CPU Cooling Challenge – Time For Water??

A case for watercooling – Joe

SUMMARY: Looks like the stars may be aligning for “mainstream” watercooling.

I’ve been observing the heatsink scene for going on five years now and every once in a while, I sit back and try to read the tea leaves to discern where things are heading.

The major event in CPU cooling has been CPU die-shrink (see “CPU Die Size – The Cooling Challenge Ahead”); basically the overarching issue confronting heatsink designers is how to cool more watts in an ever decreasing area. As the watts/mm² increase, wicking off heat becomes increasingly (maybe exponentially) more difficult.

An excellent paper by Kaveh Azar shows these trends very nicely:

Pwr Trends

Kaveh Azar, Ph.D. – paper HERE.

Note the sharp rise in power density starting 1999. The implication, as we overclockers know all too well, is that we are using alternative cooling technologies to deal with this trend, most notably water:

Power

Kaveh Azar, Ph.D. – paper HERE.

No question that aircooling is still the preferred method, considering cost and complexity; no question that how far aircooling can go is an open question (see “Heatsinks – Hitting a Wall??”). Looking at the physical properties of air vs water clealry shows these limits:

Limits

Kaveh Azar, Ph.D. – paper HERE.

If you’ve been around since ancient history (the last five years), you’ve no doubt seen heatsink increase in size. However, how far this can continue is also open to question (see “Heatsink Performance Limits”).

HS Size

Yogendra Joshi – paper HERE.

Note also from this graph by Yogendra Joshi, that as size increases, so does airflow; this translates to larger fans and, as we have seen, much more noise for aggressive performance. Many of us remember the stampede to Delta 38s, to be quickly followed by the stampede to throw them out due to intolerable noise.

One of the primary issues facing heatsink designers is spreading resistance – one technical paper found that “[When aluminum heatsink] length is changed from 50mm to 100mm, weight is changed from 133gm to 266 gm, completely double the weight, yet the performance gain is only 16%.” (Honglong Chen, paper HERE.) Doubling size by no means doubles performance.

Current materials (copper and aluminum) are about stretched to the limit of what they can efficiently handle. Until new aircooling technologies hit the market with cost efficiencies rivaling these materials, alternative CPU cooling technologies are likely to look more attractive; of these, IMHO water looks to be one of the better choices.¹

While watercooling is currently more expensive and complex, I believe the increased attention it appears to be getting from larger companies now entering the field may alter this markedly. Many “pioneer” watercooling manufacturers started as garage shop operations and some still retain these characteristics. As larger companies with more extensive financial and engineering resources enter the market, more effective solutions may see the light of day.

In addition, suppliers of key components, such as waterpumps and radiators, are beginning to take notice of this market and we can expect more products specifically geared for CPU cooling. Currently users and vendors are adapting products used for other purposes. For example, radiators designed for high pressure oil cooling may carry a cost penalty when used in very low pressure PC watercooling.

THERMOSYPHONS

One of the issues watercooling faces is the need for an active pump. One technology that is pumpless is the Thermosyphon. I tested one example, the “Calm-SV Liquid CPU Cooler”. Although performance of this particular example was not anything to get excited about, the technology is worth exploring.

The following diagram shows the basic principle:

Pic

Kaveh Azar, Ph.D. – paper HERE.

The attraction of a pumpless system is clear: waterpumps are relatively expensive and a clear “failure point” in a watercooling system. If it could be eliminated, costs drop markedly. One key point is that vaporizing the liquid is required by the CPU; this may require a liquid with a lower boiling point than water:

Props

Yogendra Joshi – paper HERE.

The design challenge then becomes selecting the right fluid for the application. Overall, an interesting technology that requires a fair amount of “art” to deliver superior cooling performance.

Conclusions

One thing about physics: It can’t be bent.

As CPU frequencies increase and size shrinks, escalating chip power densities will inevitably drive CPU cooling to more efficient solutions. Watercooling, once a geek toy, is breaking out to take a place in mainstream PC applications.

I always find it interesting that we overclockers deal with these issues a couple of years ahead of the “mass market” – in fact, we tend to pioneer these technologies. Kind of a nice place to be.

¹Apparently NEC agrees – see their watercooled PC HERE.

Additional Reading: The following sources delve into these issues in more detail:

Tags: