Results

Word of Caution:  I am not a scientist or an engineer.  The results below were measured with all other external factors (case, components, air flow, etc.)  remaining constant.  Much of the actual discussion and rational is pure conjecture as I'm not a scientist or engineer.  Yet, this is the "results" page and the discussion is secondary to the results achieved.

Overclocking Results

Summary:

With only the original stock cooling, my max oc was 345 core & 332 memory with some artifacting.  Adding the heatsink to the voltage regulator heat plate allowed an overclock to 352 core/345 memory with some artifacting & tearing. With the replacement 1u Dynatron heatsink fan and heatsink on the voltage regulator heat plate the stable overclock is a core of 364 (372 had minimal artifacting) and memory at 358 (364 had some minor tearting).  Note: These figures used 6.75mhz increments in each overclock level.  They've been rounded for ease of display.

An interesting note is that the replacement heatsink proved to better dissipate heat from the GPU and the PCB/memory/components.  As indicated in the temperatures section below, adding the replacement heatsink resulted in increased temperature in my case.  This can be attributed to the new heatsink dissipating heat from the GPU and blowing it into the case.  Furthermore, the memory heatsinks, the voltage regulator heat plate and connected heatsink, and the rear heatsink were all significantly cooler to the touch with the new heatsink attached.  This suggests that the stock heatsink is inadequate in dissipating a large amount of the GPU heat.  Consequently, it remains in the GPU and spreads throughout the PCB effecting and warming other components that are attached.  In part, this is probably why the memory overclock was assisted by the replacement heatsink.  Consequently, I can now maintain the maximum stable overclock without the rear heatsink attached to the opposite side of the GPU on the PCB (tests indicated that the rear GPU no longer assists in the overclock although it still assists in dissipating heat as it remains mildly warm whereas it use to become quite warm if not outright hot).  Arguably, I may even be able to remove the heatsink attached to the voltage regulator heat plate.  That heatsink and plate use to become quite warm if not a little hot.  As of now, it's mildly warm, if that.  I cannot say if the heatsink on the voltage plate makes any difference after adding the replacment heatsink because I haven't removed the heatsink on the regulator plate to test the hypothesis.

In conclusion, replacing the stock heatsink with a 1u resulted in an increased oc of ~20mhz for both the memory and the core (if using speeds that had no artifacting/tearing). It not only assisted in greater overclocking but helps assist in keeping the GPU heat from spreading throughout the PCB and effecting other components.  Consequently, it provides a superior overclock with a streamlined appearance as I no longer need additional heatsinks on the card.  Yet, the most significant benefit of this project for me is a silent heatsink and fan. 

Temperature Figures

Before preparing for this modification I had only 2 temperature figures: case temperature and CPU temperature from the Athlon XP CPU.  I ordered a temperature probe with my Dynatron heatsink and Zalman Fanmate.  I used this to take temperatures of the back of the PCB, opposite the GPU.  The relevant temperatures are the case temperature (which is effected by the heat dissipation of components & which effects ambient temperatures which can effect the reading of the thermal probe) and the thermal probe to measure the temperature of the GPU. 

*Source: Frostytech.com Article: Zalman ZM80A-HP VGA Heatpipe Cooler Review, Page 9

The only temperature I took with the stock heatsink/fan is with the thermal probe resting against the aluminum spacer that acts as a conduit between the rear heatsink and PCB.  It was not taped to the spacer but merely rested against it.  Unfortunately, this temperatue was also measured at a point in time when the stock heatsink/fan was dying.  It was blowing quite a bit of air still but was clearly on the out as it had become quieter as the speed of the fan slowed down.  As such, I provide these results only to demonstrate how hot the GPU can become even with a somewhat working heatsink/fan (please note that this does not effect the overclocking results with the stock hs/fan as those results were measured a few days after I purchased the card when the fan was fully functional in all it's audible whir.)  I would consider the temperature taken by Frosty Tech to be accurate and should be the temperature against which the modification is compared although the large differences in ambient temperatures should be taken into consideration.   Even without discounting the significantly cooler ambient temperature, the temperature of the GPU with the Dynatron/Cooljag compares favorably against the stock heatsink/fan.  Two weeks after obtaining the components, the thermal probe and the initial temperature, I finally got around to completing the modification.  As I went to remove the rear heatsink I nearly burned myself as it was that hot.  It was significantly hotter than when I took the initial temperature so the fan had definitely deteriorated further.  The backside hs/fan was definitely serving a purpose at this point.

Because I removed the backside hs/fan for initial testing with the replacement heatsink/fan I was finally able to take a legitimate measurement.  It measured 42c.  The case temperature increased to 35c.  There was no other reason to account for this jump in case temps.

For further experimentation I reapplied the backside hs/fan after installing the replacement hs/fan.  I thermal taped the probe to the space to create a similar environment to that in which I took the temps with the stock heatsink/fan (albeit thermal taping the probe will create better contact than merely pushing it up against the spacer).  The temperature was now 38c while idle and 40c at load.  This is much better than the stock heatsink results of 48.1 as reported by Frostytech.    Please note that this does not effect the overclocking results with the stock hs/fan as those results were measured a few days after I purchased the card when the fan was fully functional in all it's audible whir. 

Zalman ZM80 Comparison

The obvious question is "how does this mod compare to a similar replacement heatsink?" 

Temperatures and Overclockability

While I can't compare the differences in overclockability since I do not have a Zalman ZM80, I can compare general temperatures and aesthetics.  The temperature readings of  the Zalman ZM80A-HP originate from a Frosty Tech review  which used a similar method to take temperatures of the GPU.  At stock speeds and at full load, the GPU is 58.5c using a ZM80A-HP with passive cooling and 41.5c with active cooling compared to the Dynatron/Cooljag's temperatures of 42-44c.  It should be noted that not only did the active cooling for the ZM80A use a 4000RPM 92mm case fan but it was also done with an ambient temperature of 22.6C.  This is in start comparison to the environment I had when taking the temperatures.  My environment did not include a direct fan blowing on the video card (there was a 80mm 27 cfm, 1600 RPM PC Power & Cooling Silencer 80mm fan that not only hardly blows any air but it has a 15k RPM Seagate Cheetah sitting right in front of it so any air it is blowing has been mildly warmed up by the HD) and it was within a case whose temperature was 32c (my case sits by a heat vent.  Temporary situation so haven't cared to move it yet).  It is not my intention to criticize that review.  On the contrary, I'm using it because I believe it's reliable.  Yet, it is important to point out critical differences and ambient temperature is extremely important when considering air cooling results.    I recently added some side cooling which blows air onto the side of the card (although not directly on where the probe is placed).  The temperature is now 36c idle and 38c load with a case temperature of 34c (although the placement of the thermal probe reduces the temp by 2c).  As such, the Dynatron/Cooljag's results are comparable to an actively cooled Zalman in a non-overclocking environment. It is difficult to discount for the large difference in ambience temperature but I would hypothesize in an open environment the Dynatron/Cooljag would drop at least 4c.  The Dynatron/Cooljag is clearly superior to the passively cooled Zalman.  There needn't be any "discounting" of factors to make that conclusion clear.  And, with a fan contoller, the Dynatron/Cooljag is effectively as silent as the passively cooled Zalman.  It's safe to say that overclocking results are likely to be similar using either an actively cooled Zalman or the Dynatron/Cooljag.  It is also clear that most reviewers are correct in recommending that one not overclock their cards when using a passive Zalman ZM80A.  It should also be noted that many report the Zalman prevents the placement of most memory heatsinks on the card.  The Dynatron/Cooljag will not impact the placement of any memory heatsinks.  Having used memory heatsinks on both a TNT card and this 9700 Pro, I've never experienced memory heatsinks actually improving overclockability of the memory.

Noise

Using a non-actively cooled ZM80A-HP will be 100% silent but cooling results leave a lot of  wanting.  Using the Dynatron/Cooljag 1u will produce a soft whisper without a fan control device.  Using a fan control device such as the Zalman Fanmate you can turn the fan down to a point where it is silent even with the case open.

Size & Weight

There is no question the ZM80A-HP is a monster taking up a good portion of the open case above the AGP slot and taking up a PCI slot.  You will lose the closest PCI slot with either solution.  The ZM80A-HP weighs 400g versus the Dynatron/CoolJag's 252g.  This should be a consideration if you often lug your PC around.

Price

If you have the power tools necessary, the cost of this project was about $30 including S&H for the Dynatron/Cooljag fan, Zalman Fanmate, bolts, screws, drill bits, and tap.  Pricewatch indicates the lowest price on the Zalman ZM80A-HP is $34 (1 vendor) with the average around $40 after S&H.

Work

The ZM80A-HP is intended for relatively simple installation (although not as easy as typical HS/fans).  Some of the installation difficulties are discussed here.  None the less, once you get past the drilling aspect of this mod, seating the Dynatron/Cooljag heatsink is like seating any other normal heatsink and it brings a pleasure of fulfillment that buying a ready-to-install heatsink can't provide.

HOME