Sign in with
Sign up | Sign in
Your question

Write speed issue with SSD

Last response: in Storage
Share
September 18, 2012 5:35:35 PM

Hello everyone, I have a problem and Im not sure if its normal or not..

I have a Corsair Force GT 120GB SSD. Recently, I ran 2 different benchmarks to test the manufactures claim on write/read speeds.

The first program was called CrystalDiskMark and my write speeds were at 145 mb/s, read speed was at 495 mb/s..the write speeds here are way below what Corsair is telling me. They claim 550 mb/s write.

The second program was called Atto benchmark and my read speed here was 550 mb/s and the write speed was at 525 mb/s...why am i running higher write speeds with ATTO and not CrystalDiskInfo? Does this sound correct to you guys?

Thanks for the help :) 

More about : write speed issue ssd

September 18, 2012 5:48:23 PM

I use atto and it has always been correct for me, i have never used crystaldiskinfo so i cannot elaborate on it.
m
0
l
September 18, 2012 5:53:56 PM

Ive read somewhere that Crystaldiskmark uses a kind of data that is non compressible and that could be the reason that benchmark gives lower write speeds, with ATTO, everything seems to be fine.

I just want to verify that low write speeds are the norm for Crystaldisk.
m
0
l
Related resources
Can't find your answer ? Ask !
a b G Storage
September 18, 2012 6:02:09 PM

always remember that the insane 400+MB/s is from compressible data, and is not indicative of normal performance.
For normal use SSDs are not all that much faster than traditional HDDs for sequential reads and writes (well, they are faster... just not 300+MB/s faster as adds would have you belive). What you gain from SSDs is low seek time, and no real performance hit from non-sequential workloads, where a HDD slows to a crawl when you give them non-sequential work to do.

Keep in mind also that writes are typically quite a bit slower than read times, which is OK because 9 times out of 10 you are reading from the disc and not writing to it.

ChrystalDisk is more likely to be more 'true to life' performance, where ATTO is all compressible throughput, which is not at all a reflection of reality.
Think of it this way: ATTO is a drag race running in a straight line, where ChrystalDisk is more of a coastal highway race with a few twists and turns.
m
0
l
a b G Storage
September 18, 2012 6:07:04 PM

On my own computer, the minimum read performance (which is normal non-compressible data) is at 375MB/s, which when divided by my 2 drives is only 187MB/s per drive. But for compressible information I am maxing out the throughput of the interface.

There is a lot of clamoring about getting a faster interface to keep SSDs from bottlenecking, but really they could see a lot more improvement in real-world throughput before we even get close to needing faster interfaces.
m
0
l
a b G Storage
September 18, 2012 6:13:06 PM

Also, I am not sure how well Trim technology addresses the issue, but from what I recall SSD's eventually lose performance over their lifetime because of reads and writes, and thus, running bench marking software can actually be harmful to the SSD :/ . Just throwing it out there so you are aware there may be issues!
m
0
l
a b G Storage
September 18, 2012 6:23:22 PM

chugot9218 said:
Also, I am not sure how well Trim technology addresses the issue, but from what I recall SSD's eventually lose performance over their lifetime because of reads and writes, and thus, running bench marking software can actually be harmful to the SSD :/ . Just throwing it out there so you are aware there may be issues!

The SSD performance hits are 1) when the drive is just too full, or has not had time to background garbage collection (like TRIM) to kick in and do a final eresure, and 2) when the drive has had a lot of rewrites and the media starts failing. For normal to moderate use it is not really an issue for the garbage collection, and for media derogation even mid-level flash memory (like what is used in the Agility 3 drives I use) will likely have a longer lifespan than a traditional HDD (the longevity of the rest of the parts on a cheap SSD are an entirely different matter lol).

The biggest thing to avoid is keeping the entire drive pretty full, and then rewriting the same portion of flash over and over again (known as hammering). This kind of situation can prematurely damage an SSD, but it would still be pretty difficult to do. Just follow the old HDD rule of thumb for performance of not using more than 80% of the drive. The reasoning is entirely different, but it is a good rule to follow for both drive types.
m
0
l
September 18, 2012 7:33:37 PM

Cool, thanks for the info everyone :) 
m
0
l
!