Print this article

Crucial Ballistix PC3-12800 2x1GB (BL2KIT12864BA1608)


Date
: 10/15/07 – 06:41:28 AM

Author
:

Category
: Memory

Page 1 : Index

Manufacturer:
Crucial Technology (Micron Technology, Inc.)

Price:
Shopzilla: $535.50

There is a fine line between passion and obsession. I use to think I was on the edge of that line still consuming enough floor space on the passion side of the room to not consider myself obsessed. This past Friday, that all changed when I received the newest member of the 3oh6 overclocking team. Not only did I cross that line into obsession but I crossed it in a full sprint and kept on running until sometime Monday afternoon when reality kicked in and I had to stop clocking this kit of memory and go back to real life stuff like eating, showering, and sleeping.

The kit of memory I am referring to is from none other than the masters of the DDR2 universe and now, the DDR3 universe, Crucial; a division of Micron Technology, Inc.. Defending Castle Greyskull…I mean, defending the overclocking memory crown that Micron easily won in the battle for DDR2 supremacy is not something Micron or its retail subsidiary Crucial are taking lightly. With Micron DDR2 D9XXX ICs being the "must have" IC of 06/07, Micron has some big shoes to fill when it comes to DDR3 for 07/08. This is where Crucial comes into play providing todays review sample that is no doubt based on Micron ICs backed by the incredible customer service and support of Crucial.

The Crucial Ballistix line of performance memory had been tearing it up in enthusiast forums all over the internet for a long time now. Back when DDR roamed the extreme overclocking realm, the Ballistix memory and their trademark yellow heat spreaders meant business. That held true through DDR2 and the yellow, now more of an orange, heatspreaders are back again ready to rock some DDR3 based modules. I have spent an un-healthy amount of time with these modules the past week or two and there is plenty to cover so the time to begin is now.

Page 2 : Package & Contents

Crucial has always had the best package of any memory manufacturer as far as I am concerned. The brown box they ship their memory in has always shown up in just a UPS bag with no double boxing necessary to keep the memory safe. This kit of memory showed up in the same type of box…just a bit different.

It looks like the same box as previous generations of Crucial Ballistix memory has shipped in but quite a bit deeper. I thought for a minute that there might be more than one kit inside which had me get a little more excited than I already was. I am not sure if retail kits will be shipping in packages like this but from the retail packages of their new DDR3 memory, I believe these large boxes are limited to promotional samples such as my kit for this review.

This is about the size of the normal retail Crucial package that we have all become use to so the much longer box that this DDR3 kit of Ballistix showed up in was a bit of a surprise to me. Time to find out exactly what the layout inside this large box is. I am anticipating the same cardboard accordion like insert keeping the memory safe and secure.

As I suspected, the interior of the package is the exact same as the smaller packages we have seen here on O2 and elsewhere but with room for a total of eight modules. Naturally my package only has the two modules inside wrapped in their anti-static bags. I have said it before and will likely not stop saying it. I really like this packaging method because the modules are all but locked into place during transport and I have never heard of a Crucial module showing up in anything but perfect shape.

The complete package contents consist of the two modules contained within their individual anti-static bags and a small paper insert with handling and installation instructions. I don't know why the large package was used but it seems to have done just as good a job as the smaller one so no complaints from this reviewer.

Having removed a single module from the anti-static packaging, we can see that the Crucial Ballistix PC3-12800 8-8-8 have a very similar heatspreader to previous generations of Ballistix memory but with a few subtle changes.

The first of the changes is the exclusion of interlocking heatspreaders with clips at the top. Instead, these Ballistix come with two piece heatspreaders that do not converge at the top then clip together. Recent DDR2 Ballistix started showing up in users hands with heatspreaders that seemed to be too tight at the top and not fully contact ICs all the way down the modules. This is a possible reason for a split heatspreader design on these new DDR3 modules.

Another change in heatspreader design is the inclusion of a small DDR3 added to the logo to signify that these modules are not the DDR2 counterpart. I might have been expecting significantly different heatspreader design for the release of DDR3 but if it isn't broken, don't fix it. The subtle changes that were made to the heatspreader will be welcome from users I am sure.

In this photo we get a better idea of how the heatspreaders are mounted to the ICs. Crucial still uses the thicker grey thermal adhesive as they always have but when I removed the heatspreader, it seemed a lot stickier than in the past with DDR2 modules and there was no immediate indication of the woven material being present either. The other item to note from this image is that there are ICs on only one side of these modules. This means that these new DDR3 ICs are 128MB each and potentially opens the door to some very interesting 2GB modules down the road.

Page 3 : Specifications

Normally by the time we get to the specifications section of a memory review, we already know most everything there is about the modules from the sticker on the heatsink. This isn't the case with Crucial Ballistix memory as the heatsink doesn't contain the tell-all sticker that most other manufacturers have on their sticks. Instead, I will present those numbers right now:

Crucial has always been very conservative with their memory offerings. We have never and likely won't see a DDR2 module from Crucial rated over PC2-8500 while many other manufacturers have offered kits up to PC2-9600 and beyond…using the same Micron ICs the Ballistix PC2-8500. This conservative approach has served them well and this has carried over to the initial DDR3 modules. These Ballistix modules are rated at 'only' DDR3-1600 8-8-8-20 with a measly 1.80v. The timings of 8-8-8-20 are certainly not the tightest of PC3-12800 offerings using the same ICs as these modules use but the voltage also isn't rated as high as some kits from Corsair, OCZ, and others using Micron ICs.

The familiar dark Levin PCBs havent gone anywhere under the Ballistix heatspreaders. Since DDR3 is so new and long term information on voltages and how PCBs handle them isn't known yet, it is tough to say whether or not these Levin 73101 PCBs will hold their own against some of the other manufacturers choices. Before we see what is powering the Ballistix, I want to caution anyone from doing as I have and taken the heatspreader off. The thermal adhesive is very sticky and halfway through I was even thinking of turning back for fear of causing damage. There is no benefit to removing the heatspreaders as they are attached very well and doing so only voids your warranty and puts your expensive memory at great risk. With that said, let's see what is going on under the bright orange side panels.

What lurks in the heart of these modules is none other than Micron D9GTR ICs, if you look closely you can make it out. The DDR3 equivalent of D9GMH, these ICs have been lighting it up in overclocking forums on other manufacturers memory…now it is Crucial Ballistix turn. Here are the specifications of the Micron D9GTR ICs found on these modules. The full spec sheet can be found on the Micron web site, here.

Micron
D9GTR
Specifications:

• Density: 1Gb

• Part Status: Sampling
• RohS: Yes
• Depth: 128Mb
• Width: x8
• Voltage: 1.5V
• Package: FBGA
• Pin Count: 86-ball
• Clock Rate: 533 MHz
• Cycle Time: 1.875ns
• Op. Temp.: 0C to +85C
• CL: CL = 7
• Data Rate: DDR3-1066

A few things in the above list are worth noting. For starters, the part status is still listed as 'Sampling' which means it isn't a full retail part yet and likely part of the reason why the prices are so high still. Like all DDR3, JEDEC standards outline a 1.5v specification but we obviously know that it doesn't mean all ICs will only run at 1.5v. The D9GTR are no different than the D9GMH/D9GKX or the rest of the D9 family for that matter. Add more voltage and they run a lot faster with tighter timings. These D9GTR ICs are binned for CL7 operation at DDR3-1066. Clearly running at DDR3-1600 at CL8 with 1.8v is a pretty hefty jump, despite the Crucial Ballistix being very conservative when it comes to stock clocks/timings.

Page 4 : Installation

I swear the only reason I do installation sections for memory anymore is just to take some more photos. With standard size heatspreaders, memory shouldn't have an issue fitting in any motherboard these days. CPU cooling manufacturers understand that memory may interfere and design accordingly and motherboard manufacturers for the most part keep memory far enough away from the CPU socket area as to not cause problems. Still, a couple photos never hurt anyone so here they are, first in the Biostar TP35D3-A7 Deluxe and then in the test setup; an Asus P5K3-Dlx.

Clearly there are no issues in sight on the TP35D3-A7 as we can see above with this memory. The heatspreaders are no bigger than any other modules on the market and like the OCZ PC3-10666 Platinum memory I looked at a short while ago, these Crucial Ballistix really aren't much larger than modules without a heatspreader.

As with the Biostar TP35D3-A7 Deluxe, the Asus P5K3-Dlx shows no signs of causing a problem with Crucial Ballistix memory. Even with the larger Thermalright Ultra-120 hanging over the processor, the standard size Ballistix memory poses no threat to be interfered with in either slot. With that little arbitrary task out of the way, we can now move on to the most enjoyable part of the review…overclocking.

Page 5 : Test Setup & Overclocking

Before we look at the overclocking results, let's have a quick peek at the setups used for the overclocking and testing. Because I will be comparing both DDR2 and DDR3, there are two setups listed. The DDR2 results were ran on an Intel P965 motherboard while the DDR3 results were ran on an Intel P35 based motherboard. The difference in chipsets is very small but the performance characteristics can vary slightly. I have done my best to choose settings where both machines are as evenly matched aside from the memory differences but keep the fact that both systems are a little bit different in mind when going through the results on the next few pages. We'll first see what system will be powering the DDR3 machine for this Crucial Ballistix PC3-12800 memory.

DDR3 Bench Setup:

DDR2 Bench Setup:

Over the last couple of weeks I have read some reviews where they make no mention of stability testing, just show a screenshot of CPU-Z or say the system was "stable for the benchmarks". I find this troublesome as running the benchmarks on the following pages really doesn't define stable. Here at O2, we believe in full disclosure. We have nothing to prove with any given product, and are simply here providing results for the masses based on what we find. The following stability testing that I use to measure a kit of memories overclockability has proven to be all but a guaranteed method for testing stability of my 24/7 machines. Every computer system I setup, whether it be for friends, family, or myself; is overclocked at some level. Complete testing takes over a week alone after final clocks have been decided and the following format has proved to be stable running distributed computing projects at 100% load 24/7. This is for about 97% of all machines I setup. For the 3% where the following testing didn't prove to be 24/7 stable for distributed computing projects, dropping memory frequency 5MHz or less solved the issue.

Because of time restraints with reviews, testing for 24/7 reliability is impossible at the various memory timings and voltages I test at so the below testing will be used and should be considered to be stable enough for any distributed computing project such as Folding@Home, SETI@Home, World Community Grid, Rosetta@Home, or the like. These distributed computing projects rely on completely stable systems to produce results and are considered the hardest stability tests amongst enthusiasts. If you think your overclock is stable, just run Rosetta@Home for a couple days to see how many results are rejected.

My formula for testing the overclocks shown in this review consists of the following list:

• Multiple loops of 3DMark 01 / 06 (30 minutes of looping the full tests each)

• Dual 32M runs of SuperPi Mod 1.5 (ran at the same time)
• 2 hours of dual Prime95 using Orthos Beta on blend mode priority 9
• 2 hours of dual MemTest Pro in Windows using 850MB/instance

As always though, my overclocking results with this particular sample of memory will differ from any other kit of this same memory based on the modules, motherboard, and knowledge of the user. Like CPUs, GPUs, and MBs, not all memory of the same specifications will overclock the same. My results are simply one of thousands and when purchasing memory based on overclocking, you should not rely on a single set of results such as mine. Search the various enthusiast forums and other review sites to get a gauge of how well a certain kit of memory tends to overclock. Of course, even doing research doesn't guarantee an overclock…it still comes down to luck of the draw and your knowledge of the hardware.

Enough disclaimers and talk about stability testing though, let's see how my sample of Crucial Ballistix PC3-12800 memory scales with voltage at various timing sets. Again, all results listed here passed the above testing and should be considered 100% 24/7 stable for any application.

What can you say, at stock voltage of 1.8v, this Ballistix memory simply impresses. Over DDR3-1400 at CL6, DDR3-1600 at CL7, and a whopping DDR3-1800+ at CL8 means this memory is ready for any combination of FSB, CPU frequency, and memory ratio you can throw at it. On top of that, I believe my motherboard is the only thing holding back the memory at this point from reaching DDR3-2000 with a reasonable amount of voltage. No amount of tweaking or timings or adjustment of voltages could get this memory passed 940MHz with any kind of stability. This would normally indicate a motherboard limit as the memory scaled extremely well with voltage at all other timings tested up to about 930MHz, after that, things start getting tricky.

The most impressive clocks in my opinion are the near DDR3-1800 at 7-7-6 timings with only 2.01v. A hair short of 900MHz with such low voltage at this timing set being HCI Memtest and Orthos stable is quite the achievement. It is still hard to say whether these Micron ICs are going to be comfortable at 2.01v for long term use but when you consider other manufacturers providing warranty up to 2.1v and 2.2v on the same ICs as the Ballistix have, it should indicate that they are. The Levin PCBs used in Crucial DDR3 modules are still yet to prove themselves as being able to handle that kind of voltage.

DDR3-1404 6-6-5-15 1T @ 1.81v (digital multi-meter measured)

DDR3-1720 7-7-6-15 1T @ 1.91v (digital multi-meter measured)

DDR3-1832 8-7-6-20 1T @ 1.81v

After I finished with stability testing and the benchmarks that we will look at a little later, I had a free afternoon so I strapped my single stage phase change cooler to the P5K3-Deluxe and had a little fun with the Crucial Ballistix PC3-12800 modules. My focus was 32M SuperPi and previously I had no problem benching at 4.50GHz on this CPU with the phase change but today, 4.44GHz was all that seemed to want to run. Perhaps time for a re-tune of the unit, it could have just been a bad mount as well. Either way, here is a screenshot of what this memory can do at some higher CPU clocks…a nice little sub 11 minute 32M SuperPi run:

Page 6 : Memory Benchmarks

Normally I don't show benchmark results from other memory kits but DDR3 is still so very new that I thought it would be nice to give the numbers from this Ballistix kit some perspective. Included with a couple of the Ballistix timing/frequency sets are results from highly clocked DDR2 as well as the best 3.6GHz benchmarking setup I could do with the previously reviewed OCZ Platinum PC3-10666. Below is the chart outlining in fairly good detail what each color on the charts represent.

The following pages will all be coded the same with the dark grey results equaling the highest frequency I benchmarked at, up to the pink results outlining DDR2-1200 5-5-5-8 results on the P5B-Dlx. For all of the benchmarks I decided to keep the processor steady at 3.60GHz. Notice though that some of the results were run at 450*8 and some at 400*9 to get the different memory frequencies. I also kept the Performance Level at 7 on both setups throughout the benchmarking. We will look at the memory bandwidth results first. For all results seen here today, each benchmark was ran three times with reboots in-between and then averaged for the graphs we are about to see.

SiSoft Sandra Lite XII – Memory Bandwidth

Lavalys Everest Ultimate 07 v4.00 – Memory Bandwidth

ScienceMark v2.00 – Memory Bandwidth

The first thing I noticed when going over the data was that DDR2-1200 5-5-5 seems to get surpassed by the PC3-12800 Ballistix kit at the rated frequency and timings. Those two setups are about as equal as any having both been run at 400*9 and the Ballistix kit at stock timings shines brightly almost equaling or edging out the high clocked DDR2 memory in all the bandwidth tests. The other results of note are the light blue and light grey numbers. Despite being 45MHz faster and having tighter primary timings, the light blue results appear to be better than the light grey in most benches. This is a result of leaving secondary timings on AUTO and the OCZ platinum at those frequencies/timings having tighter secondary timings. Tightening up the secondary timings of the Ballistix vaulted them past the OCZ kit in that situation which goes to show, DDR3 seems to be very influenced by secondary timings; more so than DDR2 from what I have seen so far.

Lavalys Everest / SiSoft Sandra – Memory Latency

There isn't much else to really talk about here with the latency results. Everest seems to provide a bit of a staggered graph while Sandra provides a gradual decline in memory latency from top to bottom. I am leaning on the Everest results being more accurate in their representation of actual memory latency. Sandra doesn't seem to take into account the chipset latencies very much. Although, DDR3-1600 8-8-8 is showing to be a smidge better than DDR3-1800 7-6-5 and the only other difference is the FSB in which they were both ran. With that said, the dark grey result having been ran at 450FSB should have provided even a greater advantage than the primary memory timings provided but didn't. Something else is obviously at play here, either secondary timings loosened up with the FSB increase or chipset timings relax despite the Performance Level being 7 for both sets of runs.

Page 7 : System Benchmarks

Interesting bandwidth numbers could lead to some interesting benchmarks results here and in the gaming section, more so here than gaming.

Nothing has changed for the timing sets as in the bandwidth section, and rightfully so. So far the numbers seen from the benchmarks have provided some interesting insight into the way DDR3 performs in relation to secondary timings. Let's see how that correlates in benchmarks other than bandwidth and latency orientated.

Futuremark PCMark 05

WinRAR 500MB Benchmark

WinRAR seems to favor the DDR2 setup more than the DDR3 platform except for the raw bandwidth power of DDR3-1800. It is quite evident that DDR3-1800 simply crushes DDR2-1200 in WinRAR number crunching shaving a solid 20 seconds off with just higher memory clocks. The other DDR3 results didn't fair that well against their technological predecessor. PCMark 05 seems to favor the higher memory frequency as well but with the dip in the dark blue result, PCMark 05 shows a bit of an advantage to higher FSB.

SuperPi Mod v/1.5 – 1M

SuperPi Mod v/1.5 – 8M

SuperPi Mod v/1.5 – 32M

The most intriguing part of the SuperPi times is the fact that 1M and 8M calculations are all but even between the top and bottom results, despite such a memory advantage for the DDR3-1800 7-6-5 result that we have seen. On the 32M calculation, however, the DDR3-1800 time put up absolutely crushes everything in its path by 15 seconds or more. Keep in mind, the secondary timings have not be adjusted from AUTO so this 32M 3600MHz time could come down a considerable amount, as much as another 10 seconds.

Page 8 : 3D & Gaming Benchmarks

Up until this point the DDR3-1600 Crucial Ballistix have done some very impressive things at the DDR3-1800 clocks with 7-6-5 timings but with the voltage needed for those results, I probably wouldn't be comfortable running them 24/7 in that state. Dropping the timings to 8-7-6 or 8-8-8 would bring the voltage down a couple notches and make this a very nice 24/7 kit to run.

Let's see how the above timings and frequencies do in some 3D action now.

Futuremark 3DMark 01/03/05/06

It appears that memory frequency or timings isn't playing a huge role in the 3DMark results at 3.6GHz and this 8600GTS. In 3DMark 01 there are slight variations in the results that match up to our bandwidth numbers but they are so minute that it is hardly worth noting. In honestly was expecting a little wider range of numbers here but clearly was mistaken.

Gaming Benchmarks – Half-Life 2 / FarCry / Company of Heroes

Like the 3DMark results, BF2142 and FarCry show no distinguishable advantage to any memory set we ran. Half-Life 2 is another story. The DDR2 result is far ahead for some reason and after double and triple checking the setup, I could not find any reason for the sudden jump in average frames per second. Down on the DDR3 front, the higher frequencies did play a bit of a role in the 1024×768 benchmarks giving the DDR3-1800 clocks a slight advantage but nothing worth any significance.

Page 9 : Conclusion

The DDR3 revolution has only just begun and the performance is already better than a lot of people figured it would be. The clocks we are seeing from these Micron D9GTR ICs on the Crucial Ballistix modules are simply incredible and un-matched by anyone else at the moment, much like the Micron D9GMH/GKX of DDR2. Unfortunately, that also has a major drawback and that is the price.

There is no denying that the performance of high binned DDR3 like the Crucial Ballistix PC3-12800 kit that we looked at today is impressive but is it worth it yet? I can't sit here and lie to you saying that yes it is. For most users, the price just isn't going to justify the gains to be had. Of course, if you are building a new system that you want to take you through the next couple of years then perhaps going the DDR3 route is a wise course of action right now.

If you are a cutting edge user then DDR3 is your only option and the Crucial Ballistix PC3-12800 kit is a wonderful choice. They are competitively priced compared to other manufacturers PC3-12800 modules and the conservative rating provides plenty of room for overclocking. The flexibility of timings it can run will ensure that no matter what CPU multiplier/FSB/memory ratio you need to run to maximize your CPU power, the Ballistix will accommodate with excellent performance. At this point I need to grab my sunglasses because the future of DDR3 looks so bright, "that I gotta wear…shades".

Advantages

• Un-parallel performance and memory bandwidth with Micron D9GTR ICs

• Top quality packaging and customer service, as always, from Crucial
• Extremely flexible for timings available to be run at a plethora of frequencies
• Future proof for builds of today going forward

Disadvantages

• The top of all DDR3 disadvantage lists will be $$$ for some time yet…don't hate the player, hate the game

• Micron based DDR3 memory is so good right now that motherboards can't keep up

Feel free to drop into the Discussion Thread for this review in our forums if you have any comments, questions, or concerns.

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