Can someone explain random read to me?

[ensign_lee]

For instance, looking at the Samsung 830 vs the Crucial M4, I see 80k 4k Random IOPS read vs 40k. I would then think that random reads would be twice as fast on the 830 vs the M4.

And yet, when I look at anandtech's reviews, that doesnt' seem to be the case? http://www.anandtech.com/bench/Product/532?vs=355

What gives?

 

[BrightCandle]

Wouldn't be the first time that a manufacturer wasn't all that honest about the practical performance of their device.

Apparently Samsung's drive can only just reach 41k IOs read and the M4 is pushing more like 56k. I see no evidence in there that either figure is right.

 

[razel]

Each manufacturer benchmarks differently. When they are benchmarked with identical settings you'll get a picture that anandtech will paint.

 

[jwilliams4200]

The queue depth makes a big difference. Anand usually measures at QD=3. The manufacturers usually use QD=32 or 64. Probably the most important number is for QD=1, such as you can see with AS-SSD (as well as QD64).

 

[ensign_lee]

So basically what I thought was something I could compare across SSDs is not.

WTF; @(^@*(^&@%&@*&*#%&#%^ Srsly?! 😵

Stupid manufacturers...

 

[dagamer34]

So basically what I thought was something I could compare across SSDs is not.

WTF; @(^@*(^&@%&@*&*#%&#%^ Srsly?! 😵

Stupid manufacturers...

Nothing a manufacturer gives you will ever be that comparable against other products because every benchmark can be gamed in one way or another. However, the only way to really "game" all of the benchmarks at once is to build a better device.

 

[BrightCandle]

This is the reason external review sites exist. Independently testing in a fair and reasonable way with the method published so that they can be repeated by others and shown to be accurate. Its the scientific method verses the marketing methodology!