OCZ Vertex 4 on Areca RAID

[rkagerer]

I'm thinking of upgrading my 4x Intel X-25M SSD RAID0 on Areca ARC-1231ML, to four or eight 256GB OCZ Vertex 4's on ARC-1882ix instead. Already have the controller.

Any thoughts? Is anyone running the Vertex 4's in RAID?

I know firmware version 4 and 5 fixed the problem this SSD had with sequential read speed at low queue depths. It seems better than my current Intel drives in every way, and generally faster than any other consumer SSD's out there, my only lingering concerns are:


Long-term reliability
Still a minor performance issue with reads at small transfer sizes
(Go to http://www.anandtech.com/show/6074/ocz-vertex-4-review-128gb/6 and search for "Still lacks read performance at small transfer sizes")

[Nizzen]

Consider Samsung 840pro instead. Faster read, and it is not OCZ. Testet 4 pcs of V4 on the release, and it did not work very vell on Areca 1880. Noe the firmware is better, but no more OCZ ssd on Areca for me. Have 12 pcs of Plextor m3p in raid-0 and 4 samsung 840pro. And some Crucial M4s in the server on an another Areca 1880.

[rkagerer]

Turns out I can actually get similar pricing on the 840 Pro's. How's the performance you're getting from them in 4xR0 under the Areca card?

[TuKo]

I'll best posting 4x 840Pro R0 on LSI 9271-4i in a few days, keep in touch.

[TuKo]

Here is first benchmark with 4x 840Pro and LSI 9271-4i.

Configuration :
Read Policy : No Read Ahead
Write Policy : Always Write Back
IO Policy : Direct IO
Access Policy : Read Write
Disk Cache Policy : Enabled
Background Initialization : Disabled
Stripe Size : 128k

http://data.imagup.com/10/1174761526.PNG (http://www.imagup.com/data/1174761526.html)

Then I updated to the last firmware and... KABOOM ! :
http://data.imagup.com/11/1174799130.PNG (http://www.imagup.com/data/1174799130.html)

[Nizzen]

Nice results!

Can you please run Anvil with 12GB testsize? :up:

Thank YOU!

[felix_w]

Nice results!

Can you please run Anvil with 12GB testsize? :up:

Thank YOU!

9271 has 1GB cache if i remember correct, is it necessary for non-cached numbers to run that big testfile ? 2GB not enough ?

[mattkosem]

Oh noes! That firmware update really tanked QD32 write performance, and somewhat reduced QD32 reads as well.

--Matt

[TuKo]

http://data.imagup.com/11/1175069560.PNG (http://www.imagup.com/data/1175069560.html)

[Andreas]

I built a rather intense system (32 core, 256GB Ram, 6xLSI, 48xSamsung 830) last year. Before I finally chose components I tested 12 different SSDs, among them the V4. For my type of workload, the drop in write performance depending on fill factor made it almost impossible to create a predictable environment. Sometimes, mostly for the usual benchmarks, performance was superb, but with longer write activities it was the slowest drive.

when you choose to go with the V4's make sure the drives have identical write history. Be aware of the perf drop when crossing the 50% free mark.

regards,
Andy

[rkagerer]

Thanks for the info. I ended up going with the Samsung 840 Pro's and the performance is pretty good. The only thing I forgot to double-check is whether they have any sort of "host power loss protection". When I was doing research one of the items I checked off on the Vertex 4's is that they have a supercap. I've been having trouble finding whether this is true for the 840 Pro's. I was talking to the publisher of Napp-It the other day and he mentioned he doesn't think they do. Can anyone confirm this?