Table of Contents
The world of extreme subzero overclocking is a small one but for those involved, it can be fiercely competitive. The brand options and choices for extreme cooling solutions are very few and, in many cases, are not even available to the public. However, today we are evaluating a new retail product from Bitspower, which has been expressly designed to be used with extreme cooling solutions, such as liquid nitrogen or dry ice.
Liquid Nitrogen cooling on memory modules is done to increase both the frequency and in some cases also lower the timings. This can be a difficult proposition for many reasons. Due to the fact that memory ICs are positioned perpendicular to the motherboard, and also that spacing is very tight, cooling the memory with liquid nitrogen requires specialized equipment. The LN2 must be poured horizontally and the cold is delivered perpendicular to that.
To fulfill this unique cooling requirement, Bitspower introduced the universal memory LN2 container. To accomplish this task, Bitspower breaks the product up into two different pieces. They designed a special heat sink which attaches to the memory modules and also makes it possible to attach a large LN2 vessel to cool them.
Bitspower entered the extreme cooling market with two different offerings. Today we are evaluating the universal memory container; however, they also launched RTX GPU containers at the same time. If you would like to know more, check out our review of the Bitspower GeForce RTX 20 Series VGA LN2 Container.
For serious overclockers, liquid nitrogen is the only solution in order to push their hardware to the next level and get the highest benchmark scores. And for them, Bitspower presents the Universal Memory LN2 Container. Made out of a solid block of high quality copper, the Bitspower LN2 Container is precisely sculpted via CNC for a clean channel routing and most efficient cooling capabilities. Finished with nickel plating, the Bitspower LN2 Container has a great look with greater performance. –Bitspower
In the table below we examine the particular details of the LN2 container being evaluated today.
|Bitspower Universal Memory LN2 Container|
|Memory Compatibility||DDR1, DDR2, DDR3, DDR4|
|Heat Sink Specifications|
|Heat Sink’s Included||4|
Retail Packaging and Accessories
The packaging for the Bitspower universal memory LN2 container is all business. Other than the part number on the back, there really is not much on the outside of the box. While it may have been nice to include some details on the outside of the box, such as a high-quality photo, they clearly saved you some money with this minimalist box approach.
Opening the box, we see that the core components are vacuum-packed in a thick plastic bag. Your eyes are blinded by the shiny chrome finish on the pot, so you don’t notice that there is not much internal structure to protect the contents. We’ll get to the measurements and weight later, but we were a little surprised that they didn’t add much internal padding to protect the very heavy container from damaging the other contents.
Meet the container
Once out of the box, we get an immediate impression of quality. Many extreme overclocking cooling products don’t go the extra mile when it comes to fit and finish, because they know overclockers only care about results and not style. However, Bitspower clearly wants to give its users both! The bottom of the container has a protective film so no scratches will occur during shipping.
As seen in the pictures below, the inside and the bottom of the container are left as raw copper. This is something we have never seen before with extreme overclocking products, but often it’s an adaptation which the end-user will do. The process of electroplating can leave the surface uneven. In some extreme cases, surface finish variance can be up to 0.5mm. Bitspower has taken the extra step to leave the bottom unfinished, so as to avoid any unnecessary surface imperfection. The bottom is nice and flat, so no modifications are needed.
There is a second reason they have taken these steps, and it has to do with thermal conductivity. The average nickel plating has a thermal conductivity rating of 52.4 Btu/(hr-ft-F). Pure copper, on the other hand, has a thermal conductivity rating of 231 Btu/(hr-ft-F). Even though the nickel plating layer is very thin, it acts as an insulator and does not provide optimal thermal transfer. The reality is that a small nickel layer won’t hurt the realized maximum MHz too much, if at all. However, that didn’t stop Bitspower from giving you the best possible condition for optimal thermal transfer.
The total weight of just the container is 1.85 Lbs with overall dimensions (LxWxH) of 128mm x 26mm x 67mm.
The convention is to provide a round hole near the top of the heat sink for a thermal probe. However, Bitspower has taken a different approach and put the thermal read hole on the base of the copper container. This is a nice touch, given that this appears to be the only thermal probe location. Accurate thermal readings are absolutely critical for frozen memory. In many cases, it takes an exact temperature to train the bios timings/frequency, and a different temperature to enter the OS successfully.
Meet the heat sink
Any good LN2 memory container is worthless without being able to transfer the cold perpendicularly to the memory ICs. The memory stick is held in place by three very strong screws. They have also included three set-screws, which allow micro-adjustments to the mounting pressure on the ICs. All of the hardware is stainless steel, and they even included a hex key to make your life easier.
The total weight of one heat sink with screws is 67 grams, with overall dimensions (LxWxH) of 125mm x 8.5mm x 38mm.
To assemble the memory module, Bitspower has included high-quality thermal pads. The heat sink can accommodate dual-sided or single-sided memory by adjusting the number of thermal pads used. Many extreme overclockers will choose to use thermal paste on the actual memory IC, instead of the supplied thermal pad. For single-sided DDR4, we will use thermal paste on the IC and two thermal pads on the back of the memory PCB.
Highlighted by a red arrow in the picture below is a cutout for component clearance. Some memory PCBs use a large MLCC capacitor in that location, which has the potential to touch the metal heat sink if there is no clearance slot. Because half of the IC is not physically touching the heat sink, this large cutout will need to be filled in with thermal paste. For those using the supplied thermal pad on the IC side, the cutout shouldn’t be a problem.
The cutout also serves as a nice location to add a thermal probe. It creates enough of a cavity that a very small K-Type thermal probe can be inserted without interfering with the heat sink cooling.
Testing and Overclocking
The overall objective is to overclock the memory to the max and squeeze every last MHz out of it. To do this, we will use just one memory module and down-clock the CPU in order to provide more headroom for the memory frequency. Furthermore, we will dramatically increase the timings from the standard XMP profile in an effort to reach higher frequencies.
The goal is to find the maximum frequency which can be validated with CPUz using air cooling only and then cool the memory with LN2 to compare the results.
|CPU||AMD RYZEN 7 3700X 8-Core|
|Water CPU Cooler||NZXT Kraken X62 280mm – All-In-One Cooler|
|LN2 CPU Cooler||Kingpin Cooling T-REX CPU Container|
|Motherboard||ASRock X570 AQUA|
|Graphics Card||PNY GeForce GT 730|
|Solid State Drive||HP EX900 M.2 250GB PCIe 3.0 x4 NVMe|
|Power Supply||Enermax MaxTytan 80+ Titanium 1250W|
|Operating System||Windows 10 v1903|
|Memory Speeds Compared|
|Maximum frequency with air cooling = 2525 MHz (5050 MHz Effective)|
|Maximum frequency with LN2 cooling = 2743 MHz (5486 MHz Effective)|
Do Heat Sinks Matter on Air?
You might be asking yourself if adding heat sinks would actually improve the thermal load of memory. One common misconception we often hear is that DDR4 doesn’t really produce any heat and manufactures only put heat sinks on memory modules so they can add RGB LEDs or other style elements. To better understand the heat load of DDR4 memory, we conducted a quick ambient air comparison of naked memory versus the same configuration, but with Bitspower heat sinks.
- Samsung B-Die
- Open-air test bench with no direct airflow
- Stable daily overclock of 4400 MHz with 19-19-19 timings at 1.50v
As you can see in the results below, after just 15 minutes of memory stress testing we observed a reduction of 8 °C just from adding heat sinks. For memory such as the Samsung B-Die, the reduction in temperature can be a huge difference in the overclocking headroom.
Micron E-Die Air Cooling
The ASRock X570 Aqua is still in development and, thus, we were not able to increase the base clock in the OS. To obtain the results, we set everything in the bios and entered the OS for validation. Typically with memory overclocking of this nature, the bios is set as high as possible, then further overclocking is conducted in the OS for even higher frequency validations.
Heat sinks proved not to matter with Micron E-Die ICs using air cooling. With a fan blowing on the memory, we achieved the same overclocking result with the heat sink on and off.
In the screenshot above you can see that we set up the processor to run with only two cores and SMT disabled. This did not yield any extra frequency on air, but we did this in preparation for LN2 cooling. The air cooling result seen below is simply a practice run for subzero cooling.
Micron E-Die Extreme Cooling with LN2
For the best possible result, we will use LN2 on the CPU as well as the memory. The reason we are also cooling the CPU is that it can greatly help the memory overclocking capabilities. We theorize that reducing the CPU to -120 °C or lower will have a significant impact on the results.
With two elements on the motherboard being cooled by liquid nitrogen, we expect an elevated risk of damaging the motherboard and/or memory. Adding a second LN2 pot makes the motherboard exceptionally cold. If the motherboard has not been adequately insulated, there will be damaged components. We have insulated our ASRock X570 Aqua with liquid electrical tape.
Further problems with moisture can arise due to the heat output of the X570 chipset. As many of you know, the X570 chipset is very warm and needs active cooling. Placing an extremely cold memory container next to the hot chipset is a recipe for serious condensation. We will be watching this area closely for the development of moisture.
As with all subzero cooling endeavors, temperature monitoring is critical. For reading the temperature of the memory, Bitspower included a probe hole at the base of the container itself. While this would certainly be helpful, it, unfortunately, doesn’t have the accuracy we are looking for with this project.
We will ignore the suggested probe location and substitute our own method. To get the most accurate temperature possible, we are able to slip a small thermal probe inside the heat sink itself. This puts the thermal probe effectively at the IC instead of 1.5 inches away.
We feel that Bitspower missed the mark with the thermal probe location. However, it’s quite easy to simply tape a probe on the outside of the heat sink, which will be considerably more accurate than the OEM solution.
As we cool down the memory we observe a very nice thermal response. The heat output of the memory is effectively nothing, the idle temperature is near ambient. For this reason, and because the copper mass is considerable, this container holds the desired temperature exceptionally well.
We were pleasantly surprised that, with minimal LN2, we were able to bring the pot down to any subzero temperature and keep it there for extended periods of time.
To achieve the best results, we held the CPU at a constant temperature of -130 °C for the duration. This proved to be an acceptable temperature, so we didn’t have to contend with any CPU cold bugs.
Finding the ideal temperature to use for the memory proved to be a challenge. Our testing consisted of an attempt to train the bios at a given memory frequency. If the bios did not train at -120 °C, then we used a blow torch and heated up the container to -110 °C to try again. After much trial and error, we achieved the best results with -80 °C. This was a lengthy process but the Bitspower universal memory LN2 container made it easy.
After about two hours of time trying to train the bios, we ultimately ended up with a frequency of 2743 MHz (5486 MHz Effective). As we mentioned above, we are not able to increase the memory frequency in the operating system at this time. To validate the score and create the best screenshot possible we used two methods.
Firstly, we saved the CPUz validation file at 2743.8 MHz, which can be seen here: https://valid.x86.fr/7rynj3. Second, we ran Super Pi 16K using the Benchmate benchmark validation software. This produces an output that shows many details, one of which is the memory frequency during the actual benchmark.
As the motherboard continues to be developed, we hope that we will have the ability to increase the base clock in the OS. If we are able to do this, then we will be able to validate substantially higher frequencies and potentially end up over 5600 MHz. As with everything in overclocking, the quality of the silicon is a huge factor in determining the maximum overclock. Thus, we will leave you with a hint ?? that with better memory and BCLK overclocking support we will be back…
For extreme overclockers who want to push their system memory to the maximum with liquid nitrogen, we believe there is simply nothing better than Bitspower’s universal memory LN2 evaporation cooler. Crafted from a single block of copper, and produced to the highest quality standards, the container has no equal.
In our testing, we were able to achieve perfect temperature stability. Memory overclocking with LN2 can be a tricky process and requires exacting ?? temperatures, but Bitspower makes temperature regulation easy. Once the copper has reached the desired temperature, it’s nearly effortless to maintain that temperature. The only negative feedback we have is that the placement of the temperature probe hole is not desirable.
This is a premium product that also comes with a premium price tag. With an MSRP of around $275, the BP-DIMM-LN2 is substantially more expensive than the competition. Bitspower includes compatibility for 4 memory modules in the package. Looking at the competition–including support for 4 memory modules–we find the BartX RAM LN2/Dice Pot for around $175, the Kingpin Cooling Ney Pro for $89. While it may be the most expensive option, we feel that it’s simply the best product on the market today and well worth the investment for extreme memory overclocking.
David Miller – mllrkllr88