Print this article

Crucial Ballistix 2x1GB PC2-5300


Date
: 01/5/07 – 02:12:40 AM

Author
:

Category
: Memory

Page 1 : Index

Manufacturer:
Crucial Technology (Micron Technology, Inc.)

Price:
$276 USD (Newegg.com) / $379 CND (Anitec.ca)

Today I have the distinct pleasure of introducing a company for the first time to Overclockers Online readers. Crucial Technology is a division of Micron Semiconductor Products, Inc., which in turn is a subsidiary of the memory giant, Micron Technology Inc.. Micron is no stranger to myself but for those that don't know, Micron Technology Inc. manufacturers memory of almost every kind including NAND Flash, PSRAM, Mobile DRAM, DRAM and more. Micron also produces complete DRAM modules and sells them under the Micron label to OEMs but also under the Crucial brand to end-users.

In 1996, the Boise Idaho based DRAM giant, Micron Technology Inc., decided to create a direct link between the company and the end-user in the computer memory market. Up until that point, Micron had only supplied OEMs with their products. The company responsible for creating that link from manufacturer to end-user is Crucial Technology and over 10 years later, Crucial has built the brand into a world leader in the memory market. Today, Crucial markets many products including flash cards, USB flash drives, graphics cards and in conjunction with standard DRAM memory lines, an enthusiast line of memory called Ballistix.

For the first Crucial product to be reviewed here at Overclockers Online, I will be looking at a member of the Crucial Ballistix family, a 2x1GB kit of PC2-5300 DDRII memory rated for 3-3-3-12 operation at DDR667 with 2.2v. The Ballistix memory from Crucial are their best of the best with performance and reliability being top priority. The rated frequencies and timings are very aggressive and with a bit of help from my two favorite motherboards, I am sure I can make these modules run much more than their rated frequencies. Overclocking and testing will be completed on an AM2 platform as well as a Core 2 Duo based setup so we better get started.

Page 2 : Package & Contents

Like all of our reviews here at Overclockers Online, we start off with a quick look at the package that the product has come in.

Despite this being my first kit of Crucial memory, I did know what to expect as far as a package goes as I have see photos in forums. I absolutely love the fact that Crucial ships memory in a cardboard package and not a blister pack. Every other manufacturer has gone the way of the clear plastic blister pack but not Crucial.

Looking inside the cardboard package we can see that the modules are held in place in an upright position. Each module is also sealed in an anti-static bag. I received my sample directly from the Crucial retail department and all they had to do was put this box in a UPS shipping bag. Obviously the package arrived in perfect condition indicating that this type of package is designed extremely well.

As we take a closer look inside the package, it is clear that a simple cardboard piece is used to hold the modules in place. The dimensions of the box provide a very secure environment for the modules tucked inside. Again, I love this packaging as it doesn't require double boxing or a very large package at all for shipping.

The only contents of the package are the two individually sealed modules and a small leaflet with handling and installation instructions.

Each anti-static bag is labeled on one side and as I mentioned before, each modules is packed individually and sealed. Crucial is the only manufacturer that I have come across that still sells modules in a static protection bag. That again goes back to the industry wide use of the plastic blister pack.

I have pulled the modules from their protective bags and my first impression of these modules can be summed up in one word…wow. I do not think the photos will do these modules justice for you readers but I will try my best to convey just how nice they are.

Each side of the modules have the Ballistix label and one side also has a sticker with the batch number and part number of the module.

These heatspreaders really are a work of art. The black PCB of the module combined with the sharp orange color on the heatspreader sets these pieces off. If you look closely you will see that the exposed aluminum edge of the heatspreader is actually raised from the orange surface. This adds a whole new level of depth to the modules.

This close up gives us a good look at the construction of the modules and reveals a thick thermal pad between the aluminum heatspreader and the ICs on the module. The interlock at the top of the module indicates a two piece design to the heatspreader.

The top of the module has plenty of gaps for hot air to rise out and away from the PCB and ICs. The two heatspreader pieces are secured together with silver clips. I can't say this enough, these modules look simply amazing with these heatspreaders. Esthetics isnt important for memory but it never hurts to look good.

The only markings on the PCB that indicate the manufacturer are on the lower tip near the contacts. Levin 63201P is the model and manufacturer of these PCBs. I tried to find some more info on the PCBs but extensive Google searching didn't bring up any information on the Levin brand aside from Crucial seemingly being the sole user of this PCB. I was hoping for Brain Power PCBs as I know they handle voltage quite well so we will have to wait for the testing section to see how these Levin PCBs do at higher voltage. The only thing left to do is pop the heatspreaders to confirm the ICs that are hiding underneath. I do not recommend any users doing this as your Warranty is voided shortly after you do so and the possibility for damaging the sticks in the process is great.

The heatspreaders didn't come off that easily which is a good sign that the thermal tape between the ICs and heatspreaders has a good firm grip on both. Hopefully that means that thermal conductivity will be good to help keep these ICs cool under stress. I won't go into details about the ICs used just yet, you can find that information on the next page in the Specifications section.

Page 3 : Specifications

Crucial designed the Ballistix line of memory for the enthusiast that wants to get the most out of their memory. Crucial has summed up the Ballistix line best on their web site, here is how they describe Ballistix:

The Ballistix line of memory has a number of DDRII models ranging from PC2-5300 up to PC2-8000 in 512MB and 1GB sizes or matched in pairs for 1GB and 2GB kits. They even have a series of models within the Ballistix line called Tracers. These models have the same specifications as the regular Ballistix models offered but the Tracer modules come with active LEDs that indicate BUS activity and are perfect for the serious modders and system builders. This is the spec sheet for the modules that I will be looking at today:

In the contents section I mentioned that I would be discussing the ICs that are used on these modules in this section, so let's do that now.

Me and D9GMH go way back. These ICs were on my last memory samples from the Buffalo FireStix line and also the ICs used on a pair of TEAM Group Xtreem memory kits that I have recently finished overclocking reports on in our forums here and here. Needless to say, I am quite familiar with D9GMH ICs and look forward to the overclocking section after seeing them present on these modules. These are the specifications as outlined on the Micron web site for these ICs:

These ICs come out of fabrication rated at DDR667 operation with a CL of 5 at 1.8v. Obviously with more voltage they run either higher frequencies and/or tighter latencies. In the case of these modules, they are still rated for DDR667 operation but at much tighter timings with a bit more voltage. D9GMH are the second highest rated ICs from Micron that you will find on DDR2 modules, D9GKX are the highest and rated for DDR800 CL5 @ 1.8v. However, the best D9GMH will surpass most decent D9GKX in frequency with enough volts. Overclocking will be fun with this kit so lets get them installed and let the games begin.

Page 4 : Installation

Memory installation is very easy and straight forward. The only thing of note is to have the power supply unplugged and make sure the motherboard has been discharged by pressing the power button a couple times after turning off the power supply. Since the Crucial Ballistix modules are standard height and don't have heatspreaders any larger than the norm, fitting them in the testing motherboards shouldn't be an issue.

Despite the installation being simple and straightforward, I sill like to show a couple photos of the modules going in. As always, my DFI LanParty NF590 SLI-M2R/G will be used for the AM2 testing and in this motherboard, dual-channel memory needs to be in split slots. I haven't found a difference between yellow or orange slots on this motherboard with other D9GMH based modules and will be sticking with the orange slots.

There were clearly no issues getting the modules in place with my standard fan arrangement. This is how the motherboard will be setup throughout the testing period. I always actively cool my memory because it only makes sense, regardless of how much voltage I am going to run. I would always recommend all users to actively cool memory, especially if you are going to be overclocking. I will now install the modules in the Asus P5B-Dlx motherboard that will be used for Intel testing.

The memory area on my P5B-Dlx is quite an active one with the Noctua NC-U6 chipset reaching close to the modules and the Nipro-Tech.com DDR Maximizer in the middle memory slot. In case you haven't seen my last memory review, I use a DDR Maximizer for supplying voltage to the memory on my P5B-Dlx. It allows greater voltage control as well as higher voltage possibilities compared to the VDIMM the motherboard supplies.

There are clearly no installation issues with these modules and why would there be? Standard height and standard width will usually result in standard fit. This is the setup for my Intel testing portion of the review and again, active cooling is present. The 120mm fan I use also helps cool the northbridge which can come into play with some memory configurations on this motherboard. We are now ready to get started with some clocking. I will begin with the AM2 overclocking first.

Page 5 : AMD Overclocking

This is what we have all been waiting for, the overclocking sections. As I have stated, I will be overclocking these modules on both an AM2 platform and an Intel platform. This should give all users a good idea of what this memory is capable of. Here is the complete list of hardware used in my AM2 setup:

Everyone has a different idea of what stable is. For some, stability means playing their favorite game for hours on end without a crash while others feel stability is a set number of hours running dual Prime95. For many users not interested in overclocking, stability simply means having a system that works when they want it to and doesn't randomly shutdown, reboot, or BSOD (Blue Screen Of Death) on them.

Proper stability testing of memory, which includes 24 hour stints of Prime95 and other stress testing programs, takes much too long for the amount of different overclocks I will be testing. Instead, I have come up with a formula of stability tests that has proven to be very reliable for testing a set of memory. Here is what each overclock listed in the charts below will have to pass in order to be considered stable to make it in those charts:

• Dual 32M runs of Super Pi Mod 1.5 (ran at the same time)
• 2 hours of dual Prime95 using Orthos Beta on blend mode
• 1 hour of dual MemTest in Windows using 750MB/instance
• Multiple loops of each 3DMark 01 / 03 / 05

  I have spent many moons stability testing memory and the above combination is almost bulletproof in testing the stability of an overclock, no matter what your idea of stability is.

  

  I have broken each set of overclocks down by timings used and started with the sub-timings that the system chooses had I simply installed the memory in the system and booted into Windows. Many of todays modules require the end-user to enter the BIOS and adjust settings manually to achieve the advertised timings. These Crucial Ballistix are configured to boot at looser settings to ensure compatibility with all motherboards. This means that users will manually have to set the timings of 3-3-3-12 in the BIOS.

  
  3-3-3-8
  
  

  

  The specifications of the Ballistix PC5300 3-3-3-12 have it running at 2.2v but as the chart shows, my sample clearly doesn't need that much voltage to run the specified speed. I was almost able to run DDR667 at only 2.0v and at 2.2v I easily cracked DDR700 and reached as high as DDR740. The highest I wanted to run these modules was 2.3v on information provided by Crucial so that is why I stopped there. Crucial only warranties operation up to 2.2v so if you do go over 2.2v, you will be voiding your warranty. Of course, I always recommend some sort of active cooling over the modules, especially if you are overclocking your memory. A small fan will do wonders for helping the memory run cool and extend its life significantly.

  DDR734 3-3-3-8 @ 2.20v

  

  
  4-4-4-8
  
  

  

  Being rated for PC5300 and CL3 doesn't mean these modules won't run looser timings as well. The D9GMH IC's used on the these modules are used on plenty of PC6400 CL4 and PC8000/PC8500 CL5 modules from many manufacturers including Crucial so I fully expected them to scale well at 4-4-4-8 and 5-5-5-8. As you can see, the 4-4-4-8 results are mighty impressive reaching a massive DDR1000 with 2.29v and as this screenshot below shows, DDR800 4-4-4-8 wasn't an issue on my AM2 platform.

  DDR880 4-4-4-8 @ 2.00v

  

  
  5-5-5-8
  
  

  

  Things started out great at 5-5-5-8 with the Crucial Ballistix reaching a lofty DDR1080 with only 2.0v and then the 30MHz jump at 2.1v had me thinking these would be DDR1200 stable at 2.2v but they sort of ran out of steam at DDR1140 with only little 10MHz increases after that for each 0.10v increase of VDIMM. Even still, I don't think I can complain at all with DDR1160 @ 2.2v because that is just incredible for a kit of PC5300 memory.

  DDR1180 5-5-5-8 @ 2.29v

  

  
  Page 6 : Intel Overclocking
  

  The testing on the AM2 platform showed some promising results and proved that this kit is not satisfied with running just DDR667 at 3-3-3-12. We will now see if that trend continues on my C2D based Intel platform. Here is the complete list of hardware used in testing:

  CPU: Intel C2D E6300 (L630A864)
  MB: Asus P5B-Dlx Wifi-AP Edition
  GPU: Biostar 7600GS 256MB DDRIII
  PSU: Silverstone Zeus 560W
  HD: Seagate SATAII 80GB 8MB NCQ
  OS: Windows XP SP2 (with all updates)
  Hipro-Tech.com DDR Maximizer v1.2

  Memory:

  Crucial Ballistix 2x1GB PC2-5300 3-3-3-12

  Like the AMD testing, the Intel testing will go through the same strict stability testing before an overclock will be considered stable and added to the charts below. Again, I have tested memory so much and come up with this process that virtually guarantees a 24/7 stable overclock with very few adjustments required. Here again is the list of programs used for stability testing:

• Dual 32M runs of Super Pi Mod 1.5 (ran at the same time)
• 2 hours of Prime 95 using Orthos Beta on blend mode
• 1 hour of dual MemTest in Windows using 750MB/instance
• Multiple loops of each 3DMark 01 / 03 / 05

  Intel Overclocking:

  

  As we are all well aware of, the rated frequency and timings for this kit of Crucial Ballistix are DDR667 at 3-3-3-12. You will notice that I have done the overclocking at 3-3-3-8 and that is because most other memory kits I review will be or already have been overclocked on this system at 3-3-3-8. This will allow a relatively equal comparison between those kits. The clocking of this memory at 3-3-3-8 VS 3-3-3-12 is virtually non-existent in my setup as well so there is no real benefit of going down to a TRAS of 12 on my platform. Let's see how the memory did.

  
  3-3-3-8
  
  

  

  Our chart on the Intel platform looks very similar to the AM2 results with our modules running up to DDR740 at the specified 2.2v and almost reaching DDR800 with 2.3v. I really wanted to go over 2.3v and run the coveted DDR800 3-3-3-8 with these modules but with the Levin PCB, I just don't think these modules can handle more than 2.3v without some degradation of performance or outright death. I would highly recommend users of the DDRII Ballistix line of memory to stay at 2.3v or less for extended use otherwise you will likely damage your modules.

  DDR660 3-3-3-8 @ 2.00v

  

  
  4-4-4-8
  
  

  

  The overclocking results at 4-4-4-8 are eerily similar to the AM2 testing that we saw a page earlier, in fact, they are almost identical. The memory starts out reaching a fantastic DDR860 at only 2.0v and scales nicely up to an incredible DDR1000 with 2.30v. I fully expected these modules to run PC6400 speeds at 4-4-4-8 but I wasn't expecting DDR860 at 2.00v, these modules are better than most PC6400 C4 kits out there these days.

  DDR860 4-4-4-8 @ 2.00v

  

  
  5-5-5-8
  
  

  

  These Ballistix modules just didn't seem to want to stop on the P5B-Dlx. Now, I already said I was expecting them to scale well at 5-5-5-8 but I was even surprised to get almost DDR1200 out of this kit with only 2.3v. That easily surpassed my already high expectations. Even at the low 2.00v, these modules were completely stable at DDR1080 and they just scaled with volts from there. This kit definitely has something for every user whether you are looking for low frequency and tight timings or high frequency and looser timings.

  DDR1180 5-5-5-8 @ 2.30v

  

  
  Page 7 : AMD Benchmarks
  

  You can see the chart below that outlines the various CPU and memory frequencies that I will be running the benchmarks at. We have a nice array of results with the stock frequency and timings being run in yellow followed by the most I could overclock these modules at their rated 2.2v. Then I have results at the other end of the spectrum near 3GHz using dividers to adjust memory speed and timings accordingly.

  

  The one thing to note about the chart directly above is the fact that the 3-3-3-8 timings will be run at only DDR590. The reason for this is because my DFI NF590 SLI-M2R/G decided that the 533 divider didn't want to work anymore and DFI's support on this board with BIOS updates is quite pathetic. Enough of my whining, it is time to look at the benchmark results. I have always started with bandwidth numbers and today won't be any different.

  SiSoft Sandra 07 SP1 – Memory Bandwidth

  

  Lavalys Everest Ultimate 06 – Memory Bandwidth

  

  The Everest bandwidth results scale in a uniform manner and if you compare the navy and grey results, it is obvious that memory bandwidth has plenty to do with CPU speed as the much slower DDR590 is pushing more data through the memory at 2.95GHz than DDR740 does at 2.22GHz. The Sandra results are slightly different and that is because SiSoft Sandra simply calculates bandwidth differently and obviously CPU frequency doesn't weigh in as heavily.

  Lavalys Everest Ultimate 06 – Memory Latency

  

  The latency numbers are a bit interesting showing a clear advantage for the black results when compared to the white. DDR983 is substantially quicker when it comes to latency at 4-4-4-8 than DDR1180 is at 5-5-5-8. With the narrow bandwidth advantage going to the higher frequency, we may see some benchmarks results below that favor the 4-4-4-8 timing set.

  FutureMark PCMark 05 & 3DMark 01 SE

  
  

  I think the PCMark 05 results speak for themselves showing absolutely zero advantage for higher bandwidth memory. CPU frequency seems to be the only factor that was changed and had an impact in this testing. 3DMark01 on the other hand does show signs of memory frequency impacting the score. The 4-4-4-8 and 5-5-5-8 results are very close but the slower 3-3-3-8 result at 2.95GHz falls a full thousand points back.

  SuperPi Mod v/1.5 – 1M

  

  SuperPi Mod v/1.5 – 8M

  

  SuperPi Mod v/1.5 – 32M

  

  SuperPi is a very memory intensive benchmark and that is evident as seen in the large drop off in performance that we saw in 3DMark 01 for the slower 3-3-3-8 results. The difference between 4-4-4-8 and 5-5-5-8 is again, almost non-existent. I guess, 5-5-5-8 at DDR1180 is still better than DDR983 at 4-4-4-8 because they both require 2.29v in order to be stable. Had 4-4-4-8 been stable at DDR1000 with less voltage, then I would see no reason to run the higher frequency and tighter timings in a system.

  Gaming Benchmarks – Far Cry / Half-Life 2 / UT 2004

  

  The gaming benchmarks simply re-iterate what we have seen up until this point. I am almost positive that we would see virtually identical results from the bottom three sets of frequencies/timings if I had the divider working to run the memory at DDR738 3-3-3-8 instead of DDR580. We will now see if this trend continues over on the P5B-Dlx in my Intel testing.

  
  Page 8 : Intel Benchmarks
  

  The Intel benchmarks will be a repeat of the AM2 benching but of course, they will be done on the Intel test setup and at slightly different memory frequencies paired to different CPU frequencies. I have basically done the stock settings, the stock settings overclocked slightly, and then three sets of settings where only the divider changes which adjusts the memory frequency and the timings, but the CPU speed remains the same.
  

  

  As always, all benchmarks are ran three times and averaged out for graphs below. Let's see how this memory performs at the settings above.

  SiSoft Sandra 07 SP1 – Memory Bandwidth

  

  Lavalys Everest Ultimate 06 – Memory Bandwidth

  

  Both SiSoft Sandra and Everest show bandwidth increase with every increase in frequency and the results scale equally for both programs. Obviously the numbers are different and that is always going to happen because the two programs calculate the bandwidth using different methods. The key thing is that they scale about the same. The one thing to note is the increased gap between the 3-3-3-8 results and the 4-4-4-8 results. This is something we will se throughout the benchmarks here.

  Lavalys Everest Ultimate 06 – Memory Latency

  

  The Everest Latency testing shows that the sweet spot for this memory in this P5B-Dlx seems to be the 4-4-4-8 timing set at DDR983. Even DDR1180 at 5-5-5-8 doesn't make up the gap of looser timings with the higher frequency. The 3-3-3-8 results seem to fall right off the grid and that is for the same reason that the Bandwidth numbers above are low for the 3-3-3-8 timings. The P5B-Dlx doesn't like 3-3-3-8, it is actually slower than 4-4-4-8 or even 3-4-3-8 on this motherboard.

  FutureMark PCMark 05 & 3DMark 01 SE

  
  

  The PCMark 05 results provide almost identical results between DDR983 and DDR1180. DDR786 with 3-3-3-8 timings lags behind. Now, 3DMark 01 shows the slightest of slight drops in performance going from 4-4-4-8 to 5-5-5-8. Despite the much higher RAM speed, the looser timings don't make up for it and the score is lower. This is very much the same trend we saw in AM2 testing.

  Super Pi Mod v/1.5 – 1M

  

  Super Pi Mod v/1.5 – 8M

  

  Super Pi Mod v/1.5 – 32M

  

  Like every other benchmark today, the Super Pi results show the 4-4-4-8 results to be almost identical to the 5-5-5-8 results and the 3-3-3-8 times lagging slightly behind at the same CPU frequency. This pattern has been consistent with every other benchmark and that really isn't much of a surprise. This again just goes to show that DDR1000 4-4-4-8 seems to be the sweet spot on this P5B-Dlx setup for this CPU frequency that I am running. Perhaps with higher CPU clocks, the higher frequency memory would show more gains but at 2751MHz, there really isn't a difference.

  Gaming Benchmarks – Far Cry / Half-Life 2 / UT 2004

  

  All of the above benchmarks translates into an almost identical pattern for the gaming benchmarks. UT04 is the best gauge on memory performance in gaming because it is so system dependant but you can see that Far Cry shows virtually the same FPS at all of the timings and frequencies. Again, the 3-3-3-8 timings show a distinct disadvantage to either 4-4-4-8 or 5-5-5-8. It may sound funny in a review for a memory kit specified to run at 3-3-3 but if you are running a P5B-Dlx, you are better off dropping the timings down to 4-4-4 or even 3-4-3.

  
  Page 9 : Conclusion
  

  As I mentioned in the opening, this was our first experience with a kit of memory from Crucial here at Overclockers Online and I don't think we could have had better modules to look at. I had big expectations from this memory after I saw that the IC's they used were the famous D9GMH and those lofty expectations were surpassed in all aspects of overclocking.

  

  Not only is the performance excellent at the rated PC5300 3-3-3-12, but when you factor in overclocking, this memory is simply superb. The ability and almost need to run faster with relatively low DDRII voltages will allow users of even non-overclocking motherboards that are limited in their voltage settings to push this memory. For enthusiasts, I easily recommend these modules over most of the PC6400 C4 kits out there due to the incredible overclocking headroom they offer thanks to the D9GMH onboard.

  The Crucial Ballistix line with their exquisite aluminum heatspreaders and tough looking black PCB were represented extremely well here today with the PC5300 3-3-3-12 kit. The only thing I could ask for is better availability and prices of the Ballistix line in Canada so I can buy a second set to go with these.

  Advantages

  Excellent performance at PC5300
  Excellent performance at PC6400
  Excellent performance at PC8500
  The above three lines equal incredible overclocking
  Newegg.com prices are excellent in the USA

  Disadvantages

  They don't take kindly to more than 2.3v for extremists
  Canadian availability and prices are tough to swallow

  Overclockers Online would like to thank Crucial Technology for the review opportunity.