TEST ACCURACY AND RELIABILITY

1. Q. Can RAMCHECK or SIMCHECK II guarantee that my memory modules are 100% accurate?
   A. Unfortunately, no. No memory tester, including those costing $500,000 and more, can achieve exhaustive tests (that is, 100% accuracy) due to the virtually unlimited possibilities of data combinations which may be stored inside a memory device. Therefore, while we continuously strive to improve RAMCHECK's test algorithms and design, we cannot guarantee 100% accuracy of the test results.
   
   
2. Q. Well, if your tester (or any other tester for that matter) cannot guarantee the test results 100%, then what is the benefit of having it?
   A. While it is difficult to assess the overall accuracy of a tester, INNOVENTIONS testers are regarded as fairly accurate in the industry, as is attested to by the success of our pioneering products over the years. Many of our customers have been loyally using our products for many generations (Classic RAMCHECK, old SIMCHECK, SIMCHECK II, and now the new RAMCHECK).
   
   The problems that are not detected are usually rare. And in many cases, some errors undetected by our testers are later corrected with new firmware releases, as discussed below.
   
   In the few difficult situations when a technician is faced with a rare intermittent problem, our testers provide much more information than merely GO/NO GO. The technician can use the detailed speed results and structure information to compare among suspected modules and find the one that is most likely to be the culprit.
   
   With so many "mis-marked" modules in the market, the tester is an indispensable reference tool. An extremely slow module which is mis-marked at higher speed can be spotted by the tester and save you from later embarrassment.
   
   With the large variety of memory modules to choose from, it is not enough to know just the size and speed information marked on the module. Often you need to know the structure information in order to fit the module in a specific motherboard. Our testers provide you with such information.
   
   
3. Q. What is the difficulty in performing exhaustive tests of memory devices?
   A. In short, the main problem is that each bit of the memory may be shorted to quite a few adjacent bits, and there is a huge number of such bits in today's memory devices. Also, every chip die version has a different topography, even among chips with the same part number from the same manufacturer. For example, in a typical 32Mx8 chip, there are 256 million bits. Testing possible shorts between each bit to potential other close-by bits requires an astronomical number of tests, each takes a short test time. Therefore, all memory test equipment must use "optimized" short tests that provide speedy tests at the expense of missing 100% coverage. Please also refer to Application Note INN-8668-APN5 for further details.
   
   
4. Q. What is the difference between RAMCHECK or SIMCHECK II and the expensive certifying test equipment used by the memory manufacturers?
   A. In order to fully certify memory devices, an expensive tester which includes customized software and a high temperature environment chamber is used by the memory manufacturers. While even these testers cannot achieve 100% accuracy, their customized software is optimized to the internal layout of the tested device in order to detect hard-to-find bit interferences. Such a tester requires a long process of test program preparation by the software engineer, and the test itself is fairly long. Since our testers are optimized for low cost and simple operation, we do not provide true high temperature testing which is critical to certification, nor do we require you, the customer, to program the test for specific internal memory chip die. Therefore, RAMCHECK and SIMCHECK II cannot be regarded as a certifying test equipment. And yet, our product success over the last 15 years testifies to the fact that we have developed great reference tools with our improved algorithms and advanced design.
   
   
5. Q. I have a "mission critical" application. Can I rely on your testers to certify my memory modules in such an application?
   A. Absolutely not. Our testers are not sold as a certifying equipment. Certifying memory test equipment, as mentioned above, must include a high temperature chamber and advanced customized programming to exactly accommodate the internal topography of each memory chip (see also Application Note INN-8668-APN5). Such certifying test equipment costs $500,000 or more and requires qualified field engineers to operate. Even these equipment manufacturers will not give you a guarantee of 100% fault coverage.
   
   
6. Q. What is the nature of hard to find intermittent errors?
   A. In general, hard to find intermittent errors can be divided into the following groups:
   
   Temperature related: These are problems that appear only after a few hours of operation inside a hot computer. Typically, they cause an increase in the internal leakage of the DRAM cells or cause "cold soldering" to break contact at high temperature. Our patented Chip Heat test increases the temperature of the modules under tests but we do not reach the 70 degrees Celsius that is achievable only with the heat chamber of an expensive certifying memory tester.
   
   Pattern sensitivity: These problem are caused by a short (or leakage) between adjacent memory cells and create a memory failure when the cells have different data values. This problems are discussed in Application Note INN-8668-APN5.
   
   Soft errors: These are radiation related errors, which cause a reversal of data in a DRAM cell when stray radiation (e.g. alpha particles) hit the cell. These problems have been reduced over the years of memory development but they are still regarded as the prime cause for unexplained errors. This radiation can be generated from the regular ambient cosmic radiation and therefore, can attack memory everywhere.
   
   Voltage spikes: These errors can occur when voltage spikes inside the computer cause severe noise in the sense amplifiers inside the memory device. Our testers simulate voltage changes in our Voltage Bounce, Relative Spikes, and Voltage Cycling tests but you can never simulate all possible spikes.
   
   
7. Q. What other issues affect the accuracy and reliability of your testers?
   A. The RAMCHECK and SIMCHECK II test programs are fully written in Assembly for fast performance, and they include about hundred thousand lines of code. While we do our best to clean all the bugs prior to firmware release, there is no escape from such hidden bugs to be found in our code.
   
   The memory market is also huge, with many developers and manufacturers. There are new module variations introduced daily and our firmware may not support some new variations.
   
   We are also continuously improving the hardware design of our products, and offer our customers periodic Calibration and Hardware upgrades for their test equipment.
   
   
8. Q. Can operator's error interfere with test results?
   A. Although we have made the operation of our testers quite simple and automatic, the testers include advanced setup features to satisfy the most demanding users. However, some operators may set it up wrong or misinterpret the results. For example, if one operator sets the tester at a partial test of 16Mx64, another operator may be confused to see a 16Mx72 module tests at %16Mx64. This operator may not be aware that the "%" prefix indicate a setup for partial tests...
   
   
9. Q. How do you improve the accuracy of your tests?
   A. Our R&D department is constantly at work to improve our test accuracy. We are developing various algorithms and continue to improve the electronic design of our products.
   
   We also receive a lot of help from our customers. When they find a module that does not test properly, they often send it to us for evaluation. Many additional support featured have been added to our products over the years through such cooperation with our customers. If you find a module that you wish to send for our evaluation, please contact our Tech Support department.
   
   
10. Q. The test results, especially in the Test Log, provides numerous details. How do they relate to your overall test reliability?
    A. Our testers are not simply GO/NO GO testers. They provide numerous results that should be understood by the customer. For example, a customer may have a module passing our testers but the test results indicate that the module is of a special structure. Then the customer is confused to see the module failing on a certain motherboard. After contacting our Tech Support department and reviewing the test log, we have found that the RAMCHECK results indicates that the module will not work on this specific motherboard.
    
    We provide a lot of information on our Web site to allow better understanding of the test results. And of course, you are always welcome to contact our Tech Support for more information.

If you have additional questions, please feel free to e-mail us at support@innoventions.com. Please remember to include your phone and fax numbers.