SSD Raid 0 performance

[Killa|3yte]

I just got 2x512gb SSDs set up in Raid 0, and doing some quick informal benchmarks doesn't seem like the speed I was expecting/hoping for. Perhaps my expectations were unrealistic, but I'm just trying to determine if I'm missing some drivers (Win 7 x64) or made some other rookie mistake?

These are the SSDs I'm using: http://www.newegg.com/Product/Product.aspx?Item=N82E16820148821

Here is a benchmark I ran. Do these numbers seem reasonable for Raid 0 SSD?
http://puu.sh/fw4ib/b9f8d5979b.png

 

[drescherjm]

Looks better than I would expect. I mean your 4K writes show a nice 200+MB/s

Perhaps my expectations were unrealistic

raid0 does not do much to help performance in most desktop applications since they do not usually transfer GB of data sequentially or at a very high queue depth.

 

[Killa|3yte]

Looks better than I would expect. I mean your 4K writes show a nice 200+MB/s



raid0 does not do much to help performance in most desktop applications since they do not usually transfer GB of data sequentially or at a very high queue depth.

Thanks for the sanity check. After digging into the specs, it looks like my seq read/writes are actually about double the max specs, so I'm surprised to see that sort of scaling. Usually parallel hardware doesn't scale quite so linearly.

I had some work friends run the benchmark on their laptops which have single SSD and these are the results they got:
http://puu.sh/fw9YH/2603a627c1.png
http://puu.sh/fwagX/db47d7c13e.png

So it seems clear at this point that the Raid 0 is configured correctly. Thanks again.

 

[drescherjm]

Both of those laptop scores are very poor compared to a single modern day SSD.

 

[drescherjm]

actually about double the max specs, so I'm surprised to see that sort of scaling. Usually parallel hardware doesn't scale quite so linearly.

Large sequential reads and writes scale well. Small random does not usually scale well.

 

[dr/owned]

Here's what my raid 730's did (480GB model) http://postimg.org/image/jr1f3hvot/

Your results look similar so seems normal. Nowadays though I turn off the write back cache stuff so if my desktop crashes suddenly I don't get data corruption. Trades off a bit of speed for more data safety.

 

[Killa|3yte]

Both of those laptop scores are very poor compared to a single modern day SSD.

Yes, I'm pretty sure they are SSDs, but they must be old or something. Was surprised to see those results.

Here's what my raid 730's did (480GB model) http://postimg.org/image/jr1f3hvot/

Your results look similar so seems normal. Nowadays though I turn off the write back cache stuff so if my desktop crashes suddenly I don't get data corruption. Trades off a bit of speed for more data safety.

Thank you for posting your numbers, gives more data to look at. I have write-back cache enabled. Any horror-stories of having it enabled?

I'm looking at getting a fairly robust NAS incremental backup solution going here since Raid 0 makes me very nervous for obvious reasons.

 

[dr/owned]

Raid 0 on SSDs is a lot more bullet proof than Raid 0 on mechanical drives. The MTBF is just so much higher and there's fewer points of failure. I have 2 SSD raid arrays - a set of 520's and a set of 730's and have never had complete array failure.

I have gotten data corruption with write back cache thing turned on and when I was experimenting with overclocks. Would get P95 crashes and eventually it took some data with it. So unless your system is rock solid I would turn it off.

One thing I learned is to buy expensive SATA cables. The cheapie 99 cent cables are unreliable as hell.

 

[drescherjm]

One thing I learned is to buy expensive SATA cables. The cheapie 99 cent cables are unreliable as hell.

Never noticed that however I have several hundred SATA cables at work since I have built over a hundred machines so I never purchase a SATA cable. SAS to SATA breakout cables yes bout not plain SATA cables..

 

[Tsumi]

Raid 0 on SSDs is a lot more bullet proof than Raid 0 on mechanical drives. The MTBF is just so much higher and there's fewer points of failure. I have 2 SSD raid arrays - a set of 520's and a set of 730's and have never had complete array failure.

I have gotten data corruption with write back cache thing turned on and when I was experimenting with overclocks. Would get P95 crashes and eventually it took some data with it. So unless your system is rock solid I would turn it off.

One thing I learned is to buy expensive SATA cables. The cheapie 99 cent cables are unreliable as hell.

Unlike mechanical drives, it is utterly pointless to use SSD RAID 0 for a regular user just launching programs.

 

[Killa|3yte]

Unlike mechanical drives, it is utterly pointless to use SSD RAID 0 for a regular user just launching programs.

There's also the concatenation feature (2x512GB >> 1TB). Personally, I use audio software that reads 16GB cache files into memory, so the Raid 0 makes sense.

 

[drescherjm]

I use audio software that reads 16GB cache files into memory, so the Raid 0 makes sense.

16GB files. Interesting. I expect applications that assume you have large memory like this are few and far between but raid0 can certainly help and scale very well for that.

 

[Killa|3yte]

16GB files. Interesting. I expect applications that assume you have large memory like this are few and far between but raid0 can certainly help and scale very well for that.

Yes, it's fun loading instruments in under 30s. The cache files are built on your machine using settings you specify. The idea is that you configure to get the best settings you can while still staying within your RAM limit.

A rare edge case for sure though.

 

[Tsumi]

There's also the concatenation feature (2x512GB >> 1TB). Personally, I use audio software that reads 16GB cache files into memory, so the Raid 0 makes sense.

Yeah, I mentioned specifically regular OS usage for a reason. I recognize there are scenarios outside of regular OS usage that benefits from RAID 0 SSDs.