• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

How important are sequential benchmarks?

K

kazaakas

Limp Gawd
Joined
Feb 8, 2012
Messages
144
Hello guys,

Topic title pretty much explains it, after about a month of doubting, I still don't really know which SSD to get. I leaned towards the m4 for a while, then towards the corsair PP, and now towards a samsung 830.

Having read through an uncountable amount of benchmarks, I'm still not entirely sure which statistic affects what. I've mostly been focussing on incompressible random 4k reads.

I want to use the SSD for the general type of stuff: a very fast boot, a responsive system (aka firefox/winamp starting (almost) instantly) and reducing game load times as much as possible.

What statistics do I have to keep in mind for this kind of usage? I suppose random 4k NTFS reads, but at which queue depth? Do higher queue depths only tell me something about the performance when multiple applications are using the disk simultaneously? Does a bit higher write speed offer any significant boost at all when doing stuff like aforementioned? Does sequential speed even matter at all?

The 830 and the m4 (which are the drives I'm leaning to) are both around 205-210€ for the 256GB version here, so price isn't really something that matters.

Can anyone please elaborate a bit on how to read benchmarks?
I've already been through a bit of articles about this but I'm still not sure about the importance of every statistic.

I know in the end the difference is barely noticable, but hey, I still have to order one in the end and I'm not very sure.
 
Sequential benchmarks are important if you plan to copy very large (gigabytes) of files around from one drive (or set of drives in a stripe) to another. Or for large operations like remuxing videos (converting Blu-ray rips to mkvs).

I use an older generation Intel X25-M as my OS drive, and a pair of Crucial M4s in RAID0 for remuxing videos.

If you don't care about manipulating large files, you can pretty much ignore most benchmarks for newer SSDs. :p
 
Looking at benchmarks for disks is kind of like comparing a porche to a bmw m5 or something. They have different numbers, horse power, torque, etc. But you get in and drive both and it blows your standard sedan out of the water either way.

Any SSD is faster than any mechanical drive you have used. Benchmarks are nice, but for your day-to-day unless you're doing something like hosting a heavily used database or VM's that are disk intensive, it won't matter.



Que depth has to do with multiple processes trying to access the hard drive at the same time, and it may not be a good performance indicator unless you're running a bunch of virtual machines and the latency that is added by the que depth is hindering VM performance. Sequential speeds matter, in a sense, but they are generally so much faster that telling the difference between 300MB/s and 400MB/s ... you really can't.

I've bought many M4's and intels and haven't regretted any of them. The intels in particular have a 5 year warranty, which I find appealing. The samsungs are also highly regarded. Every well reviewed SSD should make your system feel sexy again. Buy the cheapest one with the most storage from a company you can feel good about.
 
I like AnandTech's reviews as they have a custom built "real life" type benchmark that focuses on common tasks. My impression is 4k performance is more important for responsiveness than sequential (unless, as stated above, you are copying large files from one SSD to another or writing large files in a remux situation).
 
Hmm okay, I was also considering putting two 128GB's in RAID 0, but as only sequential or multithreaded usage increases, that won't really help me when booting and loading applications and such, right?
 
Yup, that's right! RAID0 won't affect booting and app loading (it may even make it very very slightly slower - not that you can notice).
 
Okay, I think I'm going for the 830 256GB.
Seems like an epic deal for 208€ to me :)
 
From what I have read...

Too benchmark an SSD properly...

You first need to Fill it completely...

Which then kicks in the wear leveling etc smarts of the controller...

Only then will you get a TRUE reading of what the drive is capable of...
As benchmarking a Brand new drive will give you false readings / over what to expect out of the drive in real life usage.

I wonder if Anandtech / ARS etc peform this type of operations before benching a new drive?

I also wonder.... if when sending out a drive for review do the manufacturers send a brand NEW drive to each tester?

Or do they send a drive to say Anandtech.... let them play with it... then FORWARD the drive on to ARS etc?.... and then ARS gets a lower benchmark of the same PHYSICAL drive than Anandtech did....;)

I am betting they fire off a new one to each tester... to make their results as high as possible.

From STEC's paper on benchmarking Enterprise SSD's
<snip>
In addition, the benchmark should evaluate performance not only under optimum conditions, but over the entire life of the drive. Some SSD benchmark scores that appear very fast can be misleading because they do not reflect long-term performance. To be of real value, SSD benchmark tests need to measure the stability of consistent performance over time. This requires preconditioning the SSD under test, accomplished by writing random data patterns to completely fill all NAND blocks and engage the drive&#8217;s wear-leveling and flash management routines. Properly managing data flow and internal NAND will make the benchmark a more useful gauge of SSD performance under real-world conditions.
</snip>

http://www.stec-inc.com/downloads/whitepapers/Benchmarking_Enterprise_SSDs.pdf

.
 
Damnit, I still can't decide between the m4 and the 830, and the price difference is 3&#8364;...
Crucials have a bit better singlethreaded random reads/writes but sequential writes are a lot worse. samsung has better multithreaded reads/writes but if I cared about that I should go raid anyway..
 
kazaakas said:
Crucials have a bit better singlethreaded random reads/writes but sequential writes are a lot worse. samsung has better multithreaded reads/writes but if I cared about that I should go raid anyway..
Click to expand...

In 90+% of real world usage it's unlikely that you will even notice the difference.
 
drescherjm said:
In 90+% of real world usage it's unlikely that you will even notice the difference.
Click to expand...

Yeah, I know, if I had either one of these two lying around I would call it a day instantly, but I have to order one... and they're priced almost the same at about any shop.
 
Oh, man, you should have gone with the Crucial.

Just kidding. Great choice, and you'll be very happy with it.
 
I went from a 7200 rpm drive to an Intel 40gb ssd. It was limited to 40mb/s write speed and 160mb read speed. Write speed was less than half my HD and read speed was only 20mb/s faster. I WAS BLOWN AWAY. It was by far the single best upgrade I ever did. Everything was so much faster. Made my computer fly.
Upgraded to a Kingston HyperX 120Gb SSD, it reads and writes around 500gb/s. Thought I would be blown away again, nope... I was like meh... Windows installed faster, programs installed faster, and the programs on the SSD are a little faster. Of course I can now fit a few games on it which is nice, but as far as speed I doubt the average person could tell the difference.

Moral of the story, when it comes to SSDs go for reliability, random access speed, and price. sequential speed is really a non-issue.
 
Oh lol
Crucial just dropped 20 euro's in price and the samsung dropped 10&#8364;

:')
 
You must log in or register to reply here.
Back
Top