|by Babrbarossa | June 3, 2007|
Buffalo Firestix PC2-6400 vs. Crucial Ballistix PC2-8000 review
Buffalo Firestix PC2-6400 vs. Crucial Ballistix PC2-8000
Reviewed by Babrbarossa
If there ever was a time to rush out and buy DDR2, it is now. Prices on many modules are less than half; and in some cases a third of what they were less than six months ago. It’s hard to say where prices are going in the near future, but some sources say that manufacturers of memory chips are already beginning to hold back production in an effort raise prices again. While the tech sites around the web are currently abuzz with talk of DDR3, it will be some time before the price/performance ratio comes anywhere near that of DDR2.
Two manufacturers that are currently duking it out on this very competitive DDR2 playing field are the venerable Crucial from Boise, Idaho and the relatively new kid on the block in terms of performance memory, Buffalo Inc. based in Japan. Crucial is the retail subsidiary of Micron which is renowned for producing great overclocking memory chips, which they supply to pretty much every manufacturer that’s worth their salt. Buffalo is a subsidiary of Melco Holdings which began producing audio equipment way back in 1975.
In this review, I will be comparing the performance of a 2X1GB kit of Crucial Ballistix DDR2 1000 with EPP to a 2X1GB kit of Buffalo Firestix DDR2 800. Despite their different frequency ratings, these two kits both contain Micron D9GMH memory chips and should be expected to perform somewhat similarly.
The reason I chose these two modules for comparison is that they are reputed to perform well, they have the same chips (although they have different frequency ratings), and are the same retail price ($185 CAD at NCIX), but are regularly on sale. At the time I acquired these units, the Crucial were on sale for $139 CAD, and the Buffalo were on for $129 CAD which are amazing prices for this kind of performance.
Crucial Ballistix DDR2 1000, PC2-8000 Unbuffered x64 Non-ECC, 240 Pin- EPP enabled
Part #: BL12864AA1005.16FD3
Capacity: 2X1GB kit
The Crucial modules are shipped in a no-nonsense, but very protective cardboard box and individually wrapped in static-resistant bags.
Aesthetically, I find these modules to be downright beautiful. While the aluminum heatsinks are fairly thin and plain, they are anodized with a unique metallic orange/yellow color and have a textured silver border around the edge.
I did notice, however, that the layers of rather thick thermal tape which create the thermal interface between the heatsinks and the memory chips aren’t quite long enough to fully cover all of the chips. I also noticed that on one of the sticks, the heatsink made no contact at all with two of the chips on the end of the PCB. Despite this minor flaw, adequate airflow should keep the memory cool enough. One other thing that I noticed while handling these is that the small clips on the back of the heatsinks are not secured very well, so care must be taken not to knock one off on installation lest they land on your video card and possibly short something out.
The Crucial modules are built on their trademark shiny black, 6-layer Ballistix PCBs, which are designed by Crucial themselves. I was told by one of Crucial’s design engineers that their PCB is not a standard JEDEC design as their design team had instead designed a module that was a “pull-no-punches, cost-no-object, speed-demon” based on what Crucial was looking for in a PCB. Their designers tossed out the conventional paradigms that most modules are manufactured by and decided to build for performance first and cost second. This makes the PCB somewhat more difficult to manufacture and does make the module more expensive to build, but it also means that the only limitation is the speed of the parts on it. This module was originally designed for PC2-6400, but can and does go much faster, of course.
The Crucial modules are EPP (enhanced performance profile) enabled, which allows motherboards with this feature to automatically detect the optimal memory settings and set the timings and frequency accordingly. Most overclockers, however, will typically choose to find the stable limits of their memory on their own as it is typically faster than the EPP settings. As mentioned before, the chips on these modules are Micron D9GMH and are originally rated at DDR2 800 4-4-4-12 but are binned at a stock DDR2 10000 5-5-5-15 which is where the EPP settings will put it if you choose to enable it.
The other thing that is definitely worth mentioning is that Crucial backs their memory with a lifetime warranty, which also guarantees that they run at the advertised settings.
Buffalo Firestix DDR2 800, PC2-6400 Unbuffered x64 Non-ECC, 240 Pin
Part #: FSX800D2C-K2G
Capacity: 2X1GB kit
The Buffalo Firestix arrived in a simple clamshell package- not quite as secure looking as the Crucial, and definitely not as satisfying to open, but secure enough.
The modules themselves are also stunning to look at with their blazing red heatsinks which are slightly contoured for stylistic design purposes. I found it hard to decide which I liked better so just for a lark, I presented one of each, along with an OCZ Gold module, to my wife to see which one she thought looked best- She ranked them OCZ, Firestix, then the Ballistix.
The clips on the spine of the modules are beefier than those on the Crucial and the aluminum heatsinks are thicker.
The 6-layer Firestix PCBs are not quite as glamorous as Crucial’s (if PCBs can be considered glamorous), being the typical green color. There have been murmurings around the web about the Buffalo PCBs being of less than stellar quality which supposedly prevents them from tolerating higher voltages and so do not overclock as well as modules with beefy PCBs. For this reason, I was particularly interested in testing them against the equally priced Crucials which are supposed to have good PCBs. The recommended voltage for the Firestix is also lower, at 2.1V.
Like the Crucials, the Firestix have Micron D9GMH memory chips, but they are originally binned as PC2-5300 (DDR2 667) whereas the Crucials are originally binned at PC2-6400 (DDR2 800). Because of this I guessed that the Firestix would perhaps run very well at low latencies at low frequency settings, while the Ballistix might do better at higher frequencies and relaxed timings. So on to the testing to find out…..
Motherboard: EVGA 680i A1
CPU: Core 2 Duo e6400
GPU: EVGA 8800GTS 640MB
HD: 320GB Seagate Barracuda 7200 RPM 16MB Cache
Watercooled CPU, GPU, NB
OS: Windows Vista Home Premium
SuperPi MOD 1.4 (1M test- average of three runs)
Everest Home Edition, version 2.20.405
1) What are the lowest stable timings for each kit at DDR2 800, and what are the highest stable frequencies at stock 5-5-5-15 timings.
2) How does the performance of each module compare at these settings, and which combinations yield the best memory performance.
A setting was considered stable if MemTest successfully tested to 600% and all benchmarks were successfully run.
One advantage of conducting these tests using an NVidia 680i chipset is that the memory can be unlinked so that the frequency can be adjusted independently of the front side bus. This makes it possible to measure the differences in performance that are a product of the memory and not the CPU and memory together which would be necessary in linked mode. For the majority of my tests I left the QDR at 1600 giving the e6400 CPU a frequency of 3.2GHz. For interest’s sake, I also conducted a test in Linked and Synched mode where I clocked the CPU to 3.85GHz which translated to a DDR2 960 memory setting. I then attempted to lower the timings as much as possible for each memory kit and compared performance.
Before we begin, it is important to note that every memory kit is different, and performance varies between kits of the same product. Other people might find quite different results using another set of the same memory, so the results of a review such as this are not universal. Both of the kits reviewed here were right off the shelf- not hand picked for review, so they are real-world results.
The table summarizes the benchmark results for both kits under a variety of settings and speeds. The fastest settings for a given score are highlighted in RED and the second fastest are in ORANGE.
In the table, you'll notice that I conducted one set of benchmarks while linked/syched with a maximum CPU overclock of 3.84GHz, but in the graphical presentation of the results below, the CPU was kept at 3.2GHz so that the differences are largely a product of memory performance.
Interestingly, the Buffalo Firestix were able to reach higher maximum stable frequencies at lower voltages than the Crucial Ballistix. The highest frequency reach at 5-5-5-15 timings for the Firestix was DDR2 1120, while the Ballistix were unstable beyond DDR2 1066. However, the Ballistix at 1066 performed nearly as well as the Firestix at 1120 in the benchmarks. While the Firestix was able to reach higher speeds, the Crucial dominated the timings comparison at DDR2 800. The tightest stable timings for the Firestix were 4-4-4-5 (23) while the Ballistix ran at an amazing 3-3-3- 9 at the same frequency.
When running linked and synched with my e6400 overclocked to 3.84GHz, both memory kits performed on a par (See the Table).
When I began testing at DDR2 800, I found that I couldn’t even get the Crucial memory to boot up. I tried a variety of timings and had no luck until I dropped the voltage down below the stock 2.2V. The 3-3-3-9 timings were achieved at only 2.1V, while 4-4-4-9 timings were possible at only 1.9V.
Another interesting result was that while both kits were stable when linked and synched with a 480MHz FSB (CPU at 3.84 and memory at DDR2 960) with 4-4-4-10 (28) timings, the Ballistix would run unlinked at DDR2 1000 with 4-4-4-12 timings while the Firestix needed looser 5-5-5-12 timings to run stable at that speed.
The biggest surprise in the results was that both kits were not able to benefit from voltages over 2.3. I expected this from the Firestix which are rated at only 2.1V and are somewhat known for being voltage intolerant, but I thought that the shiny black Ballistix PCBs would allow that kit to keep climbing with increased voltage. I was equally surprised by their intolerance of higher voltages (even stock voltage) at DDR2 800. This is not really a flaw as it is better to run with as little Voltage as needed, but it might be helpful to know in case you run into the same situation.
Both of these kits have great looks, great price, great overclocking potential thanks to their Micron D9GMH chips, and are covered by a lifetime warranty. However there are some differences other than aesthetics that might help you to decide which one you will purchase.
In testing, the Ballistix required more voltage (2.2) to reach DDR2 1066 than the Firestix needed to reach DDR2 1120 (2.1), and were also warmer to the touch. I used aftermarket active memory cooling for these test, and I would recommend them to anyone who is planning to overclock their RAM and would like it to last more than two months. I would also recommend keeping voltages within the recommended range for 24/7 use.
While not capable of the high frequencies that the Firestix reached, the Ballistix were able to run at 3-3-3-9 timings at DDR2 800 with only 2.1V while the Firestix needed looser 4-4-4-5 timings at that speed and voltage. Neither one of these kits seemed to be very tolerant of higher voltages at any range but the Ballistix were very picky at DDR2 800 and would not boot at the recommended 2.2V.
With these points in mind, I would recommend the Ballistix to someone who is using a motherboard that is not capable of running the memory unlinked and will be running the memory at around DDR2 800 speeds as the Ballistix are capable of tighter timings at this speed. With 3.2GHz being the sweetspot for many Core 2 Duo overclocks, I suspect many will be running memory at this speed but it’s dependant on the available ratios in your bios. However, if you are either overclocking in the 4.0 GHz range, or (more likely) are able to run unlinked and wish to take advantage of memory that is capable of higher frequencies, I would recommend the Buffalo Firestix. The Firestix were able to reach higher speeds and the results indicate that the higher speeds/looser timings performed better than DDR2 800 with tighter timings. Both performed equally well while linked and synched at DDR2 960 4-4-4-10, but the Firestix needed less Voltage to run stable.
Keep in mind that these recommendations are based on my findings and others with the same product might have different results depending on the nature of the chips. My results are fairly consistent with other reviews that I have seen, but I have also heard of people achieving much higher frequencies than I did with the Crucial Ballistix. Again, luck of the draw. If you want to have better odds of higher overclocks, you can opt for the 8500 series which have higher binned chips, but there are no guarantees.
In summary, I have to say that both the Buffalo Firestix and the Crucial Ballistix are outstanding value, and significantly exceed their stock settings. While the Firestix were able to run faster, the Ballistix had better timings at DDR2 800. There are particular niches for the strengths of each kind, and it’s up to you to decide which better suits your system when trying to choose. Based on performance per dollar, both products are awarded our ‘DAM GOOD!’ rating.
I’d like to thank NCIX for supplying the Buffalo Firestix for this comparison, and would encourage you to visit their site for great prices on memory. As mentioned earlier, these modules are available for only $185 CAD at NCIX but can be had for as low as $130-$140 when on sale so if you’re looking for a real bargain, check their weekly sales regularly.
If you have questions, comments, or results of your own for either of these products that you'd like to share, please visit the discussion thread.
Buffalo Firestix DDR2 800, PC2-6400
Crucial Ballistix DDR2 1000, PC2-8000