Removing the cooler was a relatively simple affair but there are 12 smaller screws holding the shroud and additional heatsinks to the PCB making a total of 16 screws including the main GPU core baseplate. I found the contact with the core to be very good with the paste in the 'Goldilocks zone', not too dry, not to much - just right! Both the memory and VRM module have their own passive heatsinks which are held in place by screws and have thermal tape is a means of conducting heat.



The main heatsink is as long as the graphics card PCB and is a mixture of aluminium longitudinal fins and what appears to be copper heatpipes attached to a copper baseplate with an aluminium frame.


The four heatpipes are 6mm in diameter and I was surprised they were not spaced out evenly to dissipate heat effectively rather they are positioned towards the ends of the heatsink. I will be interested to see if this method yields better results than the other cards on test.



Rather than rely on manual labour to solder the heatpipes to the aluminium heatsink, KFA2 have employed the neater(and presumably cheaper) automatic method of dipping the heatpipes. While this sample appears to have effective contact between the heatpipes and the fins, I have seen heatsinks that have not had the best application using this method.



Arguments will rage as to whether a polished or rough base is the best method of transferring heat. Personally, I'm with the polished crowd on this as thermal interface material will never be as good as the material it bonds to for transferring heat so the less TIM the better is usually the case. Having a rough contact area means that more TIM is inevitably used to fill the minute crevices in the surface, ergo it will not be as efficient. This is of course assuming there are no other elements that will affect mated surfaces such as concave/bowed heatsinks or integrated heatspreaders. KFA2 have opted for the dull finish with this card and while it is by no means rough, it is not a perfect finish as you can see.



You could be forgiven I had gone over the top with my exposure settings when taking these pictures but the copper really is that bright, un-naturally so. I therefore scratched the surface (literally) to see what lies beneath and I am happy to report both the baseplate and heatpipes are full blown copper and not just copper plated. I can only assume the plating is there to prevent oxidisation of the copper.



The VRM module has 6 inductors and KFA2 recommend a minimum 550w PSU capable of delivering 35A on the 12v rail for the graphics card to operate. Obviously, adding subsequent cards will mean a much beefier PSU will be required. We will, as always, be testing the power consumption of the card later in the review.



The KFA2 GTX570 comes with 1280MB of GDDR5. The Samsung modules are the K4G10325FE-HC04 variant and as such are specified to run with a voltage of up to 1.5V and a maximum rated frequency of 1250 (5000) MHz.



Finally we arrive to the heart of the GTX570, the GF110 core. Having the same core as the mighty GTX580 but 32 less CUDA cores, the GTX570 has proven to be very popular thanks to its very good performance vs. price ratio. With this particular variant being the cheapest GTX570 we have tested to date we can only hope that statement is strengthened as we put the card through our batch of benchmark tests.

Let's take a look at the test setup we are going to use today...