Mushkin Ascent XP3 16000 2 x 1 GB Review

ccokeman - 2008-04-30 19:24:18 in Memory
Category: Memory
Reviewed by: ccokeman   
Reviewed on: May 11, 2008
Price: TBA

Introduction:

DDR3 memory is moving closer and closer to the mainstream marketplace. Prices are falling and speeds are on the rise. With each new performance motherboard chipset delivered by the manufacturers, the speed threshold continues to increase in a self perpetuating cycle. What that means, of course, for the power user, is that performance continues to increase as new hardware is made available. Unfortunately, one of the means for increasing that performance comes with the added cost of a higher power requirement to make the memory perform as advertised. Performance DDR3 modules voltage demands are now pushing into the 1.9 to 2.0 volt range; quite an increase from the JEDEC spec of 1.5 volts. Along side the increase in voltage requirements comes the increase in the thermal load on the memory ICs. Essentially, more power equals more heat. It's the transference of energy from one medium to another. Heat can and will kill your memory modules, I've had it happen first hand. What is needed is a means to dissipate that heat or thermal load. There have been many different heatspreader designs over the years that have met with mixed success rates. Mushkin has the Frostbyte technology, Patriot has the Viper Fin design, Corsair and other have thier own designs, and each have their merits.

With the release of the Ascent XP3 16000 series of modules, Mushkin has pushed to the forefront with another method of cooling the memory modules. Sure there is a heatspreader, but it's what lies just beneath the black heatspreader that makes the difference in this case. The Ascent modules are cooled with a technology developed by Celsia technologies, called eVCI (enhanced Vapor Chamber Interface). In a few word, this nanospreader design is a two-phase vapor chamber cooling system. Celsia describes it best, "A NanoSpreader™ is a patented copper encased two-phase vapor chamber into which pure water is vacuum sealed. The liquid is absorbed by a copper-mesh wick and passed as vapor through a micro-perforated copper sheet where it cools and returns as liquid to the wick."

Just how will this technology help cool modules designed to run at 1.9 to 2.0 volts? Will the nanospreaders do the job and keep the modules cool and operating as they should? Is this just another example of "getting more"? Let's find out!

Closer Look:

The Mushkin Ascent XP3 16000 come in the standard retail clamshell packaging that Mushkin has used. The simple understated look that says I don't need the flash. The Mushkin product card carries the "Get More" slogan, as well as the brand name on the front side, and features a brief installation and troubleshooting guide on the back. The product code for this set of modules is 996620, and it is a 2 x 1 gigabyte set of modules.

 

 

The XP3 16000 modules are designed to use 1.9 to 2.0 volts and run at an amazing 1000MHz with latencies of 9-9-9-24. Of course, a board using the 790i Nvidia chipset is reccomended. You will notice the design of the heatsinks are a radical departure from the Frostbyte technology that Mushkin has used for quite some time. The modules feature a finned heatsink that uses additional ribbing on the sides of the modules to increase the surface area, leading to additional cooling capacity.

 

 

The heatspreader design uses a thick aluminum outer shield that has fins across the top of the modules to dissipate the heat load into the air stream over the modules. The nanospreaders can be seen under the aluminum shields. These nanospreaders are in contact with the memory ICs to provide a path for the heat to reach the outer shields to be dissipated.

 

 

Here is a quick shot of how the Celsia nanospreader technology works. It looks pretty interesting as the spreader makes contact with the heat source (memory IC's) and disapates it into the finned outer shield. This provides a wider heat path to the dissipation point than a heatpipe solution. Therefore, you have a much more even distibution of the heat load, making the components run cooler across the whole face of the design.

 

Specifications:

 

Part number
2 x 1GB  996620
Frequency
2000 MHz
Latency
9-9-9-24
Voltage
1.9-2.0
Module
128Mx64
Parity
Unbuffered
Pins
240 Pins
Chips
128Mx8

 

Features:

Testing:

The way to verify that one set of memory modules is better than another, is to run a series of benchmarks and gather some basic comparison data. When all things are equal, and the only variable is the module being tested, the results are a great way to compare performance. In order to eliminate the variables, the only settings that will be manipulated will be the memory timings and voltages when overclocking. The comparison modules will be run at the manufacturer specified timings and voltages, at a speed of 1333MHz. In order to reach 2000MHz, the processor used in the test setup will have to be overclocked from 333MHz to 450+ MHz. All of the comparison modules were run at 1333 MHz, the Mushkin Ascent modules were run at both 1333 and 2000MHz. The memory and CPU were run in unlinked mode throughout the testing.

Testing Setup:

Comparison Modules:

 

 

CPU-Z: This application shows us the settings that we have chosen in the BIOS. Items shown in this application include CPU speed and bus settings, motherboard manufacturer, BIOS revisions, memory timings and SPD chip information.

 

 

Task Manager: We use this utility to show physical memory, kernel memory, page file, and processor usage.

 

Overclocking:

Overclocked settings:

Having to test modules rated for 1000 MHz (2000 MHz effective) presents some unique challenges. First of all, you need a board that can push 1000 MHz+ on the northbridge. Second, you need a chip that can run speeds of 500+ MHz to run 1:1 when testing. Well, we have the next best thing. Since my poor Q9450 just won't let me play in the 500 MHZ range, I had to look at the alternatives. The modules are designed to run on an nVidia 790i SLI chipset based motherboard, so thats what I did. Because of FSB holes, or a poor divider, I was not able to gain any speed over 2000 MHz at the rated latencies of 9-9-9-24. So instead of going higher, I decided to try and see how tight I could run the latencies. 2000 MHz at the stock volts of 1.9 to 2.0 volts and 9-9-9-24 was a cake walk. Heck, it should be. Now it got a little interesting. First up was 9-9-8-24, good so far.  Then down to 9-8-8, then 8-8-8 - I thought this looked promising. After a few hours of test and repeat, latencies of 8-8-8 were a reality at 2000 MHz. Hmmm, let me push my luck at 8-7-7-24. Surprisingly, this fell after another session of test and repeat benchmarking. "Can there be more?" I asked. Well, yes there was. CAS 7 was a no go at 2000 MHz, but 8-7-6-24 was good. However, it would lock up randomly in a few tests. With the voltage adjustments I had been successful with, they still would not allow it to happen. So, I figured I'd start low-balling the voltage to the DIMMs, and ended up at 1.82 volts for my game and benchmark stability.

 

The benchmarks used in this review include the following:

Benchmarks:

 

Testing:

PcMark Vantage: With this benchmark, I will be running the system suite, as well as the memory test suite. The measurement for the system suite will be the total score. The measurement for memory performance is the total memory score. A comparison will be made of the performance at DDR3 1333, for a comparison point, and the highest achievable speed for the Mushkin XP3 16000.

 

 

 

 

 

 

 

SiSoftware Sandra XII: In this program, I will be running the following benchmarks: Cache and Memory, Memory Bandwidth and Memory Latency. All benchmarks will be at default timings. Higher is better in all tests, except for Memory Latency, where lower is better.

 

 

 

 

 

In the PCMark Vantage 1333MHz testing, the Ascent modules finished lower on the suite score, but were over 400 marks higher on the memory score. In the Sandra testing, the scores at the 1333 MHz level were just about identical, given that the speed and latencies were again almost identical. Once the speed was ramped up, of course, the modules performed better.

 

Testing:

Company of Heroes is a real time strategy game set during World War II. The object is to occupy and control the ground you capture, while forcing the opponents to capitulate. We will use the in-game performance test to measure the performance of the system.

 

The settings used in this test are listed below:

 

 

 

 

 

 

 

 

 

 

 

 

Higher is Better

 

The Mushkin Ascent modules perform about the same as the other modules at the 1333 MHz threshold. But, move them into the range they are comfortable with and they definitely offer a performance increase. At their rated speeds and with tightened timings, the performance is nice, with a 13 FPS improvement in the lowest resolution tested, to a 6 FPS difference in the 1920x1200 testing.

 

Conclusion:

As DDR3 motherboards continue to be released, the migration to DDR3 memory has begun. The prices have started to drop as more people are starting to adopt the new standard with their latest builds. The memory manufacturers are keen to this as well, by continually releasing modules that operate at higher frequencies. With the release of the QX9770 and its 1600 MHz core clock speed (400 x 4), and XMP profiles that are being programmed into the SPD of many of the higher performance modules, 800MHz (1600 MHz effective) is the starting point. Having modules that can run natively at speeds higher than thae XMP standard is the challenge for the memory manufacturers. As with most high performance modules, pushing the limits means increasing the supply voltage to them, which in turn brings more heat. Two things Mushkin has done with the Ascent XP3 16000 modules is take care of both problems. This set of modules does indeed run at 2000 MHz natively. Getting rid of the heat is the other problem when 1.9 to 2.0 volts is the requirement for operation at this speed. To fix this problem, Mushkin is using a vapor chamber cooling assembly on the modules. This eVCI technology is provided by Celsia. It effectively removes the hot spots that a traditional heatpipe style of cooling system would impose upon the memory ICs, therefore allowing the modules to run cooler at higher voltages. The difference in module temperatures was like night and day between the HP3 10666 set of memory I looked at last year. At the JEDEC spec 1.5 volts the modules were fine. Once they started getting more than 1.7 volts the heatspreaders were actually quite warm to the touch. Not so with the Ascent modules, even at 2.1 volts they didn't scream for relief.

Once again, I am pleased with the results I have gotten from the modules Mushkin has produced. Since the XP3 16000 modules are rated for 2000 MHz operation at 9-9-9-24 using 1.9 to 2.0 volts, the expectation is that they will run and perform at this level. That they did. What I was not expecting was the ability to reduce the latencies at the 2000 MHz threshold. Normally, you will need an increase of the voltage supplied to the modules to make this happen. What it actually took was a voltage decrease to do this, down to 1.82 volts and 8-7-6-24 at 2000 MHz, was the result of this exercise. If you need a set of modules to run with that brand new E8400 or E8500, and you don't want to wonder if the modules can keep up, this just might be your set of modules. We'll just have to wait a bit longer on pricing, though.

 

Pros:

 

Cons: