First OCZ Core Series SSD results!

[Blacky]

Another review


http://www.guru3d.com/article/ocz-co...-drive-review/

[Praz]

http://forums.legitreviews.com/post116997.html#p116997

Some benches by me as well.

What chipset did you use for the benches? Seems odd that the CPU usage is being reported so high.

[Buckeye]

First off thanks for the review at legitreviews ! and first time I have seen that program CrystalDiskMark.
http://forums.legitreviews.com/post116997.html#p116997

But I think it clearly shows there is a huge differance in SSD's out there right now, depending on SLC or MLC. You take your pick on which one is better. The cost value of the newer SSD's are looking very good tho. How ever I think there is much room for improvement. I would also like to see how these new drives stack up in a real Raid using a good Raid card like a ARC-1231ML or better.

So I thought I would check out what my system came up with.

I know this is not a fair test !
1 OZC 64GB SSD vs my 8x 32GB MTRON PRO's in Raid 5

But here it is. This was with a full setup system complete with Kaspersky, MSN and just about everything under the sun all running on Vista Ultimate 64 with 4 gigs RAM



Now for a mini rant...

With one SSD in those tests the amount of noise and graphs spikes is just unbelievable. Two things come to mind, #1 the test rig was not setup correctly #2) something is very wrong with either the drivers or SSD. Even with my system running in full working config I do not even come close to the number of spikes that are shown, in fact mine are pretty much a stright line in HDTune, while HDTach does show some spiking I can get rid of most of that by turning off processes and other programs. Mine show a fully setup working system in Raid 5, not a test bench setup.

[flopper]

First off thanks for the review at legitreviews ! and first time I have seen that program CrystalDiskMark.
http://forums.legitreviews.com/post116997.html#p116997

But I think it clearly shows there is a huge differance in SSD's out there right now, depending on SLC or MLC. You take your pick on which one is better. The cost value of the newer SSD's are looking very good tho. How ever I think there is much room for improvement. I would also like to see how these new drives stack up in a real Raid using a good Raid card like a ARC-1231ML or better.

So I thought I would check out what my system came up with.

I know this is not a fair test !
1 OZC 64GB SSD vs my 8x 32GB MTRON PRO's in Raid 5

But here it is. This was with a full setup system complete with Kaspersky, MSN and just about everything under the sun all running on Vista Ultimate 64 with 4 gigs RAM



Now for a mini rant...

With one SSD in those tests the amount of noise and graphs spikes is just unbelievable. Two things come to mind, #1 the test rig was not setup correctly #2) something is very wrong with either the drivers or SSD. Even with my system running in full working config I do not even come close to the number of spikes that are shown, in fact mine are pretty much a stright line in HDTune, while HDTach does show some spiking I can get rid of most of that by turning off processes and other programs. Mine show a fully setup working system in Raid 5, not a test bench setup.

slc and mlc are different.
not much but enough to show the writes are not the strength of the drives.

Its a decision to be made,
pay the extra bucks, get better writes or go with a tad cheaper which allows a faster desktop experience even with some drawbacks.
First generations of any technology has some issues.

I go and try the ssd core version due to the price and also to find out how the experience will be using one as main systemdisk.
I just hate waiting for stuff when I click.

[Buckeye]

I hear you about waiting for things

Vista boots in ~10-12 seconds, it was ~8 seconds in Raid 0 config. Everything else is just about instant load, that includes programs like Photoshop, Illustrator etc., and gaming.

But back to the spiking issue, that concerns me. I do hate these benching programs tho as I do not feel they give us the real picture of SSD's.

All I can say is you gotta try em. Once you go SSD you will never look back

[RealTelstar]

That is because of the controller cache. Which controller did you use?

[stevecs]

post a link to a program you want, and I'll run it this weekend.

Thanks, I posted here with a script as well. This is for XDD, iozone would be good but much longer to run so we can probably skip that one.

http://www.xtremesystems.org/forums/...&postcount=124

let me know if you run into any problems/questions.

[Kobalt]

Ryder, do you know or can you comment on whether OCZ plans to release these drives with other interfaces like ExpressCard 54? My tablet could really use a SSD, but the hard drive is a 1.8" PATA so the 2.5" SATA drives wouldn't work. But it does have the ExpressCard slot, which I know Mtron produces. 4200 rpm drives are painfully slow .

[flopper]

I hear you about waiting for things

Vista boots in ~10-12 seconds, it was ~8 seconds in Raid 0 config. Everything else is just about instant load, that includes programs like Photoshop, Illustrator etc., and gaming.

But back to the spiking issue, that concerns me. I do hate these benching programs tho as I do not feel they give us the real picture of SSD's.

All I can say is you gotta try em. Once you go SSD you will never look back

great to hear.

[Buckeye]

That is because of the controller cache. Which controller did you use?

My controller card is the ARC-1231ML

[RealTelstar]

My controller card is the ARC-1231ML

Yep, my guesses are correct then. The sequential read and write are almost only the cache speed. I did a try with my HDDs with crystal and where i test the scsi drives under my PERC 4e (250mb of cache), the speeds are over what they should be for the drives.

But the good thing is that a part of that speed is true in the real world, because the controller cache is indeed used in sequential reads.
See my test with 2x 15k scsi disks in raid0.
In your case because of the greater cache (i think 1gb) also the sequential writes are influenced.

[IanY]

SSD has a long long long way to go

Look at how SAS puts SSD to shame. And its just $300 per 147GB drive.

http://uneit.com/2008/07/10/seagate-monster/

Just a two drive raidset

http://uneit.com/xpres/wp-content/up...7/76651-m1.jpg

[flopper]

SSD has a long long long way to go

Look at how SAS puts SSD to shame. And its just $300 per 147GB drive.

http://uneit.com/2008/07/10/seagate-monster/

Just a two drive raidset

http://uneit.com/xpres/wp-content/up...7/76651-m1.jpg

ssd 50x faster still in accesstime.
which is the speed difference that counts.
and you havent included 2 generation ssd, with 240/220mb read/write.

[dicecca112]

What chipset did you use for the benches? Seems odd that the CPU usage is being reported so high.

GM965 Centrino Based

[Levish]

SSD has a long long long way to go

Look at how SAS puts SSD to shame. And its just $300 per 147GB drive.

http://uneit.com/2008/07/10/seagate-monster/

Just a two drive raidset

http://uneit.com/xpres/wp-content/up...7/76651-m1.jpg

that is really great performance, but SSD's shine most in high IOPs required areas where the access time matters most.

[stevecs]

So far that can only be said for reads, and then only for sequential requests we still have not seen any write access time reports, nor have we seen any random read/write access time reports to really compare. With that in mind the known quantity of a sas drive (15K 2.5") of 2.9ms read and 3.3ms write for access times (average) has the upper hand at least until someone can produce #'s for the SSD's.

[NeedMoMegaHurtZ]

So far that can only be said for reads, and then only for sequential requests we still have not seen any write access time reports, nor have we seen any random read/write access time reports to really compare. With that in mind the known quantity of a sas drive (15K 2.5") of 2.9ms read and 3.3ms write for access times (average) has the upper hand at least until someone can produce #'s for the SSD's.

Sequential Write access time for the SLC based models seem ok @ 0.2ms. Offhand, here is Ender's results with 4x16GB Mobi on a RR3520

Original Thread: http://www.xtremesystems.org/forums/...d.php?t=191446

The MLC models seem to be poor, even for seqential writes. I'll try to find the link

[stevecs]

Thanks for that. So we're just missing info for the MLC writes and then random for both.

[NeedMoMegaHurtZ]

Yeah, here are the horrible random write access times (245ms!) (might be a bad sample...who knows)

Original article
http://www.alternativerecursion.info/?p=106

[stevecs]

wow. Thanks for that link. That was exactly the type of information I was looking for (to boot they did a decent analysis as well). With those numbers though I highly doubt their conclusion that it would be good for server/workstation use (at least for all the ones I deal with, the latency would just be killer). From this SSD's are several generations away from becoming a replacement for a good 15K drive. The only place where they would be some benefit that I can see would be in a transactional lookup database (where you batch load the data on the drives and then they are 100% reads) or things like static content servers (graphics for web pages or similar).

[Buckeye]

Nice find there with the link !

I also noticed this.

this was done on an intel ich9r sata controller in windows vista sp1 x64

It would be interesting to see these tests done on a Nvidia Motherboard.

[stevecs]

Yeah, but also remember that this is a single drive benchmark here not a raid. Still would be good to have another data point for comparison.

[Praz]

Right now I'm not relying on reported benchmarks, good or bad. Something's not right when burst rates are shown considerably slower the transfer rates.

[babalouj]

Here is a link to the official OCZ spec sheet on these drives. Look at the R Write numbers. The random writes look to be the big issue with these drives. This has been documented by OCZ themselves so its much more definitive than a couple of benchmark screen shots.
http://www.ocztechnology.com/ssd/OCZ...s_SSD_SPEC.pdf

[saint-francis]

Enough with the benchmarks. What are the opinions of people who actually own these drives and have been using them regularly? Are they snappy? Do programs open up real fast? Anything I haven't thought of?