Is my SSD slow?

M

mjohn1378

Limp Gawd
Joined
Jun 11, 2002
Messages
381
I just purchased a 60GB Vertex2 and was wondering why my CrystalDisk speeds seem right,

I am getting the following
R | W
SEQ 203.30 | 63.47
512K 191.80 | 63.26
4K 21.44 | 59.08
4KQD32 80.90 | 66.64

What I find interesting is that my 70% full WD Black edition beats the SSD @ SEQ writes.

R | W
SEQ 68.470 | 69.120
512K 31.400 | 52.580
4K 0.628 | 1.603
4KQD32 1.459 | 1.615

System:
7 x64 (trim enabled)
P5Q Deluxe
Q6600 @ 2.4 (no overclocking for summer)
2GB DDR2 1066 @ 1066
1X WD Black edition 1.0TB Storage
1x OCZ Vertex2 60GB OS
ATI HD4850
 
Man Crystaldisk made my SSD look like a dog.

ATTO shows that there really is no comparison between the two drives.

Whats up with the difference between the benchmarks (ATTO and CrystalDisk)?

ssdcomp.jpg
 
Seem to remember something from an Anand article about the Vertex 2 where it has some sort of firmware thing which makes it so it doesn't have to write everything when you write a lot of the same data. This means the Vertex 2 will do a lot better on those benchmarks than if the data itself also is random. CrystalDisk is probably writing random data, while ATTO is probably writing identical data over and over again.

In real-life applications neither case is true; the data will never be completely random, nor will it be completely identical. It would be somewhere in between, and so your real-life performance would be somewhere between 60mb/s and 260mb/s.

That's the only thing I can think of.
 
CrystalDiskMark is probably not using multiple outstanding I/O threads. ATTO is using multiple threads (try reducing the queue depth, and see if performance drops).

SandForce sucks for single-threaded I/O, but it's a workhorse when heavily loaded with multiple threads.
 
I was just wondering if it was working properly.
The SEQ write results I gor from CrystalDisk, were well below what I expected from a SSD, and even demonstrated my old HDD to be faster in SEQ writes

Anyway, I ran a few more and here is what I discovered in relation to cue depth.
I ran tests at both ends of the benchmark range for cue depth.

qc.jpg


Que depth obviously has a large impact.
 
Last edited:
ATTO uses easily compressable data; meaning you test your controller chip not the flash NAND storage. CrystalDiskMark uses junk/random data i believe, leading to lower numbers.

If you have a lot of text files on the SSD that compress well, the ATTO speeds apply. But most data is not compressable much if any at all. So for those data the Sandforce controller used in OCZ Vertex/Agility 2 does not write so fast, at all.

You should know that SSDs shine regarding random I/O, not sequential I/O. You don't pay a premium for double sequential speeds. You pay a premium for 100x higher random access performance, which are performance differences so large rarely seen in IT industry.
 
You sure you have setup your SSD correctly ? I am just mentioning that as they really do need to be setup properly for best effect, unlike a normal hdd that u just plug in and make a partition and off you go.

These things need to be setup correctly, proper offsetting etc and then the best mode in your bios etc

I had mine running like kings on my old mobo, these things flew along and when I started my pc, post was the thing that took the longest, post takes like 15-20 seconds and booting took 5 seconds.

But, that mobo broke and I swapped the ssd's into a similar mobo, only difference being that this mobo has 32 lanes for sli instead of 16 in the previous mobo, apart from that the chipset is the same, basically everything else is the same on the mobo as it was on the old dead mobo, but even after formatting my SSD's and starting over, in this mobo my ssd's run like dogs.

Your ATTO benchmarks, look decent enough, I wouldnt complain as you can clearly see the massive difference in speeds betwen hdd and ssd, you might be able to tweak it or your OS to get that touch more performance out of it.

I got fed up having to tweak them all the time, as mine didnt have trim or GC to start with and I was tweaking them constantly, a new firmware came out that sorted it out, but without trim or GC, they slowed down really quickly and I had to run freespace cleaner every week just to keep the things running in tip top condition, I still run FSC now as trim doesnt work with raid and GC takes forever to work but I dont tweak anymore as it got anoying, aslogn as they dotn slow to a crawl, I will be happy, just the weekly FSC and thats all I do to them now.

The OCZ forums are a wealth of info regarding ssd's, you will get al the info you need there to decide if your ssd's are having problems or if you think you might be able to squeeze that little bit extra out of them etc.
 
EVIL-SCOTSMAN said:
But, that mobo broke and I swapped the ssd's into a similar mobo, only difference being that this mobo has 32 lanes for sli instead of 16 in the previous mobo, apart from that the chipset is the same, basically everything else is the same on the mobo as it was on the old dead mobo, but even after formatting my SSD's and starting over, in this mobo my ssd's run like dogs.
Click to expand...

is there a step by step guide for these tweaks? What did you do to fix the SLI lanes?
 
Personally I don't like ocz ssd's. I had one and it was slow, it wasn't at all as fast as advertised. I had the vertex le with the sandforce controller. Atto said it was fine but the drive had an advertised sustained write speed of 235mb but I would get an average of sustained 95mb write speed in real file transfers. I contacted ocz and they said they use atto to measure the max, and Io meter to measure the sustained but that isn't what I would get in real file transfers. Personally I think they are lieing with those numbers that they list but some people think its fine they use benchmarks to advertise their speeds.
 
Happy Hopping said:
is there a step by step guide for these tweaks? What did you do to fix the SLI lanes?
Click to expand...

changed from a 16 lane sli motherboard to a 32 lane sli motherboard

head on over to ocz forum for lots of advice on how to setup a ssd correctly, basically if you are running windows 7 and you make the partition on the ssd during the win7 install procedure, it should be setup correctly, but if you use any other OS, then you will be better manually setting up the ssd as their is some weird shit that has to be done to get the best out of it.

Munkle, I think it really all comes down to the motherboard you are runnign them on and the way they are setup, on my last motherboard they were smoking fast, but on thsi one where the only difference is it has more lanes , 32 compared to 16 on the old mobo, they run like shit and the mobo wouldnt see both ssd's when connected, it would only ever see one until I flashed the firmware on one and was running 2 different sets of firmware on my ssd's, that way the mobo would see both but they still dont run nowhere near as fast as they were on the old mobo.

So I think it all boils down to which motherboard you hook them upto, some mobos and ssds play good with each other and some dont.
 
Last edited by a moderator:
You must log in or register to reply here.
Back
Top