Solid-state storage is the new hotness. Of course, that hotness is relative—this is still the storage market we’re talking about. Think less Scarlett Johansson and more Tina Fey. Nonetheless, the recent almost-affordability of blazing-fast flash drives has been an interesting development in the storage world. And it’s getting even more interesting, because in addition to an entire family of flash drives, the SSD market also has a crazy uncle.
The focus of most solid-state storage discussions centers on flash-based solutions, but they’re not the only game in town. DRAM chips offer solid-state storage, too, and they’re widely available on plain old memory modules. What’s more, even Intel’s fastest X25-E flash drive is limited to 250MB/s reads and 170MB/s writes, but your average DDR2-800 DIMM can read and write at up to 6.4GB/s. That’s like lining up a 40-horsepower motor ripped from a Vespa against the W16 monstrosity that powers the Bugatti Veyron.
Of course, you still need a way to get DRAM chips to behave as a hard drive. The hardware to do so was initially offered only for high-end servers and workstations, and it was priced accordingly. Then, about three years ago, Gigabyte went out on a limb with an i-RAM storage device that allowed users to plug four DIMMs into a PCI card that hooked up to a standard Serial ATA port. The i-RAM was ridiculously fast, and for most intents and purposes, it behaved exactly like a normal hard drive.
The original i-RAM’s total capacity was limited to 4GB of DDR SDRAM, so it’s not particularly useful by today’s standards. However, a company called ACard has done one better with the ANS-9010 RAM disk, which has eight DDR2 DIMM slots and support for up to 8GB of memory per slot. The ANS-9010 also features a pair of Serial ATA ports, allowing it to function as a single drive or masquerade as a pair of drives that can easily be split into an even faster RAID 0 array. Can this latest RAM disk live up to its staggering performance potential, and perhaps more importantly, could you live with it as a primary hard drive?
The difference DRAM makes
Before taking a closer look at the ANS-9010, I should date myself by noting that the RAM disk concept is really nothing new. Those who have been around for a while will remember that sectioning off a chunk of system memory for storage was all the rage back in the day. RAM disks were a software solution back then, so they used memory already plugged into your motherboard. The move to hardware-based RAM disks has made things much easier by allowing users to access memory modules through a standard hard drive interface without the need for drivers or even an operating system.
So why use DRAM over the flash memory everyone and their mother seems to have their fingers in? Speed, for one. I’ve already mentioned DDR2’s 6.4GB/s transfer rates. However, that 6.4GB/s is just a theoretical peak. The Serial ATA interface is currently limited to 300MB/s, and even with both its SATA ports active, ACard only claims transfer rates up to 400MB/s for the ANS-9010.
Another benefit DRAM has over flash memory is that there’s no limit on the number of write-erase cycles it can endure. Effective wear leveling algorithms and single-level cell memory can greatly improve the lifespan of a flash drive, but they just prolong the inevitable. DRAM’s resiliency does come with a cost, though. While flash memory cells retain their data when the power is cut, DRAM is volatile, so it does not. To keep DRAM data intact, you have to keep the chips juiced.
Evolving the RAM disk
Storage solutions generally aren’t much to look at, but the ANS-9010 is a little different. Instead of being tucked away in an internal hard drive bay, it slides into an external 5.25″ optical bay where you can see it.
The device’s front panel has a few ventilation slits and is littered with indicator lights for power, drive access, and battery life. ACard equips the ANS-9010 with a 7.4V, 2400 mAh battery that keeps the drive’s memory powered when the host system is shut off. Our battery was able to keep 16GB of memory powered for more than four hours, which isn’t nearly good enough if you want to turn your system off at night, but more than enough to get through the occasional reboot or hardware upgrade.
For those who would rather not trust a battery with preserving their data, the ANS-9010 also features a CF card slot up front. With the push of a button, the drive can back up or restore the contents of its memory to a flash card. This backup capability might seem like a small thing, but it greatly improves the RAM disk’s usability in the real world. ACard quotes a CF backup time of 21 minutes and 34 seconds for a 32GB configuration and a restore time of just over 14 minutes for the same config.
Busting the ANS-9010 open reveals eight 240-pin DDR2 DIMM slots. ACard says module densities up to 8GB are supported, bringing the drive’s total capacity potential up to an impressive 64GB. Keep in mind that 8GB memory modules are extremely expensive, though. For most folks, cost alone will limit the drive’s total capacity to 32GB or less. ACard doesn’t yet list any 8GB modules on its memory compatibility list, either. You’ll want to pay attention to that compatibility list, because registered and ECC DIMMs aren’t supported. It also took a couple of firmware revisions for the ANS-9010 to get along with the OCZ DIMMs we used for testing.
The RAM disk’s DIMM slots are arranged in pairs of four, with spacing similar to what you’d find on a motherboard. This leaves just enough room for modules with normal heatspreaders, although taller DIMMs may bump their heads on the 5.25″ drive bay ceiling. You can still run taller DIMMs by removing the ANS-9010’s top panel, and as long as you don’t have a 5.25″ drive in the bay directly above, most cases should be accommodating.
Between the rows of DIMM slots sits the RAM disk’s battery and really the only silicon of note. Hidden under a passive heatsink sits a FPGA chip that ACard has programmed to make the memory modules behave as a hard drive. As Intel’s X25-series SSDs have illustrated, a good storage controller architecture can significantly improve performance. ACard isn’t saying much about the ANS-9010’s storage controller design, but we’ll soon see whether it’s made the most of the prodigious bandwidth potential inherent to DDR2 memory.
We do know that ACard’s storage and memory controller has at least one interesting trick up its sleeve—it’s able to split installed memory between two physical Serial ATA ports. Even with only a single DIMM installed, the ANS-9010 can present itself as two physical hard drives via two SATA ports located at the rear. This feature allows users to wring even more performance from the drive by pairing it with a RAID controller for a little striped RAID 0 love. And because solid-state storage is immune to the mechanical failures that make running RAID 0 dicey with traditional hard drives, you don’t have to worry about a head crash hosing the array.
Switching the ANS-9010 between single-drive and RAID modes requires flipping a jumper, which is accessible from the rear. Otherwise, there isn’t much to see. The drive pulls juice from a standard Serial ATA power cable, although it only seems to draw from the 12V line.
Bring on the DRAM
The ANS-9010 arrives empty, so you have to provide your own memory. Today we’ll be testing the drive loaded with 16GB of DDR2-800 memory courtesy of OCZ. DDR2-800 is cheap and plentiful these days, and it’s often less expensive than slower speed grades that should still provide more than enough bandwidth for a RAM disk. These Gold Series modules can be found in two-DIMM, 4GB kits for as little as $44 online, pegging the cost of 16GB at well under $200. Thanks to OCZ for providing these modules for testing.
Our testing methods
We’ve assembled motley crew of competitors to line up against the ANS-9010 today. In addition to the fastest mechanical desktop drives from Hitachi, Samsung, Seagate, and Western Digital, we’ve thrown in a few flash-based SSDs from Intel and Samsung. And since the i-RAM started this whole business of inexpensive hardware RAM disks, we’ve included it, as well.
To keep the graphs on the following pages easier to read, we’ve color-coded the bars by manufacturer. The ANS-9010 appears in bright red, and we’ve tested it in single-drive and RAID 0 configs. In RAID mode, we relied on the RAID feature built into our test system’s Intel south bridge.
All tests were run three times, and their results were averaged, using the following test system.
|Processor||Pentium 4 Extreme Edition 3.4GHz|
|System bus||800MHz (200MHz quad-pumped)|
|Motherboard||Asus P5WD2 Premium|
|North bridge||Intel 955X MCH|
|South bridge||Intel ICH7R|
|Chipset drivers||Chipset 220.127.116.113
|Memory size||1GB (2 DIMMs)|
|Memory type||Micron DDR2 SDRAM at 533MHz|
|CAS latency (CL)||3|
|RAS to CAS delay (tRCD)||3|
|RAS precharge (tRP)||3|
|Cycle time (tRAS)||8|
|Graphics||Radeon X700 Pro 256MB with CATALYST 5.7 drivers|
Seagate Barracuda 7200.11 1TB
Seagate Barracuda ES.2 1TB
Samsung SpinPoint F1 1TB
Hitachi Deskstar E7K1000 1TB
Western Digital Caviar Black 1TB
Western Digital RE3 1TB
Western Digital Caviar SE16 640GB
Seagate Barracuda 7200.11 1.5TB
Samsung FlashSSD 64GB
Intel X25-M 80GB
Intel X25-E Extreme 32GB
Gigabyte i-RAM with 4GB DDR400 SDRAM
ACard ANS-9010 with 16GB DDR2-800 SDRAM
|OS||Windows XP Professional|
|OS updates||Service Pack 2|
Thanks to NCIX for getting us the Deskstar 7K1000 and SpinPoint F1.
Our test system was powered by an OCZ PowerStream power supply unit.
We used the following versions of our test applications:
- WorldBench 5.0
- Intel IOMeter v2004.07.30
- Xbit Labs File Copy Test v1.0 beta 13
- HD Tach v3.01
- Far Cry v1.3
- DOOM 3
- Intel iPEAK Storage Performance Toolkit 3.0
The test systems’ Windows desktop was set at 1280×1024 in 32-bit color at an 85Hz screen refresh rate. Vertical refresh sync (vsync) was disabled for all tests.
All the tests and methods we employed are publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.
WorldBench uses scripting to step through a series of tasks in common Windows applications. It then produces an overall score. WorldBench also spits out individual results for its component application tests, allowing us to compare performance in each. We’ll look at the overall score, and then we’ll show individual application results. You won’t find Gigabyte’s i-RAM in the graphs below because its 4GB maximum storage capacity is too limited for WorldBench to run.
The ANS-9010 gets off to a good start in WorldBench, besting the X25-E Extreme by a single point. Running the drive in RAID mode bumps its overall score by two points, which is less than one might expect. However, keep in mind that WorldBench runs through common desktop applications that tend not to be bound by storage subsystem performance. Let’s step through the individual application results to see how the ACard fares in each test.
Multimedia editing and encoding
Windows Media Encoder
VideoWave Movie Creator
Although it stays at the front of the pack through most of WorldBench’s multimedia editing and encoding tests, the ANS-9010 is a hair behind Intel’s fastest SSD in the one benchmark that spreads the field. In the Premiere test, the X25-E comes out ahead of the ANS-9010 by one to four seconds depending on whether the RAM disk is running in single-drive or RAID mode.
Another member of the WorldBench suite that benefits from faster storage is the ACDSee test, which has the ANS-9010 making an SSD sandwich out of the X25-E.
Multitasking and office applications
Mozilla and Windows Media Encoder
The field doesn’t spread much in WorldBench’s office and multitasking tests, although the ANS-9010 continues to turn in some of the quickest completion times of the lot.
Among all of WorldBench’s individual application tests, Nero and WinZip are the most demanding of the storage subsystem. In the latter, which appears primarily to stress read performance, the ANS-9010 just edges out Intel’s SSDs. The RAM disk stretches its legs even more in Nero, with our single-drive config outpacing the X25-E by 13 seconds. The RAID 0 setup is an additional 20 seconds faster than that. This test appears to have a mix of read and write requests, as evidenced by the performance gap between the X25-E and the X25-M; the X-25M’s write speeds are 100MB/s slower than those of the X25-E Extreme.
To test system boot and game level load times, we busted out our trusty stopwatch.
The i-RAM makes its first appearance in our boot time test, and it’s roughly as fast as ACard’s RAM disk. However, both are lodged in the middle of the pack and much slower than not only Samsung’s FlashSSD, but also a few mechanical hard drives.
The RAM disks claw back to the front of the pack in our level load tests. The i-RAM and ANS-9010 share the lead in Doom 3, but they’re not quite as quick as the X25-M and FlashSSD in Far Cry.
File Copy Test is a pseudo-real-world benchmark that times how long it takes to create, read, and copy files in various test patterns. File copying is tested twice: once with the source and target on the same partition, and once with the target on a separate partition. Scores are presented in MB/s.
To make things easier to read, we’ve separated our FC-Test results into individual graphs for each test pattern. We’ll tackle file creation performance first.
The ANS-9010 easily outguns the competition in FC-Test’s file creation tests. Only the i-RAM and X25-E Extreme give it much competition here, and they’re really only within striking distance of the single-drive config in two of the five test patterns. Striping makes the ANS-9010 even faster, although it’s hardly the doubling of performance one might expect—or at least hope for—from such an arrangement.
Interestingly, the ANS-9010 stumbles a little when FC-Test moves from writes to reads. ACard’s RAM disk is by no means slow, but it’s not as quick as Intel’s X25-series SSDs, which take the first two spots through four of five test patterns. Only with the Windows test pattern, which is made up of a large number of small files, does the ANS-9010 eke out a win, and then only when running in RAID 0.
Next, File Copy Test combines read and write tasks in some, er, copy tests.
The ANS-9010 returns to glory in FC-Test’s suite of copy tests, although it takes RAID to outrun the X25-E Extreme with the ISO test pattern, which is made up of a small number of very large files. Note that the gap between the RAM disk’s single-drive and RAID configs is significant with some test patterns. The ANS-9010 is also notably faster than the i-RAM across the board.
Next, we turn our attention to FC-Test’s partition copy tests, which copy data between partitions on each drive. The i-RAM’s 4GB maximum capacity precludes it from running these tests, so you won’t find Gigabyte’s RAM disk listed in the graphs below.
The partition copy tests shake out much like the first wave of copy tests, except this time the single-drive ANS-9010 sweeps the X25-E Extreme across all five test patterns. Keep in mind that the Extreme is Intel’s most expensive SLC-based solid-state drive. The X25-M uses less expensive MLC flash memory, whose comparatively slow write speeds really hurt the drive’s performance here.
We’ve developed a series of disk-intensive multitasking tests to highlight the impact of seek times and command queuing on hard drive performance. You can get the low-down on these iPEAK-based tests here. The mean service time of each drive is reported in milliseconds, with lower values representing better performance.
Our iPEAK workloads were recorded using a 40GB partition, so they’re a little big for the 4GB i-RAM, 16GB ANS-9010, and even the 32GB X25-E. The app had no problems running, but it warned us that I/O requests that referenced areas beyond the drives’ respective capacities would be wrapped around to the beginning of each drive. Since there should be no performance difference between the beginning and end of an SSD, the results should be valid.
When configured in a RAID 0 array, the ANS-9010 runs the table in iPEAK, turning in the quickest mean service times across all nine workloads. This RAID setup is notably quicker than running the ANS-9010 in single-drive mode, too. If we average the response time of each config across all workloads, ACard’s RAM disk turns in a 0.18-millisecond service time in RAID and a 0.32-millisecond service time in single-drive mode, making it nearly twice as fast when you use both SATA ports.
Of course, the single-drive config’s slower performance leaves it vulnerable to the X25-E in some workloads. In four of our multitasking tests, the Extreme just edges out the single-drive ANS-9010. ACard’s take on the RAM disk concept is consistently faster than Gigabyte’s four-year-old i-RAM, though.
IOMeter presents a good test case for both seek times and command queuing.
Wow. That’s, uh. Wow. With the exception of the web server test pattern, which is made up entirely of read requests, the ANS-9010 completely dominates our IOMeter transaction rate tests. Even in single-drive mode, it’s quicker than the i-RAM and much faster than the X25-E. The Intel SSDs prove worthy competition with the web server test pattern, but they’re still trumped by the ANS-9010 running in RAID mode. Splitting ACard’s RAM disk into a striped array just about doubles its transaction rates at higher load levels.
Of course, there’s a price to be paid for those high transaction rates. The ANS-9010 also consumes quite a lot of CPU power in the process. To put these results into perspective, we’ve whipped up another set of graphs that illustrates the transaction rate per percent CPU utilization. Since our mechanical hard drives don’t deliver anywhere near SSD levels of performance here, we’ve left them out of the equation, with the exception of the VelociRaptor.
The ANS-9010’s RAID config may be exceptionally fast, but there’s a greater CPU utilization cost for each transaction than with a single-drive setup. Interestingly, the i-RAM and the single-drive ANS-9010 are evenly matched here. However, Intel’s SSDs have the highest transaction rates per CPU cycle overall.
We tested HD Tach with the benchmark’s full variable zone size setting.
So much for DRAM’s peak bandwidth advantage over flash memory. In HD Tach’s sequential transfer rate drag race, the ANS-9010 only manages to read at 173.6MB/s and write at 144.1MB/s. What’s more, it’s slightly slower when running in RAID 0. Those sustained transfer rates do put ACard’s RAM disk ahead of the i-RAM, but the X25-E Extreme is much faster. Even the X25-M is quicker in the read speed test.
The ANS-9010’s burst performance isn’t as quick as one might hope, either. The drive doesn’t even manage 200MB/s with one SATA port plugged in, although it does eclipse 300MB/s when running in RAID.
Solid-state drives do exceptionally well in random access time tests, and the ANS-9010 is no exception. The drive’s seek times are essentially instantaneous, and interestingly, they’re just a little bit quicker when running in single-drive mode. I’m hesitant to make too much of the 0.1-millisecond difference between single-drive and RAID modes, but it was consistent across three test runs, suggesting that there’s a hint of a latency penalty associated with splitting the RAM disk’s memory between two SATA ports.
HD Tach’s margin for error in the CPU utilization test is +/- 2%, which narrows our results considerably.
Noise levels were measured with an Extech 407727 Digital Sound Level meter 1″ from the side of the drives at idle and under an HD Tach seek load. Drives were run with the PCB facing up.
Because it has no moving parts, the ANS-9010 is completely silent. The 42.6-decibel scores you see for the SSDs represent the ambient noise levels of the test system itself.
For our power consumption tests, we measured the voltage drop across a 0.1-ohm resistor placed in line with the 5V and 12V lines connected to each drive. Through the magic of Ohm’s Law, we were able to calculate the power draw from each voltage rail and add them together for the total power draw of the drive. Because the i-RAM pulls its power from a PCI slot, we were unable to measure its exact power draw and include it in the results below.
Regardless of what it’s doing, the ANS-9010 pulls just under 12 watts with 16GB of memory spread across all eight of its DIMM slots. The drive’s power consumption is significantly higher than that of flash-based SSDs and even most of our hard drives, but it’s not entirely unreasonable.
I have to admire ACard for coming up with the ANS-9010 RAM disk. Not only does it make the solid-state storage market a little more interesting, but it occasionally offers truly inspiring performance, as well. The drive easily handled our disk-intensive iPEAK multitasking and multi-user IOMeter loads, besting the fastest SSD on the market not only when running in RAID mode, but more often than not, in a single-drive config, too. However, the ANS-9010’s dominance wasn’t universal. Sure, it was the quickest drive of the lot in WorldBench, but not by a significant margin. And although it exhibited the fastest write and copy speeds in FC-Test, Intel’s SSDs proved quicker when it came time to reads.
The battle between the ANS-9010 and Intel’s X25-E Extreme SSD was certainly an intriguing one to watch, in particular because the Extreme’s storage controller seems to be the best in the business. A relatively poor performance in HD Tach’s sustained and burst transfer rate tests suggests this latest RAM disk still has room to improve on that front. Sure, it handily beat the i-RAM across the board, but Gigabyte’s take on the inexpensive RAM disk concept is several years old.
Despite leaving some DRAM performance potential on the table, ACard has done a number of smart little things to make the ANS-9010 easier to live with and more attractive than the original i-RAM. With eight DIMM slots and support for configurations up to 64GB, there’s plenty of storage capacity to be had. The CF backup feature is really nice given DRAM’s volatility, too. And let’s not forget the ANS-9010’s hint of memory mitosis, which splits available capacity for RAID configurations that can, in some cases, almost double performance.
So how much does all of this cost? $380 for the ANS-9010, plus the cost of memory. The 16GB of DDR2-800 DIMMs we used today will set you back roughly $176, bringing the drive’s total cost up to about $555. A 32GB configuration made up of 4GB modules, which run around $100 each, will set you back just under $1200 when all is said and done. This latest RAM disk isn’t cheap, then. But neither are SSDs that offer equivalent performance. Intel’s 32GB X25-E Extreme runs around $600, for example.
For most applications, the fact that the X25-E Extreme comes in a smaller form factor, consumes much less power, and actually has a lower cost per gigabyte than the ANS-9010 makes the RAM disk a tough sell. However, the ANS-9010 can be much faster under the right circumstances, and there’s less reason to worry about the drive eventually wearing out.
Even if it may ultimately be relegated to a niche of the ultra-high-performance storage market, I quite like ANS-9010. ACard should be applauded for doing something a little different. And speaking of a little different, you can also get an ANS-9010B RAM disk for about $250. This B revision is limited to six DIMM slots and 24GB of memory, and it lacks the second SATA port for RAID, but it’s nice to have the option of a cheaper alternative with fewer frills.