Cooling Options Information

Performance or Noise? Generally speaking whatever cooling choice you make there is a trade off between noise and performance.  Inevitably for a given size of heatsink inside the machine air needs to move over it that is of a lower temperature in order to cool it.  In the simplest terms this means just locating the heatsink in a cool place in the case and allowing convection currents (naturally swirling currents of air warming and cooling) inside the case to waft over it, known as Passive cooling.  However, as heat build up has become higher and the density of components that produce heat higher inside the case this has resulted in the need to introduce Active cooling in the form of fans blowing air through the case.  We can combat this with clever case design, careful cable layout (so as to promote air movement and not restrict it) and using bigger and more efficient heat sink designs.  On top of that we can reduce fan noise by making them bigger which means we can reduce rpm while still moving the same volume of air, and, by clever fan blade design that is more efficient causes less drag and turbulence and therefore generates less noise.

Our noise target is for all our systems to be under 30Db under all loads, and while idle or doing little to be around 20dB.  A high street system typically emits well over 30dB and an office air conditioning system can be as loud as 40dB or more, so as you can see our systems are very, very quiet.  Here is a guide to reference how loud different day-to-day noises are:

• The quietest sound you can hear: 0 dB
• Whisper: 15 – 25 dB
• Quiet background: about 35 dB
• Normal home or office background: 40 – 60 dB
• Normal speaking voice: 65 – 70 dB
• Lawn mower: 90 dB
• Chainsaw: 100 dB
• Live rock band: 120 – 150 dB
• Jet aircraft taking off: 140 – 180 dB

Below is a summary of our cooling systems:

• Air Cooling - Our standard high performance air cooling uses large heatsinks and quiet medium flow fans.
  We also offer the option of Super Silent cooling which uses much more expensive fans with specialised bearings, blade design and fairings to actually match or even surpass the air flow rate of our standard fans but generate less noise.  We also
  
• Water Cooling – Can be the ultimate in quiet cooling or a step-up from air cooling in performance depending on what you want to achieve from your Cryo extreme performance computer.  When optimised for silence we use either totally passive radiators, i.e. no fans whatsoever, or very low flow rate fans that generate now audible noise beyond 1m distance.  Alternatively we can optimise your system for maximum performance (as we do as standard for the Cryo Velox) then we will use a large active radiator, large diameter half inch hose and a high pressure water pump.  This can be noisier than our air cooling solution as the rapid air movement through a large radiator is quite audible, though not offensive.
  
• Hybrid Thermoelectric / Water cooling (TEC) - thermoelectrics make use of a property of semiconductors where when a current is passed across the gate it causes heat to be exchanged from one side of the plate to the other.  In effect the thermoelectric plate acts like a low power refridgeration unit exchanging heat from one side to the other (The Peltier effect).  However, water cooling is still required as the heat moved from the hot side of the plate to the cool then needs to have water passed over it and the heat lost through the radiator.  We offer TEC cooling as an add on to the high flow water cooling system on machines such as the Velox, or, as a stand alone option on other units.  Using TEC cooling you usually gain more processing power as the processor is cooled below ambient temperature, reduce noise emission (with the standalone units) but increase power draw on the power supply (TEC’s aren’t very efficient).
  
• Phase change cooling - If phase change sounds like some alien science fiction to you then i ask you to give us one minute for an attempt at an explanation in one paragraph without sending you to sleep.  Firstly semi conductors and electrics in general work much better the cooler they are as the atoms vibrate less and move much closer together tightly packed and ‘lined up’.  In fact, at close to absolute zero (-273C, as discovered by the great scot and pioneer in thermodynamics William Kelvin) resistance in the conductor is reduced to zero and they are known as superconductors.  There are also some other effects of even temporary super cooling that we take advantage of with other permanent cooling techniques.  For our purposes thereduction in resistance has a virtuous cycle in performance boosting as it causes less heat to be generated by increased voltage and increased CPU speed while at the same time less heat builds up allow us to boost still further.  Phase change will typically hold your CPU down at around -30C to -50C while in use boosting overall performance by as much as 70-80%, that’s like owning a computer today that wont exist for anyone else until probably 2012.  Take a look at the latest results from Intel on a Cascade phase change system.  Phase change systems are not trivial to setup and can easily kill your computer with the condensation that forms around them.  Definitely for the extreme overclocking pro only!
  
• CO² and Liquid Nitrogen (LN2) - We’ve put this in for completeness but for anyone not in the liquid gas business or some related field its just a no go area.  There are a lot of risks attached to handling liquid gasses due to the extremely low temperatures (-196C), the damage it can cause, and its potentially explosive properties if it evaporates quickly in a confined space.  To say nothing of how difficult it is to get it onto your motherboard and components and refill it as it evaporates.  Definitely a choice for the pure benchmark record hunters!  Amazing results are achievable with extreme cooling like this record breaking 5.8GHz from an Intel QX9650 Quad Core Extreme.  By the time I’ve finished typing this someone else will have broken the world record so if your bored and want to see what crazy things these guys are doing with CPU’s browse the overclocking world records database to see where its got to.
  

How much memory do I need? Different applications need more or less memory.  Photographers and video editors have always needed a lot of memory to enable them to maximise how much work they could keep in memory at any one time.  The more that can stay in memory the faster you can work.  Ever wondered why the hard drive light was flickering while you were editing your work or running your game, or why everything just seems to hang for a few seconds then come back to life.  That’s probably you running out of memory.  If the computer runs out of real memory then it uses virtual memory, i.e. memory on your hard disk, and that is a lot slower and takes time for the CPU to get at as mechanical hard drives are an awful lot slower than memory chips.  In the mean time what is currently in memory has to be swapped back out to disk, a slow process, hence the ‘hanging’.  DDR2 memory is now very cheap and very fast buying as much as you can afford up to 8GB (the current practical limit).

• 3GB+ memory support - Like the Millennium bug there is a hangover in PC design from the early days that means the effective maximum memory limit is 4GB (or 4,294,967,295 bytes to be exact!), this corresponds to the maximum number available in 32-bit binary representation (i.e. 2^32).  All programming code and CPU internal design has fundamentally been based on a 32 bit width interpretation of numbers.  So why can we see only 3GB?  Well you can see slightly more than that, the practical limit is about 3.3GB depending on your graphics card.  Your CPU can’t address the full 4GB AND also get to another 768MB or more in your graphics card so what it has to do is overlay the graphics memory over physical memory, so you effectively lose the memory you have by the amount in your graphics card (plus a little bit more for other reasons we wont go into).  The problem is obviously solved in 64-bit operating systems like Windows XP and Windows Vista 64bit versions.  Read the notes on operating systems though for the downsides!
  
• Memory speed and Low latency memory – Memory has two factors effecting its speed, the clock speed it can run at and the preset delay timings required for it to read and write data.  To boost its performance we need to tune both they cannot be adjusted in isolation if you want to achieve the best results.  Better quality memory sometimes runs at a lower frequency but also lower latency (lower delays!) and hence in practice runs faster!  As standard we use a balance of high bandwidth low latency memory trading that off against the steep rise in cost as you try to achieve both goals.  If you select the appropriate option we will use the best memory available with exceedingly rare properties of both very low latency and ultra high frequency.  It’s rarity in the manufacturing process makes it a lot more expensive (sometimes two or three times the price) but it gives your processor Wings!
  
• DDR2 & DDR3 memory - Dual Data Rate memory has been around for sometime and its evolution from DDR to DDR2 and now DDR3 in motherboard chipsets has pushed bandwidth up and up to were we now have 2GHz memory available due to the doubling in size of the prefetch buffer (from DDR2 of 4 bits, to 8 bits).  So what you ask does PC3-12800 1600MHz C7 mean??  Its simpler than it looks.  PC2 is DDR2, PC3 is DDR3, 1600MHz is the raw data transfer speed (subject to latency as discussed) so that’s 1600 million transfers a second!  Probably the most important numbers to us (us being overclockers) is the 12800M/bit peak transfer rate (bandwidth) and CAS latency of 7.  CAS?? So now we are really getting down into the weeds of memory technology.  CAS is the Column Address Strobe and you will remember from the bullet above that memory requires a delay timing.  The Column Address Strobe is probably the most important one of these that in simple terms tells the memory controller how long to wait before it can come back to the memory modules for another read request, clearly a lower CAS number gives us faster memory.  So now you can see that low latency DDR2 could be faster in real terms than higher latency DDR2, and even DDR3.  Again when we choose our memory modules we balance these factors out and choosing memory we know we can get a lot more out of than it says on the packet.  3D graphics cards require such phenomenal bandwidth that DDR3 isn’t enough and they have been using DDR4 for some time and are now moving on to DDR5.
  
• FB-DIMM memory – fully buffered memory is used in mission critical enterprise serverapplications.  Our server strength builds including the Octane with the Skulltrail motherboard require this memory.  Unfortunately due to the way the memory controller works it has a generally lower ‘paper’ performance than enthusiast memory reaching only PC2-6400 8000MHz standards, however in practice due to the way the memory controller works it is comparable with higher speed conventional DDR2 memory.  It has another advantage the decoupling of the memory bus directly from the CPU means that the PC can support much larger memory modules.  Typically a FB-DIMM memory PC will support double that of a conventional DDR2 memory PC (16 or 32GB as opposed to 8GB).