OC Report :: Crucial Ballistix 2x1GB PC3-16000 CL9

3oh6 · #1 (**permalink**) May 17, 2008, 06:54 PM

"Still rock my khakis with a cuff and a crease..." Guess who's back? That's right, it is the original OC Report rearing its dual 32M running head and the star of the show is once again a kit of Ballistix. Last time around the test platform was quite a bit different than this session but at the heart of it, the 240 pin DDR3 DIMM slots haven't changed...just everything that powers and supports them including motherboard, chipset, and processor. Make sure you are buckled in and please keep your arms and hands inside the ride at all times, things are about to get crazy.

Crucial, or should we say Micron, continues to run the DDR3 SDRAM game and it doesn't look like that is about to change any time soon. Since my last Ballistix OC Report, a few things on the hardware landscape have changed, most importantly the label on the ICs of the highest rated DDR3 kits money can buy. As we will see in a short while, the D9 variant of Micron IC's powering this kit is not like the last and despite some differences, the primary ass kicking aspects of this memory are still the same. Micron continues to push the envelope and develop ICs that keep things interesting for use enthusiasts and keep the debate hot whether one is better than the other. Remember the DDR2 D9GMH/D9GKX arguments that populated every major overclocking forum not that long ago, including this one? Well get ready for round two because the DDR3 equivalent is already here and I am going to be fanning the fire with these results.

Directory:

The format is the same, the methodology is the same, and the goal is the same...make the memory scream. As I always say, let's get this horse and pony show on the road!

3oh6 · #2 (**permalink**) May 17, 2008, 06:55 PM

I think the majority of users here are more than familiar with DDR3 specifications and likely already know the details about this kit. So I wills imply present those specs for you and not dummy it down with an explanation.

Like I said, this memory simply stated is rated for DDR3-2000 or 1000MHz @ 9-9-9-28 w/1.90v. Now it doesn't say anywhere on the Crucial web site which chipsets these ratings are for so we are left to assume that this would be for any chipset. Unfortunately I highly doubt that to be correct. For sure this memory is rated for 1000MHz operation on NVIDIA 7 series chipset based motherboards due to the EPP2.0 profile or in marketing terms 'SLI-Ready' when in fact it has nothing to do with SLI. It simply means NVIDIA based motherboards can read the EPP profile and adjust settings accordingly to the specifications, including voltage. This is very much the same as the Intel XMP profiles. Here is a look at the SPD table as well as the default timings when the EPP2.0 profile is enabled. Which, by the way, works perfectly on the EVGA 790i NVIDA reference motherboard with timings, memory ratio, and voltages being adjust correctly.

The EPP2.0 profile sets the secondary timing tRAS at 28, tRC at 50, Command Rate to 2T, and tRFC at 110. All of these timings are extremely loose, as are the main timing sets of 9-9-9. This will ensure that the memory can run at 1000MHz with only 1.9v but it will also all but kill performance on the 790i chipset because my testing has found that tighter secondary/primary timings will usually beat out higher frequencies a lot of time...the frequency really needs a big jump to overcome loose timings. Of course this is based on SPi 32M testing but since SPi relies heavily on the memory sub-system, it seems the most logical test. Bandwidth doesn't seem to be effected as much by the loose secondary timings though so real world difference in programs other than SPi may or may not show a huge disadvantage.

In a few moments we are going to look at the modules and what we will find is that the ICs populating these modules are Micron D9GTS (MT41J128M8BY-15E). These ICs are the big brother, if you will, of the famous Micron D9GTR ICs that have been populating PC3-14400 CL7 kits for months now. The family part number is MT41J128M8BY and encompasses both D9GTR and D9GTS ICs. Digi-Key no longer has any MT41J128M8BY parts listed on their site, only Arrow electronics does but I find their prices to be highly inflated. For example, on D9JNL which both Arrow and Digi-Key stock, the price difference is nearly $9USD/unit higher at Arrow. With that as our reference, D9GTS should be selling for about $10USD/unit because Arrow has it listed for $18USD/unit. This could be way off though as we can't find any IC supplier selling D9GTS aside from Arrow at this point. Of course, Crucials cost is going to be lower because their parent company Micron is the producer. This puts the IC cost of each module at approximately $80USD (based on my quasi math which may be completely off as D9GTS might be closer to the $18USD/IC that Arrow lists), so when considering two modules per kit, we are up to $160USD. That doesn't include binning costs, PCB costs, or the cost to manufacture; which isn't exactly cheap but doesn't exactly justify the price when you consider kits of D9JNL binned for 1000MHz for $240USD or $270CND from OCZ.

Either way, here are the full specifications outlined, the information has been pulled straight from the Micron web site:
D9JNL Specifications

Density: 1Gb
Part Status: Production
RohS: Yes
Depth: 128Mb
Width: x8
Voltage: 1.5V
Package: FBGA
Pin Count: 86-ball
Clock Rate: 667 MHz
Cycle Time: 1.5ns
Op. Temp.: 0C to +85C
CL: CL = 9
Data Rate: DDR3-1333

The only difference in the listed specs of D9GTS and D9GTR is the difference in rated frequency and the CL at that frequency at 1.5v. This D9GTS is rated for operation at 667MHz with a CL9 @ 1.5v while D9GTR is rated for operation at 533MHz with a CL7 @ 1.5v. This is very similar to the difference between D9GMH and D9GKX in the DDR2 world. It would make sense then that D9GTR is normally found on PC3-12800/PC3-14400 CL7 kits of memory and this D9GTS is on the PC3-16000 CL9 kit because of these ratings. Other manufacturers who have PC3-16000 kits aren't using these IC's though. It has been reported that a number of manufacturers are using D9JNL for their DDR3-2000 kits so it was a bit of a surprise to find D9GTS on these modules. Here now is a list of retail outlets and their prices where this memory can be purchased at the time of posting:

Crucial.com - $499.99 USD
NCIX.com - $619.99 CND

At this point, the pricing on this kit is un-acceptable to be honest. Any on-line outlet that has this memory available has this kit priced at well over $600 and weren't even worth posting, I simply listed NCIX.com as a reference for Canadian users. I understand this is a highly rated kit of DDR3 but OCZ has a similar binned kit, likely using different D9JNL ICs but still binned for DDR3-2000 CL9, for $270CND/$240USD. It would be impossible to recommend the Ballistix over the OCZ because the difference between the ICs is so minimal that the cost offsets those differences quite easily. It is hard to say if all PC3-16000 kits coming from Crucial are D9GTS or will be going forward but I have a hunch we will see a price drop and a change to D9JNL/D9JNM/D9JZF for this kit in the future.

3oh6 · #3 (**permalink**) May 17, 2008, 06:55 PM

Click for full size...

The retail package for Crucial hasn't changed in a long time and that is fine with me. Personally, I really like the cardboard box approach as opposed to the molded plastic clam shell. The box makes for easy transport, storage, and protects the memory better than anything else on the market as far as I am concerned. The accordion inserts keep the memory separated and standing up, unable to move in any direction. In addition to Crucial being the only manufacturer that ships modules in a cardboard box, they are also the only manufacturer that ships memory in ESD sensitive bags. These bags are sealed and the only way to open them is cutting or ripping. This is another nice feature of their shipping package...you always know if you have brand new memory from the factory. With the plastic clam shell, it can easily be ran, tested, used, and abused; then put back in the clam shell for sale as new.

I got a little trigger happy with the camera and this kit but I just like the looks of them. Crucial moved to a clip less heat sink a while back now and relies solely on the thermal adhesive to hold them in place. This is a definite upgrade from the interlocked and clipped heat sinks of old that sometimes had too much pressure at the top and actually lifted the heat sinks off the ICs towards the bottom of the modules. Overall though, the Ballistix heat sink is very basic being a simple aluminum design but the way Crucial finishes them, they look very nice. The size obviously won't hinder or complicate installation in any setup, or in 4 X module configurations.

Crucial continues to use the grey spongy thermal pad and adhesive with their modules and it seems to work just fine for them. The application is always bang on and as we can see in the first photo, almost the entire IC is covered with the thermal material on this kit. The debate about heat sinks on memory and thermal materials used will go on forever but I still think with heat sink designs like this, it is more for marketing than for cooling. I believe that only when you get into the Corsair DHX heat sinks do you actually get noticeable benefits from them.

As we talked about in the specifications page, these modules use D9GTS which was completely unexpected. At this point, it was my understanding that all 1000MHz kits from all manufacturers are using D9JNL. I am going way out on a limb here but I have a feeling the market for D9GTR/GTS dried up quickly when D9JNL came out that can do almost everything the earlier parts could but for a lower cost. This might be Micron getting rid of D9GTS before switching this kit over to D9JNL or even D9JNM/D9JZF? Again, 100% unconfirmed pure speculation on my part but I just couldn't believe it when I saw D9GTS under the heat sinks.

3oh6 · #4 (**permalink**) May 17, 2008, 06:56 PM

Click for full size...

Many of you will already be familiar with this setup despite this being the first OC Report on my EVGA 790i motherboard. At this point, this motherboard has already taken the 32M world record for E8400 processors, was the first to break 15K read bandwidth in Everest, and absolutely clocks memory like a beast. Everything that has touched this motherboard has been considered gold so hopefully that streak continues with this kit. Here is a break-down of the hardware that was used for the results we are about to see.

Code:

Setup:
MB:          EVGA 790i Ultra SLI (vMCH / vDROOP / vCORE mods)
CPU:         Intel C2D E8400 (Q808A476)
CPU Cooling: Chilly 1 SS
GPU:         2 x BFG 9800GTX / Biostar 8600GTS 512MB DDR3
NB Cooling:  Swiftech MCW30 / PA 120.2 / DD5 
SB Cooling:  Thermalright HR-05-SLI
PWM Cooling: Swiftech MC14s
PSU:         Ultra ULT-750P
HD:          Seagate SATAII 80GB 8MB NCQ
OS:          Windows XP Pro SP2

Ambient Temperature: 22-23C

All of the dual 32M screenshots were done with the dual 9800GTXs as there are two GPU temps in the Everest on screen display. Since then though, the 9800GTXs had to move on and I am back with my trusty 7600GS for any further benchmarking. There may be a point where my HIS HD3870X2 is used for some 3D benching with this kit but can't say for certain.

3oh6 · #5 (**permalink**) May 17, 2008, 06:57 PM

We have finally arrived at the heart of the 'Original OC Report', the dual 32M SPi overclock testing. As the name suggests, all results listed in the chart below have passed dual 32M SPi using HyperPi 0.99b. I am not claiming any further stability than what is shown in the screen shots. Single 32M SPi will likely be higher and full Prime Blend, HCI Memtest, or 3D stability may be lower. I have found that dual 32M SPi stability testing is a nice and quick way to find out where a kit of memory will fall. Further testing for more or less stability can then be conducted from these results.

Result Caveat #1 - Obviously not all retail samples of this memory are going to clock as good, at the same time, some will clock better.

Result Caveat #2 - This sample was supplied by Crucial for a review. I have confirmed with Sam Harmer (Crucial/Lexar Public Relations Manager) that all review samples from Crucial go out to reviewers are random retail samples. The marketing department simply puts an order through their retail department and the sample sent out is completely random and from the exact same pool of samples that would go out to any other customer ordering memory from Crucial. This memory is 100% not handpicked in any way.

Result Caveat #3 - Memory clocking results will vary from sample to sample of 790i motherboards we are finding. My 790i reference board seems to be a pretty decent one. On top of that, I have obviously modified the SPP voltage as well as the cooling. Keep this in mind when looking at results. The vSPP displayed in the Everest on-screen display is accurate with actual voltage being supplied as measured from the output coil with a DMM.

Result Caveat #4 - There are a lot of issues with 24/7 stability/corruption and the 790i motherboards, particularly 3D gaming and watching movies of a certain type. Don't read more into these results than they are, I am only showing dual 32M stability, nothing more. If you are considering a 790i board for a daily machine, then please be sure to visit manufacturers forums like EVGA to see exactly what the issues are before spending your hard earned money.

Result Caveat #5 - The vDIMM listed for each overclock is accurate with a DMM measurement from of the open DIMM slots at the first VDD pin beside the key in the DIMM. The ground used is the lower left hand ground plane at the motherboard mounting hole. The DMM used is UEI DM393 that is calibrated yearly and was last calibrated in March of this year.

Now that I have that out of the way, here is what this monster kit cooked up:

6-6-5-X 1T:
NVIDIA 790i:
Click link for full size screenshot...
780MHz :: 1.90v | 797MHz :: 1.95v | 813MHz :: 2.00v | 830MHz :: 2.05v | 846MHz :: 2.10v

Well there you have it. This kit obviously doesn't mind 6-6-5 and going to 6-6-6 didn't gain much over a couple MHz if that. At stock voltage of 1.90v this kit ran a comfortable 780MHz and climbed to a very respectable 846MHz with 'only' 2.10v. The 24/7 ceiling for DDR3 and these D9 based Micron ICs has been all but agreed upon to be 2.10v. Even then, we are not 100% sure they will last, 2.00v is the safer 24/7 voltage. For benching, however, 2.10v seems to be more than fine and even 2.20v appears to be pretty acceptable for certain periods of time. With this kit though, I will stay at 2.10v until all of the other testing for the review is complete, then perhaps we will see an update at 2.15v and 2.20v.

7-6-5-X 1T:
NVIDIA 790i:
Click link for full size screenshot...
905MHz :: 1.90v | 925MHz :: 1.95v | 945MHz :: 2.00v | 967MHz :: 2.05v | 981MHz :: 2.10v

Again, going from 7-6-5 to 7-6-6 gained next to nothing for frequency which is consistent with other kits of D9GTR/D9GTS that I have clocked. I was kind of hoping for a bit of headroom with a tRP of 6 as the 790i is suppose to take off at 7-6-6 versus 7-6-5 but I just didn't find that with these modules. This kit is definitely going to be 7-6-5 1000MHz single 32M stable, but the real question is going to be whether or not tRFC will be able to come down. As it stands, for dual 32M stability, it had to go up to a rather high 75 for the maximum frequency hit. In fact, for any decent clocks at 7-6-5, tRFC had to come up to 65 which really hurts 32M SPi performance.

8-7-6-X 1T:
NVIDIA 790i:
Click link for full size screenshot...
1021MHz :: 1.90v | 1045MHz :: 1.95v | 1068MHz :: 2.00v | 1088MHz :: 2.05v | 1112MHz :: 2.10v

This is where things got really fun. I just recently posted my G.Skill PC3-12800 (D9GTR based) results which showed an 1100MHz single 32M run and 15K Everest Read result on this motherboard. This memory took that a big step forward cracking out an 1112MHz dual 32M SPi run which up to this point has never been seen to my knowledge. To severely understate these results one would say they were nice. To appropriately state these results, one would have to use terms like incredible, un-believable, and a few adjectives that will simply show up as bananas. At some point I will be showing the max single 32M run with this kit at 8-7-6 but I want to save that for a really crazy run at max frequency of this setup...so look for that down the road a little bit.

3oh6 · #6 (**permalink**) May 17, 2008, 06:58 PM

1M SPi // 15K+ Everest :: 1130MHz 8-7-6
click for full size...including memset

Ahhh, the breaks from banging my head against the wall with an un-named X48 DDR3 motherboard make the whole day seem like it wasn't a waste. This setup is absolutely a treat to work with and despite hitting a FSB wall at 595MHz, I continue to find ways to make things happen with this board...and boy are the Ballistix willing instruments.

3oh6 · #7 (**permalink**) May 17, 2008, 06:58 PM

There you have another OC Report proudly brought to you by myself and 3oh6.com. I would like to thank Sam and Kelly from Crucial/Lexar for their fantastic support of my addiction and supplying the modules tested here today. Clearly there is some huge potential with D9GTS ICs when paired with the 790i chipset, but that was pretty much a known already. These numbers may just re-ignite a debate over whether D9GTS is better than D9GTR because these results are a pinch higher than what I have achieved with any D9GTR kits I have had. At the same time, this is also a new chipset being the 790i and I haven't had all my kits put through their paces on it. So for now, let's just chalk it up to a good kit of memory and forget the D9GTS/GTR debate for now.

No matter how you look at it, this kit of memory did incredibly well and will be a nice addition to my fleet of benching memory. For 32M SPi though, I hope I can bring the tRFC down as my G.Skill can bench 32M at these frequencies with a tRFC of 58 which has helped tremendously in my efficient times.

DDR3 prices are coming down but unfortunately this kit is priced rather high at this point. Hopefully that changes very soon though because I know a lot of people that are going to be interested in modules like this. Thank-you for reading and don't forget to keep your eyes on this thread when it comes to the top of the pile as there should be some updates coming in going forward. Have a great long weekend Canucks!

sswilson · #8 (**permalink**) May 17, 2008, 07:27 PM

Heh.....

You're bound and detirmined to switch us all over to DDR3 before the end of this year aren't you!?!?!?!? :)

Great to see some positive press for Ballistix, and looking forward to seeing what kind of 32M super-pi you can squeeze out of them once you get the timings down.

As always, great review!