Memory Diagnostics Shootout – Round 2

Pages: 1 2 3

On To Round 2…

In order to streamline the testing a bit (I allowed a 12-hour or longer run for each program to detect at least one error), I decided to only keep two of the Ultra-X products for Round 2 – RST and RST Pro, since all Ultra-X products use much the same test patterns, and these are the memory specific utilities. So, the four products vying for the title of World’s Best Memory Diagnostic were RST, RST Pro, MemTest86 and GoldMemory.

The next test suite thrown at the products were three defective DDR DIMM modules, provided again by Micron Technology specifically for this article. One of the modules had bad cells, which all of the diagnostics identified. The second produced errors only when CL2 was specified, but not at CL3 (this is a CL2 spec’d module), and all products identified errors in at least some of the test runs.

The last module (PC1600 CL2) was by far the most difficult to identify. During the first test run, none of the diagnostics detected an error at 100MHz and CL2. I was a bit puzzled by this, as Micron had marked it as defective. I switched motherboards and ran it again at 100MHz, without getting any errors. I then decided to run it at 133MHz CL3… again, no diagnostics detected any errors. Finally, I set the CAS Latency to 2, and some errors began to pop up. RST, RST Pro and GoldMemory all identified the errors fairly quickly (though RST passed it in about half of its runs). MemTest86, however, never identified an error, even after running for 24 hours solid.

Even though MemTest86 had failed the previous test, I decided to include it in the ultimate memory diagnostic test I could find – the 3-RIMM i820 chipset error detection test. After the i820 data problem was identified, most manufacturers pulled their 3-RIMM designs from the market, however AOpen continued to offer theirs for customers who wanted the additional memory capacity and felt that the data corruption issue was unlikely. After all, only very specific and unusual data patterns were reported to cause the problem.

First, I tested with only 2 RIMMs installed, just to make sure that the board itself and the memory would pass in an officially sanctioned configuration. None of the diagnostics detected any errors. I then installed the third RIMM, and threw the diagnostics at it. Both GoldMemory and MemTest86 ran for 24 hours without detecting a single error. RST and RST Pro both showed data errors within several minutes, giving the crown of ‘Best Memory Diagnostics’ to the Ultra-X products.

In the final analysis, all of the products that made it to Round 2 of this shootout do an exceptionally good job of identifying memory errors. MemTest86 is certainly more than adequate for the home user and hobbyist that does not keep extremely critical data on his/her system… and it is free. GoldMemory costs a little more, but also seems to be a bit better. This is an excellent tool for smaller shops and support techs that don’t have much of a budget to spend on diagnostic tools – however, note that absolute data integrity cannot be guaranteed. In my opinion, the Ultra-X products are top-of-the-line, and should be in the toolbox of any real support professional. Businesses that rely on their systems for mission critical work are generally willing to pay for top-notch support and service, and the RST products will allow support professionals to provide the level of service expected by these types of companies. While RST is excellent for most PC systems, it will not detect ECC or parity errors. RST Pro is the most expensive diagnostic I’ve seen, but it is also undeniably the best system memory tester available on the market, from my experience. I have heard comments from QA and support people within large memory module suppliers who swear the RST products are as good, or even better, than the multi-thousand dollar hardware testers they have… and are faster and easier to use.

The results of these tests also point out that memory problems are not necessarily related to physical DRAM chip defects. There may be data bus, address bus, memory controller or other memory subsystem component problems that would not be identified by a typical hardware RAM tester. Even those who manufacturer or sell quantities of memory modules would be served well by using one of these tools for compatibility testing in various motherboards and/or systems. The bottom line is, in my opinion, the price of RST or RST Pro is money well spent for the support professional.

Pages: « Prev  1 2 3  

Be the first to discuss this article!