Given the atrocious VRM cooling on the Sapphire Vapor-X cards I figured I would put together a how-to on how you can bring the VRMs under control. The process requires patience and a moderate amount of skill but not anything terribly difficult. A understanding of thermodynamics would also be beneficial. First up, lets look at the problem with the VRM cooling; Note the areas highlighted with a red box, these are the actual locations of the VRMs (as can be seen in the image below). In the case of the small cluster of VRMs the problem should be immediately noticeable. The VRMs are not actually heatsinked at all despite appearances. Instead what has happened here is that Sapphire have just slapped a thermal pad underneath the anti-warp bar to prevent shorting, as such the VRMs will get incredibly hot under full GPU load, leading to unnecessary heat production and buildup as well as reducing the lifespan of your card and possibly decreasing stability and limit OC potential. Needless to say this design is entirely unacceptable especially from a manufacturer that is supposedly meant to be reputable. Q: How could Sapphire of fixed the problem without significantly adding to costs? A: Simply adding some fins to the VRM area of the anti-warp bar would of been sufficient. The second problem with the VRM cooling is not noticeable until you remove the main VRM heatsink. With the heatsink removed you will see the design of the VRM heatsink is horrendous. Sapphire failed to take into account the forest of resistors (as you can see below) which are taller than the VRMs themselves. How did Sapphire “fix” this problem? They added standoffs to the heatsink and used a dirty fat thermal pad (a good quarter of a inch thick!), doing this causes the following problems; - Extremely poor thermal contact - Extremely poor thermal conductivity due to the botch-up fix of standoffs and a fat thermal pad leading to a (extreme) “hotspot” - Due to the above, the very high VRM temps will reduce lifespan of the card, possibly effect stability and limit overclocking. Q: How could Sapphire of fixed this problem without significantly adding to cost? A: Sapphire could of avoided all these problems by removing the standoffs and cutting a groove into the underside of the VRM heatsink to accommodate the higher sitting resistors. A normal thickness thermal pad could of then been used which would of greatly improved thermal conductivity. Alternatively Sapphire could of made the heatsink slimmer but taller with more fins in a standard criss-cross design. The added height of the heatsink and additional fins of a standard criss-cross design would of been more than enough to make up for the lesser surface area, and indeed would of in all probability surpassed the cooling efficiency of the design in use. So, the question now is what does all of this mean and how do we fix these problems? Due to the very poor job done cooling the VRMs even in a well ventilated case running anything that will fully utilise the GPU will soon see those VRMs soaring to 90c – 100c (VRM2) or in excess of 110c – 120c (VRM1). To remedy this problem; You will need: Dremel Drill bits Vice Jr. Hacksaw SK 81 / 100 SA Heatsink (VRM1) Any chipset heatsink with lots of small fins (VRM2) Diamond tip burs Fine grit sandpaper Spring loaded push-pins OR some small screws with nuts Phobya thermal pads 0.5mm (7W/mk) Step 1: First measure and mark out a template for the small cluster of VRMs, once you have a cut to size heatsink you will want to mark out the positions for the holes and drill. When finished you will have something similar to this; Now just cut part of one of the Phobya pads to size, place over the VRMs and push on the new heatsink. Step 2: This is where things might get tricky if you have decided to make completely new VRM heatsinks as I have. The reason for why things may now get tricky is due to the “Z” shape of the hole positions, although its nothing that should really cause any headaches just a little care required. Your task should be made easier still if you got the SK 81 / 100 SA heatsink like I did; SK 81 / 100 SA heatsink The great thing about this type of heatsink is that you can design the new VRM heatsink so one of the fins sits directly over the VRMs, giving them a “dedicated” fin to drawing heat away which will help cool them even more. If you feel so inclined you could also saw through the fins stopping at the base which will aid cooling further as in the case of air coolers the airflow could more easily pass across the heatsink. I chose to cut fins 1 inch apart from each other. Measure the correct width and cut what you need, the width needed if going for tall and slim will be 4 fins, but cut right at the edge of the 5th fin so you will have ample space for drilling the holes. With that done, cut a Phobya pad to size, place over VRMs and attach the heatsink. The final result with everything attached will look similar to the below image, note that I took the individual part of the anti-warp bar and attached it with a few screws and nuts – this will prevent the PCB from warping. Custom VRM heatsinks all fitted Side View Close Up Note in that final image how tall the VRM heatsink is, I could make the heatsink this tall because I’m also using the excellent (far superior to the Vapor-X stock cooling in both performance and noise) Gelid Icy Vision A. You are now all done, if you so choose you can finish the heatsink in a different colour or just to rid it of any scuff marks. The VRMs are now cooled by a properly designed thermal solution and you now have the following benefits; - Significantly reduced “hotspots” temps under full GPU load - Extended the life of your card - Improved stability - Improved overclocking potential - Reduced overall generated heat Final maximum load temps before and after for me with 99% GPU load are as follows; VRM2 100c (stock) VRM1 120c (stock) -------------------------- VRM2 78c (OC 1125 / 1565) VRM1 88c (OC 1125 / 1565) These temps were taken with ambient temps of about 24c (its been pretty warm today) using GPU-Z, however my temp gun strongly disagreed with the GPU-Z readings. I've just used the GPU-Z results as far more people have access to GPU-Z than a temp gun. Bare in mind I chose to use aluminium to make new heatsinks, if I had used copper I’d think you could take off a further 5c or so. In case anyone is wondering, the reason for the Phobya pads instead of actual TIM is because I tried using actual TIM and this resorted in a short, plus the Phobya pads are excellent and very close in capabilities to actual TIM so why not go with a easier, far less messy option?