OCZ has given us the opportunity to look at their new Solid State Drive (SSD), the Vertex 3. This drive is supposed to be ‘Breaking through performance, cost, and maintenance barriers, OCZ enterprise and consumer SATA solid state drives have taken the storage landscape by storm.’ Let’s see if OCZ made good on their promise.
Specifications and Features
Below you can see the technical specifications for the Vertex 3 line of drives. Note that there are speed differences throughout the lineup with the best performance being on the 240 GB drive with a slight drop at the 480 GB level on writes. Take a look at the Max 4k random writes on this drive at 60k! Some key differences between the Vertex 2 and 3:
- SF2281 controller
- 25 nm NAND
- Up to 60k IOPS (4k random read aligned)
The Sandforce 2141 and 2181 are slated to essentially replace the older Sandforce 1222 controller. Don’t worry, the the older controller will still be in the market for at least another year. The entire Sandforce 2100 line produces enhanced random read/write performance. While the 2181 is able to support capacities higher than 64 GB. Both controllers are still SATA 3Gb/s parts, but they support newer versions of the toggle-mode and ONFi 2 interfaces. As manufacturers stop making older flash, SSD makers will have to move to new flash interfaces. Thus, companies like OCZ will have to use a controller compatible with those technologies.
Why talk about a controller that is not in this drive? That’s because the Sandforce 2281 controller is physically the same as the Sandforce 2582 (OCZ’s Vertex 3 Pro) and the Sandforce 2682 (SAS). All three SATA 6Gb/s controllers support sequential read and write speeds of up to 500 MB/s and random 4 KB writes up to 60,000 IOPS. The difference is that the 2282 controller has capped the random write throughput of 20,000 IOPS. This is purely a firmware limitation, as SandForce believes that 20K is high enough for PCs in a client environment.
A lot of what makes Sandforce drives perform so well comes from data compression. For example, if I write a large, originally uncompressed file to the drive, it’s easy for the controller to compress this type of data. On the flip side, a ZIP file is already compressed, and consequently cannot be easily made smaller and therefore takes longer to write to the drive. Your computer still thinks that a xx MB .doc file is being written as xx MB, but once that data hits the controller, it ideally ends up as half that value. This compression helps lower write amplification and extend the life of the drive by using fewer program-erase cycles.
The Drive (Pictures)
Please note on the shot of the drive from the front that it is actually a black and silver label on it as opposed to the gold you see here. It doesn’t look much different at all compared to previous generations.
Test System, Methodology, and Results
Test System
- i7 2600k (stock)
- Asus Maximus IV Extreme
- 2x2GB OCZ Ripjaws DDR3 2133 7-10-7-27
- Asus GTX580 DirectCU II
- Seasonic X750
- Windows 7 Pro 64bit
- Latest AHCI and Intel chipset drivers
Method
- Secure erase the drive (used OCZ’s toolbox).
- Restart into Windows.
- Run benchmark (sometimes formatting it first is necessary, depending on the bench).
- Restart into Windows and Secure Erase the drive.
- Repeat ad nauseam.
Results
AS SSD
AS SSD is one of the most consistent click-it-and-forget-it SSD benchmarks. Regularly, they are within one or two points of each other with multiple runs.
AS SSD Benchmark (Read Tests)
Drive | Sequential | 4k | 4k-64 Thrd | Acc. time | Score |
OCZ Vertex 3 | 512.47 | 19.83 | 189.16 | 0.052 | 260 |
OCZ Vertex 2 | 206.17 | 18.53 | 119.37 | 0.078 | 165 |
AS SSD Benchmark (Write Tests)
Drive | Sequential | 4k | 4k-64 Thrd | Acc. Time | Score |
OCZ Vertex 3 | 285.62 | 79.32 | 236.90 | 0.208 | 345 |
OCZ Vertex 2 | 133.41 | 65.96 | 107.93 | 0.52 | 187 |
AS SSD Benchmark Score
Drive | Total Score |
OCZ Vertex 3 | 734 |
OCZ Vertex 2 | 431 |
To nobody’s surprise, the OCZ Vertex 3, when placed on a proper SATAIII port, struts its stuff all over the previous generation Vertex, which was never a slouch to begin with.
CrystalDiskMark
CrystalDiskMark x64 (Random Data – Read Tests)
Drive | Sequential | 512K | 4k | 4k QD32 |
OCZ Vertex 3 | 511.7 | 461.7 | 36.09 | 196.4 |
OCZ Vertex 2 | 213.0 | 201.3 | 19.26 | 128.5 |
CrystalDiskMark x64 (Random Data – Write Tests)
Drive | Sequential | 512k | 4k | 4k QD32 |
OCZ Vertex 3 | 306.5 | 303.0 | 101.2 | 233.4 |
OCZ Vertex 2 | 142.0 | 139.5 | 72.25 | 123.8 |
CrystalDiskMark x64 (0Fill – Read Tests)
Drive | Sequential | 512k | 4k | 4k QD32 |
OCZ Vertex 3 | 493.1 | 449.1 | 30.31 | 136.9 |
OCZ Vertex 2 | 273.1 | 257.8 | 23.35 | 136.5 |
CrystalDiskMark x64 (0Fill – Write Tests)
Drive | Sequential | 512k | 4k | 4k QD32 |
OCZ Vertex 3 | 374.4 | 329.1 | 22.7 | 127.4 |
OCZ Vertex 2 | 255.2 | 250.6 | 73.76 | 163.8 |
CrystalDiskMark x64 (1Fill – Read Tests)
Drive | Sequential | 512k | 4k | 4k QD32 |
OCZ Vertex 3 | 493.8 | 449.3 | 30.17 | 135.7 |
OCZ Vertex 2 | 272.9 | 258.4 | 23.41 | 133.4 |
CrystalDiskMark x64 (1Fill – Write Tests)
Drive | Sequential | 512k | 4k |
4k QD32 |
OCZ Vertex 3 | 379.2 | 330.4 | 23.43 | 126.8 |
OCZ Vertex 2 | 255.5 | 250.9 | 71.98 | 164.9 |
Here, with CrystalDiskMark, you can see the Vertex 3 pulls ahead of its previous generation in most tests. I’m not sure why we are seeing the Vertex2 beat out the Vertex 3 in the 0/1 fill tests however.
ATTO Disk Benchmark
This benchmark is what OCZ uses to rate its SSD’s speed. It consists of highly compressible data making for a best case scenario for this drive/controller.
Yet again the Vertex 3 takes a commanding lead from the 4k mark forward, but seemed to ramp up to that point a bit slower in this set of testing. Just note when testing the Vertex 3, I used the default settings as opposed to the Vertex 2 review which used a Queue Depth of 10. My results with 10, for some reason, were noticeably lower on the higher end, but a bit faster under 4k (beating the Vertex 2). Although I have tried re-running this many times (of course after a Secure Erase), the higher end was still limited with a QD of 10.
IoMeter
The gold standard of SSD testing, IoMeter takes a bit more finesse to run. Thankfully, after some trial and error, we have established the way this benchmark should be run to get the best results for the drives tested. This means setting the correct amount of threads. Any reviewers at Overclockers.com now use a sort of template/configuration file to help ensure accurate results across different drives and reviewers. The configuration used for the tests were Queue Depth 32 with an LBA of 8GB (per OCZ).
Iometer 4K Random Read
Drive | IOPS | MB/s |
Average I/O Response Time |
Maximum I/O Response Time |
CPU Utilization |
OCZ Vertex 3 | 64360.11 | 263.62 | 0.0619 | 22.3456 | 19.63% |
OCZ Vertex 2 | 43608.98 | 170.35 | 0.7334 | 84.6327 | 10.58% |
Iometer 4K Random Write
Drive | IOPS | MB/s |
Average I/O Response Time |
Maximum I/O Response Time |
CPU Utilization |
OCZ Vertex 3 | 58904.67 | 241.27 | 0.5429 | 233.282 | 13.00% |
OCZ Vertex 2 | 40612.55 | 158.64 | 0.7876 | 44.3103 | 17.38% |
Iometer 2MB Sequential Read
Drive | IOPS | MB/s |
Average I/O Response Time |
Maximum I/O Response Time |
CPU Utilization |
OCZ Vertex 3 | 268.28 | 562.63 | 238.407 | 241.910 | 1.43% |
OCZ Vertex 2 | 134.48 | 268.96 | 237.871 | 261.616 | 2.36% |
Iometer 2MB Sequential Write
Drive | IOPS | MB/s |
Average I/O Response Time |
Maximum I/O Response Time |
CPU Utilization |
OCZ Vertex 3 | 250.69 | 525.74 | 255.120 | 315.238 | 1.52% |
OCZ Vertex 2 | 125.11 | 250.21 | 255.632 | 299.382 | 1.27% |
You can see in the 4k random write the Vertex 3 is darn close to hitting its maximum specifications of 60k 4k writes (aligned). In the reads side of the house, we have breached that number and are pushing 65k IOPS’s 4k reads.
Boot Time
We use an application called Boot Timer to time how long a system takes to come up. It’s easy as can be; you double-click on it and it restarts your computer while timing how long it takes for the OS to boot and startup programs to load.
There isn’t going to be much of a direct comparison here as the system used to get these results in with the Vertex 2 is significantly different than the test bench I use. You can see the results for yourself, the Vertex 3 nearly cuts the boot time in half off a Vertex 2. My boot time was after all benchmarks and applications I use were installed and left sitting at the desktop overnight.
Conclusion
Leading the pack in overall performance with their SATA III SSD’s, OCZ has done it again with their Vertex 3 line. Offering 25 nm NAND as well as the new Sandforce 2281 controller powering the drive to those numbers. In order to achieve the rated performance on these drives, you must use a SATA III controller. This drive will easily saturate a SATA II port and you will top out around 280/260.
Pricing on the drive in hand is currently $499.99 at newegg.com (not including a mail in rebate) with the 120 GB version coming in at $258.95 (also not including mail in rebate). Although, we still haven’t hit the magical $1 per gigabyte number yet (no other manufacturers have yet either), performance for the dollar has been going up. The pricing for these drives are a bit market inflated due to its performance and recent introduction. I would imagine in time, like everything else, prices will come down a bit. Still if you want the best, you have to pay a premium for it. Even with pricing, this drive is still absolutely worth it. Performance is off the charts with application loads, and boot times beating out anything I have used (the majority of the Vertex series).
An SSD is arguably the most noticeable upgrades one can make to a PC, even a slower model. If you add a OCZ Vertex 3 to your system, I am certain you will be simply giddy with its performance, especially if coming from a mechanical hard drive. I just hope pricing will come down from all vendors on Solid State Technology. With that said, there really isn’t a choice here but to give this the Overclockers Approved stamp. OCZ has done it again!
Discussion
What XP Startup time do you get in PCMark05 with one of these? Probably exceeding 220MB/s...
P.S. This review showed up on top under SSDs in Sparks in Google+ :thup:
Not sure on those CDM results, but they were highly repeatable, SE, after SE, after SE, after SE. Peculiar. All other results appear spot on though..
As far as PC05 times. I havent tried it. Its got to be an absolute bear, like was the Revo x2 to try to get it under 220MB startup.
"A lot of what makes Sandforce drives perform so well comes from data compression. For example, if I write a large, originally uncompressed file to the drive, it’s easy for the controller to compress this type of data. On the flip side, a ZIP file is already compressed, and consequently cannot be easily made smaller and therefore takes longer to write to the drive. Your computer still thinks that a xx MB .doc file is being written as xx MB, but once that data hits the controller, it ideally ends up as half that value. This compression helps lower write amplification and extend the life of the drive by using fewer program-erase cycles."
Okay all that right there pretty much puts this squarely into an OS drive position. I have a couple of drives in my system, and on my non-OS drives there lives a high amount of non-compressable material, rar's, video, pics, audio.
So two questions:
1. Am I really getting the space advertised on the drive? IE, is a 120G drive really close to 120G (I know there will be data loss to overhead, formatting, backup sectors ect..). So an 80+ gig rar file could fit on it right?
2. Is this the direction all SSD's are going to take? The only market I really see for the 120+ gig drives is in laptops. Most can only handle a single drive, so if you need speed and storage then bigger is needed. In a desktop machine an SSD and mechanical drive work fine as a team. (Obviously when speed is important such as video/photo editing or if you are an avid gamer, SSD's are better.) As an OS drive it's great, but I see this compression thing as a real limiting issue depending on how it is marketed.
2. I really do not know. You can see though from the benchmarks outside of ATTO, that in-compressible data is still written/read much faster than a mechanical drive.
The biggest issue with SSD's, to me, are the price / GB.
I agree with that above anything else.
Well it's the use aspect/data type vs. real world size. Obviously on a non-compressing drive like standard rust stuff data size is data size. But say if you used one of these drives to store a fairly compressible database. Does that mean you could store more data than the actual size of the database.
-Note I'm not arguing about the formatted empty drive size vs. what the package says. I know that varies some by manufactures, I'm only talking about the data being written to the drive and how the drive handles that in terms of compression vs. free space vs. real physical bytes free. So take the 120 gig drives formatted at 100 gigs free space as an example, not a real world thing.-
120 gig drive formatted to 100 gigs, 20 gigs of text documents, but now with 85 gigs still free or 80 gigs free?
Or say you did video work:
same 100 gig formatted drive, 20 gig movie, now 80 gigs free or 75 gigs free?
It reminds me of something like Drivespace from eons ago. I had a friend with 120 meg Conner drive that actually had more physical data than could fit on the disk. Obviously when you are dealing with software it's a whole lot trickier than dealing with a piece of hardware that is handling it all for you.
I am giving OCZ and everyone else that will use these Sandforce controllers the benefit of the doubt here and when it says 120 gigs you get close to that in real uncompressed bytes. I'm just pointing out, if SSD's continue to scale then marketing needs to understand that even if you could fit 3 terabytes on a 2 terabyte drive then it still must be marketed as a 2 terabyte drive. Or market them like LTO tapes and say 120 gigs uncompressed / 160 gigs compressed.
And after reading the review, it seems that this sucker is well worth the cash. I guess its time to start saving up.