Friday 12 June 2015

Overclocking: Risk Vs Reward

Overclocking

Should you go down this road or not? James Hunt looks at the pros and cons…

Whether you’ve bought new hardware that you want to push to the limits or you’re trying to squeeze extra performance out of your existing components, overclocking might just be the way to do it.

If you’re reading Micro Mart, it’s a safe bet you’ve at least heard of overclocking, but a surprisingly small percentage of people have ever tried it. That may be because of the associated risks – the idea that your hardware could be permanently damaged, but it might also be because the rewards aren’t quite obvious. Sure, your hardware will run faster, but how much faster? And at what cost?


Addressing And Minimising The Risks


The thing most people are concerned about when overclocking is that they’ll end up with a permanently damaged piece of hardware that needs to be replaced, wasting both money and time. But even though it’s possible to fry hardware by overclocking it, it’s also quite easy to avoid doing that as long as you’re careful.

The risk of overclocking damage is rooted in the chemical makeup of your components. CPUs, GPUs and RAM chips are made out of millions of transistors, which are themselves mostly constructed from semi-conductive silicon. The reason silicon is used is because its electrical resistance decreases when the temperature rises, which is the opposite of what happens in more common metals.

This means that if silicon gets hot enough, it becomes fully conductive, which results in an electrical short that can damage your hardware permanently. That’s why you have to keep it cool. Ideally, you want your processor (CPU or GPU) to be kept in the region of 70-80°C, but that’s only a rule of thumb. Some chips will have much higher tolerances, some a little lower, and you should be able to find a fairly reliable range for your hardware if you see how other overclockers got on with it.

In practical terms, overclocking involves raising voltage and multipliers, and the more you do that, the greater the likelihood of failure becomes. To mitigate the risks, start by making small increments in the multiplier and only step up the voltage when the chip becomes unstable because it’s not getting enough power (you can tell when this happens, because the chip will be unstable at lower temperatures). Remember that even if someone else claims to have pushed the same model of chip to 4GHz, you shouldn’t start there – every individual chip has different tolerances and if you go in too high, the damage might be instant.

A good way to check your system before (and after) you overclock is to run a stress test, which will tell you whether it’s running correctly under its current settings. To do this, all you have to do is run a benchmarking tool with your CPU’s default configuration so you have a baseline with which to compare future changes. AIDA64 (www.aida64.com) has a built-in Stability System Test, which you should run for a few hours to see how your system fares. If the hardware doesn’t crash, you know you’re safe to at least try overclocking – and if it does, you know you’re due a replacement!

The other thing that can negatively affect overclocked systems is, perhaps a little surprisingly, the PSU. If your power supply unit is cheap, old or otherwise inefficient, it might be providing a ‘dirty’ stream of electricity. A dirty supply gives an uneven voltage, which can spike and drop, leading to instability (or even damage!) in systems where the tolerance is tight – overclocked systems, for example. There’s no particularly easy way to test this without specialist equipment, but if you notice otherwise unexplained problems, your PSU might be the source of it. If you’re building a new system, try to buy a high-apacity, 80 Plus-rated PSU to ensure a clean supply of electricity.

CPUs


CPU overclocking is the most common kind. It’s quite simple to get to grips with but carries the most danger because processors are fragile, expensive and essential for system use. RAM can be cheaply replaced and graphics cards can be bypassed, at least temporarily, but if your CPU goes up in smoke, you basically have no choice but to buy another.

But overclocking a CPU is also a good way to improve a performance hike proportional to whatever the increase is, and with the right cooling, some chips can be made to work exceptionally well. Even now, the average overclockable Core i5 can be made to perform as well as the best Core i7s if you manage it correctly and the hardware tolerance is good enough. We’re not talking about small leaps here.

Gaming is the main area where you can see speed improvements, but any CPU-heavy tasks will benefit. Graphics rendering and video processing times will be cut as a result of the higher speeds, for example. As with all overclocks, the effect is most visible in area where the CPU speed is limiting your system’s performance, and those situations are rarer than you might think. If you want faster speeds in games, then it normally makes more sense to upgrade your SSD or GPU before you overclock your CPU.

Either way, the important thing when overclocking is that you keep the chip from getting too hot. That means that you’ll probably want to replace the OEM cooler and heatsink with a high-end version before you start. In any case, if you make sure the automatic temperature shut-off in your BIOS is enabled at a value about 5-10% lower than the temperature tolerance level for your CPU, you won’t have to worry about your hardware reaching it. If it gets too close, it’ll shut down and therefore be safe from damage.

RAM


Of all overclocking opportunities, RAM is probably the one most people don’t bother with in any specific way, largely because standard XMP profiles match the RAM to any overclocked CPU. But as with CPU and GPU overclocking you can get performance improvements by doing it manually, especially in gaming, where high frame-rates and huge data requirements mean RAM throughput gets taxed on almost any system. Part of what puts people off is that by default, RAM doesn’t come with its own cooling, so it’s tempting to imagine that it’s running at its high tolerance already. This isn’t really the case.

RAM itself is very low temperature even when overclocked. High-end modules might have heat-spreaders, but even they’re mostly just there to look cool. If you’re planning to overclock, you might want to look into an active cooling device, whether that’s a memory-specific fan or an exhaust fan placed above the slots on your motherboard, but as long as you don’t raise the voltages too, high there’s very little chance of damage to your RAM; it’s more likely that the rest of the system just won’t be able to keep up.

The important thing to realise about RAM overclocking is that even though your system architecture might be rated at 1600MHz, that doesn’t mean you’re restricted to that speed. Intel and AMD guarantee minimum RAM speeds, but higher RAM speeds are possible with most chips. As with all forms of overclocking, it’s a matter of tolerance. All chips are guaranteed to work at the minimum rated speed, but a significant number will fare no worse even with a much faster RAM clock speed. If you’re really unlucky your hardware might be from a batch that can’t go beyond 1600MHz, but it’s far more likely that you can push 2000MHz and beyond.

Indeed, some tests have shown that every Intel i7 4770K CPU can handle RAM at speeds of 2800MHz. If you were to buy modules rated at that tolerance and install adequate cooling, you shouldn’t see any system instability despite a 75% increase in performance. Lower-end chips are likely to have worse support, but even Core i3 CPUs are almost completely stable with RAM run at 2000MHz.

GPUs


Overclocking your GPU used to be a huge hassle, but as the popularity of it grew, manufacturers decided to embrace people’s enthusiasm. Now, overclocking your GPU is even easier than overclocking your RAM or CPU: all you have to do is load up your graphics card’s helper app and tweak the settings there. No reboot required!

Overclocking your GPU is a little riskier than overclocking your CPU, since the hardware operates on the same principles but doesn’t have the same protection from the motherboard’s temperature gauges. On the other hand, most GPUs come with two or even three active cooling fans, so at least you know that the temperature shouldn’t get out of control unless you let it.

As with CPU overclocks, you need to get a benchmark for your system before you do it (if only so you can make sure you’re actually getting better performance), so use a program like 3D Mark (www.3dmark.com) before changing any settings so you know what your hardware is currently capable of.

When overclocking your GPU, remember to keep an eye on your card’s temperature and look out for glitches and artifacts that suggest not everything is going well. An overclocked CPU will crash your system when you demand too much of it, but an overclocked GPU will just struggle to provide proper visuals well before it gets into the real danger zone. You might also see the benchmarking app crash, slow down or even start returning lower results as the card reaches its limits, so as soon as you get to that point you should start to slowly drop the core clock speed.

Note that overvolting can improve a card’s performance even further, but it’s also likely to trash them. Even a small ‘safe’ overvolt could cause instability, increased wear on components and higher GPU temperatures. Unless you’re desperate or prepared to kill your card, stick to stock voltages; in this case, the performance improvement you can get is vastly outweighed by the risk of trying to get it.

The improved performance of an overclocked GPU will probably only be visible in graphically intensive tasks (typically gaming or when viewing high-definition media), so if your benchmarks are stable, try playing your favourite came and see if you notice any difference. The speed increase will only be proportional to the speed at which you increased the clock, so don’t be surprised if it seems like a small amount. Many graphics cards are even sold with a  slight overclock, so you might not be able to push them as far as you initially think, because some of the legwork was done for you!

Hopefully that’s given you some idea of what to expect when you try to overclock your hardware, and what the risks of trying it are. If you’re planning to  overclock, remember that this isn’t an exhaustive guide to the procedure, just a general overview. If you want to go ahead with it, you’ll a more detailed guide. And it almost goes without saying that if you choose to overclock, then it’s with full knowledge that your system can be permanently damaged as a result. Try it at your own risk!

Do I Need To Overclock?


It’s a simple question, but in a way it’s really at the heart of this article. Most systems can be overclocked to some degree. But does every computer need to be overclocked? Clearly not. So how do you know whether you qualify or not?

The way to tell is to use benchmarking and loadmonitoring software to see whether or not your components, at their current clock levels, are actually holding you back. Some tasks – gaming or media encoding, for example – use the CPU at full load, and an overclocked chip will definitely improve the times on that. If you’re only doing word processing, then you could create the best, most high-end overclock configuration and you’d still see very little improvement in the way your system runs.

The same is true of graphics cards. Games tend to use the full availability of a GPU, but almost no other task does. One of the things to be careful about when you investigate overclocking is that just because it improves a benchmark, that doesn’t mean it’ll improve your day-to-day system usage. Benchmarks  represent the high capability of your system, and if you never reach that anyway, then an increased benchmark is practically meaningless.

As we’d said elsewhere, in most average systems the real bottleneck is the hard drive, so don’t think that overclocking your CPU will solve your slow-PC problems overnight. It’ll improve performance in specific tasks where a faster CPU, faster GPU or more responsive RAM matters, but remember, that’s not the same thing!