52GiB write-speed suddenly reduced, this is not mean cache size. cache size is much smaller than this value. on writing cache will be have spaced already written this is empty space, this emtpy space will be occupied by new transfered data. If Cache has no space to receive data, Speed-Decreasing occured.

 

Read/Write operation works like below. Behavior of cache is different by on/off cache. this figure is when cache ON

 

Cache work flow

When writing, by empty cache(part also available), cache can be occupied high speed, This shows fast performance in intial parts.

 

Last Note

USB-stick Write-speed is important for move data to other. We spend money to reduce time, by distorted benchmark result, I post this report to Testers that report help for reducing time and monety to others to be more accurate result.

Nand-memory is already devided to quad, read/wrie different will be yawn, I don't figure out of coverage by controller's power...

 

Download

 

https://osdn.net/projects/crystaldiskmark/

 

CrystalDiskMark Project Top Page - OSDN

CrystalDiskMark is benchmark software that measures the transfer speed of media data storage drive such as HD, SSD, USB memory, SD card and NAS. Important Notice Both CrystalDiskInfo and Crystal...

osdn.net

 

Testing device

Sandisk Ultra-Flare 32GiB : Most manufactures don't describe write-speed if write-speed is very slower than read-speed.


*1 : Currently No Devices have over 16GiB Cacahe memory. but, Can be exists, Experts only deal with these situation, This value is reasonable for most case.

Replies
Reply Write