Does Cooling Matter With Core 2 and AM2?

In a recent review of the upcoming Core 2 Duo E4300, Anand found that the new 1.8GHz E4300 could be easily pushed to 3.37GHz with stock cooling, an astounding 87% overclock. Why then, would anyone want to worry about enhanced cooling? The answer is really simple. If you want to reach even higher overclocks - to potentially around 4 GHz - you need to look at improving CPU cooling.

It is absolutely true that Core 2 Duo is not Pentium 4 when it comes to heat generation. The extremely hot and poor performing NetBurst processors have been replaced by Core 2 Duo, which runs very cool at stock speeds and overclocks like few chips we have ever seen in the evolution of the CPU. As a result life is good right now for Intel CPU users.

This year we will see if AMD can accomplish something similar with AM2 at 65nm. Current top 90nm AM2 processors have almost no "head room" at all, making concerns about overclocking relevant only for lower end AM2 chips. Early 65nm chips are also not stellar overclockers, but it is still early, and everyone expects that AMD can achieve overclocking results similar to C2D when their 65nm manufacturing is fully rolled out. It will be interesting to stop the speculation and actually see what AMD rolls out this year.

With relatively cool CPUs and massive overclocking capabilities it is worth looking at what happens with a stock processor and a stock heatsink/fan, at rated speed and when overclocked. We monitored CPU temperature on an NVIDIA 680i motherboard running an X6800 CPU with the stock Intel HSF - both at the stock speed of 2.93GHz, and at the highest stable overclock that would still allow gaming of 3.73GHz.

X6800 CPU Operating Temperature with Standard Heatsink/Fan
(Ambient Temperature 21C (70F)
  CPU System
IDLE 2.93GHz 41C (106F) 32C (90F)
GAMING 2.93GHz 56C (133F) 33C (91F)
IDLE 3.73GHz 56C (133F) 37C (99F)
Gaming 3.73GHz 71C (160F) 38C (100F)

While temperatures are dramatically improved compared to P4 NetBurst processors, it is clear that temperatures do increase rapidly as the overclock increases. At the highest overclock of 3.73GHz idle temperature had increased to 56C (133F) and stressed temperature rises to 71C (160F).

At this 3.73GHz overclock we suspect that the biggest handicap to further overclocking is CPU cooling. However, this is a question that can only be answered with a closer look at the performance of the Tuniq Tower 120. In the end we buy improved coolers for better cooling in most performance configurations. This improved cooling should theoretically allow even higher overclocks of the CPU. If the cooler you are considering does not cool better and/or improve overclocking, there is absolutely no justification at all for buying the cooling solution.

Index Sunbeam Tuniq Tower 120
Comments Locked

50 Comments

View All Comments

  • LoneWolf15 - Monday, January 15, 2007 - link

    Thanks Wesley, for the review.

    I'd love to see a complete "cooling tower" review. There are a lot of similar products out there, such as the Sycthe Ninja/Scythe Infinity family, a couple by Noctua, the CoolerMaster Hyper 6+, Arctic's Freezer Pro line, and (sort of similar) Zalman's 9500/9700 line. A lot of us would like to know how they stack up against each other on both Intel and AMD platforms, with the following notes:

    Weight of each cooler (lighter being better if it doesn't sacrifice performance)
    Ease of mounting on both AMD and Intel systems (with a key to whether the heatsink can be mounted properly for fan exhaust direction, regardless of socket orientation on the mainboard)
    Noise level
    Cooling effectiveness
    Smoothness of CPU mating surface

    I've been tempted to buy a tower heatsink/fan unit, as I think it will cool better than my good, but aging Swiftech MCX64-V, since if chosen correctly, a new model should blow air straight to the 120mm exhaust fan at the back of the case. However, I haven't seen a good side-by-side comparison yet that tells me everything I'd wish to know. I hope Anandtech will look into this. Thanks!
  • AlabamaMan - Monday, January 15, 2007 - link

    More pictures would have been very helpfull. I have an e680i mobo and Lian-Li A10 case I still can't fiure out if I can put in the Tower without hitting the top-mounted 120mm case fan.
  • Wesley Fink - Monday, January 15, 2007 - link

    If the fan is really at the top, and not at the side of the case (in a tower design) the concerns are different. Considering dimensions toward the physical top (long dimension or height of a tower case), the Tuniq width in that direction is 110mm or 4.3". You can measure from the center of the socket 775 to see if you have 55mm (2.165") clearance to your fan from the center of the socket to the edge of the fan. Just for info the top of the Tuniq is 110mm x 128mm (4.3" x 5").
  • Wesley Fink - Monday, January 15, 2007 - link

    The Tuniq Tower dimentions are on page 3. A normal case is 19cm (7.5") deep in the cooler dimension, while the Tuniq is 15.5cm (6.1") tall. I can tell you for a fact the Tuniq clears all components on the 680i baord and can even be used with tall memory mounted (Corsair Diminator for example). You will need to determine if 1.4" is enough clearnace for you fan. Also since the fan would be blowing down on the solid top of the Tuniq, it likely will accomplish nothing as far as cooling is concerned and should be removed.

    Additional information about the size of the Tuniq can be found at www.tuniq.com
  • mackintire - Monday, January 15, 2007 - link

    I m pretty sure the Tuniq Tower 120 is the best air cooled heatsink.

    I have a Intel Core 2 Quad QX6700. I do have a Scythe Ninja on it with a scythe 1600rpm S-fan. Running 2 instances of Orthos on all four cores gave me a full load temp of 74C and I know I have too much artic silver 5 on it. So realistically my load temps should be 73-72C.

    I just purchased a Noctua NH-U12F which I believe to be really close to the performance of the Tuniq Tower 120.

    I think we definately need the Noctua NH-U12F in the best cooler roundup.

    The ninja need to be there for reference. And the Scythe Infinity would be a good idea too.

    Another item of note is that these larger heavier CPU's show a different delta when loaded up with super hot CPUs.

    My quad core running at 3.2 produces mega heat that few heatsinks can deal with.

    The reason I switched to the NH-U12F was that it dropped my temp by 8C at stock speeds. 2 C of that I blame on properly application of artic silver.

    Either way, this heatsink was well worth the money to me.

  • mackintire - Monday, January 15, 2007 - link

    I changed over to the Noctura this past weekend. Just clearing up my previous post.

    The Ninja is an excellent heatsink, but I think that there is a limit as to how much heat a heatsink can deal with. I have noticed more recently with the current round of newer solutions that the coolers with more mass tend to deal with larger heat loads better. This is not an exact trend, I m just saying that I haven t found any 300g heatsinks that can perform as well as the better 700g heatsinks under heavy heat loads.
  • baronzemo78 - Monday, January 15, 2007 - link

    Does anyone know how the thermal grease that comes with the Tuniq compares to Arctic Silver 5? Also I would love to see an Andantech article about lapping. I have seen some articles that say that lapping doesn't really improve your temps that much.

    http://www.overclockers.com/tips458/">http://www.overclockers.com/tips458/
  • Araemo - Monday, January 15, 2007 - link

    Lapping was much more important on Pentium 3s and Athlons that had an exposed CPU Die, and no heat spreader.

    On modern Intel and AMD CPUs with a heat spreader, the thermal transfer point that needs the highest efficiency is between the heat spreader and the die - you can't do anything about that. The heat spreader gives you much more surface area to transfer the heat with, so the small increase from lapping isn't nearly as important as it used to be.

    Perhaps if you're using a Tuniq Tower w/ a super high speed fan and an AC unit blowing at your case inputs and you're trying to break 4Ghz.. it might help. But for us real people, I'd say you're better off spending the time/money on a case with better airflow dynamics or just tying your cables out of the way to maximise airflow.

    I have lapped a couple heatsinks, one I did a real half-assed job on, and temps were disappointing(lower than OEM heatsink, but higher than I would have liked).. my last one was used to overclock an AthlonXP mobile from 1.8 to 2.4, and I daresay it helped, but as I mentioned, those AthlonXPs had exposed CPU dies, so you had to transfer the full 60W or whatever using .5 square centimeters of surface area, so improving the transfer rate in any way possible helped.
  • mino - Monday, January 15, 2007 - link

    Please consider doing the tests on Quad CPU(Kentsfield) as well as on a Duo CPU.

    Many coolers while beeing great for C2D perform poorly on Quad and vice versa.
  • Avalon - Monday, January 15, 2007 - link

    I'd like to see the Coolermaster Hyper TX included if possible!
    Good review!

Log in

Don't have an account? Sign up now