Vertex 3 specs and benchmarks

[Anvil]

The Vertex 3 has finally started shipping to retailers as of Tuesday according to Ryder @ OCZ.

The specs are also official and as expected they are looking really great.

There is one I noticed, random reads vary a lot by capacity.

Link to product sheet (requires PDF)

v3_sheet.PNG

I've got a few on order and will put them to test asap, hopefully within next week.

Should be really exciting, maybe the most important release of the year.

[Callsign_Vega]

Interesting. Do you think the double of IOPS on the 240 vs 120 on the 4KB random read would render much better performance? I was going to get the 120s but if the 240s are much faster I might have to reconsider.

[johnw]

For most desktop usage, it will make little difference. For QD=1, the 4KB random read is only about 5000 IOPS. You can only get to 20K or above at relatively high QDs.

[Brahmzy]

Interesting. Do you think the double of IOPS on the 240 vs 120 on the 4KB random read would render much better performance? I was going to get the 120s but if the 240s are much faster I might have to reconsider.

That's a hrad one to answer - depends totally on usage. Also, those are marketing specs (very round numbers) until we see benchmarks and some IOMETER action on the 120's we won't know.

I'm in the same boat. I will probably get the 120's.

[Computurd]

two questions;

why is 4k random write lower on the 480gb v the 240?

and secondly...

will there be throttling as in the previous gen of sandforce devices?

[Brahmzy]

two questions;

why is 4k random write lower on the 480gb v the 240?

and secondly...

will there be throttling as in the previous gen of sandforce devices?

According to the 5 people that got V3's for testing, the throttling is completely a non-issue on the V3's. One of the guys basically said he had to run CDM 9 times x4000mb to get it to look like it was throttling and then the drive came back to previous performance right away. These are different controllers with different firmware and SF looks like they're making sure the same throtling issues don't manifest in the same way on the V3s.

[johnw]

will there be throttling as in the previous gen of sandforce devices?

Yes, the persistent Sandforce flaw of reduced write speed after heavy use is still there in the 2281 controller and Vertex 3 firmware. As usual, the place to go to see a good test is bit-tech. The pre-release 240GB Vertex 3 sequential write speed goes down from 284 to 220 MB/s.

http://www.bit-tech.net/hardware/201...-ssd-preview/3

[Brahmzy]

Get yours now - just grabbed one off Newegg - LIMIT 1 PER PERSON. Argh.

[Anvil]

Looks like youre getting the first samples, nothing in stock here yet.

I'll be watching this space

[Brahmzy]

Looks like youre getting the first samples, nothing in stock here yet.

I'll be watching this space

It going in and out of stock on the Egg - just like when the ASUS B3's started going up.

[Callsign_Vega]

Back in stock if you are quick:

http://www.newegg.com/Product/Produc...82E16820227706

[F@32]

$600 to get 1GB/s from 2R0 About $240 more expensive than it should be... 3R0 with these puppies just for $360 including higher 4K QD1 and you're all set. Waiting for lower prices and lower capacities.

[Callsign_Vega]

You get more space with the 2 120GB vertex 3s in Raid 0 and they are also faster individually in other respects besides maximum throughput to the C300.

[mbreslin]

they are also faster individually in other respects besides maximum throughput to the C300.

Which other aspects are these?

[F@32]

You get more space with the 2 120GB vertex 3s in Raid 0 and they are also faster individually in other respects besides maximum throughput to the C300.

Yes, 240 is larger than 180. Single C300 64Gb have the same theoretical 60K IOps at 4K random read, which translates to the similar ~30MB/s. C300 64 is slower at writes, but it's not an operation where desktop drive spends most of the time. QD1 4k random/sequential read is the majority of storage ops. Hint: read the findings in hIOmon thread

If you are looking for serious DAW or DV solution, I'd never touch MLC drive with 9 foot barge pole. For those applications it's either R5 with X25-E on 9260 w FP or better yet Acard 9010BA or io-Xtreme.

[Khoral]

A new review @Benchmark Reviews

http://benchmarkreviews.com/index.ph...=717&Itemid=60

[Brahmzy]

A new review @Benchmark Reviews

http://benchmarkreviews.com/index.ph...=717&Itemid=60

Pretty lame review. Of course the V3's are going to look good when everything QD32+. Almost every bench they ran is unrealistic to 99% daily usage.

[He-ColdFinger]

Few questions, I will be running the Vertex3 drive or drives on the 6GB controller of my Rampage III Extreme. I can either purchase one 240GB or two 120GB and run a RAID 0 array. Will I see a big difference in performance with the RAID array compared to a single drive and have all the prior issues with lack of TRIM support in RAID been resolved?

I was also reading that people with some of the first drives are experiencing alot of freezing and lockups from the "Intel Rapid Storage Technology driver software". Can anyone here confirm this?

[Ao1]

Pretty lame review. Of course the V3's are going to look good when everything QD32+. Almost every bench they ran is unrealistic to 99% daily usage.

I saw your Iometer results for 4K random @ QD1 over at the OCZ forum. Boy that sucks. What type of I/O transfer data did you use? Pseudo random or repeating bytes? Hopefully pseudo or those results are even worse.

It seems that unless you get a 240GB drive you are better off with a V2 drive using 3xnm. Does that seem about right?

I wonder how much of a performance hit you get from a smaller capacity G3 drives. Anyone tested them yet?

[Brahmzy]

Yeah, looking forward to Anand's review of the 120 today.
Might be sellin a couple of 120's cheap.

[Ao1]

Yeah, looking forward to Anand's review of the 120 today.
Might be sellin a couple of 120's cheap.

Let’s see what Anandtech say, but I think it’s a bit of a push to classify different capacity drives under the same product name if they are using different density NAND chips that effectively half the channels. It’s not dissimilar to the difference between an X25-M and X25-V.

That said there must be something more to it than that....

OCZ Specs V3
120 GB = 20,000
240 GB = 40,000
480 GB = 50,000

Intel Specs
40 GB = 30,000
80 GB = 38,000
120 GB = 38,000
160 GB = 39,000
300 GB = 39,500
600 GB = 39,500

[Manicdan]

i assume those are 4k random read max iops you listed there?

[Ao1]

yep, sorry I should have stated that.

EDIT:

lol, even the X25-V can managed 25,000 random 4K read iops.

[Manicdan]

yeah if the 120GB is half the speed, its a total waste to me.

i know this isnt exact science, but looking at the number of newegg reviews on products, one can guess the sales ratios
for the vertex 2 there are:
391 for 60GB
287 for 120GB
63 for 240GB

so yeah, the SSDs being reviewed seem to make up ~10% of the sales, and the lack of lower capacity reviewing lets companies get away with giving out really horrible performing drives.

[Ao1]

.....so yeah, the SSDs being reviewed seem to make up ~10% of the sales, and the lack of lower capacity reviewing lets companies get away with giving out really horrible performing drives.

Nothing new there then
4K aligned and repeating byte Iometer benches are also misleading benchmark stats.

(Brahmzy)
Did you run iometer on repeating bytes to get that score?