OCZ Core SSD review

[8steve8]

http://www.alternativerecursion.info/?p=106

check it out, bump/comment if you like.

 

[scaryfast]

Thank You.

Are there methods of write buffering that would help this disk out?

For instance, hardware RAID ? If so, would software RAID help.

I read that there is acceleration software for helping this disk with writes. Is that possibly true ?

😀

 

[Dorkenstein]

Nice review, very helpful. I really want one of these for a boot drive.

 

[ScottFern]

I went with a Velociraptor........seems like you still get more bang for your buck.

 

[taltamir]

mmm.. so while the price is certainly nice, it is not a velocy replacement.. only their 500$ SSD is.
but this will be a VERY nice and relatively inexpensive upgrade for a laptop...

You said it is YOUR review? can you please add power consumption to it?

 

[scaryfast]

Any idea why it has such a dramatic change in the graph concerning writes when it becomes 28%, or so, full? I saw another bench using the same tool and that one also had that same drop off; where it starts at about 60 to 80 and then suddenly drops off into that wild oscillating patteren that averages, I'm guessing, about 30 for the rest of the way.

 

[frostedflakes]

Originally posted by: scaryfast
Thank You.

Are there methods of write buffering that would help this disk out?

For instance, hardware RAID ? If so, would software RAID help.

I read that there is acceleration software for helping this disk with writes. Is that possibly true ?

😀

Managed Flash Technology (MFT). Unfortunately, licenses are like $125 (the software is really aimed at the server market, not regular users). There are a few log-structured file systems available for Linux, but I don't think they are designed to handle large flash disks.

Hopefully SSD manufacturers will start to implement write buffering on the controller level soon. Not only does it increase performance, but it also increases the life of the flash memory (especially important for MLC). Surely it wouldn't be that expensive, you'd just need a controller designed to do write buffering and then some DRAM.

 

[chizow]

Very nice bench and write-up, saw some interesting tests done in there that I haven't seen in published reviews. So it looks like the OCZ Core series meets their Read specs but misses their Write specs quite badly. Hopefully a firmware update can fix this as these were the first SSDs I had considered buying.

 

[taltamir]

chizow, actually it is because he gave a single figure. They do very well on writing, except they fail miserably in writing specific sized files, where the velociraptor and the otherwise "slower" but twice as expensive SSD they sell both do much better.
So it is better then what he said in some types of writing, and much worse at others...

I don't think that type of performance crash is acceptable to me, it was very tempting before it.
But it is an interesting drive for a laptop now. Very interesting.

 

[frostedflakes]

Exactly, sequential write is what OCZ advertises. This review focuses on random write, which all SSDs are poor at, due to the fact that flash has to be rewritten in large blocks. So for example, writing 2KiB of data requires deleting a 2MiB block and rewriting the entire thing with the old and new data; it isn't as simple as just sticking the 2KiB of data in there. This situation reduces the effective file transfer by 1000 times (i.e. writing 2KiB of data effectively requires writing 2000KiB of data), so it's easy to see why writing a bunch of small files really slows things down. Despite the poor random write performance, SSDs still usually outperform spinning disks because they are far superior in about every other metric. Or at least that's the case with SLC, MLC has higher access times which hurts random read and write.

And one solution (that will hopefully be adopted soon) people mentioned earlier is the use of a write buffer. For example, you could use a high-speed DRAM chip to store write operations until you have 2MiB of data that can be sequentially written to the SSD. Not only can this increase write performance, but it can also significantly increase the life of the flash memory.

 

[chizow]

Actually I was going by the Disk Bench graphs which I believe are random read/writes and historically mirror published drive specs, at least for mechanical drives. Good to know that SSD speeds are for sequential only though, certainly need to factor that in when looking at the published specs.