Vertex LE vs Crucial C300

[jcool]

Nice, what sort of pricing did you get them for?
Tempted to get one (& maybe a 2nd later) instead of an x25m 80gb, or one of the other sandforce offerings.

thank-you.

Around 150€ net (+VAT for end users) from Wave Computer Germany.
The X25-M costs around 10€ more (net), but offers higher capacity.

I'll soon find out about 2R0 LE vs 2R0 C300.

I could tell you that 2R0 LE are the fastest drives I've ever tried but that would be unfair as I haven't tried the C300 array yet.

I've got a feeling that it could be a close race, at least using 3Gb/s interface.

I'll share my benchmarks and initial findings by this time tomorrow.

@jcool

Looks like the 50GB SF drives are performing quite nicely.

Personally I prefer small stripes, initially I used my LEs on a 128KB stripe but they are now at a 16KB stripe.
I'll be doing some tests this weekend using 4-8KB stripes on both the LEs and C300s.

Yes, I think so too. Even though I don't notice any real world gains (OS) in performance vs. my single Indilinx drive... maybe it needs a reinstall of Win7 instead of a clone

Something else: I heard from my source at Compustocx that the SF-1200 based drives will get a FW update to unlock more space
50GB will be made a 60GB, 100GB a 120GB and so on - if true that would be awesome

[GullLars]

@natedog

For files compressable to 2:1, the 2R0 100GB LEs will give you roughly 500-540MB/s both read and write, 2R0 C300 will give you 500-550MB/s read and 280MB/s write. IOPS will be ca 60K read 60K write for the LEs and 100K read 60K write for the C300s.
As long as you're sticking to the mobo SATA 3Gbps controller, i'd go with 2x LE 100GB, benchmarks indicate they perform equal to or beat the C300 in almost all possible scenarios on 3Gbps controllers, the exception being incompressible data.
Be carefull to check that the mobo can use NCQ in RAID mode, if it's the same as for consumer mobos, you'll need RAIDXpert software to enable it.

If you go for that system, pleese post PCmark Vantage scores for the setup


EDIT: @JCool: makes sense. SF-1200 aren't targeted at enterprise, and should get by with either RAISE disabled if the NAND grade is high (doable in FW for manufacturers using higher grade NAND), or RAISE can still be enabled by using some of the dynamic spare area gained from compression (and TRIM). Very few people will fill an SSD completely with data compressible to less than 80-90% of original size.

[jcool]

.. looks loke cloning my OS with Acronis resulted in the offset getting ed up.. ASSSD shows 104422K - Bad
No wonder I didn't see any speed improvements there. Argh... time to reinstall I guess :/

[onex]

i'll just add a little bit here,
i havn't seen any 128GB C300 benches so far, and got too much into it,
yet, found this thread here :
http://forums.overclockers.co.uk/sho...8#post16177361
might give you some more information in the meanwhile

[Natedog]

@natedog
As long as you're sticking to the mobo SATA 3Gbps controller, i'd go with 2x LE 100GB, benchmarks indicate they perform equal to or beat the C300 in almost all possible scenarios on 3Gbps controllers, the exception being incompressible data.
Be carefull to check that the mobo can use NCQ in RAID mode, if it's the same as for consumer mobos, you'll need RAIDXpert software to enable it.

Here is the motherboard ive got on backorder. Supermicro H8QGi-F...AMD SP5100 SATA controller??

http://www.supermicro.com/Aplus/moth...x0/H8QGi-F.cfm

[onex]

The codenames of the chipsets "RD890S", "RD870S" and "SB700S" has also changed to "SR5690", "SR5670" and "SP5100"

the SP5100 is apparently a server chipset, same as the SB700.

SP5100

* All features of SB700
* Super I/O
* Target for future server chipsets [21]
* Originally named SB700S ,[38] later renamed as SP5100.

http://en.wikipedia.org/wiki/AMD_700...s#Southbridges

maybe that should do it?
http://drivers.softpedia.com/get/Oth...24154026.shtml
apparently, the 100GB LE functions way better then the C300 128GB, at least on the SATA 2 interface:
that's from anvil, LE 100GB:


that's the 128GB C300 on the SATA2 from the thread above:

[Natedog]

I just realised something...since the motherboard is on backorder i can still switch the order to the H8QG6-F which has a 6gb/s LSI 2008 controller onboard (for $100 extra). hmm, i think i might be worth it for future SSD drives.

http://www.supermicro.com/Aplus/moth...x0/H8QG6-F.cfm

[Anvil]

onex,
That's got to be the wrong image linked or a bad setup.

This is Nizzens C300 128GB

c300__Crystaldiskmark_IDE_mode.PNG

[onex]

yeah, you can't really trust these benches when you'r unaware of the setup, hurdles and who is using it,
these are fabulous numbers, from a brief look, it seems the marvel controller isn't being shy next to the sandforce one,
yet this is still only brief,
very impressive.

that's from the same guy, with SATA3 (C300 128GB):

obviously, something is terribly wrong here (at first picture).

[GullLars]

Keep in mind, CDM and AS SSD gives close to no compression at all. For a usage pattern where data will be compressible 2:1 you can double the sequential numbers for the LE (up to the SATA 3Gbps bottleneck).

[onex]

so MAYBE, shall we go using IOMeter?
why people are still using these benching softwares??

anyhow, a new IOMeter version should come along soon where the test file isn't compressible and so can test drives worst bandwidth scenario.

it's a shame no software allows you to manipulate the test file compression percentage or at least to allow IOMeter to use any file available else then the iobw.tst,
if they would allow it to use any file for testing, i.e - DLL's, executables, image file and batches of files that would give much better view of the drives, especially the new compression using ones such as the SandForce.
and also would allow us to have a better view of the drive design.

[Anvil]

onex,

You can always generate your own iobw.tst file, it should make a difference to read tests.
I have no issues using the current "iometer" model but I do welcome the new option for random data or not.
If they only could make the data not so random, somewhere in the middle or maybe 3 options : the current way, 50% compressible or 100% random.

I asked the author of AS SSD but he wasn't interested in making any changes, he want's a user friendly bencmark without any options.
(a bit far fetched as most people that do benchmarks are more than willing to make adjustments and tweaks)
OTOH, it would make it almost impossible to generate a fair score for the benchmark.
SF based SSDs would be the only ones that would benefit from such changes.

edit:
I've just picked up the second C300 but I have to finish some work before I can start testing.

[onex]

I've just picked up the second C300 but I have to finish some work before I can start testing.

your semi-nuts man,
to figure you just bought an 800$ (1000 euro prices?) SSD simply because you wanted to test it's RAID function ,
i wish i could have such access to hardware,
some 9260 or 9211 and 4 LE's and/or 4 X25-M's, a 4x12 core magny cours system with some 16 or 32GB of RAM, just for the kicks,
a few SSD's from different manufacturers to test they're function, maybe some PCI-e drives etc, a nice SS for a desktop system benchmarks
and anand's access to NDA'd software,
even a pass to IEEE USA forums would be nice as a start,
just need someone who refer you inside, as they have canceled 2 registration attempts already, probably due to a generic e-mail .

[GullLars]

You have special IO-tracing programs used in the industry for tracing speciffic workloads and using real world access patterns for benchmarking, like the one Anand has used for his ligh, heavy, and gaming benchmarks.

I've also seen a build of CDM 3 that allows you to use the partern 0x0000, i think it was anvil that posted a screenshot of his LE with it earlier.
CDM 3.0 and AS SSD are really good benchmarks for charting SSD performance quickly and fairly accurate for the specs meassured, for all SSDs. Since they don't allow compression, you get RAW performance for drives using compression. To get real life performance, simply multiply by data compressability (up to the next bottleneck). Max performance can be found with benchmarks like ATTO, HDtune/tach and IOmeter. Real world performance will be somewhere in between dependent on the data.

I also have a request people start filling in the fields of the benchmark programs where they are avalible, like CDM, AS SSD, and ATTO have dedicated fields to list the info. I'll illustrate with my own screenshots what i mean.

[onex]

i don't know gullars, i really don't like these programs except from IOMeter yet it is too, limited.

the idea of adding system data withing the benching program or thread, is -well about time?-
people coming from out side, not reading the whole thread can barely ever participate in it, the members who discuss the thread usually expect any new comer to already understand what is being discussed or either they like to keep the same users watching the thread or they're "friends" which is a bit uninviting.

ps - whats the story with this Metron 1MBps write? and why ATTO is showing almost 150MBps for it?
i really don't like this program, it seems so inaccurate!
HDTune, the same, u seen probably it's benches over the 1231ML 8xX25-M's thread (400MBps..)

[Anvil]

your semi-nuts man,
to figure you just bought an 800$ (1000 euro prices?) SSD simply because you wanted to test it's RAID function ,
(.

LOL, I know
It just happens to be work related and a hobby as well.

BTW the C300 256GB is quite cheap in Norway compared to other 200-256GB SSDs.

[onex]

hm,
found not long ago someone at e-bay who was willing to sell an asus p6t deluxe with an I7 920, 6GB of memory (some ocz or corsair) 2xEVGA's GTX285, an 160GB G2 X25-M new boxed WIN 7, a nice cooling fan (forgot which), new (forgot which) nice mouse and keyboard and some other stuff for around 2000$,
everything was new, i've calculated it to be around 2600 or 2900, it was a really nice deal,

another who was selling a dual gainstown E5540 8GB ram, MB, a whole setup for around 1600$ and
another one as the above with 2xGTX 285, some 6 146GB SAS HDD which was a presentation computer for around 2000 euro,
they all never went in the end, some issues.
hardware over here is VERY expensive and most users don't it so much so you can barely find anything useful through the second hand market.
SSD's are not even an option, the cheapest ones you find are above the US prices and whenever you buy them from a store you obviously pay premium,
2nd hand is almost impossible to find,
through the last 3 or 4 months there were about 6 of them through all the second hand sites, 2 OCZ core, one patriot torqx, an OWC 50GB which came last week for 270$, just because you can't find it here in stores and an X25-M 160GB for 470$ new from some seller who buy them in the US or cheaper at e-bay and sells it 50$ more just because it is hard to find,
they cost about 630$ in the stores.

raid controllers are impossible to find, unless they are cheap old ones, hi-end hardware as well,
you open a post which is saying "very strong computer for editing and professional use" and you get some q8200 with a 9600 nvidia card,
you would like to buy a server?
you find a dual processor dual core xeon running 2.8 or 3.2 saying "a very strong server" and they cost more then a full 860 or 920 system.

sometimes it seems impossible to get something decent here .

[alfaunits]

If it's 104422K (and not bytes) then it probably created a GPT/dynamic partition
Create the partition manually, then restore the files from the Acronis image instead of restoring the partition (you can use the same image)

.. looks loke cloning my OS with Acronis resulted in the offset getting ed up.. ASSSD shows 104422K - Bad
No wonder I didn't see any speed improvements there. Argh... time to reinstall I guess :/

[GullLars]

ps - whats the story with this Metron 1MBps write? and why ATTO is showing almost 150MBps for it?
i really don't like this program, it seems so inaccurate!
HDTune, the same, u seen probably it's benches over the 1231ML 8xX25-M's thread (400MBps..)

Mtron, not metron.
AS SSD and Crystal shows 4KB RANDOM at QD 1/4/32/64.
ATTO shows all the block sizes at SEQUENTIAL with the selected QD.

AS SSD writes 1GB of 4KB random, so in total 250.000 IOPs at QD 1, and another 250K at QD 64. Crystal just writes 4KB random for 10 sec in a xMB testfile, wich gives a bit better numbers since the drive don't run out of clean spare area. In real OS usage, the numbers will likely be closer to CMD than AS SSD.

The numbers are pretty accurate for what they are meassuring, but if you don't know what you're meassuring it may seem inconsistent and conflicting between benchmarks.
EDIT: BTW, notice my Mtrons get 35-40MB/s at 4KB random QD 1, that is about 9-10K IOPS at QD 1. Most SSDs are around 5-6K, and Acards are around 11-13K.

[onex]

AS SSD and Crystal shows 4KB RANDOM at QD 1/4/32/64.
ATTO shows all the block sizes at SEQUENTIAL with the selected QD.

AS SSD writes 1GB of 4KB random, so in total 250.000 IOPs at QD 1, and another 250K at QD 64. Crystal just writes 4KB random for 10 sec in a xMB testfile, wich gives a bit better numbers since the drive don't run out of clean spare area. In real OS usage, the numbers will likely be closer to CMD than AS SSD.

o.k, that's sublime !

EDIT: BTW, notice my Mtrons get 35-40MB/s at 4KB random QD 1, that is about 9-10K IOPS at QD 1. Most SSDs are around 5-6K, and Acards are around 11-13K.

yep, 35MBps is very nice !

E:though still, very low write speeds and high access time..

[jcool]

If it's 104422K (and not bytes) then it probably created a GPT/dynamic partition
Create the partition manually, then restore the files from the Acronis image instead of restoring the partition (you can use the same image)

I just used Acronis 2010 (V13) and it kept the offset. Now it shows 103424 and green again
And yeah it's notably faster than before. The offset is due to Win7 creating a 100MB "pre-partition" before the C: one I guess...

Anyway, pretty happy with 2x SF-1200 for now

[GullLars]

E:though still, very low write speeds and high access time..

Keep in mind the Mtron 7025 series was released in 2007, and i bought mine in august 2008. Before the Intel G1 drives were avalible. The only alternatives were first generation JMicron and Samsung. Back at that time, >250 IOPS sustained random write was good, and it's better than the newest Velociraptors and 15K SAS drives (if you don't short-stroke). And the CMD numbers (10 sec with random writes) show closer to 1000 random write IOPS.
When i was doing the AS SSD test, it started out at 5MB/s, dropped to 3MB/s in 3-5 seconds, and gradually worked its way down to 1,1-1,15MB/s where it stabelized the last 5 minutes of the test. In real world scenarios, i've never experienced it hanging or becoming slugish, so clearly there's seldom large continous bursts of random writes.
The only part i'm not 100% satisfied with is the poor scaling of write performance i got on SB650. A single drive benched to 90-100MB/s, but the RAID never went over 150MB/s even when new. There's been no performance degrading either, it performs just as it did 1,5 years ago.

[onex]

The only part i'm not 100% satisfied with is the poor scaling of write performance i got on SB650.

it could be the SB650, people seem to complain on AMD's onboard RAID controllers.

p.s - as for the 3 second access time, it could be that the Metron controller has the same illness as the Jmicron JMF602 with lack of cache or dram buffer.

[Razmatazz]

I just used Acronis 2010 (V13) and it kept the offset. Now it shows 103424 and green again
And yeah it's notably faster than before. The offset is due to Win7 creating a 100MB "pre-partition" before the C: one I guess...

jcool, please excuse my ignorance. What do you mean by the offset in using Acronis to clone your OS? I have used the system image backup in Windows 7 which has actually saved my a$$ once already. Thankfully everything worked out fine. I have been using Genie Backup manager for data backup to a seperate HD. This has also worked out well after a HD crash. I see many people recommending Acronis. I'm not quite sure what I would gain by using this program.

[GullLars]

it could be the SB650, people seem to complain on AMD's onboard RAID controllers.

p.s - as for the 3 second access time, it could be that the Metron controller has the same illness as the Jmicron JMF602 with lack of cache or dram buffer.

It's 3 MILLIseconds, not 3 seconds. It's at least a 100x difference in bad case write accesstimes to the JMicron JMF602.