The OCZ RevoDrive 120GB is a high performance solid state drive that utilizes a pair of SandForce controllers in RAID 0. To alleviate any bandwidth limitations, it is housed on a PCI Express 4x card, offering a much fatter pipe than SATA 3/6 Gbps.
March 15, 2011 by Lawrence Lee
Product | OCZ RevoDrive 120GB PCI-E SSD |
Sample Supplier | |
Street Price | US$270 |
Interface speed has always been an advertising gimmick in the age of magnetic storage, as hard drives never delivered enough performance to push the limits of the pipes to which they were attached. At the time they were released, ATA-133 did not provide any faster performance than ATA-100, and neither did SATA 3 Gbps over SATA 1.5 Gbps. However, with the rise of the SSD, bandwidth is starting to become an issue. High performance models are pushing boundaries, particularly when performance junkies combine two or more in RAID configurations.
The OCZ RevoDrive is a high performance storage product that uses the PCI Express bus to go beyond the limits of SATA. As a PCI-E 4x card, it can handle up to 16 Gbps — one PCI-E 2.x lane can push 4 Gbps. The card/drive uses a pair of SandForce controllers which are used in many popular budget SSDs such as the OCZ Vertex 2, Corsair Force, and G.Skill Phoenix series (all 3 Gbps models). The secret to SandForce’s success is a fast compression scheme that lowers the amount of data to be written, resulting in a net performance gain. To boost speeds further and actually make use of the extra bandwidth, the RevoDrive uses a Silicon Image 3124 software RAID controller, allowing the two SSD partitions to work in tandem in RAID 0.
|
|
Not only does the RevoDrive use a bigger pipe than SATA, it frees up SATA ports to increase the total amount of storage in your system. Being a PCI-E 4x card, it requires a 4x or higher slot, so if your current motherboard has just one such slot occupied by a graphics card, you’re out of luck. It should be noted that many current boards, even budget models, have a pair of 16x slots, often with the second one running at a reduced 4x speed. The RevoDrive is an expensive proposition, with the 120GB model retailing for about US$300 compared to just over US$200 for a 120GB Vertex 2.
Specifications: OCZ RevoDrive (from the product web page) | |
Available Capacities | 50GB, 80GB, 120GB, 180GB, 240GB, 360GB, 480GB |
Capacity | PCI-Express interface (x4) |
Usage | Primary boot drive or data storage |
RAID | Internal RAID 0 |
Dimensions | 181.07 (L) x 21.59 (W) x 125.08mm (H) |
Shock Resistance | 1500g |
Seek Time | 0.1 ms |
Operating Temp | 0°C ~ +70°C |
Storage Temp | -45°C ~ +85°C |
Power Consumption | 3W Idle, 8W active |
MTBF | 2,000,000 hours |
Warranty | 3-Year |
50-80GB Max Performance | |
Read | Up to 540 MB/s |
Write | Up to 450 MB/s |
Sustained Write | Up to 350 MB/s |
Random Write 4KB (Aligned) | 70,000 IOPS |
120GB-480GB Max Performance | |
Read | Up to 540 MB/s |
Write | Up to 480 MB/s |
Sustained Write | Up to 400 MB/s |
Random Write 4KB (Aligned) | 75,000 IOPS |
TRIM / Garbage Collection
The storage portion of SSDs is composed of NAND Flash cells divided into smaller units, blocks, that are separated into pages. A problem with the way SSDs work is that only entire blocks can be deleted, not individual pages. If you delete a file, the pages it occupies are marked for deletion and the SSD continues on its merry way. When new data comes in that needs to be stored on these pages, the other pages in the block storing good data (not marked for deletion) are moved to cache, the block is wiped, and the good data is rewritten (a process called write amplification) along with the new material. To make matters worse, SSDs do not keep track of deleted pages, and doesn’t recognize their presence until it attempts a write to the block in question. As you can imagine, this can slow down performance considerably over time as the drive fills up.
Most modern drives support TRIM commands in Windows 7 that tell the SSD to perform write amplification preemptively at the time of deletion, i.e. when you empty the recycle bin or format the partition. A TRIM compatible drive has nothing but fresh pages to write to, and it never has to perform that extra read/write cycle on blocks to clean out deleted data, speeding up the write process. As there are many non-Windows 7 users out there, some drives have a garbage collection feature that actively hunts for blocks with junk data and cleans them up when the drive is sitting idle so as not to impact performance. Manufacturers can also provide a utility that can be manually run from time to time to accomplish the same thing.
Unfortunately, the RevoDrive’s Silicon Image RAID controller does not support TRIM, nor does OCZ offer any form garbage collection. For some tech savvy users, this can be a deal-breaker. When we inquired OCZ about this, they claimed that the performance degradation over time should be negligible as the RevoDrive performs so well to begin with — an argument that does have some merit as SSD performance degradation is usually not substantial. However, it is noteworthy that the higher-end REVO X2, which uses four SandForce controllers rather than two, does support idle garbage collection.
CORRECTION: According to a spokesperson from OCZ Technology who contacted us after this review first went live, “neither the REVO or the REVO X2 support TRIM; however, because the drives are SandForce based, both have garbage collection.” More on this later. |
PHYSICAL DETAILS
The RevoDrive 120GB is a full-sized PCI Express 4x card (can be used in 4x, 8x, and 16x connectors) and measures just over seven inches in length. It has two SandForce controllers placed in RAID 0 using a Silicon Image RAID controller.
Though large and well padded, the RevoDrive package doesn’t hold much. Aside from the card itself, the only accessories are a quick installation guide and a silly ego-boosting sticker. Drivers must be downloaded from the OCZ website and put on a disc or USB drive to be pre-loaded during Windows installation. |
The hardware is placed on a black PCB with the NAND Flash chips on the right hand side. The large IC on the left with the “R” monogram is a Silicon Image 3124 software RAID controller chip. There are also a series of blue activity LEDs near the PCI bracket. |
Two SandForce SF1200 series controllers are used, with each partition recognized as 55GB in the BIOS. |
There are 32 NAND chips, 16 on both side, with a total capacity of 128GB. Like most SSDs, a portion of its space is reserved as a contingency plan as cells will eventually wear out over time. |
The trace side of the PCB. |
TESTING
Our samples were tested according to our standard
hard drive testing methodology. As of mid-2008, we have been conducting most acoustics tests in our own 10~11 dBA anechoic chamber, which results in more accurate, lower SPL readings than before, especially with <20 dBA@1m SPL. Our methodology focuses specifically on
noise, and great effort is taken to ensure it is comprehensively measured
and described. Performance is covered only lightly, for reasons discussed in detail in the methodology article.
Two forms of hard drive noise are measured:
- Airborne acoustics
- Vibration-induced noise.
These two types of noise impact the subjective
perception of hard drive noise differently depending on how and where the drive
is mounted.
Both forms of noise are evaluated objectively and
subjectively. Airborne acoustics are measured in our anechoic chamber using a lab reference
microphone and computer audio measurement system. Measurements are taken at a distance of one meter from the top
of the drive using an A-weighted filter. Vibration noise is rated on a scale
of 1-10 by comparing against our standard reference drives.
Summary of primary HDD testing tools:
- IOMeter – I/O
subsystem measurement and characterization tool for single and clustered systems.
It is used as a benchmark and troubleshooting tool. - HD Tune Pro
– Benchmarking tool for storage devices and used to check/set Automatic Acoustic Management. - SPCR’s Audio Audio
Recording/Analysis system using SpectraPlus
and other utilities - SPCR Anechoic Chamber
- Custom-built HDD power
measurement and Vibration test tools
Performance Test System:
- AMD
Athlon II X4 630 processor – 2.8 GHz - Gigabyte
GA-MA785GM-US2H motherboard – 785G chipset - DDR2 Memory – 2 x 2GB DDR2-800
- Microsoft
Windows 7 Ultimate operating system – 64-bit
Performance Test Tools:
- Call
of Duty: World At War – PC game - Far
Cry 2 – PC game - ESET NOD32
– anti-virus scanner - ExactFile –
file integrity verification tool - TrueCrypt
– file/disk encryption tool - 3DMark06 –
3D benchmarking tool - Cyberlink PowerDVD
– Video player
Benchmark Details
- Boot: Time elapsed from pressing the power button to the desktop with all system tray icons loaded (minus the average time to get the “loading Windows” screen, 16 seconds)
- COD5: Load time for “Downfall” level.
- Far Cry 2: Load time for initial screen plus one level.
- ExactFile: Creating a MD5 check file of our entire test suite folder.
- TrueCrypt: Creating a 8GB encrypted file container.
- 3DMark06: Install time, longest interval between prompts.
- PowerDVD: Install time, longest interval between prompts.
- Small File Copy: Copy time for 1,278 files ranging from 10KB to 4MB
in size (HTMLs, JPEGs and MP3s). - Large File Copy: Copy time for 4 files, 2 x 700MB and 2 x 1400MB
in size (AVIs).
A final caveat: As with most reviews, our comments
are relevant to the samples we tested. Your sample may not be identical. There
are always some sample variances, and manufacturers also make changes without
telling everyone.
Ambient conditions at time of testing were 10~11 dBA and 21°C.
HD TUNE RESULTS
While not entirely indicative of real world performance, HD Tune does give us a glimpse into a drive’s inner workings. Its read transfer rate benchmark measures speed across the entire drive, giving us an idea of how fast it is over its entire span.
HD Tune measured a promising maximum transfer speed of 253 MB/s for the RevoDrive, and an average of 192 MB/s. This is well ahead of the other other solid state and hard disk drives we’ve tested thus far.
Access time was low as is the case with most SSDs, but at 0.3 ms, it was a little bit slower than the OCZ Vertex 2 and Intel X25-V.
REAL WORLD PERFORMANCE
A Windows 7 image loaded with our test suite was cloned to a 50GB partition
at the beginning of each drive and our entire
test suite was run start to finish 3 times. A reboot and defragmentation
was performed in between runs except for SSDs and the Seagate Momentus
XT (defragmentation resets optimizations made to the XT’s flash memory).
Average times were collected for comparison.
Load times were relatively poor for the RevoDrive, slightly slower than other SSDs when loading both Far Cry 2 and Call Of Duty: World at War. Boot time was particularly poor, though being a PCI-E storage device, it added a short delay after the initial POST screen, pausing to recognize the drive and offering the user a chance to enter the RAID array’s configuration menu.
The RevoDrive dominated our application performance tests, breaking records in both. This is the first time an SSD has decisively defeated performance hard drives in our TrueCrypt test which consists of creating a large encrypted file container. The RevoDrive also excelled in our ExactFile test which involves creating a MD5 checksum file of our test suite folder.
REAL WORLD PERFORMANCE (Continued)
The RevoDrive also performed superbly in our file copy tests despite our test folders containing many files that are already heavily compressed (JPEGs, MP3s, and AVIs). Overall, the RevoDrive was 35% faster than the next closest competitor, the WD SiliconEdge Blue.
In timed installs of PowerDVD and 3DMark06, the RevoDrive was less impressive, coming in third behind the Vertex 2 and Kingston SSDNow V series Gen2.
OVERALL PERFORMANCE
To gauge the overall performance of the drives, we assigned a score of 25 to the drive that excelled most in each benchmark series (loading, application, file copying, and installation) and the rest proportionally, giving each benchmark set an equal weighting. If a drive is the fastest in all categories, it receives a perfect 100 score.
The RevoDrive completed our test suite with the best overall score of just under 89, noticeably higher than the next fastest drive. However, compared to the Vertex 2, the performance increase was only 15%. The RevoDrive has about 37% faster than our best hard drive, the WD VelociRaptor 600GB.
Power Consumption
We derived power consumption figures from the RevoDrive indirectly by measuring its overall power draw on a system compared to an identical test configuration with a previously tested drive.
While the 120GB RevoDrive is essentially a pair of Vertex 2’s in RAID 0, using the PCI-E interface is no where near as energy efficient as SATA. By our estimates it consumed approximately 5.1W when idle and 7.5W when seeking, more than “green” 2TB hard drives like the Samsung EcoGreen F4 and WD Caviar Green (EARS). Compare to a single Vertex 2, the RevoDrive used eight times as much power.
Noise (or the lack thereof)
As solid state drives have no spinning platters or moving parts of any kind, they are effectively silent storage devices. It is possible that there could be a tiny bit of electronic noise (typically a high pitched squeal) being emitted, either intermittently depending on task, or continuously, but we never heard a thing, even in the anechoic chamber with ear pressed right up against it. Frequency analysis also didn’t reveal any tones that might have been generated at the edges of our audible spectrum.
Drive Comparison Tables
As always we present to you our most recent vibration, noise, and power consumption tables for previously tested hard drives. Tested SSDs were not included as they generated no vibration or audible noise. Use a perfect vibration score of ’10’ and our anechoic chamber’s noise floor of 10~11 dBA@1m for acoustics if you want to compare any of the drives below to an SSD.
NOTEBOOK HARD DRIVES | ||||
Drive Mfg date firmware version | Vibration 1-10 (10 = no vibration) | Activity State | Airborne Acoustics (dBA@1m) | Measured Power |
Hitachi Travelstar 5K320-250 250GB May 2009 firmware FBE0C40C | 8 | Idle | 13 | 1.08 W (0.85 W heads unloaded) |
Seek | 13~14 | 2.15 W | ||
Seagate Momentus XT 500GB ST95005620AS June 2010 firmware SD22 | 7 | Idle | 13~14 | 1.34 W (1.16 W heads unloaded) |
Seek | 14~15 | 2.08 W | ||
Hitachi Travelstar 5K500.B 500GB February 2009 firmware PB4OC60G | 8 | Idle | 14 | 0.7 W (0.6 W heads unloaded) |
Seek | 15 | 2.2 W | ||
Seagate Momentus 7200.4 500GB ST9500420AS March 2009 firmware 0002SDM1 | 7 | Idle | 14 | 1.1 W (0.9 W heads unloaded) |
Seek | 15 | 2.35 W | ||
Seagate Momentus 750GB ST9750420AS October 2010 firmware 0001SDM1 | 8 | Idle | 15 | 1.06 W (0.85 W heads unloaded) |
Seek | 15~16 | 2.60 W | ||
WD Scorpio Blue 500GB WD5000BEVT September 2008 firmware 01.01A01 | 7 | Idle | 15 | 0.95 W (0.8 W heads unloaded) |
Seek | 16 | 2.5 W | ||
Seagate Momentus 5400.6 500GB ST9500325ASG February 2009 firmware 0001SDM1 | 8 | Idle | 16 | 0.8 W (0.65 W heads unloaded) |
Seek | 18 | 2.2 W | ||
WD Scorpio Blue 640GB WD6400BEVT September 2009 firmware 01.01A01 | 7 | Idle | 15~16 | 0.87 W (0.74 W heads unloaded) |
Seek | 16 | 1.95 W |
5,400 ~ 5,900 RPM DESKTOP HARD DRIVES | ||||
Drive Mfg date firmware version | Vibration 1-10 (10 = no vibration) | Activity State | Airborne Acoustics (dBA@1m) | Measured Power |
Samsung EcoGreen F2 500GB HD502HI February 2009 firmware 1AG01114 | 8 | Idle | 12 | 3.2 W |
Seek (AAM) | 15 | 5.3 W | ||
Seek | 16 | 6.1 W | ||
Seagate Pipeline HD .2 1TB ST31000424CS April 2009 firmware SC13 | 7 | Idle | 12 | 4.3 W |
Seek | 16 | 7.3 W | ||
WD Caviar Green 2TB WD20EARS August 2010 firmware 01.00A01 | 9 | Idle | 12~13 | 2.8 W (2.4 W heads unloaded) |
Seek | 6.5 W | |||
WD Caviar Green 1.5TB WD15EARS March 2010 firmware 80.00A80 | 7 | Idle | 13 | 4.4 W (2.7 W heads unloaded) |
Seek | 5.8 W | |||
WD Caviar Green 1.5TB WD15EADS November 2009 firmware 01.00A01 | 9 | Idle | 13 | 4.5 W (2.8 W heads unloaded) |
Seek (AAM) | 14 | 5.8 W | ||
Seek | ||||
Seagate Pipeline HD .2 500GB ST3500414CS March 2009 firmware SC13 | 8 | Idle | 13 | 2.7 W |
Seek | 14 | 6.7 W | ||
WD Caviar Green 2TB WD20EADS February 2009 firmware 01.00A01 | 7 | Idle | 13 | 6.4 W (4.0W heads unloaded) |
Seek (AAM) | 6.5 W | |||
Seek | 13~14 | 6.5 W | ||
Samsung EcoGreen F4 2TB HD204UI August 2010 firmware 1AQ10001 | 7 | Idle | 13 | 4.0 W |
Seek | 15 | 5.6 W | ||
WD Caviar Green 2TB WD20EVDS November 2009 firmware 0.100A01 | 8~9 | Idle | 13~14 | 3.9 W |
Seek | 6.5 W | |||
Seagate Pipeline HD 500GB ST3500321CS July 2008 firmware SC14 | 6 | Idle | 14 | 4.1 W |
Seek | 15 | 6.9 W | ||
Seagate Barracuda LP 2TB ST32000542AS June 2009 firmware CC32 | 6 | Idle | 14 | 4.7W |
Seek | 17 | 7.9W | ||
WD Caviar Green 3TB WD30EZRS September 2010 firmware 01.00A01 | 8 | Idle | 14~15 | 4.1 W (3.7W heads unloaded) |
Seek | 7.5W | |||
Seagate Barracuda Green 2TB ST2000DL003 November 2010 firmware CC31 | 8 | Idle | 14~15 | 4.6 W |
Seek | 17~18 | 7.3 W | ||
Samsung EcoGreen F3 2TB HD203WI May 2010 firmware 1AN10003 | 7 | Idle | 15~16 | 4.7 W |
Seek (AAM) | 17~18 | 6.1W | ||
Seek | 18~19 | 6.9 W |
7,200 RPM DESKTOP HARD DRIVES | ||||
Drive Mfg date firmware version | Vibration 1-10 (10 = no vibration) | Activity State | Airborne Acoustics (dBA@1m) | Measured Power |
Hitachi Deskstar 7K1000.C 1TB HDS721010CLA332 February 2010 firmware JP4OA39C | 5 | Idle | 13 | 4.6 W |
Seek (AAM) | 15~16 | 6.4 W | ||
Seek | 17 | 9.6 W | ||
WD Caviar Blue 1TB WD10EALS August 2010 firmware 05.01D05 | 7 | Idle | 14 | 5.2 W |
Seek (AAM) | 16~17 | 6.6 W | ||
Seek | 20 | 8.2 W | ||
Seagate Pipeline HD Pro 1TB ST31000533CS September 2008 firmware SC15 | 4 | Idle | 15 | 7.1 W |
Seek | 16 | 10.1 W | ||
Seagate Barracuda 7200.12 500GB ST3500418AS September 2009 firmware CC37 | 8 | Idle | 18 | 4.7 W |
Seek | 19~20 | 7.9W | ||
WD Caviar Black 2TB WD2001FASS August 2010 firmware 01.00101 | 6 | Idle | 16 | 6.3 W |
Seek | 21~23 | 10.5 W | ||
Seagate Barracuda XT 2TB ST32000651AS May 2010 firmware CC13 | 7~8 | Idle | 17 | 7.0 W |
Seek | 18~19 | 7.9 W | ||
Seagate Barracuda 7200.11 1.5TB ST31500341AS October 2008 firmware SD17 | 4 | Idle | 17 | 8.8 W |
Seek | 19 | 10.7 W | ||
WD Caviar Black 1TB WD1001FALS July 2008 firmware 05.00K05 | 4 | Idle | 21 | 8.5 W |
Seek (AAM) | 10.9 W | |||
Seek | 25 | 11.0 W |
10,000 RPM DESKTOP DRIVES | ||||
Drive Mfg date firmware version | Vibration 1-10 (10 = no vibration) | Activity State | Airborne Acoustics (dBA@1m) | Measured Power |
WD VelociRaptor 600GB WD6000HLHX August 2010 firmware 04.05G04 (bare drive) | 7 | Idle | 14~15 | 4.2 W |
Seek | 20~22 | 5.5 W | ||
WD VelociRaptor 300GB WD3000GLFS May 2008 firmware 03.03V01 (bare drive) | 7 | Idle | 15 | 3.9 W |
Seek (AAM) | 20 | 5.7 W | ||
Seek | 22 | 6.2 W |
FINAL THOUGHTS
The OCZ RevoDrive 120GB is easily the fastest SSD we’ve tested and was particularly dominant in our file copy and application tests. However, as it is essentially composed of a pair of SandForce-based SSDs running in RAID 0, we were expecting more than a 15% performance increase. It is true, however, that large portions of our test suite consist of pre-compressed data which may have had a negative impact on SandForce performance. We were also disappointed that it lacks both TRIM and garbage collection to prevent performance degradation over time; this lack may be the fatal flaw which cuts short the life cycle of this product in the marketplace. Its energy efficiency isn’t great either, as we estimate it uses about the same amount of power as a typical 7200 RPM hard drive.
The most interesting thing about the RevoDrive is the interface. SSDs with internal RAID are not a new development but putting it on a PCI Express 4x card eliminates any possible bottlenecking by the 3 Gbps interface common to most SATA SSDs. The recent announcement of the SATA 6 Gbps compliant Vertex 3 line by OCZ may negate the advantage of PCI-E though. According to OCZ, the RevoDrive has a maximum read speed of 540 MB/s which shouldn’t be able to fill a 6 Gbps pipe. If you own or are planning on purchasing a SATA 6 Gbps board in the near future, a pair of Vertex 3’s may be preferable.
The OCZ RevoDrive 120GB is an incredibly fast drive, but with a street price of about US$270, it has one of the worst performance to dollar ratios in the business, representing a classic case of diminishing returns — you’re paying a lot more for a marginal improvement. Is it worth the extra cost over a single Vertex 2? For many potential buyers, the answer is probably no, but that can be said of most computer components in its price range. A super-fast drive is a luxury that every user would like to have, but not many would be willing to plunk down that much cash for one.
If your pockets are extremely deep, you might go one step further and try the OCZ RevoDrive X2 instead. It is basically a RevoDrive with four controllers in RAID 0 rather than two and garbage collection, which may be critical if you want to retain the performance you’re paying for. OCZ also has a line of enterprise drives, the Z-Drive R2/R3, which use an PCI-E 8x interface with either a four or eight way RAID.
ERRATA / COUNTERPOINT: According to a spokesperson from OCZ Technology who contacted us some hours after this review first went live…
|
Many thanks to OCZ Technology and the OCZ RevoDrive 120GB sample.
* * *
(Editor’s Note of Historical Interest: This is not the first slot-card SSD we’ve tested. SPCR veterans may recall the Cenatek Rocket Drive, a solid state PCI card based on SDRAM that emulates a hard drive, which we reviewed back in late 2002. It had many limitations and the cheapest board without any RAM was $800 but easily ran many rings around a 7200 RPM performance drive of the day.)
SPCR Articles of Related Interest:
Seagate Barracuda Green 2TB & Momentus 750GB
Hitachi Deskstar 7K1000.C 1TB vs. WD Caviar Blue 1TB
New high areal density 2-and-3 TB Greens from WD
Consumer SSD Battle: WD, Kingston, OCZ, Intel
WD Caviar Black 2TB & VelociRaptor 600GB
Seagate Momentus XT: The Best of Both Worlds?
* * *