www.supertalent.com The Official Super Talent Support Forum

FAQ MembersRegister

SandForce SSD 4 - IO METER (Full Random Data Pattern)

SandForce SSD 4 - IO METER (Full Random Data Pattern)

Postby SSD-James » Tue Mar 27, 2012 8:54 am

Now, let's talk about more with IO Meter test result.
All of SSD makers were using IO Meter 2008 RC2 for a long while in the past.
But we now start to use IO Meter 110 becuase it has a feature to select the Data pattern.
IO Meter 2006, 2008 were using Pseudo Data pattern which is absolutely not the Random data pattern.
But IO Meter 110 has "Data Pattern Selection" such as:
1. Repetitive
2. Pseudo random
3. Full random

As you can guess, Full random is the right one we have to test with.
Full random does not necessarilly mean "In-compressible"
It is still compressible but much less compressible data than Pseudo Random.
As you can easily guess, there is no Pseudo Random data pattern in the real world.
So, it is absolutely correct to test the SSD with "Full Random Data pattern"

And, let's talk about the SSD status.
If SSD is just bought and it does not have any data in it, it is Clean status.
Most of end customers using PC, Desktop, Notebook, WorkStation,... use the SSD as clean as possible.
How? use the TRIM feature...

But in a mass storage application, it cannot be "Clean"
So, we need to check the "Steady Status" performance.

First, I compared CLEAN status Standard FW and GOLD FW performance difference.
Most importantly, Queue Depth = 1 is the arena you need to look. QD=2, 4, 8, 16, 32 are meaningless.
Here's screenshot of summary.

1. 120GB CLEAN STATUS (Higher number is better)
* I marked BOLD for the important comparisons.
IOM CLEAN.PNG (66.5 KiB) Viewed 3704 times

2. 120GB STEADY STATUS (Higher number is better)
* I marked BOLD for the important comparisons.
* It is more meaningful data to have the test data with the RAID cards to see the differneces.
(I have tested both Standard and GOLD with various raid cards, but the test results are almost same)
The best raid cards with SSDs is LSI 9260-8i or LSI SS 2108 ROC controller base raid card.
I know that LSI released new 2208 controller which is using 1.2GHz clock speed. It is very expensive and it gets too hot. I think we'd better have a little downgraded firmware...
In a long run, HEAT will be resolved.
In that sense, Standard Firmware is much better theoretically.
IOM Steady State.PNG
IOM Steady State.PNG (75.29 KiB) Viewed 3704 times

** this is MASS STORAGE
IOM Mass Storage.PNG
IOM Mass Storage.PNG (334.73 KiB) Viewed 3704 times

From the above 2 testing result screenshots, you may now understand what the differences would be:
Personally and to represent the whole of my engineering team, Standard Firmware is better.
1. Performance is almost same. no difference.
2. From the Endurance perspective, Standard is better. This is major difference between Standard FW and GOLD FW.
3. But in a real world, 99% of SSD makers are using GOLD firmware. Why?
It is the victory of Marketing.
Big and major makers are doing a good marketing to advertise GOLD Firmware is premium/higher performing... bla bla bla... Whatever.

At Next Session, we will talk about the Good and Bad thingies of Sand Force controller base SSD and RELIABILITIES.
Posts: 2076
Joined: Fri Jul 16, 2010 12:41 pm

Return to James SSD Benchmarking

Who is online

Users browsing this forum: No registered users and 1 guest