After a brief break from overclocking reports I have managed to find time to put together another overclocking assessment of some fine looking memory from the good folks at Crucial, AKA Micron, AKA The Business, AKA The Supplier, AKA...alright; you get the idea.
Since the early days of DDR2 memory, Micron has been the only game in town when memory manufacturers were looking for a fix of the good stuff. It started with 'fatties' and then escalated to the now famous D9GMH/GKX crazy of 06 that spilled into 07. Now that 07 is winding down, a new kid is on the block (DDR3), and guess who everyone has on speed dial?
Today we take that step into the future of memory with a look at the Micron 'in-house' offering from the Ballistix line of DDR3 memory from Crucial. Crucial is the retail outlet for the memory monster that is Micron. Anyone can attest that Crucial offers top notch service and we all know what the product performs like. Here are the specifications of their only DDR3 Ballistix modules right now, the PC3-12800 (800MHz / DDR3-1600).
To demonstrate what the above specifications show, here is how the modules show up in Windows with the timings left to Auto in the P5K3-Dlx BIOS and the system adjusted to 400*9 / 1:2 / 266 Strap = 800MHz (DDR3-1600).
Obviously the system doesn't set the stock 8-8-8-24 timings but this is because of the motherboard. There is an XMP profile for the default timings and voltage (X38/X48 chipset feature much like EPP or SLI memory on NVIDIA chipsets). In order to run this memory at the stock timings/frequency...
A) you must be running at 400FSB
B) you must set the timings manually
To ensure compatibility and that all modules delivered to your door meet the standards outlined above, Crucial currently tests each dual-channel kit on two motherboards. Here is a chart outlining the test beds pulled from the Crucial Ballistix web site:
As for availability of this memory, you can find it pretty much anywhere you can buy Crucial memory...except in Canada, it is listed but not available. Here are the places I was able to find the memory and the prices at time of posting:
- Crucial.com - $569.99 USD
- Mwave.com - $492.25 USD
- NewEgg.com - $549.99 USD
- NCIX.com - $599.99 CND
- DirectCanada.com - $540.75 CDN
Package:
Click for full size...
Now what is wrong with these images? That's right, I too thought that I was getting multiple kits when this behemoth of a box showed up but alas, there was only a single kit inside...it's okay Sam, I won't hold it against you . I have seen some reviewers get this big box, and some get the normal sized two module box with this kit so I am not sure what the deal is but the big box certainly makes your heart race when you pull it out of the UPS bag. Aside from the size, nothing else has changed from the last dozen or so kits of Ballistix I have received from Crucial directly or purchased from on-line retailers.
The accordion cardboard stand inside holds the modules firmly in place and the anti-static sealed bag keeps the modules safe and cozy. There was also the standard memory handling and installation sheet accompanying the modules on their journey up north.
The Goods:
Click for full size...
The hardware underneath the heatspreaders has changed drastically from the Ballistix DDR2 counterparts but the heatspreader has all but remained identical with a couple subtle adjustments. First off, the heatspreader is no longer inter-locked at the top therefore the clips at the top have disappeared. This will surely alleviate the issue some Ballistix kits saw with un-even heatspreader contact on the ICs from top to bottom. This un-evenness was a result of the heatspreaders being too tight at the top of the modules. The other adjustment is the inclusion of the small DDR3 under the X in Ballistix.
The ICs used on these modules are naturally going to be the good stuff and if you look closely you can make out the D9GTR under the Ballistix stamp. Only the best Micron for Micron modules after all. In case you have been living under a draw bridge with a troll for the last few months, you can find out more about Micron D9GTR at the Micron web site, here. As for the PCB, I cannot find any info aside from what is stamped on the edge in the image above (Levin 73101). Crucial has stuck with the Levin PCBs it appears for their first round of DDR3 Ballistix. Let's hope they are ready for the abuse that these modules are going to see.
All Hail the P5K3-Dlx:
Click for full size...
This OC Report is a little bit more dis-organized than I am usually able to pull off so this memory had some testing done with just air cooling, while other timing sets were tested with the phase slapped on the P5K3-Dlx. You will also notice from the above photos that some clocks were done on the motherboards stock cooling while others were with aftermarket cooling on the north and south bridges. In addition to this, both an E6600 and E6850 were used for some clocks. Things have been crazy lately for me so I had to shelf my OCD for this report and get the clocks done when I had time regardless of what the setup was. Not all Dual 32M runs are fully optimized either. Again, time was a factor so I wasn't able to fully tighten secondaries like I would have liked.
Code:Setup: MB: Asus P5K3-Dlx Wifi (vMCH mod) CPU: Intel C2D E6600 (L629B383) / Intel C2D E6850 (L724A479) CPU Cooling: Thermalright Ultra-120 / Chilly 1 SS GPU: Biostar 8600GTS 512MB DDR3 NB Cooling: Stock / Noctua NC-U6 SB Cooling: Stock / Thermalright HR-05-SLI PWM Cooling: Stock / Naked PSU: Silverstone Zeus 560W HD: Seagate SATAII 80GB 8MB NCQ OS: Windows XP Pro SP2 Ambient Temperature: 22-23C
I don't have a particularly amazing P5K3-Dlx so I had to help it along to make sure I can maximize the memory I test. The vMCH mod has helped get the last little bits out of this kit and other DDR3 kits so keep that in mind when looking at results. P35 chipsets are almost like CPUs these days where some are great, others not so much. This has led to the X38/X48 binning that we are seeing in the next generation of Intel chipsets already. Great, now not only do we have to pilfer box after box of CPUs looking for a decent clocker with high FSB capabilities...we now get to test motherboard after motherboard to find one that clocks memory well. It's not like they cost enough as it is, let alone buying 6 only to sell 5 off at a loss. I digress...
Maximus Extreme is in the house!:
Click for full size...
I originally posted this report with just P5K3-Dlx results but the Maximus Extreme has been added to the family and I have begun adding results from this setup as well. Here is a complete breakdown of components used, it is pretty much the same list as the P5K3-Dlx:
This motherboard is clearly more stable at higher memory frequencies, as we will see, but doesn't appear to be the better setup for high FSB sub-zero benching. I am still not sold on X38 a there appears to be plenty of issues with memory ratios and the performance in SPi seems to be about the same at a lot of the clocks I have tested.Code:Setup: MB: Asus Maximus Extreme CPU: Intel C2D E6850 (L724A479) CPU Cooling: Thermalright Ultra-120 GPU: Biostar 8600GTS 512MB DDR3 NB Cooling: Stock SB Cooling: Stock PWM Cooling: Stock PSU: Silverstone Zeus 560W HD: Seagate SATAII 80GB 8MB NCQ x 2 OS: Windows XP Pro SP2 Ambient Temperature: 22-23C
P5K3-Dlx Dual 32M SPi Overclocking:
As usual, a few notes before we see the numbers. The voltage listed in the open text file in each screenshot is measured with a DMM from a test lead point in-between the memory slots and grounded to a ground spot on the board. P5K3s are known to over-volt it appears and this particular board is usually .06v above what I set in the BIOS. You can also see what was set in the BIOS in the AISuite window that is open in all screenshots.
Second, I am only interested in dual 32M SPi stability for my overclocking reports. I have found dual 32M stable with tight secondaries usually means Prime / Windows HCI Memtest stable within a handful of MHz at the same voltage or with a small bump in voltage. Don't confuse dual 32M stable with 24/7 stable but it is reasonably close most of the time.
Lastly, yes, this is a review sample sent directly from Crucial but is simply a random sample that anyone would receive who ordered from their web site. Memory manufacturers know better than to send cherry picked sticks these days...we all know how that turns out in the forums. Enough chit-chat, on to the first part of the show:
(B) = BIOS selected voltage
6-6-5-X 1T:
Intel P35:
Click link for full size screenshot...
710MHz :: 1.75v(B) | 732MHz :: 1.80v(B) | 752MHz :: 1.85v(B) | 764MHz :: 1.90v(B) | 782MHz :: 1.95v(B)
Yes I am going to be a pansy and not go over 2v with this kit just yet. There has been very few people with Ballistix DDR3 and I don't feel like being the guinnea pig that figures out what voltage is safe with these Levin PCBs. I have plenty of benching to get done on this kit before the die on me and lets be honest...they seem to be clocking quite well at 2v or less.
7-6-5-X 1T:
Intel P35:
Click link for full size screenshot...
808MHz :: 1.75v(B) | 838MHz :: 1.80v(B) | 866MHz :: 1.85v(B) | 880MHz :: 1.90v(B) | 900MHz :: 1.95v(B)
I have also done some clocking at 7-6-6-x and found that an extra 10MHz or so can be had in dual 32M stability by changing tRP to 6 from 5 at the same voltage. I have posted the 7-6-5-x results for now but will likely put the full range of 7-6-6-x dual 32M clocks in the relatively near future. Getting to 900MHz under 2v dual 32M stable puts this kit right up there with some of the better clocking kits I have seen. Of course everyone else just posts single 32M or 1M shots so it is hard to judge where this kit really stands.
8-7-6-X 1T:
Intel P35:
Click link for full size screenshot...
920MHz :: 1.75v(B) | 930MHz :: 1.80v(B) | 940MHz :: 1.85v(B) | 950MHz :: 1.90v(B) | 954MHz :: 1.95v(B)
Clearly the results above are the result of a motherboard running out of juice. Many P5K3-Dlx users can relate to what is going on above. As you can see, MCH voltage is jacked right up at almost 2v for most of the higher frequency dual 32M runs and getting past 950MHz is quite a battle. I am hoping to get a water loop up and running on the NB so perhaps that will help get a little more stability at the higher frequencies. For now though, anything over 950MHz is going to be relegated to single 32M or less.
Maximus Extreme Dual 32M SPi Overclocking:
As we saw in the P5K3-Dlx results, the BIOS selected voltage was listed but noted that the motherboad over-volts by 0.06v. The Maximus Extreme also over-volts but it reports the correct actual voltage in Windows through Everest so every screen shot below has the correct voltage listed so I have simply listed the results with their actual voltage.
8-7-6-X 1T:
Intel X38:
Click link for full size screenshot...
944.1MHz :: 1.90v | 970MHz :: 1.95v | 990MHz :: 2.00v | 1014MHz :: 2.06v | 1032MHz :: 2.10v
Hahaha, what can I say, these sticks are absolutely incredible at 8-7-6 timings on this motherboard. Clearly the limiting factor on the P5K3-Dlx setup was the motherboard and once again, the motherboard seems to be the limit of these clocks here again. I am still working on getting a little more out of this beast but it looks like 1030MHz might be close to the max on the 266 or any strap for that matter strap.
Final Thoughts:
So far this Ballistix kit is really holding its own. Once I grow 'a pair' I will creep over the 2v mark and put up some bigger numbers. I also hope to have some single 32M results posted along with a whole section dedicated to bandwidth but as we have seen, all of the screenshots above have the Everest bandwidth listed in them. Be sure to check the PerformanceLevel in Memset and the FSB of the motherboard as the bandwidth numbers are heavily dependant on those two things. There is quite a mish-mash of bandwidth results because of the sloppy organization on my part in the testing.
I hope this report, like all my previous reports, are found to be useful for those making purchasing decisions or those users looking to clock their own kits. Keep in mind, however, that this is simply a single sample and memory all clocks differently. I encourage you to hunt out results from other users and not to base a buying decision solely on my findings. Eva2000 from i4memroy.com has also played with a kit of these Ballistix so there is another great source of results.
Of course, feel free to ask any questions about the memory, the P5K3-Dlx, or comment on the report in this thread. Also watch for the more exciting results that I will be posting in the second post of this thread periodically as they happen. Thanks for your time and keep clocking
Bookmarks