OCZ Revo 3 X2 240GB Review
First up is of course ATTO. ATTO is universally used by SSD manufacturers to show what the drives maximum read/write capabilities are under ideal circumstances. The bandwidth results here are not necessarily entirely reliable for reporting the maximum drive speeds, but it does help convey results of the bandwidth for different file sizes.
Not only does the drive meet the performance specifications of up to 1500 MB/s read and up to 1225 MB/s write it exceeds the specifications by quite a bit coming in at an astounding 1602 MB/s reads and 1487 MB/s writes! The highest performance we have ever seen from a single drive.
Closer Look
The OCZ Revo 3 X2 is entirely unlike any other SSD we’ve seen (other than the previous OCZ Revo series drives) since it’s technically not really a drive. It looks and operates more like a “card” than a drive; call it a Solid State Card if you like. It’s about the same size as a large audio card, if that’s any indication. The major difference here from all other SSDs is the interface: the Revo 3 X2 installs directly into the motherboard in a PCI-E x4 slot, rather than sitting in a drive cage and tethered to the motherboard by a SATA cable.
The advantage of the PCI Express x4 interface is it has some insane bandwidth available, and won’t constrain the drive speeds on the Revo 3 X2 like a SATA interface would. Further, this setup could potentially save space inside a case by making the drive cage unecessary. However, the downside is you need an available PCI-E x4 slot (or an x16 in a pinch if you want to sacrifice a graphics expansion). If you’re running a high-end system with dual graphics cards and a discrete audio card, some motherboards might not have enough PCI-E slots left over for the Revo 3 X2.
The difference between the OCZ Revo 3 and the Revo 3 X2 is the former has just one PCB with two partitions, whereas the latter has two PCBs and four partitions. One moves at super speed, the other moves at ludicrous speed.
Of course, there are a couple downsides to the RAID 0 setup on the Revo 3 X2. While the Revo 3 technically supports TRIM, it doesn’t function in actual usage.
Test Setup and Methodology
Before we begin, there are a few important caveats regarding test procedures. Testing a Solid State Drive doesn’t use quite the same as the approach as the methodology for a mechanical hard disk drive.
SSDs come with an internal disk controller present for wear-level management, which then interacts with the motherboard controllers, so the multitude of combinations can be staggering, and the performance can certainly vary. Additionally, since SSDs are flash-based products, there is also a degree of variability between using high performance or inexpensive DRAM chips that can also theoretically bias the results.
Drive performance tests are also normally developed for mechanical hard drives, which are designed to sample traditional sectors for performance along the length of the drive platter (both the inner and outer rotational portions of the drive). The issue, however, is that SSDs don’t read or write from the same sector, so we end up with testing suites that are attempting to perform a job they’re not specifically designed for. It’s similar to using a wrench to bang in a nail where a hammer should be used; the wrench will do the job but it won’t be pretty.
Consequently, we’ll be running a variety of tests, that when viewed together as a whole, should provide a “big picture” result. Therefore, take all SSD benchmark results with a grain of salt and remember to take them into context. There’s really no point in arguing the minutiae of test results since test tolerances can account for that; we prefer to focus on the bigger picture and what that means to consumers.
We’ll be using a Z68 chipset motherboard , along with the following software to try to paint as accurate a picture as possible for SSD testing.
Results: ATTO
First up is of course ATTO. ATTO is universally used by SSD manufacturers to show what the drives maximum read/write capabilities are under ideal circumstances. The bandwidth results here are not necessarily entirely reliable for reporting the maximum drive speeds, but it does help convey results of the bandwidth for different file sizes.
Not only does the drive meet the performance specifications of up to 1500 MB/s read and up to 1225 MB/s write it exceeds the specifications by quite a bit coming in at an astounding 1602 MB/s reads and 1487 MB/s writes! The highest performance we have ever seen from a single drive.
Results: Crystal Disk Mark
Next we will test with Crystal Disk Mark. Crystal Disk Mark is a lightweight benchmark that has two parameters, Sequential and Random read/write, to evaluate the performance of a drive. Crystal Disk Mark’s default test uses random data and the writes in particular can be very punishing on Sandforce drives and double punishing on asynchronous nand equipped drives and triple punishing on lower GB drives such as our 240GB model. First up we will test with default random data.
Here we see the penalty incurred with random data, asynchronous nand, and a smaller sized drive. While the sequential R/W does suffer, the supremely important 4k performance is absolutely astounding! Here again, the highest we have ever scored with any drive. Keep in mind, the 4K test is indicative of how well the drive can perform with what is typically found in daily use on an operating system drive. All I can say is WOW! Next let’s see what the same test looks like using compressible data.
This is where the SandForce controller really flexes its muscle….compressible data. Again looking at the all-important 4K scores, they are nothing short of amazing and, are the fastest we have ever scored from a single drive.
Results: IOMeter
Finally we will test using IOMeter 2010. While marketing hype from companies tends to focus on big sequential numbers that seek to impress consumers, we know that the reality is that random read/writes, particularly with small file sizes, are critical for an OS drive. Consequently, we use IOMeter with a custom workers setup to simulate these conditions to assess the suitability and performance of an SSD as an OS drive.
IOMeter is a subsystem tool that was originally developed by Intel, then later discontinued but given to the Open Source community, which has been working to continue the development of the application. It is a powerful tool that can simulate just about any disk scenario that can be imagined.
For our test, we use 100% random commands of 4K size, in a custom blend of Read and Write operations. In our opinion, this will simulate much of what an average user would encounter in a daily setup during normal computer use.
Summary
As we’ve said before and we’ll continue to hammer any chance we can get, the single biggest bottleneck in modern systems is the traditional plattered hard drive, and using an SSD in your system will make such a massive difference that you’ll want to cry if faced with having to use a traditional hard drive afterwards. Make no mistake, the speed boost is that pronounced and alluring.
The SandForce controller is behemoth, and with a RAID configuration, the results are nothing short of astounding.
Of course, blazing speed doesn’t come cheap. Costing about $710 USD, the 240GB version of the Revo 3 X2 is almost stratospheric in price. However, four 60GB OCZ Vertex 3 drives would cost about $550 and offer the same capacity and potential speed, but really would need a RAID card to get the most out of that setup, so that would add to the overall cost. Of course, there is a price premium on the Revo 3 X2 for the convenience of having a single, thin drive card.