Techgage logo

Mushkin 2GB XP2-8500

Date: December 13, 2006
Author(s): Rob Williams

It’s been a few months since we’ve had the opportunity to play with a Mushkin kit, but we now have their XP2-8500 on the hotseat. We compare its performance to Super Talents and Kingstons top kits to see who comes out on top.


Who produces the best memory kit out there? It’s useless to even mention a name, since within days another company will come out and stomp that one to the ground and regain the crown. That said, Mushkin is right up there with the rest, although they fall slightly behind OCZ and Corsair in regards to top offerings. However, the kit we are taking a look at today offers some incredible speed with reasonable timings.

Like all other memory kits available with 8000 speeds and beyond, you can expect to slap over at least $400USD for these. Checking out a popular e-tailer, the $399 price tag fits right in between everything else, so it’s not overpriced. Corsair offers an 8500 kit for $10 less, but has 5-5-5 timings as opposed to Mushkins 5-5-4. This is not a big deal regardless. My recent favorite kit, the Kingston 8000 also has 5-5-5 timings but retails for around $435. That said, Mushkin is comfortably sitting in the middle price-wise, but has the best timings of the bunch.

We’ve only taken a look at Mushkins value line-up in the past, so I was interested to see how their XP2s stacked up to the other kits I’ve recently looked at. Is this kit worth your cash in the end? You won’t find out until after we’ve taken a closer look.

Closer Look

Since Mushkin performed a redesign a few months ago, their packaging has benefited… it looks far better now. Clean white background in a plastic blister pack that keeps your modules safe during travel. Despite the photo looking blue, it actually is a bright white background.

The modules themselves are pure black, which to me is the best possible color. It’s only too bad that the PCB weren’t also black… that would be one pimp looking module. These are the “FrostByte” heatspreaders which Mushkin introduced a while ago. The top is completely opened for better air circulation. Functionality aside, these are straight out some of the best looking modules on the market. Nothing flashy, but there doesn’t need to be.

The sticker here tells you absolutely everything you need to know, except for required voltages. I hate these stickers, but they are needed for warranty reasons, so it’s understandable. At least Mushkin designed the stickers to be reasonably good looking.

Here is a better, albeit dark, look at the rings atop the modules. There is plenty of breathing room here, but the rings look better than nothing at all.

If you are drooling, you can stop because it’s time for the technical details and then some benchmarking.

Testing Methodology & Technical Details

Because these are from Mushkins performance line, we should expect some great results in our testing. PC2-8500 is somewhat of an odd number, as it results in DDR2-1066 speeds, the same as Corsairs modules released earlier this year. As I mentioned, we are dealing with 5-5-4-12 timings, which is tighter than the standard 5-5-5-15 that I have seen everywhere else.

Mushkin recommends 2.2-2.3v for these and after initial testing I found that MemTest for Windows would only become error free with 2.3v. Depending on your hardware and luck, 2.2v may be quite feasible for you. Of course I didn’t keep stock speeds for very long. I first tested both DDR2-800 and DDR2-900 and also my theoretical maximum overclock of DDR2-1100. As mentioned in my recent memory reviews, my 4600+ is lackluster in that DDR2-1100 is the most I can push out of ram, since moving into the 2.80GHz area is sketchy, and will bring on errors in any benchmark after a long run. 2.76GHz on the other hand proves 100% stable, so I stick to it for our max overclocks.

I will be comparing these modules to my Kingston PC2-8000 and Super Talent PC2-8000 kits, which both use Micron D9GKX chips… the cream of the crop. These Mushkins however use the lesser binned D9GMH chips. The primary differences between each of these two chips is minor… one is clocked for 333MHz and the other for 400MHz. The GKX chips cost more and tend to increase prices of the end modules. This has not really been a problem lately though, as GMH and GKX is priced similarly on the market simply depending on the overall stock speed of the modules in question.

Both of these chips are amazing, there’s no question about that. So, don’t take my DDR2-1100 overclock as being the max capable. I am sure with a killer CPU, all of the modules I have around would be pushed at least to DDR2-1200 speeds, which is becoming a norm in the enthusiast community. That said, here are the settings I chose for benchmarking.

One thing you may be asking yourself is, “Does this guy believe in even numbers?” and the answer is no. That aside though, the DFI M2R/G motherboard has FSB options in multiples of 2, therefore 275FSB is not possible, but 274 or 276 is. When I run into a spot where I’d like an odd number, I just use the next best setting. In that case, our max overclock was accomplished using 276FSB and a 10x multiplier.

The system we are using for testing:

Regardless of what product we are benchmarking, our testing machine goes through optimizations before we begin. All testing is performed on a stand-alone version of Windows XP Professional with SP2. System services not needed are closed down, in addition to unwanted programs in the system tray. Computer is cleaned up of scrap files, which includes emptying of the recycle bin. Directly before any testing occurs, the OS hard drive is defragged using Diskeeper 2007 Professional. Finally, we assure that the computer has adequate cooling and airflow.

As I mentioned, I will be using also the Super Talent 2GB PC2-8000 and Kingston HyperX 2GB PC2-8000 for comparisons sake. The results from both kits were used from earlier reviews using the same setup. However, since the Mushkin kit uses a stock speed of 1066MHz, I re-tested that speed on both of these kits. As usual, since we are running all of the kits of ram at the same speeds, we don’t expect either one to come out on top. The goals are to make sure the modules in question can keep up to what’s standard.

Sandra, Everest, Sciencemark

Sandra is one of those applications that enthusiasts are not foolish enough to leave out of their toolbox. It’s not only a memory benchmarking tool, but can test the performance of anything else in your system. They have recently updated their application to XI, which resulted in a small redesign of the app.

In all of the tests here, the XP2-8500s actually came in last except for one.

The same goes for our unbuffered tests, for the most part. All of the results are still tight.

Everest 3.5

Everest is another popular benchmarking tool, but primarily handles memory and CPU, in addition to some disk testing. One thing to note with Everest is that with each new version, the results in the tests can change drastically. That being the case, results from 3.0 cannot be compared to results from 3.5.

The situation is a little different with our Everest run over the Sandra one. Our 800MHz and 1100MHz results are very close to one enough, between modules. The middle sets sway a little bit, but is nothing worth noting.

Same for our latency tests… very close results all around.

Sciencemark 2.0

Without a doubt, Sciencemark has proven to be one of my favorite CPU and memory benching tools. Oddly enough, the developers behind the program seem to have lost their domain, so it’s unlikely we will ever see a Sciencemark 3. It’s too bad too, considering it gives far more in-depth results than anything else out there.

Once again the results are in line with what we expect, and there is no clear superior.

Super Pi, Real World, Final Thoughts

Super Pi is a great tool for wagering the performance of your CPU. The faster the chip, the faster your computer can compute 1 million digits past the decimal in a not so simple Pi equation. Memory latencies and speeds can play a big part in the success of low scores (that being a good thing) so it’s good to include it in our tests.

Ditto. The Kingston proved better capable of handling the lower times overall though.

Real World Multi-Media

In order to provide more accurate results in our memory and CPU reviews, we are currently in the process of reconsidering which tools we use. We will no longer use PC Mark 05 or 3D Mark 01 in our memory reviews simply because the scores fluctuate far too much in between each test, making it unreliable.

That said, one area of computing that will benefit from faster CPU and memory speeds is multi-media work. This can include DVD Ripping, video conversion, audio conversion and even image conversion. The latter three will be included in our results below. These tests will be refined in the coming weeks and we are hoping to have a solid multi-media benchmarking suite to use in our future reviews.

Before the results though, allow me to explain which each process consists of:

The results are quite interesting here, and prove that there must be some factor that helps one of the kits be faster overall regardless of timings. All of the kits here had identical settings, yet the first three tests vary considerably. The Audio #3 on the other hand, doesn’t move between either kit.

Final Thoughts

Thanks to my gimped CPU, it’s hard to be impressed by any new kit that flies onto my desk, since most of them can handle the same max overclock that my system can handle. This Mushkin kit was no surprise, it handled it just fine. However, it did throw us a few curveballs that I will bring up.

First off… the voltages. That’s the only thing that really differs this kit from the others I tested with. The stock voltages are 2.2-2.3v, but 2.2 was not possible for me because MemTest would throw errors around. 2.3v took care of that entirely though, so should you choose to pick this kit up that’s what you can expect. You may get lucky with 2.2v though, who knows. It doesn’t matter in the end really, it’s not a significant difference.

At our max overclock though, the same problem came into play. While the Kingston and Super Talent kit handled DDR2-1100 speeds with only 2.4v, the XP2-8500 required 2.53v. Even 2.50v would give me some errors, but 2.53v went through a 1000% test error free. So, these are voltage hungry, but so are enthusiast overclockers ;-)

Price wise, these modules retail for $400 which places them in line with the competition. Corsairs kit costs $10 less and Kingstons kit costs $40 more. Take this as you will. At this price range, you cannot go wrong, at least with the modules mentioned. All of them have plenty of overclocking headroom, depending on your CPU. If you have a beefy AMD that can hit 3.0GHz or a nice Core 2 Duo, you should be able to hit DDR2-1200 as long as you are willing to pump up the voltages. I am awarding the XP2-8500 an 8 out of 10… a great showing from Mushkin.

Discuss in our forums!

If you have a comment you wish to make on this review, feel free to head on into our forums! There is no need to register in order to reply to such threads.

Copyright © 2005-2021 Techgage Networks Inc. - All Rights Reserved.