Prioritization of component cooling between CPU and GPU.
Prioritization of component cooling between CPU and GPU.
I'm currently using a Nzxt Kraken X61 with the CPU at i54690k OC'd and a GTX 970. I'm considering adding another cooler, specifically a 120/140mm model with a G10 adapter, to the GPU. The GPU usually gets hotter, so attaching the 280mm rad to it seems logical. Does anyone have experience with this setup and know which cooler works best?
Due to their heat output (greater area for heat generation), liquid cooling for the GPU proves much more efficient for the money spent. If you wish to overclock the GPU, a CLC equipped with the G10 adapter will be essential. Also, your CPU, even without overclocking, is less likely to become a bottleneck than the GPU, meaning you might gain more practical performance by optimizing your GPU than by upgrading your CPU.
Even a modest radiator with water cooling on the GPU works well. For better heat management, opt for a larger radiator on the CPU, since even with water cooling and overclocking, temperatures can easily hit their limits. The 970 series has limited overclocking options due to VRM constraints, restricting it to around 1.3v with a BIOS upgrade, and even then only up to 1.256v with stock BIOS. Running a single 120W GPU will still struggle to reach 50°C+ on stock BIOS, especially with high room temperatures.
GPU usually has a higher TDP, often doubling that of a CPU in many cases. Its thermal limit is also greater, meaning a weak cooling solution on a GPU doesn't pose the same risks as one on a CPU.
Currently CPU's heat generation is reduced to half or less compared to typical GPUs, not even reaching Titan X levels. This trend could shift with upcoming 16nm GPU generations, but it seems unlikely given their massive transistor count of 8 billion, which would make achieving under 150w challenging.
Bloomfield features 731 million transistors with a 130w TDP, while Haswell offers 1.4 billion transistors and 88w TDP. A Titan X (8 billion transistors) and a GTX 480 (3 billion transistors) both share a 250w TDP. Simply adding more transistors doesn't always mean a higher TDP.
Specific heat intensity is also included in the formula. GPUs distribute their heat uniformly across the entire chip. CPUs usually have more concentrated heat sources, leading to higher cooling demands compared to a more evenly spread source. The intensity of the heat plays a crucial role in cooling efficiency, and because GPUs are less dense, cooling them offers more advantages than cooling a CPU of the same size.
The G10 bracket demonstrates this point. Even mediocre CPU coolers perform exceptionally well on GPUs when used with this bracket.
Look...this is quite straightforward to grasp.
Boost a GPU submerged, particularly a 970, without heavy hardware upgrades to raise the voltage significantly, and even with strong cooling, you won't always reach the necessary intensity.
You'll never push it hard enough to require a large radiator.
Overclock a CPU underwater, any CPU, with only minor BIOS tweaks to raise voltage, you can easily hit the thermal limits. Keeping things cooler becomes more crucial.
CPU = big rad
GPU = small rad
This is just simple loop design stuff...