Swiftech MC14 BGA Memory Heatsinks Review :: Installation & Performance

12-09-2005 · Category: Hardware - Cooling

By Tulatin

As expected with any heatsink that features a form of thermal tape, installation was as simple as removing any existing solution present, cleaning off old thermal grease, then peeling back the protective paper and applying the sink. It should be taken to note however, that the thermal pads included with these heatsinks are nowhere near as forgiving as 3M's ubiquitous "frag tape", and once they're installed, they definitely won't be removable with the hope of re-application. Now, for most users installation of these heatsinks would be problem free, but unfortunately, due to the location of two memory chips beneath the reference cooler's fan, the bulk of the stock cooling solution had to be removed in order to properly apply these heatsinks. In lieu of the reference fan, an 80mm fan was affixed to screws and washers so to keep the stock heatsink cooled, as well as provide some airflow over the memory chips. To be fair, the card was tested in this manner with and without memory heatsinks, in order to show the effectiveness and gains that can be afforded by Swiftech's heatsinks, in opposition to the standard thermal solution for cooling the memory.

Swiftech MC14 BGA Memory Heatsinks Review
Swiftech MC14 BGA Memory Heatsinks Review
Swiftech MC14 BGA Memory Heatsinks Review

Before we start into the over clocking section of the review, let's take a little look at our testing environment, as well as our methodology. All of our testing of these memory heat sinks was completed on the following:

  • Athlon 64 3000+ Winchester
  • 2x512MB Modules of Crucial Ballistix PC4000
  • Enermax 465W PSU
  • NVIDIA Reference 256MB 6800GT.

Test bed in mind, it's time to address our rather simple methodology. Rather than do things the good old fashioned way - that being to tap clocks up a few MHz then benchmark for a while, we took advantage of W1zzard's ATITool - which, satisfyingly enough, can now automatically overclock NVIDIA cards while scanning for artifacts. Once ATITool found the card's memory plateau, we'd leave it scanning for artifacts, and pending a successful half hour run, the value would be recorded, and the next stage of overclocking would begin - efficient no? Methodology in mind, let's get on to the results shall we?

In our initial testing of the card, we left the reference cooler completely intact, with standard memory heat pipe solution and all installed. Without any extraneous airflow across the card's face, we were able to top the memory out at a meager 515MHz (or 1,030 as NVIDIA's driver says). When we added a 120mm fan capable of pushing 55CFM (the tradeoff for the low airflow is noise), the chips found themselves capable of topping out at 528.5 (1,057), which was a fairly nice improvement for just adding airflow, one that proved that the stock fan isn't capable of moving sufficient air.

Swiftech MC14 BGA Memory Heatsinks Review Stock, no air flow

Stock, no air flow

Swiftech MC14 BGA Memory Heatsinks Review Stock, focused air flow

Stock, focused air flow

With the standard thermal solution as a reference point, it was time to get down to business - we unmounted the standard thermal solution, and gave the BGA chips hiding underneath a good scrubbing with some cotton balls and rubbing alcohol, then after a brief period of drying, fired the card up once again. With no airflow across them and no sinks to aid in heat dissipation, the chips registered a pathetic 503MHz over clock (1,006). After a short period of testing, we added the 120mm fan once again, and found that it helped clocks vastly, allowing these sweltering little chips to push up into the 515MHz region (1,030) - the same as they had achieved with passive heat sink cooling.

Swiftech MC14 BGA Memory Heatsinks Review Naked, no air flow

Naked, no air flow

Swiftech MC14 BGA Memory Heatsinks Review Naked, focused air flow

Naked, focused air flow

Finally it was time for the show - we popped open the package, took out Swiftech's sleek little sinks and applied them to the memory chips. Now, this was definitely a strange moment, as either with or without airflow across them, these little heat sinks allowed us to top the memory out at an impressive 536MHz (1,072) - which was a nice improvement over the stock solution. The only problem came in the way of re-worked cooling eating up more space in the system; after all we couldn't leave two of the poor chips uncooled and out of airflow's way.

Swiftech MC14 BGA Memory Heatsinks Review No airflow

No airflow