Water Drops

Some thoughts on taking the plunge to watercooling – Joe

SUMMARY: Some thoughts on taking the plunge to watercooling.

I get emails just about every day asking me about watercooling – “Should I” or “I’m interested, what do I look for?” seem to be the most common.

Answering question such as these are always difficult, as it’s impossible to put yourself in another person’s shoes. However, sharing some thoughts about watercooling, from the perspective of one who gets to play with a lot of gear, might shed some light on “Should I or shouldn’t I, and how?”

First some basics:

The basic principle of CPU cooling, whether air, water or whatever, is to move heat from the tiny CPU core to a larger surface to dissipate heat (a radiator). The more efficient the transfer and the larger the radiator, the better.

The inherent limitations with air cooling are spreading resistance and surface area. Copper, the best heatsink material from a cost/efficiency standpoint, can only move so much heat so far. In addition, the footprint for CPU cooling can only take so much area and weight. The combination of these two factors limits heatsink (radiator) size – and we’re pretty much at the limit now.

Water does this job very well; watercooling is about 100 times more efficient than air. How well a particular system does this job depends upon how the following components do their jobs:

  • Waterblock: We’ve done a fair amount of rigorous testing; taking a close look at results over the past year reveals the following:
    • The temperature range among blocks tested to date is fairly narrow: 7ºC at 100 watts CPU power @ 1 gpm waterflow; for a CPU radiating 75 watts, it’s a little over 5ºC. The absolute difference attributable to one waterblock vs another will most likely be less than 10ºC.
    • Copper does make a difference; at a minimum, a waterblock with a copper base will be a better choice than one of aluminum.

    • Small fittings exact a performance toll. Flow rate is a key performance determinant, and narrow, ¼” fittings severely restricts flow; opt for ½” fittings throughout the system.
  • Radiator: It’s pretty simple: In this case, all other things being equal, size matters. But note that the same radiator but twice as large does NOT mean system performance is two times better. One series of test I did showed that doubling radiator size dropped system temps by 3ºC (@ 100 watts). Further, don’t expect another doubling to equal the same improvement – diminishing returns will rear its head.
  • Waterpump: Flow rate matters: In all waterblocks tested to date, increasing flow increases performance. However, within the range between 0.5 gpm to 1.5 gpm, almost all waterblocks will vary by +/- 1ºC. Waterpumps rating something like 300 gph and 60″ head should deliver flows in this range, assuming other components in the system are not unduly restrictive.

  • Fittings: Small fittings restrict flow rates. It only takes one ¼” fitting in a system where other components are ½” to radically slow things down (an example HERE).

Putting It All Together

Following the basics outlined above will yield a high-performance system. But when you “glue” all these things together into a system, what performance range might you expect?

A good clue it to take a look at Watercooling Kits we’ve tested to date.

These are system tests – it’s the result of how the waterblock, radiator, fans, waterpump and fitting size all work together. If you take a look at the pieces in each system, you’ll see that the “entry level kits” typically are lower priced systems that use small pumps and radiators. Their don’t perform as well as high end systems that typically feature larger radiators, pumps and fittings.

Considering what we’ve found to date, this should not surprise anyone.

The performance range of all kits tested is 21ºC @ 100 watts. The range, excluding “entry level” kits, is 12ºC @ 100 watts.

Assuming careful selection of components, I doubt you’ll see more than a 5ºC @ 100 watts variation among users.

Temps At The Desktop

The performance numbers I’ve indicated above are under full load conditions. Are you a “power user?” Odds are that you’re not, especially if 95% of your time is spent on browsing the net, emailing and maybe some spreadsheet work. Under these conditions, a good watercooling system will most likely keep CPU temps at not more than 5ºC over ambient temps, with fan noise very low; more than likely your hard drive or power supply will make more noise.

My “work” system is a modified Koolance case, and running the system 24/7, I have never seen temps more than 4ºC over ambient; noise is barely noticeable three feet away.

Some Practical Advice…

You can tell a lot about a system’s performance by looking at it:

  • Tubing Size: The bare minimum is ¼” ID tubing – 3/8″ is preferred. Any larger, routing the tubing through the system becomes more difficult.
  • Radiator Size: If an 80mm fan just fits over it, it’s too small; minimum look for a radiator that uses a 120mm fan for airflow.
  • Fittings: All fittings should equal the size of the tubing’s ID.
  • Waterblock: Copper base construction.
  • Waterpump: 300 gph, 60″ head.

The Upgrade Path….

OK – so you bought a system already and want to “trade up” to better performance:

  • First place to look is the radiator – doubling the size will yield a performance boost.
  • Second, the waterpump – increasing flow is an easy performance gain.
  • Third, the waterblock – if it’s an aluminum block, trade up to copper.
  • Fourth, tubing size – lowering system resistance increases flow, etc. As you trade up components, make sure the fittings are not choke points.

Can I do as well with aircooling?

Yes – but NOT at the same noise levels. There are very good low noise aircooling solutions – any high performance heatsink (lots of surface area) with a low rpm fan will safely cool a CPU, but it will be almost impossible to equal the same performance, at the same noise levels, as watercooling.

And so…

In my book, the most effective balance between performance and noise favors water, and by a large margin.

Email Joe

Discussion

Leave a Reply

Your email address will not be published.


*