Both Nvidia and AMD have densely packed product stacks, with GPUs separated by as little as $30 increments. Let’s just say that Nvidia didn’t rack up a revenue of $12 billion in 2018 by letting overclockers undercut its bottom line. There’s a method to Nvidia’s apparent madness.
Leveraging GPU Boost with Liquid Cooling
Enter Nvidia GPU Boost. On the surface it delivers the benefits of automatic overclocking to the masses. It also conveniently hobbles overclockers by severely limiting control over important overclocking parameters such as voltage and power draw. That makes the maximum achievable clockspeed with Pascal (GTX 10xx series) and Turing (RTX 20xx series) GPUs largely dependent on the available thermal headroom.
In short, your painstakingly-achieved manual overclock matters little when GPU Boost dictates clock speeds based on GPU thermal parameters. That’s precisely why Nvidia was brave enough to launch Turing graphics cards with barely any performance advantage over their predecessors. Nvidia GPU Boost ensures that you can’t close the performance gap between the two GPU generations with a simple overclock. However, if you own a last generation Nvidia Pascal card, there’s a cheaper alternative to upgrading performance without having to buy a new Turing GPU. The idea is to drastically improve GPU cooling to leverage the propensity of Nvidia GPU Boost to allow for higher sustained clock speeds when thermal headroom is available. This can be achieved by adding a low-cost AIO liquid cooler onto the GPU.
How Does a 20-Percent Faster GPU Sound?
Instead of losing money upgrading to a new Turing GPU, I added Corsair’s H80i v2 AIO liquid cooler onto my year old Nvidia GTX 1070 Ti Founders Edition GPU with the help of the NZXT Kraken G12 mounting kit. To find out if this works, I put the stock, overclocked, liquid-cooled versions of the same GPU through the paces using the 3D mark Time Spy graphics benchmark. The results were astonishing.
Even after overclocking the GPU to the hilt, I could only manage to get 5.7 percent improvement in performance with the stock cooling setup. This is because Nvidia’s GPU Boost prevents the GPU from achieving the full OC potential owing to strict thermal restrictions. However, the AIO liquid cooler mod netted a massive 20 percent performance improvement compared to the stock GPU. In fact, the modded GPU’s 3D Mark Time Spy GPU score of 7900 points put it a class above the newer Nvidia RTX 2060 with its relatively measly Time Spy score of 7600 points. It’s safe to say there’s a genuine incentive in implementing this mod. This is especially true when you can achieve better than RTX 2060 performance at a fraction of the upgrade cost.
What You Need: Bare Essentials
Before we can get started, here’s a list of everything you’ll need to successfully cool your GPU with an aftermarket AIO liquid cooler. Do note that this guide expects some degree of competence and familiarity with PC assembly and disassembly from the user. If you are not comfortable taking a screwdriver to your GPU, this mod isn’t for you.
NZXT Kraken G12 mounting kit Compatible AIO liquid cooler Phillips head screwdriver with tips/bits in PH1 and PH00 sizes 4mm hex socket driver Alcohol wipes or isopropyl alcohol (minimum purity of 70 percent. 90 and 99 percent preferred) Thermal compound Lint-free wipes or microfibre cloth
You can choose an AIO liquid cooler of any make, model, or size as long as it is listed in the compatibility list for the NZXT Kraken G12 mounting kit. In fact, AIO coolers from brands and models not in the official compatibility list also work, provided they use the same Asetek mounting system found in the NZXT Kraken X-series of AIO coolers. The Corsair H80i v2 liquid cooler used in this guide isn’t officially supported, but it works absolutely fine. Depending on the make and model of your AIO cooler, you may or may not have to buy thermal compound separately. You can purchase the Philips head screwdriver tips in PH1 and PH00 sizes along with the 4mm hex socket driver separately. These are also available along with other useful drive bits in the iFixit Mako precision screwdriver set.
The Optional Extras
These parts are strictly optional. If you are on a tight budget, you can safely skip these components without worry. These optional extras are only required for improving the aesthetics and further stretching the overclocking potential of the GPU. The M2.5 hex nuts will come in handy for retaining the GPU backplate. The GPU fan adapter and Y-splitter cable are indispensable if you want the AIO cooler’s radiator as well as VRM fans to be controlled by the GPU itself. Otherwise, you can save money by connecting these fans to the motherboard fan headers.
M2.5 hex nuts 4-pin PWM fan adapter/splitter for GPU VRAM heatsink (14mm x 12mm x 5.5mm) MOSFET heatsink (22mm x 8mm x 5mm) PWM controller heatsink (10mm x 10mm x 10mm) Either 3M 8810 thermal adhesive or Arctic Silver thermal epoxy Anti-static wrist strap Cable ties
The VRAM, MOSFET, and PWM controller heatsinks will help you push overclocks much higher. Mounting these tiny heatsinks onto the VRM (or power delivery) components can be achieved with 3M’s 8810 thermal adhesive tape. There’s always a risk of these electrically conductive heatsinks falling off the GPU and into the case with the 3M thermal adhesive tape. This is especially true if you choose heavier copper heatsinks, which are far more effective than lighter aluminium ones. In that case, it is recommended to permanently fix the heatsinks onto the components with a thermal epoxy such as Arctic Silver. Mind you, once you epoxy heatsinks onto your graphics card, you can kiss your warranty goodbye since they aren’t going to come off. Ever. Now that we have that out of the way, we can move onto the actual modding process. Note: Make sure you’re using an anti-static wrist strap that is grounded to a large enough conductive material such as the PC case. Avoid standing on a rug and work barefoot, if possible, to prevent build-up of static electricity.
The Steps
Use the PH00 screwdriver bit to remove every single screw securing the backplate. Take off the backplate and keep it safe along with the screws. You’ll need these later.
Switch over to the PH1 screwdriver bit to remove the larger screws holding the stock heatsink onto the GPU. Don’t forget to remove two screws securing the heatsink to the rear I/O shield marked by the arrows.
You’ll need the 4mm hex socket driver to unscrew the standoffs, which serve as backplate mounting points as well as screws that hold the heatsink onto the PCB.
Carefully separate the stock heatsink from the PCB. Don’t yank the whole thing off just yet, or you’ll risk damaging the GPU fan and logo LED cables. Gently pry both cables off and separate the heatsink from the PCB.
Skip to the next step if you don’t intend to add aftermarket heatsinks to the VRM modules and instead prefer to rely on the VRM cooling fan provided with the Kraken G12 mounting kit.
Attaching the VRM heatsinks using the 3M 8810 thermal adhesive is a simple matter of cutting the adhesive to size, placing it onto the modules, and pressing the heatsinks firmly in place. Make sure you clean every surface (heatsink and VRM modules) with isopropyl alcohol and wait a few minutes to let it evaporate.
Using a more permanent method such as Arctic Silver thermal epoxy is a bit more complicated. Both epoxy compounds must be mixed thoroughly in small batches and applied quickly. This epoxy hardens fairly fast, so break the entire heatsink application process into smaller batches. Apply a small drop of epoxy, taking care that it doesn’t overflow onto the PCB and short components. The heatsinks must be pressed onto to VRM modules for a few minutes to ensure a thin and thermally-efficient application. It is advised that you use clamps to secure the heatsinks onto the VRM modules at least for the first 15 minutes. This ensures good adhesion and a thin epoxy layer. 6. You can skip this step if you don’t want to bother reinstalling the GPU backplate. Retrieve the backplate mounting standoffs from Step 3. Insert them from the rear of the PCB, and secure them from the front using M2.5 hex nuts. You can now reattach the backplate onto the PCB.
- Screw in the NZXT Kraken G12 mounting brackets onto the PCB using the provided spring-loaded screws as shown in the photograph.
Now would be a good time to clean the old thermal paste off with isopropyl alcohol and lint-free wipes or a microfibre cloth. Apply a small rice grain-sized drop of thermal paste if your AIO liquid cooler doesn’t have thermal paste pre-applied to the GPU block. 8. Twist the AIO liquid cooler GPU block into the correct position on the Kraken G12 mounting bracket. Depending on the brand and model of your AIO cooler, you might have to remove a few mounting brackets off its GPU block. Referring to the AIO cooler and NZXT Kraken G12 manuals should give you an idea. The Kraken G12 will have already come equipped with a 92mm VRM cooling fan.
Now would be a good time to attach the 4-pin PWM fan adapter for GPU. Split this fan header into two using Y-splitter PWM fan cable. This should allow the GPU to control the speed for both radiator as well as VRM cooling fans.
Align the four mounting holes around the AIO cooler’s GPU block on the Kraken G12 mounting bracket with the corresponding holes on the mounting bracket we installed onto the PCB in step 7. Place the entire assembly down on the GPU. The AIO liquid cooler’s GPU block will have made contact with the GPU die if you did this right.
Secure the assembly using four spring-loaded thumb screws around the GPU block. Do not tighten the thumb screws just yet. Only take up the slack for now. Moving diagonally from one screw to the next, gently tighten each screw with just your fingers; taking care not to apply an unequal amount of torque across the four screws. Keep tightening with your fingers the diagonally opposite screws until you can’t anymore. Use a screwdriver to further tighten the screws while taking the same diagonally opposite path.
Connect the 92mm VRM cooling fan to one of the fan headers on the Y-splitter PWM fan cable we installed in Step 9. Now would be a good time to use some cable ties to manage the cables in a more streamlined fashion. It also pays to doublecheck all fan connections while you still can at this stage. 11. Install the GPU into the designated PCI-E x16 slot on your motherboard, followed by installing the AIO cooler’s radiator and fan onto the PC chassis. Don’t forget to connect the radiator fan to the last remaining fan header on the Y-splitter PWM fan cable. Alternatively, you can connect the radiator and VRM fan to motherboard fan headers if you opted out of the GPU fan adapter route. Connect the GPU power cable, and you’re done.
Unlocking the True Potential of Your GPU
I was able to unlock a whopping 20 percent extra performance over the stock version of my Nvidia GTX 1070 Ti with the Kraken G12 AIO liquid cooling mod. What’s more impressive is how this is still 13.5 percent faster compared to the maximum achievable overclock without this mod. The GPU temperature doesn’t exceed 50°C even after several hours of gaming. That is precisely why my GPU boost clocks stayed well over 200MHz higher compared to the stock GPU and 180MHz higher compared to the overclock achievable without this mod. This can easily match (or even exceed in my case) the 10 to 15 percent performance improvement promised by upgrading to the new Turing GPUs. This isn’t bad at all for a mod that costs less than $100.
