Determining and Testing Stability - A Beginners Guide

Add Your Comments

. How to see

Introduction

So, you’ve just built your sweet machine. If you’re planning on overclocking (and I’m assuming you are if you’re reading this guide), then I’m sure you’ve made a variety of great decisions regarding the hardware that you’ve chosen. If your experience was like mine, you poured over technical papers and the forums themselves looking to compare and contrast not only theoretical performance but real-world performance as well.

Finally, taking your fellow overclockers’ advice into consideration, you’ve chosen a good deal of (at least semi) expensive computer hardware. This will be your “Beast” or whatever you decide to call it when you finish assembling it. (The Beast is taken, by the way 😉 )

After (or perhaps during) the process of assembly, you’ll get the nervous jitters. “Did I pick the proper hardware?” “Am I getting the best deal for my money?” “Am I going to fry this part and flush some greenbacks down the toilet?” I assure you, the more nervous you are, the more mistakes you’ll make. Calm down. This guide will officially pick up after you’ve assembled your computer and are tweaking it for optimum performance.

The problem is – most overclocking guides tell you how to do it, but they don’t tell you what to do if something goes wrong, or how you tell if it’s even STABLE. This guide will attempt to tell you how to find if your hardware is running in stable, blissful harmony, or if it’s held together “by bubble-gum, some tape and a prayer.”

And now, on with the guide!

The Brain

If you picture your collection of computers as a football team, then your good ol’ CPU would be the quarterback. Sure, he gets a lot of glory and he’s foremost in the eye of the public, but he’s not the MOST important part of your team. We’ll cover him (or her, it doesn’t matter to me) first anyway. Whether you’ve got a Celeron 300A overclocked to 450 MHz, an Athlon 1.0 GHz AXIA at 1.4 GHz, or a Duron 600 MHz at 1.0 GHz, the same basic information applies.

I won’t get into cooling – a LOT of great information can be found both on the boards and in the other guides here. Heat is BAD for your processor. Depending on what kind of processor you have, you’ll have to worry about heat to different degrees (We won’t compare something like a Celeron to a Thunderbird). Suffice it to say, heat and lack of juice (or power) will make your CPU unstable. On the other hand, too much power just may fry your CPU!

Anyway, grab yourself the latest version of WCPUID. You can do a search on the ‘Net or you can probably find it on one of the forum member’s FTP servers (Pinky and Shadow come to mind currently). This will tell you a little bit about your chip, most notably your Front-Side Bus and your CPU Clock Speed.

It’s not a big deal now, because most computers boot up with it showing. However, in the future AMD might be switching to a PR system for determining their processor performance (Instead of a MHz/ GHz rating, they’ll show an “equivalent” rating compared to the latest Pentiums out). Bah I say. But what’s done will be done.

So now, if you couldn’t tell your CPU vitals before, you can tell now. Next we’ll move on to some stress-testing.

So your 1.0 GHz AXIA chip is clocked at 1.5 GHz? But is it stable? If it’s clocked that high and it crashes in Unreal Tournament after playing for 5 and a half minutes, then it’s not worth it, is it? (I hope you say no to that question) This is especially important to all our members on the SETI and Folding@Home teams!

Even if you think your system is ROCK-SOLID, you may be generating errors in the data you’re sending back. You may be raising team standings, but you’re NOT helping the effort. I’m sure most of you agree that sending proper data and being in second place is better than being the top-output team!

You’ll want to stress-test your CPU to make sure that it’s not making miscalculations. Before you move on, download the latest version of SiSoft’s Sandra. There are a plethora of benchmarking utilities contained within this lovely little package. Oh, and I’m sure you’ll want to go benchmarking your system to prove how big your…… hard drives are, so you can do that too.

Have you got Sandra now? Excellent. First, open it up. You can run the CPU and the CPU Multi-Media benchmarks and compare them against equivalent processors from their databases. You’ll probably want to alt-printscreen these and then open paint and save them somewhere. You’ll need these if you want to prove your massive overclocks, and heck, save them for posterity too. There’s nothing like squeezing a few extra megahertz out of your chip and then seeing what the difference makes in raw equation processing. 🙂

Benchmarking is all well-and-good, but the real reason that you want the Sandra package is for it’s Burn-In Wizard. Whenever I want to test my CPU, I open this lovely little program and set it to run the CPU and CPU Multi-Media benchmarks. Loop it for about 99 times. If you’ve completed these without some sort of blue screen or other problem, well then you might just be about halfway there!

Now it’s time to find another little stressing application. You can search the forums for a program entitled Burn6. There will probably be a new version circulating the forums by the time you’re using this guide, but for now this is what’s offered. It works excellently on Athlon Thunderbirds, but I’m not sure as to how well it performs on Pentium 4s.

It’s a pretty ugly little program. But who cares how pretty it is when it gets the job done? If you’ve got it running on some sort of NT based machine, open up your Task Manager and check out that CPU usage. It should be somewhere near 100%! Lovely. This program will also terminate if any errors are detected in the calculations (or at least it mentions this in the README file).

Great, so now you know if you’re sacrificing accuracy for speed (which is a BAD thing). If you can run this little gem of a program for 2 or 3 hours, well then you should have a pretty stable CPU.

Our SETI and Folding@Home members have (I believe) their own little methods of testing for miscalculations. I believe they’re some kind of test blocks that your computer will crunch and report results to you. After this is done, you compare them with the official results to make sure your processor is not making floating point errors. I’d appreciate more information on this from any of our SETI or Folding people, but until then you can ask in the forums.
{mospagebreak}

Bill Mason

The Memory

Grr, memory! So vital, yet such a pain in the rear. Due to Front-Side Bus increases, this can (many times) be the cause of an unstable overclock. While processors are sometimes able to be run 200 to 300 MHz out of spec, most RAM dislikes being run even 10 MHz over! Of course, if you went ahead and purchased some quality RAM, you’ve weighted the proverbial dice in your favor.

DDR appears to be much less overclockable than SDR, and I’m not sure about RAMBUS, though many claim that the setting of 100 MHz is very conservative and that todays will do 133 MHz at least. Bah I say. RAM is very picky.

So for stress-testing your most finicky piece of hardware, you’re going to again make use of the Burn-In Wizard provided by Sandra. I simply run the Memory benchmark in a loop about 30 times – 50 if I have the time. Sometimes it will simply generate an error and close, and sometimes it will reboot your computer for you if your RAM is unstable. But in any case, you’ll know it’s not. Usually if your RAM survives 30 loops, it’ll survive 99 (Test if you like to make sure) This is the pickiest RAM test I’ve seen and run to date.

Oh, you can also benchmark your RAM to see how your bandwidth fares with other reference chipsets.

If anyone has any other recommendations for testing RAM stability, please drop me a line.

Power Supply

Oftentimes the culprit behind many “mysterious” overclocking instability is your power supply! If you have to get a Duron 750 MHz instead of the Duron 1 GHz simply to get a respectable power supply, DO IT. Today’s processors need around a 250W power supply (before you count overclocking – because you’ll want the option of raising the voltage to your components) to function properly. Power-hungry Athlon users will want to get at least a 300W. Myself, today I would go with a 400W. I’ve gone with a 400W Antec and am very pleased with it so far.

It seems to be a good deal harder to tell if the instability is a result of your power supply. I’d recommend finding the latest version of Motherboard Monitor (I believe MotherBoard Monitor 5 – MBM5 – is out currently). This will allow you to monitor temperatures, fan speeds, voltage ratings, and even your CPU speed.

With regards to your power supply, you’ll want to watch your voltage ratings CAREFULLY. Most notable among these is the 5V line. Currently it appears to be the largest indicator of a weak supply. If your computer is idle, and your 5V line is below 4.9V, then you’re going to have a problem.

If you want to verify this problem, keep Motherboard Monitor up and run Burn6, or another heavy CPU-testing program. With a weaker power supply and at 100% processor usage, you’ll see the 5V line dip down-down-down until your system becomes unstable and either crashes or reboots.

Video Card Instability

Afraid your video card is causing instability? This could be possible for a variety of reasons. First, if your Front-Side Bus is overclocked, then your AGP or PCI slots are going to be running out of spec. PCI and AGP slots run on some sort of divisor system.

For example, at a Front-Side Bus of 133 MHz, PCI slots use a 1/4 divisor. 133 * 1/4 = ~33 MHz. This is where they want to run. At any rate, some accessories are more picky than others. Graphics cards are generally pretty good about running in a slightly out-of-spec AGP slot.

If you’d like to test and make sure – download 3DMark2000 and 3DMark2001 for starters. Running these will give you a nice little benchmark for you to compare against other people, but more important, it will tell you if your graphics card MADE IT THROUGH the tests without crashing the system.

This is important: Looping this for 10 or 20 times should be sufficient, but if you’re not sure – go ahead and loop it 50 times. It takes a long time, but it’s better to be safe than sorry.

These are also great at telling you if your card itself is running too high. Most video cards offer options of raising the Memory and Core of the on-board components. This can cause instability as well. These applications will be able to tell you if you’re pushing too hard.

General Instability – The FUN Part!

Now it’s time for the best part. If you’ve done all the above and you’re pretty sure that you’re in the clear with regards to those tests, you can begin the fun testing. Simply take your favorite game and play it for 2 or 3 hours.

It sounds kind of crude, doesn’t it? No matter, it’s fun and it’ll test your components in real world situations as well! If you experience problems that are out of the ordinary (a lot of games ARE buggy these days) then you can test the suspect components once again.

In addition, this’ll be a good way of determining stability of your various other components, like your sound card or ethernet cards – as they’ll be working on the out-of-spec PCI slots.

Conclusion

In any case, I hope the information above has helped you. If you have any comments or suggestions for some sort of updates please let me know. Use something like “Stability Guide” for your subject, if you would. Thanks, and happy, stable overclocking!

APPENDIX

General Processor Solutions:

  • Improve cooling
  • Raise voltage
  • Lower Clock Speed

General Memory Solutions:

  • Improve cooling (Ram sinks controversial)
  • Raise voltage
  • Lower Front-Side Bus

Power Supply Solutions:

  • Name-Brand Power Supply
  • Large Power Supply
  • Lower voltage/speeds of components

General Solutions:

  • Remove all but necessary components – replace one at a time until problem occurs again
  • Switch PCI slots, some components are strange (esp. SBLive cards)
  • Lower Front-Side Bus (Makes your PCI and AGP out of spec)

Again I hope this has helped. If you have comments and suggestions, drop me a line. Thank you!

Bill Mason

Leave a Reply

Your email address will not be published. Required fields are marked *