Raid 0 Vertex 2 benchmarks

Hey everybody.. Just upgraded my setup from raid 0 / 3 Intel 40gb ssds to these two vertex 2 60gb. At $130 dollars a piece couldnt pass it up. And they are so far smoking the intels.:cool:

I will be adding some more benchs so far but wanted to show something that might be interesting to some here. I did 5 quick benchs of ATTO with different settings under windows with some interesting results that I know many have been asking about and talking about for sometime in different forums.. Maybe this will give some more light on the situation.

Raid 0 stripe is at 16k on an ICH10R just like the 3 Intel SSDs. With testing that was the strongest setup I found. Gullars had recommended this also. Was also confirmed stronger on the Vertex 2's. Link below...

http://www.ocztechnologyforum.com/fo...10R-benchmarks

Ok on to some benchs you all should find interesting if you have SSD's.

Atto test 1 = Enable write caching on the device not checked (disabled)

http://img714.imageshack.us/img714/6477/offfs.png

Atto test 2 = Enable write caching on the device checked (enabled)

http://img192.imageshack.us/img192/9606/69506728.png

Atto test 3 = Enable write caching on the device checked (enabled) / Turn off Windows write-cache buffer flushing on the device checked (enabled)

http://img514.imageshack.us/img514/8066/onflush.png

Atto test 4 = Enable write caching on the device not checked (disabled) / All cpu powersaving modes and states in bios disabled

http://img713.imageshack.us/img713/2...owersaving.png

Atto test 5 = Enable write caching on the device checked (enabled) / Turn off Windows write-cache buffer flushing on the device checked (enabled) / All cpu powersaving modes and states in bios disabled

http://img697.imageshack.us/img697/5...owersaving.png


Ok first thing first...

Atto 1 settings seemed good if you write alot of larger files and you dont want to disable your powersaving states in bios. In the big blocks Atto 1 seemed to have some of the best write results.

Atto 2 settings reflect a doubling of performance in the small blocks due to your system memory being used as a buffer. This is great for smaller writes but is effected negatively as you go to larger size blocks. By using your memory as a buffer you add a variable that can effect your drive performance positively and negatively depending on settings from your bios for your system power states.

Atto 3 is same as Atto 2 but by turning off the flushing helps to push the performance up a little more on the smaller blocks but leaves you with the same problem as Atto 2. Limited write ceiling.

Atto 4 is a unique one to described. It seems to be what ATTO 1 is but with out the powersaving bios restrictions. With no power states enabled it has the highest and the most over 540MB reads. It starts low with the writes but once it reaches block 16 it is equal or faster than the others. And by the time it reaches block 64 it is the fastest writer after that and doesnt look back. It is actually almost reaching the limits of what 2 vertex in raid 0 theoretically can be. 285read/275write is 570read/550write.:shocked:

Atto 5 seems most balanced from these tests. It doesnt top out as high as ATTO 4 but the average of the writes and reads seem be higher. CPU, memory, and generally the system as a whole is running full speed with no regards to clocking down.

The summary here is that Atto 4 and 5 settings are what you should be looking at having or using. Powersaving settings in the bios seem to be the biggest impact. Writing large files and reading or streaming large files seem to lead more to the Atto 4 pattern. Atto 5 seems like it would be best for general usage and OS usage, while you lose a little top end write speed. The other Atto 1,2,3 should really not be used, unless you want to use your powersaving features in the bios then i would go with Atto 3.

I will be putting up some more benchmarks soon.. was just getting started and trying to shed some light on these settings and how they seem ot effect performance since there didnt seem to be any clear answers.

Forgot to mention that these 2 vertex in Raid 0 I just ran through these tests are my windows 7 boot drive. So if you subtract a little bit of overhead from windows doing things while I was benchmarking it is very possible these damn things are hitting right at the theoretical limits. Very impressive for 2 drives. :)

Also because someone always asks.. i used the newest 6/8/2010 RST drivers from intel.. Not the ones from march. Version 9.6.4.1002.

Using ATTO on the SandForce is only for marketing purposes :)

No doubt about it, the SF controller is a great controller but those numbers aren't realistic using real data.
It's almost impossible to benchmark the SF controller using the available benchmarking applications.

Keep the benchmarks coming :up:

We''ll se how it goes comparing the Intels to the LE using other benchmarks :)

Benchmarks are coming. ;)

I did this more to see the effects of those settings and powersaving bios settings against the SSD. Any comments on the differences I found. All benchmarks I post now will be more real world type tests, (if there is anything you all want to see let me know, as long as it is real world type bench) I dont like the synthetic tests very much but am interested and intrigued in the results it gave. So I will be benching with the Atto 5 settings. Any comments on this let me know!! :up:

I still prefer the standard option.

Enable write caching (Enabled)
Turn off Windows write-cache buffer flushing... (unchecked)

Unless the real life gains are substantial I wouldn't touch the buffer flushing on any modern SSD. (unless you've got an UPS)

Ok starting the Vantage HD tests now.. Will post in a few moments. I will be testing with all the same different settings as above. I will call them ATTO 1, ATTO 2, and so on so you know how i am configuring and what effects those settings may have to the results of Vantage.

Below will be a new setting I missed which Anvil mentioned he prefers.. Actually it seemed almost equal to flushing checked with a few MB here and there lower. But as he mentioned for the security of your data this one would seem the best.

Atto test 6 = Enable write caching on the device checked (enabled) / Turn off Windows write-cache buffer flushing on the device not checked (disabled) / All cpu powersaving modes and states in bios disabled

http://img801.imageshack.us/img801/1...owersaving.png

These are just the synthetic marketing type tests as Anvil mentioned. I am interested in seeing the effects these settings have on more real world type tests.:)

Ok here are the PCMark Vantage scores... They are sort of following the same patterns as the ATTO Tests. The top performers are the ATTO 5 and 6. ATTO 2 and 3 would be the next set. With ATTO 1 and 4 with the lowest scores.

ATTO 1 and 2 settings

http://img709.imageshack.us/img709/412/atto1.png http://img88.imageshack.us/img88/5247/atto2.png

ATTO 3 and 4 settings

http://img694.imageshack.us/img694/4749/atto3.png http://img227.imageshack.us/img227/4572/atto4.png

ATTO 5 and 6 settings

http://img693.imageshack.us/img693/1200/atto5.png http://img143.imageshack.us/img143/2206/atto6.png

To simplify the fastest setting is (ATTO 5) no powersaving in bios enabled, writeback enabled, with flushing.

Second is (ATTO 6) no powersaving in bios enabled, writeback enabled, no flushing enabled.

Third is (ATTO 3) powersaving in bios enabled, writeback enabled, flushing enabled.

Fourth is (ATTO 2) powersaving in bios enabled, writeback enabled, no flushing enabled.

Fifth is (ATTO 4) powersaving in bios disabled, writeback disabled

Sixth is (ATTO 1) powersaving in bios enabled, writeback disabled

I wish I could get my hands on Anandtech's Storage Bench. :shakes:

Looking for some other type of realworld tests. Will happily take some recommendations. If someone wants to see something.

Ok here is a AS SSD bench below..

http://img827.imageshack.us/img827/1315/asssd.png

Anybody know how to setup up Iometer to get some meaningful tests from it? I've not used it before.:confused:

Oh just a quick comment for some that had concerns about wear impact using small stripes on the 3 intel raid setup I had before. When I broke it down it was still showing all drives at 99 wear level. It only moved down 1 point from within the Intel Toolbox when I checked each one. I saw others that were down to 93 or so with other setups. So just thought I would mention that since I had that setup for a few months. ;) Those were system drives also. Looks like it is more impacted with what you do with your drives and not how you set them up, in regards to Raid 0.

P.S. Anybody have any info on the crystalmark bench in regards to how they effect sandforce drives. I have not done a bench with that yet because the OCZ mods pretty much say that it hammers the controller and takes awhile to recover. Specifically because the amount that it writes and how it writes.

here is some profiles linked you can download for iometer:
http://www.ocztechnologyforum.com/fo...l=1#post523697

and a short guide on how to use those profiles with the program that i made on another site:
http://www.ocztechnologyforum.com/fo...l=1#post525472

could you run pcmark05 for giggles?

PcMark05 coming up..

Thanks Computurd.. Let me read up in there a bit.. Be right back with a PcMark05 bench..:up:

Here is the PcMark05 bench. I dont have any current references to compare on this one, but found an older review from anandtech where he was getting about 7000+ with a 150gb raptor. He was impressed at the time. Thought that was funny with where we are now... :rofl:

http://img844.imageshack.us/img844/5195/pcmark05.png

This was tested with ATTO 5 settings from above.

Working on the Iometer now...

Here is a post of the AS-SSD Copy Benchmark..

http://img651.imageshack.us/img651/1978/asssdcopy.png

I have been reading alot and have noticed that these and probably all SSD's are very performance limited by the platform you are running. I have seen some other benchs of my vertex 2's and the reads others are getting are quite a bit higher. Only difference I am seeing between them and me is platform. For example I have a highly overclocked q6600 with ddr2 on water.. but some of the others have i720's clocked to 4ghz with ddr3 or six core Phenoms like below. The CPU and memory seems to effect the read speeds. But my writes are way higher than them on 4k. :confused: Like below

http://www.ocztechnologyforum.com/fo...rce-RAID-0-x-2

My question is with the way i have my cpu overclocked.. What is more of an impact on the performance of an SSD.. THe memory or the CPU? How much more powerful is an overclocked PHenom x6 than what I have my CPU clocked at. I am thinking it is the DDR3 but his reads on AS-SSD are quite a bit more. ALso some of his AS-SSD writes.

because there is nearly no CPU usage when using these drives the CPU overclock should have little to no effect.
its like a hardware raid vs software raid.. so to speak.

i think its a case of every single drive is slightly different, every windows install is different every piece of hardware, even two identical batches are different and each one is going to act different.

you are not THAT much slower/faster than those others... not that is saw
so to say they are better or youare is completely subjective.
i think you are more or less equals and in real world, every day computing and gaming you will never know the difference.

@lestat

Can you explain why there is such a boost in performance than with the CPU powersaving features disabled. You can see with many of the benchmarks that letting the CPU run full tilt with out letting clcok down is hitting the top scores. Guess I will clock it down to default right now and see what the benchmarks show. :)

nice work.....i am seeing similar effects on PCMV in my recent testing

still trying to work out why my Adaptec 21245 is performing so poorly against ICH10 though

have you played with stripe sizes?

Ok finished a complete run of everything. Here are the settings.

I have been using ATTO 6 settings from above... FOr me that is the fastest. Only thing I changed is downclocked the q6600 back to default 2.4ghz. Default FSB, Memory. Left the powersaving off in bios.

So with everything running at default here are the results...

http://img19.imageshack.us/img19/8518/defaultcpu.png

You can compare with my results above.. and you can see that in the ATTO and AS-SSD the writes are effected the most. So writing to these does have some relation to your CPU speed.

But more interesting and possibly damaging is in the Vantage and PcMark05 results which are suppose to be more reflective of real world type usage. You can see most tests there had a negative result compared to the results with the high clock speed. Even worse many of my test suite scores were almost cut in half just by going to stock CPU clocks. :eek: So the CPU you have does effect the performance of your SSD's. I would assume this to be the same for other SSD's also. Productivity of what these drives can fully do is impacted by the speed of your processor.

Now to put in perspective, even these scores above at default CPU clocks are beyond any hard drive.. but the impact on you would be more related to what you do with your SSD's. I would think if you used them in servers or did alot of writing that consisted of many small transaction writing you would want a strong CPU to get the most from them (or if your a benchmark record breaker :cool:). Also going by Vantage and PcMark05 would seem to confirm that your system would be completing tasks much faster and generally more responsive with a faster CPU. :up:

BRB need to overclock everything back!!

Quote:

---

Originally Posted by dinos22

nice work.....i am seeing similar effects on PCMV in my recent testing

still trying to work out why my Adaptec 21245 is performing so poorly against ICH10 though

have you played with stripe sizes?

---

I think Gullars and others have mentioned that using 1-3 SSD's is best on the ICH10. Intel made a good southbridge and it scales well with the way these SSD's move data.

Whats the difference in performance numbers you are seeing between the 21245 and the ICH10?

Havent played with stripe sizes yet. Been still trying to figure out how to set tests on Iometer. Downloaded a profile from computurd but it wasnt giving me results I was seeing from other raid0 vertex 2 setups.:shakes:

It is the QPI and DMI lanes which handle the traffic=higher cpu clock, less internal latency.
higher CPU clock equals higher IOPS, period!

@dinos- there arent going to be much you can do with that raid card outside of a pci-e bus overclock and maybe throwing some volts at your IOH (if you have the cooling for it, mind you! safe with water....otherwise at your own peril), also you might try larger pci-e packets, but you have to be careful with size selection with that if you are running SLI. the bus sends info across in packets of varying sizes, and the devices on the bus wait for each packet to pass before sending their own, so it is like rapid fire and they have to be paced correctly. if your packet size goes to large and you are running tri-sli (usually, slower mobos it shows up quicker with dual SLI) then your video cards will have to wait to send their packets, thus introducing stuttering and lagging into your video. there will be a sweet spot. You just have to test.
OCZ forums is full of guys who had adaptecs, and are running 6GB/s cards now. the adaptecs just cant handle the load. imo your better off on ICH10R, but even by the numbers you see that by now.

here is a copy/paste from elsewhere that i have found invaluable over time. It really is written to optimize GPU performance, but really it is going to apply to any device on pci-e bus. if you are patient enough to tweak, this is with a classified motherboard. info is attributed to shansmi btw, just to give him credit :)


http://i517.photobucket.com/albums/u...e1/shansmi.png

this is written for EVGA mobos, but might apply to you, in part, or in full. this is really kinda advanced tweaking guide for any pci-e device....

this from evga forum right? i swear ive seen it hehe

yeah that is where shansmi is from :)

Quote:

---

info is attributed to shansmi btw, just to give him credit

---

haha awesome! in the Overclocking or bench mark section yeah...