Vertex 4 F/W 1.3 vs 1.4 across ich10r, pch, Areca 1261 and 1880 and ... [Archive]

SteveRo

06-05-2012, 04:18 PM

Good evening all,

Just getting started here - looks like drives were delivered with F/W 1.3 -

http://img252.imageshack.us/img252/650/dsc05387rw.jpg (http://img252.imageshack.us/i/dsc05387rw.jpg/)

http://img268.imageshack.us/img268/2839/dsc05389k.jpg (http://img268.imageshack.us/i/dsc05389k.jpg/)

http://img507.imageshack.us/img507/8885/disk1and2cdicrystaldisk.jpg (http://img507.imageshack.us/i/disk1and2cdicrystaldisk.jpg/)

:)

SteveRo

06-05-2012, 04:23 PM

Vertex 4 128GB (f/w 1.3) initial runs on 2600k/m4e @ stock - on pch - (2xR0 is at 64K stripe (thanks JR)).

http://img507.imageshack.us/img507/7016/onevertex4ahcionpchfw13.jpg (http://img507.imageshack.us/i/onevertex4ahcionpchfw13.jpg/)

http://img254.imageshack.us/img254/5599/2xr0vertex4pchwbconfw13.jpg (http://img254.imageshack.us/i/2xr0vertex4pchwbconfw13.jpg/)

That's all for tonight the basketball game is on - go Celtics!! :)

Computurd

06-05-2012, 07:04 PM

nice stuff, man i wish i had a few of these to play with! these new marvells are UBER...

Anvil

06-06-2012, 01:10 AM

Marvell or marvel :)

I ordered last night, normally takes 2-3 days, could be here Friday or Saturday.
(I did go for the 256GB, could be an interesting compare)

SteveRo

06-06-2012, 02:37 AM

^^ Nice!, yes that will be a good compare.

SteveRo

06-06-2012, 04:07 AM

I will post some more results in a couple hours - looks like the vertex 4 has a conflict with the areca 1880.
Tried 2xR0, tried as pass through, tried with ncq and all read/write aheads turned off - same result - tries for a while then drops the drive. :(
Lets hope this is fixed in 1.4 :)

http://img560.imageshack.us/img560/5346/areca1880v4weird.png (http://img560.imageshack.us/i/areca1880v4weird.png/)

Anvil

06-06-2012, 07:55 AM

SteveRo

06-06-2012, 08:08 AM

^^ sounds good - below is 1x Vertex 4 (f/w 1.3) on ich10r non-Raid (should act the same as ahci, giga x58 exteme with 920@4) -

http://img703.imageshack.us/img703/3484/1xvertex4nonraidich10rf.jpg (http://img703.imageshack.us/i/1xvertex4nonraidich10rf.jpg/)

SteveRo

06-06-2012, 08:09 AM

Below is 2xR0 Vertex 4 (f/w 1.3) on ich10r, wbc on (64K stripe, giga x58 exteme with 920@4) -

http://img210.imageshack.us/img210/8985/2xr0vertex4ich10rwbconf.jpg (http://img210.imageshack.us/i/2xr0vertex4ich10rwbconf.jpg/)

SteveRo

06-06-2012, 08:13 AM

Below is 2xR0 Vertex 4 (f/w 1.3) on Areca 1261 2GB cache (4k stripe, 2600K on m4e no oc) -

http://img43.imageshack.us/img43/6417/2xr0vertex4areca12612gb.png (http://img43.imageshack.us/i/2xr0vertex4areca12612gb.png/)

trans am

06-06-2012, 08:24 AM

nice results steve.

Anvil

06-06-2012, 08:33 AM

The old Areca is doing just fine :)

Are you cleaning the drives between each run and are you consistently using small partitions?
(I'll duplicate the tests as close as possible)

Nizzen

06-06-2012, 08:54 AM

Nizzen had issues on fw 1.3, I don't think he got to try 1.4. (impatient :))

Did it fail as a single drive as well?
(in that case I could try one of mine as mine are already on 1.4)

I don't think it's worth playing much with 1.3 though.

It looks like Plextor M3P is the closest competitor to the V4 128GB.
I'll post some benchmarks I ran last night on my X79 a bit later. (2R0 Plextor M3P 128GB on fw 1.03)

I had problems with 4 V4s. No problems with 2 as far as I remember. I did not try 3... Problem with 4 v4s was confirmed by OCZ. And as far as I know, the problem is not fixed.

SteveRo

06-06-2012, 12:38 PM

The old Areca is doing just fine :)

Are you cleaning the drives between each run and are you consistently using small partitions?
(I'll duplicate the tests as close as possible)

Yes, I am using 10GB partition sizes, standard 4k cluster but not cleaning.
Waiting 5 minutes or more between tests but not cleaning.
More power to you Mr. Anvil if you are cleaning between each test! :clap::clap:

SteveRo

06-06-2012, 12:44 PM

Upgrading from f/w 1.3 to 1.4 either works or it doesn't.
I could not get the toolbox to update on x58/ahci at all - it would say it was done and successful but after reboot - still 1.3.
On P67 I had no problem, each drive updated the first try.

Looks like updating is worth it though - here is 1x Vertex 4 128GB f/w 1.4 on pch, ahci -

http://img846.imageshack.us/img846/9558/1xvertex4ahcipch2600km4.png (http://img846.imageshack.us/i/1xvertex4ahcipch2600km4.png/)

:) :)

Anvil

06-06-2012, 01:29 PM

Yes, I am using 10GB partition sizes, standard 4k cluster but not cleaning.
Waiting 5 minutes or more between tests but not cleaning.
More power to you Mr. Anvil if you are cleaning between each test! :clap::clap:

I was just asking whether you did :), I normally do try to do a quick format or something if I'm testing FOB performance.

Here's the Plextor M3P, looks like the Plextor is stronger on reads and the V4 is doing a little better on writes.
(this was on 2011/RIVE using a 3 series driver so it's not a direct comparison to your testing)

127413

127414

Anvil

06-06-2012, 01:43 PM

This is something I'm playing with right now, it's not set so no point in looking at scores.

It's a Vertex 4 256GB, almost full. (what I call almost full)

127415

SteveRo

06-06-2012, 03:17 PM

^^ extreme edition, interesting!

Below is 2xR0 Vertex 4 (f/w 1.4) 64K stripe - pch - wbc on, 2600K on m4e.

Seeing some interesting drops in Atto - most obvious in 16K and 32K reads - maybe due to ssd internal ram cache not keeping up with the bench? :shrug:

Still though - seems like great performance for just 2x128's. :D

Looks like the new firmware gives up a little in reads and gains a lot in writes? (however, 4MB sequential read stays same or higher)

http://img151.imageshack.us/img151/5574/2xr0vertex464kstripepch.png (http://img151.imageshack.us/i/2xr0vertex464kstripepch.png/)

SteveRo

06-06-2012, 06:49 PM

Looks like the Vertex 4 128GB is still broken on the Areca 1880 even at 1.4 F/W - :( -

http://img836.imageshack.us/img836/6885/v41880broken13and14.png (http://img836.imageshack.us/i/v41880broken13and14.png/)

Config is 2xR0 Vertex 4 (f/w 1.4) 4k stripe with areca settings as follows -

http://img27.imageshack.us/img27/4467/areca1.jpg (http://img27.imageshack.us/i/areca1.jpg/)

http://img546.imageshack.us/img546/2268/areca3.jpg (http://img546.imageshack.us/i/areca3.jpg/)

http://img59.imageshack.us/img59/7762/areca2.jpg (http://img59.imageshack.us/i/areca2.jpg/)

SteveRo

06-06-2012, 07:49 PM

Vertex 4 (f/w 1.4) continues to play well with Areca 1261 -I hope a future f/w update will fix the drives for the 1880 -

http://img59.imageshack.us/img59/6417/2xr0vertex4areca12612gb.png (http://img59.imageshack.us/i/2xr0vertex4areca12612gb.png/)

Anvil

06-07-2012, 12:44 AM

So it's still not playing with the 1880, I'll check the LSI 9265 when I get more drives. (ETA Friday)

Hopefully they'll fix it shortly.

Nizzen

06-08-2012, 08:23 AM

So it's still not playing with the 1880, I'll check the LSI 9265 when I get more drives. (ETA Friday)

Hopefully they'll fix it shortly.

I waited since the first day they came.. And still waiting. It took Crucial 6 months to fix they`re problem, so do not expect faster fix from OCZ :p

Anvil

06-08-2012, 10:27 AM

The new drives were already on 1.4, so it looks like a new batch.

Incompatibility with some of the most popular raid controllers is no good, I'm expecting a quick fix.

Anvil

06-08-2012, 03:59 PM

Some initial tests on the X79, 3 series RSTe

127455

127453 127454

Writes are great, reads are so so...

Anvil

06-09-2012, 02:51 AM

... and the fix might be here already

interim fw 1.4.1.3 (http://www.ocztechnologyforum.com/forum/showthread.php?102159-Interim-Firmware-1-4-1-3-Sleep-and-Cold-Boot-Fix)

A two step update (read the first post in the link)

Improvements since version 1.4:
•Enhancement: Improved RAID card compatibility.
•Fixed: Corner case issue where drive would not resume from S3/S4 properly.
•Fixed: Rare issue where some platforms would not respond correctly after a cold boot.

I'll be updating shortly.

Nizzen

06-09-2012, 12:35 PM

JR.

06-09-2012, 12:48 PM

Hi,

For interest here are a couple of benchmarks I've run this evening on 2 x 256GB V4s in R0, with the latest 1.4.1.3 fw.

127465

127466127467

Some observations I'd like to share based on my personal testing and use of 2 x 256GB V4s in R0:

1. Just like a single V4 my array has not dropped in performance regardless of how hard I bash it
2. I've noticed that a few seconds after deleting a large file (e.g. an 80GB file full of random data) there is a burst of continuous disk activity (HDD LED glows continuosly) - this appears to me to be just the same as is evident for a single V4 with trim pass through in action. It seems to me that the V4 invokes some form of on the fly GC activity regardless of whether trim pass through is in action or not. With V4s in R0 my feeling is there is no need to hold our breath waiting for trim to be enabled for arrays.
3. I've also noticed that the peak of performance on my array does not occur straight after a Secure Erase. I have found that the V4 hits a peak in benchmarks when all blocks have been written to at least once and the pool of available blocks is relatively small. For example, my best results have come when I have filled the drive with random data files and have left just 50GB free and then have written and deleted another 30GB random data test file (all in one Windows session). Of course all this is artificial/synthetic nonsense to hit an extreme result in a benchmark and does not really have much to do with the day to day real world - but as this is the extreme storage forum I thought you may be interested :)

Regds, JR

p.s. here's a link for a handy little test file creation utility http://www.softpedia.com/get/System/File-Management/FileFiller.shtml

Anvil

06-09-2012, 12:59 PM

If you get a lucky run you can get full scaling in raid-0. (at sequentials and high QDs)

I'm playing with strip sizes in W8 on SB.

SteveRo

06-11-2012, 11:20 AM

JR - wow! - very nice numbers! :up::up:

SteveRo

06-11-2012, 12:02 PM

f/w 1.4.1.3 is a little faster on the 128s :)
Next up - see if they work on the 1880 now.

http://img17.imageshack.us/img17/5255/1xvertex4ahcimch2600km4.png (http://img17.imageshack.us/i/1xvertex4ahcimch2600km4.png/)

http://img585.imageshack.us/img585/2272/2xr0vertex4pchstripe64c.png (http://img585.imageshack.us/i/2xr0vertex4pchstripe64c.png/)

SteveRo

06-11-2012, 12:35 PM

Looks like f/w 1.4.1.3 is still broken on the 1880 :( :(

RyderOCZ

06-11-2012, 01:22 PM

I let the devs know Steve.

SteveRo

06-11-2012, 01:30 PM

I let the devs know Steve.

Now I am not sure - I can't seem to format 2xR0 C300/64s now in the same setup - might just be me :( - I will need to research some more!

RyderOCZ

06-11-2012, 01:58 PM

How many Vertex 4 drives are you running, just 2? 128GB, correct?

SteveRo

06-11-2012, 03:53 PM

^^ correct just 2x128s :)

SteveRo

06-11-2012, 04:02 PM

Someone else should confirm but looks like Vertex 4 f/w 1.4.1.3 is still broke on the areca 1880 :(

http://img835.imageshack.us/img835/3371/stillbroke.png (http://img835.imageshack.us/i/stillbroke.png/)

RyderOCZ

06-11-2012, 04:13 PM

The devs and test team have gotten back to me, they have used FW 1.49 on the 1880i with no issues.
64k stripe, all cache settings disabled.

Please list your full config for me.

SteveRo

06-11-2012, 04:30 PM

Mr Ryder sir, thanks for the help - config as shown here - http://www.xtremesystems.org/forums/showthread.php?281374-Vertex-4-F-W-1.3-vs-1.4-across-ich10r-pch-Areca-1261-and-1880-and-...&p=5109283&viewfull=1#post5109283

Also tried with latest areca f/w dated early 2012 - no go :(

I wasn't able to get them to work with all cache settings disabled either - but even if Vertex 4 did work that way - I think most folks would say that would be an unacceptable "perminant" fix.

RyderOCZ

06-11-2012, 05:25 PM

I meant your system config please.. what motherboard, CPU, how much memory, etc you are running.

RyderOCZ

06-11-2012, 06:17 PM

After you updated to 1.4 and then after you updated to 1.4.1.3 did you try secure erasing the SSD's?

Computurd

06-11-2012, 09:07 PM

it is the Areca IX versions that have the problem with marvell controllers that are used on several types of ssds. testing with the I version will not show the problem, as it does not have the expander that is creating the issue.

Anvil

06-12-2012, 01:23 AM

I'll check on mine later tonight.
(need to get to the computer with the Areca)

Will check the LSI as well. (9265)

SteveRo

06-12-2012, 03:20 AM

Good morning Mr. Ryder -

I have tried on 2 different systems with the same incompatibility result -

system 1 - 2600K on P67 asus maximus IV extreme with kingston hyperx KHX2000C8D3T1K3/6GX 2x2GB
system 2 - 920 on x58 gigabyte ex58-extreme with corsair CM3X2G1600C8D 6x2GB

much thanks for the help! :)

SteveRo

06-12-2012, 03:28 AM

After you updated to 1.4 and then after you updated to 1.4.1.3 did you try secure erasing the SSD's?

No just reformat and go - secure erase via vertex tool box?

Hondacity

06-12-2012, 05:06 AM

yehey :D

Anvil

06-12-2012, 01:51 PM

I'm seeing the same thing as Steve and I'm using the Areca 1880IX-16 4G

2R0 Vertex 4 256GB on fw 1.4.1.3

127551

ATTO pretty much shows what's going on in other benchmarks as well. (and so I won't post others)
ATTO works it's way slowly through the steps, very slow.

---

I tried a single drive in PassThrough mode as well

127552

SteveRo

06-12-2012, 02:09 PM

^^ thanks for confirming ...

edit - What Paul said - probably the expander - I think the expander was the reason why acards could only connect to 1880ix at sata1.
How about the 9265?

Anvil

06-12-2012, 02:25 PM

I'm about to test on the LSI, 10-15 minutes

raid-0 created, all is OK so far

NRA = No Read Ahead
DIO = Direct IO
WT = Write Through

127553

I'll change some parameters and run with caching ++
(just to see if there is something weird)

Changed from WT to WB (WriteBack)

127554

SteveRo

06-12-2012, 05:59 PM

^^ very nice numbers - about right??

Anvil

06-12-2012, 11:22 PM

Hi Steve

I'll have to do some more tests, it's not acting up like on the Areca.

(I'll compare to a 2R0 V3 240GB later tonight)

Computurd

06-13-2012, 03:35 PM

LSI has seemed to be on the ball with the fixes, with the latest FW it may not be an issue anymore, which is a very positive thing!

Anvil

06-14-2012, 02:59 PM

I've ran a few tests using ATTO (for sequentials only) on the LSI 9265, w/o caching. (Direct IO, Write Through, No read ahead)

Keep in mind that the V3 is working with compressible data, they are in a "heavily used" state though, several TBs written to each drive w/o any cleaning (no TRIM as they are in raid) for 3-4 months, maybe more.
The m4's are set up in a raid 00 (a span of 2x raid-0) so there are 4x 128GB drives, also in a used state.

127607

127608

The V4's could do with some tuning on reads. (which can be seen in other tests as well, coming later...)

SteveRo

06-15-2012, 03:50 AM

Yes, agree writes are looking nice, reads - so so ....

http://img525.imageshack.us/img525/2113/2xr0vertex4pchwbconfw14.png (http://img525.imageshack.us/i/2xr0vertex4pchwbconfw14.png/)

SteveRo

06-19-2012, 03:33 AM

For fun cache run with old areca 1261 :)

2xvertex4, f/w 1.4.1.3, 1261/2gb, 4k stripe, disk write cache mode disabled, array as boot drive, 2600k @ 4.8ghz on m4e -

http://img19.imageshack.us/img19/7568/2xvertex412612gbfw1413d.png (http://img19.imageshack.us/i/2xvertex412612gbfw1413d.png/)

Anvil

07-10-2012, 02:29 PM

A quick test on FW 1.5 shows that things are looking much better on the Areca 1880IX-16 4G

There was a strange reading in ATTO (at 1KB) but compared to previous tests this is something that's worth playing with.

128271

128272

(the results are cached, the point is that it's working, on earlier fw's it wouldn't finish any benchmark)

I'll play a bit more too see if the ATTO anomaly at 1KB can be reproduced.

edit:
64KB strip size, Volume cache mode -> Write Back

johnw

07-10-2012, 03:03 PM

For ATTO, try "Neither" instead of "Overlapped I/O"

Also, try increasing the "Total Length" to 2GB (or whatever is the max allowed)

Anvil

07-11-2012, 01:39 AM

I'll do some more testing later this week/weekend.

My goal was just to see if it was working, at all :)

Tiltevros

07-11-2012, 02:12 AM

try to set 1mb stripe size on LSI controller and see what is happening on read write sequential and 4k@64 on sandforce controller

(4xR0 1mb stripe size no cache on, direct io, no read ahead 3k kingstons)
128302
128303

SteveRo

07-11-2012, 04:23 PM

WOW! 1.5 f/w is looking good - I will try!

edit - Hey there Mr. Tilt - nice AS SSD run there! :up:

SteveRo

07-12-2012, 03:55 AM

Looks like vertex4 compatibility with areca 1880 might be fixed with F/W 1.5.
Probably needs some time to check out completely but looking ok for the most part so far.
I did have a glitch running Anvil Storage Utility, needed to rerun the read test and then all looks as expected'
These are cache runs - 2xVertex4 128gb F/W 1.5 R0 Areca 1880 4GB on crosshair V formula with 3xcpu Phenom II X2 555 BE @ 4ghz -

http://img193.imageshack.us/img193/8590/atto2xvertex4r0fw1564ks.png (http://img193.imageshack.us/i/atto2xvertex4r0fw1564ks.png/)

http://img849.imageshack.us/img849/7189/2xvertex4r0fw1564kstrip.png (http://img849.imageshack.us/i/2xvertex4r0fw1564kstrip.png/)

For comparison here is cache run - 2xVertex4 128gb F/W 1.5 R0 Areca 1261 2GB on X58 with W3520 @ 4ghz -

http://img40.imageshack.us/img40/1733/2xvertexr064kstripefw15.png (http://img40.imageshack.us/i/2xvertexr064kstripefw15.png/)

Interesting that the Anvil benchmark keeps the name "ARC-1261" when switched from 1261 to 1880 - the array name is maintained as I did not recreate the array just moved the drives from 1261 to 1880 :)