Logo   Banner   TopRight
Gigabyte I-RAM Storage Device
Date: Mar 28, 2006
Author: Robert Maloney

Although performance has been greatly enhanced over the last few years through the use of faster spindle speeds, larger caches, and newer interfaces, hard drives still tend to be a bottleneck in a typical PC. Mechanical devices, by their very nature, require time to seek the data requested, read it, and then transfer that information back to system memory. Also, while SATA technology has increased transfer rates from the paltry 33MB/s of the original ATA spec to a healthy 300MB/s on the latest SATA-II controllers, there's still room for improvement, although the drives themselves will still struggle to saturate the bandwidth on this interface.

One idea, although hardly new, is to create a solid-state drive with no moving parts whatsoever.  Past attempts, such as Cenatek's RocketDrive, have been very costly and as a result have not found much acceptance in the retail market.  The optimal result is to achieve the right price vs. capacity ratio; a drive large enough to install files or applications to that won't cost much more than the rest of the system's individual components.  What we've received from Gigabyte may just fulfill those requirements.  The i-RAM, or GC-RAMDISK as it is officially known, is a single PCI card that has support for up to 4GB of memory, using nothing more than standard DDR DIMMS, regardless of their respective speeds.  Recognized as an ordinary drive by the system, it's a simple, straightforward approach at increased drive performance.  Let's find out if this drive is "i"-RAM tough, or just a sheep in wolves's clothing.

Gigabyte's i-RAM GC-RAMDISK
Specifications & Features
Power Interface
- PCI 2.2 mechanical compliant slot

Data Transfer Interface
- One onboard SATA connector supports 1.5GB/s data transfer rate

Memory Interface
- 4 DIMM slots
- Supports Unbuffered / Non-ECC DDR 200/266/333/400MHz
- Total Capacity: 4GB (4x 1GB modules)

- Memory Height: up to 4.2mm

Backup Battery
- One 1600mAh lithium battery

Form Factor
- 22.0 x 10.4cm

System Requirements
- One available PCI 2.2 compliant slot
- One available SATA 1.5Gb/s compliant connector
- Motherboards based on the following chipsets:
Intel: ICH6 / ICH6R / ICH7/ ICH7R
VIA: 8237R
SiS: 964 / 965L
NVIDIA: NF4 SLI / NF4 4X / NF4 SLI Intel Edition
ULi: M1689

Gigabyte I-RAM



Gigabyte I-RAM



The Bundle:

There's nothing fancy about the I-RAM's package contents, but with a new product such as the I-RAM that doesn't have much competition or require many accessories, that can be somewhat expected.  The manual is actually larger than you might think considering installation consists of nothing more than installing DIMMs onto the i-RAM, and then sliding the card into an available PCI slot and connecting a cable.  The manual quickly covers additional topics, such as Data Loss Prevention including the use of Gigabyte's proprietary backup tool.  The Driver CD covers one scenario: installation onto one of NVIDIA's chipsets, although these drivers come standard with every nForce based motherboard.  Typical installations do not require any drivers at all, as the i-RAM is immediately recognized as a standard hard drive in the BIOS and in Windows.  The last item in the box is a short SATA cable that has a unique bend in one connector, which works perfectly with the vertical connector in the upper corner of the i-RAM.


A Closer Look and the Test Setup

A Closer Look at the Gigabyte i-RAM
No moving parts = no sound



The Gigabyte i-RAM is not a fancy looking card at all, with an aesthetic that look more like a basic daughter-board than a super-fast drive.  But, as the story goes, it's not what's on the outside but rather on the inside that really counts.  The "brains" behind the operation come from the XILINX Spartan-3 chip, which acts as both the DDR memory controller and the SATA controller as well.  It also must perform the translation between the two, taking in data requests from the SATA bus, relaying that info to the memory and then back.

The i-RAM's face is mostly taken up by the four DIMM slots on the front of the card and by the large lithium battery placed directly to the right.  The i-RAM's state is relayed to the user by small LEDs directly above it, telling the user whether or not the card is properly detected by the system, if its actively powered, and whether the battery is charging or already at full charge.  Another LED is placed on the other end of the card, useful for displaying the battery level - the more LEDs are lit, the more power remains in the battery.  In normal operation, the i-RAM receives power from the PCI slot with the backup battery providing peace of mind for any lost data -- for about 15 hours, that is.


The hard part when installing the RAM partly comes from choosing the right sticks.  With slots placed so closely together, we found that we couldn't install some of the RAM we had laying around the lab.  Most of these had thick heatspreaders on them which wouldn't fit side-by-side and we wound up ordering some new Kingston 1GB modules instead.  This, of course, is a big caveat to any would be i-RAM buyers: the total cost one needs to consider should include not only the i-RAM, but possibly a full allotment of RAM to populate it.  Size might also be a concern for some users, especially SFF owners, as the i-RAM is easily as large, if not larger, than most graphic cards including Gigabyte's own GV-NX78T256V-B (pictured above).

Test System Specifications
It's a Gigabyte to chew on
Gigabyte GA-8N-SKI
(NVIDIA nForce 4 SLI X16)

Intel Pentium 4 550 (3.4GHz)

2x512MB Corsair DDR2
CL 4-4-4-12

Gigabyte GV-NX78T256V-B (GeForce 7800 GT)
On-board Ethernet
On-board Audio

Gigabyte I-RAM v1.2

4x 1GB Kingston KVR400X64C3AK Modules
Western Digital SE16 250GB HD X2
7,200 RPM SATA 300 16MB Cache
Seagate Barracuda V HD X1
7,200 RPM SATA 150 8MB Cache

Windows XP Pro SP2
nForce 4 Drivers v6.85
NVIDIA Forceware v81.98
DirectX 9.0c

Our test methodology for all drives we benchmarked was fairly straight forward.  The procedure for all drives was to set them up as the primary boot device, format them as a standard Windows XP NTFS partition, and install Windows XP SP2 onto them.  For the RAID array we utilized nVIDIA's RAID BIOS menus to configure a striped RAID 0 array with the default "maximum performance" 64K stripe size.  After the drivers and benchmarks were installed, we ran a quick defrag to optimize them further.  The only test in which we varied from this course was when testing Far Cry - a game that took up more space than the i-RAM could handle when combined with the OS files.   For this test only, we installed the OS on a separate drive, and then benchmarked Far Cry from the secondary location. 

SiSoft SANDRA Benchmarks

We began our testing with SiSoftware's SANDRA File System benchmark module. This test's method of hard disk performance analysis it what we would consider a "light duty" consumer-level evaluation tool.  The folks in IT would have your head for recommending a drive based solely on SANDRA File System test results.  However, the benchmark is a popular utility within the performance PC enthusiast community and it does give a decent quick swag at high-level throughput characteristics of the total storage subsystem, which of course includes HD controllers and other associated system components.

Preliminary Benchmarks With SiSoft SANDRA 2005 SR2

SANDRA's Read Performance tests show some surprising levels of parity for the Gigabyte i-RAM.  The reading scores were 135-136 MB/s for each test, placing it well above the results for a single SATA drive.  Comparing the i-RAM to a pair of Western Digital SE16's in RAID 0 is a mixed bag, however.  Random Read speeds were double for the i-RAM over the RAID configuration, yet barely over 10MB/s better in sequential testing, and not even comparable performance in buffered reads due to the fast 16MB cache found in these drives.

Write Performance Tests were, for the most part, a mirroring of what we found with the read scores.  Again the i-RAM runs pretty consistently at all three tests, and easily outpaces the Seagate Barracuda.  And once again, the RAID 0 combination of SE16 drives has numbers that fall to either side of the i-RAM's performance levels.  Random write scores are lower, but sequential and especially buffered writing bandwidth is better with the RAID configuration.

SANDRA's Drive Index is the final measurement of the drive's performance.  Overall, the Gigabyte i-RAM came out on top at 131MB/s - almost exactly the same score that can be found in SANDRA's internal database.  Our RAID 0 setup was 28MB/s lower, roughly a 20% lower score than the i-RAM.  A single SATA drive may be the cheapest and easiest solution when building a PC, but it only scored a meager 37MB/s in SANDRA, almost 100MB/s slower than the i-RAM.

PCMark05 Hard Drive Tests

Next up is PCMark05 from FutureMark Corp.  We specifically used only the HDD Test module of this benchmark suite to evaluate all the drives and configurations we tested versus the Gigabyte i-RAM.  We consulted Futuremark's white paper on PCMark05, for an understanding of what this test component entails and how it calculates its measurements.

Futuremark's PCMark05

Courtesy, FutureMark Corp. -
For these tests, we use RankDisk, an application developed and copyrighted by Intel. In our testing, we found RankDisk to be suitable for a neutral benchmark. RankDisk is used to record a trace of disk activity during usage of typical applications. These traces can then be replayed to measure the performance of disk operations for that usage. RankDisk records disk access events using the device drivers and bypasses the file system and the operating system's cache. This makes the measurement independent of the file system overhead or the current state of the operating system. In replaying traces, RankDisk always creates and operates on a new "dummy" file. This file is created in the same (or closest possible) physical location of the target hard disk. This allows the replaying of traces to be safe (does not destroy any existing files) and comparable across different systems. The traces contain different amount of writing and reading on the disk; total ratio in the HDD test suite disk operations is 53% reads and 47% of writes.  The following input traces are used:

Windows XP Startup:  This is the Windows XP start trace, which contains disk activities occurring at operating system start-up. The test is 90% reading and 10% writes. This trace contains no user activity.

General Hard Disk Drive Usage:  This trace contains disk activities from using several common applications.
These are:
- Opening a Microsoft Word document, performing grammar check, saving and closing
- Compression and decompression using Winzip
- Encrypting and decrypting a file using PowerCrypt
- Scanning files for viruses using F-Secure Antivirus.
- Playing an MP3 file with Winamp
- Playing a WAV file with Winamp
- Playing a DivX video using DivX codec and Windows Media Player
- Playing a WMV video file using Windows Media Player
- Viewing pictures using Windows Picture Viewer
- Browsing the internet using Microsoft Internet Explorer
- Loading, playing and exiting a game using Ubisoft™ Tom Clancy's Ghost Recon
The General Usage trace is 60% reads and 40% writes.

Virus Scanning:  Virus scanning is a critical task in today's PC usage. As the major bottleneck of scanning
viruses is in hard disk activity, it is reasonable to include virus scanning as a HDD test. The test consists of HDD
activity of scanning 600MB of files for viruses. The Virus Scanning test is mostly disk reading (99.5%).

While SANDRA's tests were generally mixed when it came to comparing the i-RAM with a RAID 0 setup, PCMark05's results were much more aligned in Gigabyte's favor.  The Virus Scanning benchmark relies almost solely on reading performance, and as such, the cache found on the Barracuda and SE16 drives helps make up some ground on the i-RAM.  The General Usage and WinXP Startup tests, on the other hand, are simply blowouts.  It used to be enough to look at RAID vs. single drive setups and point out the 30-40% increases.  When looking at the i-RAM's performance though, the differences were more on the magnitude of 1500% over a single drive's general usage and 1000% better than RAID 0.  

PCMark05's overall performance score is huge victory for the i-RAM.  We're quite familiar with seeing performance numbers in the four digit range, but to see 33067 show up at the end of the test ensures a double-take.  As with the General Usage and WinXP Startup tests, there's really no way to compare standard hard drives to the Gigabyte i-RAM.

HD Tach 3.0.1 Performance

The three HD Tach metrics that we sampled for each drive setup are Random Access, which is largely dependent on spindle speed of the drive, Read Average, and Burst Read throughput.  Again, Read Average scores are what we would consider the most meaningful to the end user, whereas Burst Read is typically tied into the drives on-board cache and SATA interface speed.

Single / RAID 0 Performance
Performance with HD Tach 3.0.1

If the results in HDTach 3 look somewhat familiar, it's possibly because they look very close to the SANDRA Read Performance scores.  The main difference would be the random access score, as it is reported in time rather than bandwidth.  As such, the numbers are particularly low, and in the case of the i-RAM it's a big fat zero.  When it comes to access speed, the i-RAM truly is king.  The Burst and Average Read performance for the i-RAM are up close to the limit of the SATA interface, both of the scores coming into the 130+ MB/s range.  The RAID 0 setup is roughly 25% slower when it comes to the average read score, but moves up well past the i-RAM in burst speed, partly because of the large amount of cache, but also since it is unhindered by SATA I limitations.

Real-world Timings

Now that we've ran the i-RAM through the typical drive benchmarks, we're going to get to the part that we think many of you are most interested in: real world usage.  Meaningless points and performance speeds are great for comparison, but the real fun is watching it perform.  As such, we've rigged up a few tests that we deemed the i-RAM would be ideal for: installing as a boot partition or gaming drive.  The size limit of the i-RAM prevents it being used for much more than either application, and even then could probably only support one modern game at a time.  For the boot drive tests, we timed how long it took to install Windows XP SP2 from start to finish, and then how long it took for our GA-8N-SLI to completely boot up into Windows. 

I-RAM Real-World Performance - Windows XP
Installation and Boot Times Measured

Installing Windows XP was a little tricky as we had to make sure we stopped our watch for any user-input screens and other sections that would lead to variances in the timings.  For example, formatting a 4GB drive compared to a 250GB drive would ensure the 4GB winning.  As such, we timed only the install process and nothing else and stopped when the first Windows XP screen was completely loaded.  The i-RAM came in at the fastest time, but it wasn't really a runaway with the RAID 0 array only a minute and half behind.  Other factors such as reading files from the CD come into play here as well, and we figured the boot time would show more of a difference.  Percentage-wise, it did, as the i-RAM finished first, a full four second faster than the RAID 0 array, and 6 seconds faster than a single Barracuda.  That's approximately 11% faster boot times than the RAID setup, and 16% faster than a single drive.

Far Cry Level Loading Times
Tired of waiting for the next level

4GB only gets you so far, as many of today's games can take up nearly as much, and in the future they will take up even more space.  For games with extremely long loading times (read Battlefield, Doom, etc.) the i-RAM could come in handy by decreasing load times.  We took that premise to task by loading in the Catacombs level on all three drive setups:

Initial loading of this level on our usual drive, the Seagate Barracuda V, took almost 35 seconds of waiting before we could get in and play.  That's a lot of thumb-twiddling and mouse-swirling between levels.  RAID 0 arrays didn't fare all that much better, shaving off just over a second and a half.  Using the i-RAM really sped things along, however.  It still took 26.3 seconds to get to fragging, but that's over 8 seconds quicker that the original timing, almost a 25% improvement.

Benchmark Analysis and Our Conclusion

Benchmark Analysis: Looking back at the benchmarks, its obvious that the Gigabyte i-RAM is fast.  There's really no comparison in many tests, except for the buffered/burst reads which are more of a gauge of the performance of a hard drive's cache and interface than anything else.  A PCMark05 score of 33067 is simply unheard of with standard hard drives, and almost seems to good to be true.  Of course, synthetic benchmarks don't always tell the whole story.  By looking at the real-world performance results, however, we see that there are actually some measurable benefits for using the i-RAM as a Windows or game installation drive.

Gigabyte made some lofty promises with the i-RAM, and they seem to have made good on just about all of them.  Installation of the card is a simple process, as it is recognized just like any other standard drive and doesn't require any special drivers or other software.  There's also a backup battery that can preserve data for up to 15 hours, even if the card has been removed from the system. This feature should help alleviate some worries that user's will lose data during a power outage, or when shutting a machine down for the night.  And most importantly, the i-RAM's performance made everyday single drives look pale in comparison, and it typically bettered our RAID 0 setup in everything but burst reads as well.

There are some major drawbacks to the i-RAM, however.  The current version has only 4 DIMM slots, and can only support up to 1GB per slot of ordinary DDR RAM, and the RAM cannot have any heatspreaders.  While it's possible that users might have some RAM lying around, it's doubtful that they would be large 1GB modules, so potential customers will likely have to purchase a few DIMMs as well.  Power-users that have already spent $1000+ on SLI setups will probably have no issue throwing down an additional $375 or so to get their hands on a fast solid-state device like the i-RAM, but the price of the drive itself (~$140), with some RAM, combined with the i-RAM's relatively small capacity will surely limit this product's appeal. Still, it's great to see Gigabyte moving in a new direction, and we hope they introduce a larger capacity version of the i-RAM in the future.  Our final grade: an 8.5 on the HotHardware Heat Meter.

_Fastest single drive.  Period.
_Backup battery saves data
_Powered by PCI slot
_No moving parts = quiet use
_4GB limit on RAM
_DIMM slots are too close together

Discuss This Article In HotHardware's PC Hardware Forum Right Now!

Content Property of HotHardware.com