Vertex LE vs Crucial C300

[Mungri]

OR maybe Asus can somehow sneak out a magic update that make Raid work great on the Marvel ports

One can dream

[Anvil]

Last time I upgraded my UD7 there was an update to the Marvell chip, I havent had the time to test it yet.

[Budwise]

first run of my new 128GB C300 on ICH10 AHCI.

[Vapor]

2R0 Vertex 2 50GB, FW 1.10, RST 9.6.0, write caching disabled, 32kb stripe

Boot drive and steady state (been in use for a couple months now)

[Vapor]

Enabled write-caching (first checkbox only), rebooted 6 times, wrote 100GB of random data, and rebooted again. Looks like a few small changes in performance in both directions, will probably keep it enabled though.

2R0 Vertex 2 50GB, FW 1.10, RST 9.6.0, write caching enabled, 32kb stripe

Boot drive and steady state (been in use for a couple months now)

[jadeskye]

can't wait until i get my C300s =D

[Johnny87au]

first run of my new 128GB C300 on ICH10 AHCI.

Nice scores mate, loving these ssd's.. Raid these babys!

[jalyst]

SATA3 on Intel ICH can't come soon enough.
C'mon Intel, get with the program already!

[Zloyd]

Great to see crucial getting something out thats gotten everyone a little excited If I had to get a drive now I think it would be a c300.

[Ao1]

Warning (not for the faint hearted)
The following screenshot shows in bytes the amount of data read/written on these few tests.
It seems CDM is a bit of a killer wrt writes, these few benchmarks compares to a full weeks worth of writes. (I'm investigating this matter as it writes more than one should think)

I ran AS SSD and it generated 3GB of writes, so that is exactly in line with your writes if you ran it twice. I thought the test file size on CMD 3 would be the size of the chosen selected test size (?) but looking at the size of the writes that would not even work out if 1GB of files were written for each subtest

Also I guess if you use CMD on default and then on 0fil it would be possible to see how much data the controller compresses by looking at the reduction in the writes that are seen by the OS

EDIT:
Also I noticed that your private memory set use for AS SSD was significantly higher than that of CMD

[Mungri]

SATA3 on Intel ICH can't come soon enough.
C'mon Intel, get with the program already!

But likely a new motherboard and socket, and if it isnt triple channel memory, those of us with 6 modules are screwed.

[Ao1]

Yep, it looks like a really expensive upgrade is round the corner. Don't know if I should be happy or sad. (Happy that it seems like quite a big all round jump in performance is on its way, sad that it will cost an arm and a leg)

[Praz]

Also I guess if you use CMD on default and then on 0fil it would be possible to see how much data the controller compresses by looking at the reduction in the writes that are seen by the OS

The operating system is blind to the compression that takes place on the NAND side of the FTL.

[Ao1]

Are, so the OS registers the full write even thought it has not actually be written to the SSD?

[Praz]

Are, so the OS registers the full write even thought it has not actually be written to the SSD?

Yes. The operating system know no different.

[Vapor]

Anybody have steady state C300 R0 performance numbers? From what I've seen from C300 sans TRIM, things aren't too pretty, but maybe that's an extreme case?

[Anvil]

I think that is an extreme case, it might be possible to recreate in real life but I doubt it.
They don't degrade more than than other SSDs I'd say.

I've been using my C300s (256GB) in raid-0 for months, no visual signs of degrading performance at all. (running 3-4 VMs, a mix of W2008 Server R2, W7)

I've prepared them for benchmarks on the Areca 1880 and I simply forgot to benchmark them before cleaning.

mbreslin might have some drives in steady-state?

[Anvil]

4R0 Vertex 2 on the 9211-8i (software raid, non-bootable)

as-ssd-bench ATA OCZ-VERTEX2 28.08.2010 00-27-07.png

atto_4R0_Vertex2.PNG

CDM (default left, 0-Fill right)
cdm_4R0_Vertex2_60GB_swraid_default.PNG cdm_4R0_Vertex2_60GB_swraid_0-fill.PNG

Pretty great results, great for running VMs, games,...

[Computurd]

yes, very very nice, compared to the charts we put up way back when..with tilt and yours and mine...lemme go necro-thread that...



looking back through those threads to find that brings up a lotta good testing we have done good stuff!

you should post up a revised chart, that would blow those out of the water!

[Anvil]

Yes, I'll do that!

It's about time to update those charts

[jalyst]

But likely a new motherboard and socket, and if it isnt triple channel memory, those of us with 6 modules are screwed.

Oh well just sell-up and move on to the next platform.
No matter what you'll have to pay extra $ for the privilege.
But at least you can make a little bit back from the sale.

[damha]

Enabled write-caching (first checkbox only), rebooted 6 times, wrote 100GB of random data, and rebooted again. Looks like a few small changes in performance in both directions, will probably keep it enabled though.

2R0 Vertex 2 50GB, FW 1.10, RST 9.6.0, write caching enabled, 32kb stripe

Boot drive and steady state (been in use for a couple months now)

I am surprised you are using 9.6.0. 9.6.2(+?) gives better reads with trim.

[Gilgamesh]

What is best ssd's raid 0 pair for sb850?
I can get corsair f60 or realssd c64 for same price.
Or, 4x f40 for a little premium price, but how scale performance with that sb? It' waste of money?
Thanks

[Anvil]

It does scale on the SB850, I've done a few tests using Vertex 2 60GB on the ASUS CHIV.

I'll see if I can get to those screenshots/benchmarks later today.

As the SB850 fully supports 6Gb/s drives, I'd probably go for the C300s.

[Gilgamesh]

It does scale on the SB850, I've done a few tests using Vertex 2 60GB on the ASUS CHIV.

I'll see if I can get to those screenshots/benchmarks later today.

As the SB850 fully supports 6Gb/s drives, I'd probably go for the C300s.

many thanks.