The SOS Guide To Keeping Your PC Cool, Quiet, and Stable

PC Musician

Technique : PC Musician



Even a few degrees of extra heat inside your computer could damage your CPU and shorten the life of other components. Fortunately, there are steps you can take to bring down your PC's temperature while still ensuring the quiet operation you need in the studio.

Martin Walker


If you let your expensive CPU (Central Processing Unit) overheat, did you know that your PC will shut down or crash, possibly at an inopportune moment? Even worse, are you aware that letting this happen can shorten the CPU's life, permanently damage it, or burn it out completely? High-street PCs avoid this scenario by fitting lots of noisy fans, but this isn't a welcome option for musicians. Specialist music retailers, who can't rely on fans to the same degree, have to tread a fine dividing line between operating temperature, noise, and reliability, and they use software stress testing to determine how reliable their machines are likely to be.



So if you've just built yourself a new PC, have upgraded an existing one by fitting quieter CPU or case fans, or are simply concerned that your shop-bought model is too noisy or too hot, it's important to do the same tests on your machine that the music PC specialists do. Running electronic components at lower temperatures can increase their longevity by a considerable margin, and once you know how to measure the temperatures of the various components inside your PC, how to stress them for testing purposes, and what's regarded as safe, you could end up with a far more reliable computer that runs as quietly as possible. Let's find out how.

Top of page

Measuring Temperatures



The first requirement is to discover how hot your PC gets. With most machines it's easy to find out via the BIOS (Basic Input Output Subsystem), which you can usually enter by pressing the Del or F1 key during boot-up. The majority of PCs provide a 'Hardware Monitor' page that displays two temperatures — one for the CPU and one for the motherboard — and occasionally a third for the PSU if it includes a suitable temperature sensor (most don't). This page should always be your first port of call after fitting a new processor or heatsink/fan combination, since any problem will soon show up as a rapid rise in temperature. If the CPU reaches 50 degrees Centigrade within a couple of minutes, something is almost certainly wrong and you should power-down and check.

PCmusician-2

Using a dedicated utility (such as the Asus one shown here) bundled with your motherboard is the easiest approach to temperature monitoring. Here you can see the effect of continuously running the CPU at 100 percent of its processing capability. The temperature rises from about 30 degrees Centigrade on the left of the top graph to the high 40s on the right after just a few minutes.



If you've somehow managed to plug the CPU cooling fan cable in the wrong way around it won't spin up at all and the cable should be reversed (although most connectors are keyed to prevent this), or you may have plugged it into the wrong motherboard connector, or simply have forgotten to plug it in at all. However, if the fan's spinning normally when you next boot, the most likely cause of excessive temperature is bad contact between the CPU and its heatsink (see 'Replacing a Heatsink or CPU' box) or a misguided attempt at overclocking.



The BIOS temperature readouts often monitor the output of a thermal diode incorporated within the core of many modern processors, which theoretically allows them to be quite accurate. However, these diodes are sometimes not that close to the main heat sources within the CPU, while some motherboards may abandon them in favour of a temperature sensor, which may not be in particularly good contact, mounted underneath the CPU. Combined with the fact that the external temperature-measuring circuitry of many motherboards often under-reads the absolute value, the result of all this is that the actual CPU temperature will very likely be higher than the reported value by a constant amount ranging from a few degrees to as many as 20 degrees. The motherboard's BIOS usually adds a suitable correction value to compensate for this, so under normal operating conditions the temperatures as displayed by the BIOS are accurate enough to measure idling temperatures.



However, while running Windows applications we can't access the BIOS to measure system temperatures, and must therefore run a suitable Windows utility that reads and displays the same motherboard information. There are lots of freeware and shareware utilities that can do this for you, but fortunately many motherboard manufacturers make your life easier by bundling a suitable utility on their support CD-ROMs (Asus, for instance, include their Probe utility), which you can generally trust to incorporate sensible temperature compensation values if necessary.



If your motherboard doesn't include a suitable utility, the two most popular generic temperature-monitoring utilities are the shareware Hmonitor (www.hmonitor.com) and Motherboard Monitor (http://mbm.livewiredev.com). However, when using any generic utility you may well have to establish a correction value for yourself, following its instructions. In most cases this simply means loading in a preset for your particular motherboard, or looking up the correction value from a table.



Don't be worried if your Windows idling value ends up measuring a few degrees lower than the BIOS figure — Windows NT, 2000 and XP run a system idle process using the processor's HLT (Halt) instruction when it's otherwise unoccupied. This cools it down, and on my machine I've noticed a slow drop in temperature of up to five degrees after switching to any Windows-based utility after spending some time monitoring temperatures in the BIOS. If, for any reason, you have to manually calculate a correction value, for a quick and dirty calibration you can normally monitor the CPU idling temperature in the BIOS, load Windows and immediately launch your temperature-monitoring utility, find out how many degrees lower the CPU temperature is, and use the difference as a correction value.

Top of page

Other Sources Of Heat



I'm mainly concentrating, in this feature, on CPU temperature, since the CPU is the most expensive and at the same time vulnerable component, and the most likely to get significantly hotter under load. However, there are various other heat-producing components inside the average PC that you should occasionally monitor. This is particularly true when you're first trying out a new DIY machine, or after installing new expansion cards, a new power supply, more RAM, a new processor and/or associated heatsink and fan. Temperature checking is also advisable after any major change in cooling arrangements, such as fitting a new case fan or changing the speed of any existing fans.

PCmusician-3

For a one-stop guide to the maximum allowable case temperatures of almost every CPU ever released, as well as their power dissipations, pay a visit to Chris Hare's excellent web site of Processor Electrical Specifications (see main text for link).



The motherboard itself has to supply power to all the expansion cards, as well as its own circuitry, and you may find several heatsinks on it, and even a cooling fan on some models. You can monitor its running temperature using the BIOS or a Windows utility, and it normally runs considerably cooler than the CPU. Other individual components without integral temperature readouts can be monitored using some sort of thermometer (an electronic one with an external probe would be ideal if you have one) but for rough and ready measurements here's a handy finger-temperature guide, courtesy of Peter Cyriax (www.pcysys.co.uk), who regularly contributes to the SOS PC Music Forums:

40°C Pleasantly warm 50°C Hot 60°C Painful 70°C Burnt fingers



As always, take care when poking about with your fingers inside any computer. The highest voltages you'll normally encounter outside the computer's PSU are +/-12 volts, which can't give you an electric shock, but you should make sure you earth yourself before making contact with any heatsink, by touching the metalwork of the case. Doing so will dissipate any static that might otherwise destroy delicate electronic circuitry. Also, avoid touching any components other than metal heatsinks unless you really know what you're doing.



Bearing the above in mind, it is worth opening up your PC and using your fingers to find out how hot various components get after heavy use. Check heatsinks on the motherboard, graphics card, any audio DSP cards, hard drives, and in some cases even RAM (some of the fastest chips now require heatsinks of their own!). Inside my PC neither of my Silentdrive-encased hard drives exceeded a safe 50 degrees using the finger test, but the heatsink of an elderly Matrox G450 graphics card ran at about 60 degrees until I installed a rear case fan.



A sure sign of a component that needs to dissipate a lot of heat is one with a cooling fan already mounted on it. Many modern AGP graphics cards have such fans, as they can no longer be cooled passively using a heatsink alone. Unless you anticipate running games or heavy-duty graphics applications, as a musician you should be quite happy with the performance of cheaper and slower graphics card that don't require a potentially noisy fan. Alternatively, you can now buy specialised heatsinks, from companies such as Quiet PC (www.quietpc.com), to replace the fans.

Top of page

Replacing A Heatsink Or CPU

PCmusician-1

There are now many cooling products and solutions available for PC owners, ranging from fans, heatsinks and controllers to thermal pastes, ducts, brackets and mountings, all available in the UK from suppliers such as this one. However, without a systematic approach like that outlined in this feature, you may not achieve the best compromise between low noise and long-term reliability for your PC.

For optimum transfer of heat from the CPU to its heatsink, you need good thermal contact between the two using some sort of TIM (Thermal Interface Material), and there are two main types: thermal grease (aka heatsink compound or thermal adhesive) or a thin pad (aka Thermal Interface Pad) that's often already attached to the heatsink when you buy it. Both provide a thin, squashy layer with high thermal conductivity, which fills the inevitable gaps when mating two otherwise flat surfaces, so that the heat from the CPU is efficiently transferred to the heatsink.

The pads are the easiest solution, but are only really suitable for single use — if you want to upgrade your processor and re-use the heatsink, you'll probably have to remove the pad and either replace it with another or use thermal grease instead. You'll need to clean both the heatsink and CPU to remove all traces of the pad, either with a plastic scraper or a solvent such as nail polish remover (which contains Acetone). Detailed instructions on how to do this can be found on the Internet (I came across a good example at www.overclockers.com under the title 'How To Remove the TIM from a Stock AMD Heatsink and CPU').

The stock pads fitted by manufacturers to bog-standard heatsinks are generally of the inexpensive grey/silver graphite variety with fairly indifferent performance, but you can buy versions made from better-quality material. However, although thermal grease is messier to apply, it provides better results than a pad if you do it properly, and most specialist heatsinks are shipped with a plastic syringe containing a small amount of grease, made from silicon and zinc oxide. This is fine for most applications, but if you want to get the best results it's worth investing in a specialist thermal grease such as Arctic Silver (www.arcticsilver.com) which, as its name suggests, contains silver particles that result in a lower 'thermal resistance' (ie. they conduct the heat from CPU to heatsink more efficiently). Available from lots of PC component suppliers at under £7 a tube, and loved by overclockers everywhere, such products can sometimes result in a processor running at a 10-degree lower temperature with a particular heatsink.

You only need a tiny amount of grease to do the job (generally similar in size to a grain of rice), smeared out to a thin layer using something flexible such as a credit card — the idea is to have just enough material to fill any voids between the two mating surfaces. Follow carefully any instructions provided by the heatsink and/or grease manufacturer when applying the grease and fitting the heatsink and, finally, don't worry if your newly fitted heatsink feels hotter than its predecessor. This means that your CPU is actually running cooler, and that its heat is being transferred to the heatsink more efficiently.

Top of page

Safe CPU Temperatures



So what's a safe maximum temperature for your CPU? Well, for the official maximum operating temperature recommended by the manufacturer for your particular processor model you can visit the AMD or Intel web site and download the appropriate technical datasheet. However, it's much easier to visit Chris Hare's excellent page at http://users.erols.com/chare/elec.htm, where he's posted the specs of just about every CPU ever released, from all the major manufacturers (see above). Figures available here include Thermal Design Power (the maximum number of watts that need dissipating when your processor is working hard) and Maximum Case Temperature.

PCmusician-4

Passmark's Burnin Test lets you run multiple tests to stress different aspects of your PC simultaneously. Here I'm running the CPU really hard, as well as testing my RAM, both hard drives, graphics card and soundcard.



Typically, the maximum temperature of the CPU's metal casing is specified at somewhere between 70 and 95 degrees Centigrade. If the temperature of your case ever exceeds this, your PC may randomly crash (particularly in the case of many AMD models) until it cools down again. Most Intel processors will automatically switch themselves to a slower mode until the case cools. Many users might not even notice this, but musicians would instantly spot it as they would suddenly be unable to restart the song that, moments before, was running perfectly at about 90 percent CPU overhead.



Allowing your processor regularly to reach high temperatures will shorten its life, or even permanently damage it. One rule of thumb that's widely quoted is that each 10-degree temperature rise will halve the life of the processor. The sensible answer is to make sure your cooling system is sufficient to prevent the processor ever reaching such temperatures. This is fairly easy to do with a general-purpose PC. As Intel themselves state, "Analysis indicates that real applications are unlikely to consume maximum power dissipation for sustained periods of time." Unfortunately, this is exactly what most musicians are aiming for when trying to run as many plug-ins and soft synths as possible, which is why we need to run stress tests to find out just how hot everything can potentially get.



So what's a good maximum CPU temperature to aim for? Opinions vary, but I personally think the hottest peak that you should allow is 60 degrees Centigrade, which, unless you are using temperature-controlled fans (see later) will probably mean that your CPU will idle somewhere between 30 and 40 degrees. I prefer to aim for a peak in the mid 50s and around the low 30s idling. If you're really concerned about CPU longevity, a maximum temperature not exceeding the high 40s should be the ultimate goal. The ambient (room) temperature is also part of the equation, so your PC could end up at a different maximum temperature in winter than in summer. Most commercial systems available worldwide should guarantee safe operation for external ambient temperatures of up to 35 degrees Centigrade, although in the UK I somehow doubt that this top limit will ever be reached! Winter DIY PC builders should bear in mind that their peak CPU temperature will rise slightly in summer, while summer DIY builders already face the worst scenario. Anyone overclocking their CPU should also be aware that overclocking generates a lot more heat and potentially reduces the processor's life significantly.

Top of page

Chill Skills: Cooling Hints & Tips

Most fan-cooled systems rely on cool air being sucked in at the bottom of the front panel and warm air being exhausted from the PSU and any case fans. So, if you're unsure which way around your fans should be mounted, a front or side chassis fan should suck, the rear one should blow, and the CPU one should blow air down onto the processor.

If you're thinking of updating or upgrading your PC's cooling system, a useful rule of thumb is that larger fans operating at lower speeds will move the same amount of air as smaller ones at higher speeds that make significantly more noise. If you want to try some larger fans, you can buy adaptors to attach an 80mm fan to a 60mm fitting, or an 120mm one to a 80mm fitting. Among other suppliers, Kustom PCs (www.kustompcs.co.uk) sell these in the UK for about £5.

Dust and accumulated muck on your cooling fans, air filters and heatsinks reduces airflow and will eventually cause temperatures to rise unnecessarily, so don't forget to clean them occasionally (about once a year is normally sufficient unless you smoke or burn open fires in the vicinity). Some people have even opened up their PCs to find the CPU heatsink completely solid with dirt between its fins. While you're at it you can give your computer keyboard a clean out as well. After powering down the PC you can use a soft brush to dislodge the dirt (some people use those brushes with a squeezy air bulb sold in photographic shops). Many people rely on a few squirts from a can of Airduster (compressed air) to blow the muck away, although this may result in moisture inside your PC, so be careful. A vacuum cleaner may be used to suck the dust out if you're careful, although beware of static. You can either use a standard household model with an attachment; one of those mini battery-powered vacuum cleaners sold specifically for computer, video and photographic maintenance; or a slightly more powerful version sold for in-car use.

You can buy rounded cables for IDE hard drives, floppy drives, and so on, which may improve airflow compared with the standard ribbon cables and thus reduce internal temperatures slightly. Although a few people were initially concerned about possible performance degradation due to their increased crosstalk (and they certainly won't improve performance), the only important guideline is not to buy any that exceed the recommended length of 18 inches for hard drives, as this can definitely result in reliability problems.

Take care when dressing internal cables to keep them in a tidy loom as far away from the motherboard, and particularly the CPU, as possible. Cables that are too long should have their excess folded up and tied with a cable tie to make them shorter. Not only will such measures allow plenty of airflow to the processor, to help keep it cool, but will also help to reduce EMI (Electro-Magnetic Interference).

The BIOS Hardware Monitor page may display fan speeds as well as temperature information. If you've spent a long time perfecting your cooling regime, note down the final fan speeds, as this will enable you to duplicate the results in a clone PC. This is what the professionals do.

Top of page

Stress Testing



As anyone who read our recent Manufacturer's Round Table (SOS September 2004) will already know, specialist music retailers put each PC music system through stress or burn-in testing for a period of between eight and 48 hours before shipping the unit to the customer. This means that any weak components in each custom-built machine will be weeded out before it's shipped, which is considerably better than finding them a few months after you've bought a PC.



Whether you've just bought a new 'high-street' PC, just finished building one yourself, or indeed just bought a specialist music PC, it's well worth doing some stress testing for yourself. You could simply run your choice of music software with a particularly complex song that pushes the CPU to its limit, but it's easier to use dedicated software that continuously runs various components at up to 100 percent of their capabilities for hours at a time, while you monitor their temperatures.



The most popular such utility for stress testing is Prime95 (www.mersenne.org/prime.htm). This freeware application was primarily written to search for new prime numbers, but since that pursuit involves loads of mathematical calculations the software can occupy your CPU at 100 percent capacity for as long as you run it. However, since it runs at a low priority it will let you carry on performing normal tasks, only using otherwise idle CPU cycles.

PCmusician-5

Products such as the Zalman Fanmate, for fine-controlling fan speeds, noiseless resistor cables to simply drop them to a fixed but significantly lower speed, and a Multi Fan Speed Controller to tweak them all from any convenient drive bay, are all available from suppliers such as Quietpc.com.



Select the 'Torture Test' from the list of Options, click on the OK button to run the default Blend test, and then leave the software running for several hours (or days if you like), while monitoring the CPU and motherboard temperatures. You don't need to be present to spot any problems either, since Prime95 writes its results to a text file (results.txt), and will let you know if any hardware problems occurred. When you're happy that your CPU is staying within safe limits, use the Stop option in the Test menu to stop the calculations, then exit the application.



It's also worth specifically testing the RAM on a new PC, or after upgrading your memory, to discover any intermittent faults that may not have been picked up by the manufacturer. Such faults can cause random crashes or spontaneous reboots, but a good thrashing with a specialised utility will eventually weed them out. The two most popular ones are Chris Brady's donationware memtest86 (www.memtest86.com) and Michal Tulacek's shareware Gold Memory (www.goldmemory.cz). The latter costs just $29 after the initial 30-day trial period.



Both utilities can boot from floppy or CD-ROM, so they can test all of your RAM rather than only what DOS or Windows isn't already using, and they use a variety of test methods to expose weaknesses. Details of all tests are provided in the documentation, but most people will find it enough to just create the floppy disk or CD-ROM, boot their PC from it and leave the default test sequence running for several hours, or preferably overnight. Any memory errors will still be displayed on the screen when you return to it.



I find that Prime95 generally pushes my CPU to its highest temperature, but for an all-round workout that stresses your PC in a way more similar to a MIDI + Audio application, a utility such as Passmark's BurnIn Test (www.passmark.com; see screen opposite) is ideal. I first mentioned version 2.2 in SOS November 2001, and the software is still going strong, now at version 4. You can select which hardware items to include in your tests; I generally disable the CD-RW/DVD and floppy drive tests, add my audio data and sample partitions to the default drive list, and (after a trial run to determine the current processor overhead as a percentage) increase the 'duty cycle' for the CPU tests so that it's just reaching 100 percent. This ensures that the other tests are not starved of CPU power but still runs the processor hard. Any errors can be logged, and Passmark even provide a downloadable white paper on using BurnIn Test in a production-line environment. In addition, they explain how to include a number of reboots into your test cycle, to check system stability on startup, how to place the PC automatically into sleep or hibernation modes, and how to write script files that run a certain sequence of tests in order — ideal to provide a thorough automated workout for anyone who regularly builds PCs.

Top of page

Stabilising



Whichever stress utility you're running, it will probably take half an hour or less for your CPU to reach a maximum stable temperature (hopefully of 60 degrees or less). If the stable temperature is higher than this, or if your processor hasn't reached a stable temperature and is still rising beyond 70 degrees, you really need to beef up your cooling arrangements, either by increasing the speed of the CPU fan if you have a Fanmate controller (about £5 from component suppliers, but bundled with Zalman heatsinks), speeding up the existing case fans, or adding an extra one.



If the stable temperature is less than 50 degrees, you can probably turn down the CPU fan's speed a little and make your PC a bit quieter. The safest approach with a DIY PC is to initially leave the Fanmate controller at full speed until you measure the highest stable CPU temperature under stress conditions, and then slowly reduce fan speed a little at a time, leaving the system for at least 10 minutes each time you change speed, until it reaches a new stable temperature. The aim is to determine the lowest (and therefore quietest) speed that will never let your CPU temperature rise beyond 60 degrees.



If you have case fans running, leaving the CPU fan at a higher speed and slowing the case fans down instead may result in even lower overall system noise (since the CPU fan is buried within the computer while the case fans are mounted on the outside). You can buy an extra Fanmate controller or simpler speed-reduction cable incorporating a dropper resistor to do this (cables cost around £2 from specialist suppliers such as www.kustompcs.co.uk, www.overclockers.co.uk and www.quietpc.com). Mind you, it may result in your motherboard and expansion cards running hotter, so use some common sense and try to keep everything at a reasonable temperature.

Top of page

Temperature Control



The ultimate approach to fan cooling is to monitor temperature and adjust the speed of fans automatically in real time. When the PC is idling (after all, this is when you notice fan noise most) the fans then just tick over quietly, but when the CPU is going flat-out running your latest musical masterpiece the temperature sensors pick up the increase in overall system temperature and speed up the fan rotation to keep it in check. Since musicians are listening to the music at this point, the increased fan noise is generally less obtrusive.

pc musician widget.s

Zalman's Fanmate fan-speed controller in a little more detail.



You don't need to be a rocket scientist to automatically control fan speed. Some PSUs (including the SilenX range) already incorporate temperature-controlled fans that do it for you. If this isn't made obvious, the usual giveaway in the spec is an exceptionally low noise level at switch-on and a higher one at full load when the fans speed up. It's also possible to buy case fans with temperature sensors from manufacturers including Acousti Products, SilenX, Vantec and Zalman, at between £10 and £15 each. These either have a temperature sensor built-in or on a flying lead so that you can place it where you wish. The secret with the latter is to put the sensor somewhere close to the CPU, so that the case fan's speed becomes directly linked to CPU temperature.



One knock-on effect of using temperature-controlled fans is that the system's idling temperature will be higher than when using fixed-speed fans. For example, when I recently reviewed Inta Audio's Opteron-based PC, which relies totally on its temperature-controlled PSU fan for system cooling, its CPU idled at a seemingly high 46 degrees, but only rose to a safe level of 59 degrees when fully stressed.


You can also buy tiny fan controllers, such as Cobalt 3's Pyramid V, for about £15 (www.quietpc.com is one UK supplier of these), that you mount using a self-adhesive fixing somewhere near the PSU, and to which you can attach up to eight case fans to temperature-control all their speeds simultaneously. UK specialist music retailer Philip Rees (www.philrees.co.uk) has designed similar proprietary central airflow control circuitry, and his tower case systems use this to control seven fans for an 'astonishingly quiet computer system' (I'll be reviewing one of these shortly). You can now even buy a wide range of drive-bay fan controllers with rotary knobs for speed control of each system fan (for instance, Zalman's ZM-MFC1 Multi Fan Speed Controller is about £20 and can control the speeds of up to six fans). Personally, I think this is going over the top for all but the most inveterate tweaker. After all, the PC musician ideally wants a totally reliable system that's as cool yet quiet as possible, but without any more flashing lights and things to tweak — we have far too many distractions from our music making already!  

Published in SOS December 2004