Heatkiller V & Delta Mate MPII: Watercool and Thermal Grizzly Take On Four Other AM5 Water Blocks
The summer is hot, the CPU needs to stay cool, and two German manufacturers promise the perfect solution. We test their new water blocks alongside four proven models. This article is freely available from PC Games Hardware 08/2026.
Test Add as a preferred source on Google
In this article
Price developments in 2026 have shifted priorities: when new hardware remains unaffordable, existing components need to be used as effectively as possible. Good cooling is one part of that equation — and generally beneficial to a component's lifespan — which ideally means water cooling. Although the graphics card still produces most of the heat in a gaming system, CPU water blocks have their place as well. Overclockers need them because tuned AMD CPUs in particular run very hot and can benefit from every bit of additional cooling. Graphics enthusiasts can integrate the CPU into a modular loop with little extra effort if they are already building one for the GPU. For case modders, meanwhile, an elegant water block simply looks better than a high-end air cooler that has to cover the entire upper half of the motherboard to deliver comparable performance.
Test platform
When discussing "hot AMD CPUs", it is worth reiterating the distinction between a large amount of heat and a high temperature. The former is a quantity of energy converted into heat when electrical power is consumed. The latter is the result of that heat being removed poorly. A processor can therefore be efficient, as many Ryzen models are compared with similarly powerful Intel counterparts, yet still run hot if heat builds up inside it. This is a particular problem with AMD CPUs because they consist of two types of chips. The larger IOD handles all external interfaces, including the memory interface, and consumes relatively little power. The actual processing units sit on the CCD or the CCDs, with one being used up to eight cores and two CCDs for Ryzen 9 models. As much as 85% of the heat is produced here, concentrated within a tiny area at one end of the processor. This heat consequently transfers poorly from the silicon that generates it to the externally accessible heatspreader, resulting in high temperatures.
Because this bottleneck lies inside the CPU package, it persists regardless of the cooler being used, making AMD CPUs a second-choice platform for cooling tests. They convert less electrical power into heat, so the coolers have less work to do. The temperature difference between a strong and a weaker cooling solution is less pronounced on a 200 W PPT AMD CPU than on a 253 W MTP Intel counterpart. At the same time, the Ryzen's absolute temperatures are higher, meaning weak cooling solutions may be impossible to test because the processor starts throttling. But impressive results with Socket 1851 CPUs are of little use when everyone is looking for AM5 coolers. The asymmetrical heat distribution of these processors also requires different optimisation. We therefore bite the bullet and conduct our tests with a Ryzen 9 9950X.
Cold-Plate Details Compared
Unlike its 16-core predecessor, the 7950X used in our 2025 comparison, the 9950X operates at 200 W package power out of the box when fully loaded. It also transfers heat out of the package more effectively overall, or at least places its temperature sensors in cooler locations. Every returning contender records temperatures 12 kelvin lower than on the Ryzen 9 7950X, despite slightly poorer flow in the new, larger loop. We did ask AMD about the technological background for this observation, but have not received a reply before the editorial deadline of this article. Similar questions in the past stayed likewise unanswered. However, the cause makes no difference to throttling behaviour: maximum boost clock speeds are maintained up to a reported temperature of 95 °C. Whether the reading is lower because the silicon is cooler, because measuring is done at a cooler spot, or because the readout routines have been optimised, is irrelevant. The Ryzen 9 9950X has considerably more thermal headroom than the Ryzen 9 7950X before it reduces its clock speed.
8- versus 16-core temperatures
Ryzen 9 processors are good at exposing differences between coolers because of their high overall heat output, but AMD's top Ryzen 7 models are considered even harder to cool. The eight-core chips concentrate their processing units, and therefore a large proportion of their heat, on a single CCD. With only half the heat-transfer area, heat build-up is particularly severe. If, for example, a Ryzen 7 9800X3D were to reach its theoretical PPT of 162 W, it would produce a heat density more than 50 per cent higher than that of the Ryzen 9 9950X tested here. When we simulate this situation by applying an extremely one-sided 170 W load to the Ryzen 9, Tctl rises by 9 kelvin to an absolute 85 °C in an exemplary test with the Delta Mate MPII at 60 l/h. In games, however, even a Ryzen 7 9850X3D is unlikely to exceed 120 W package power. That corresponds almost exactly to the per-CCD heat output, and therefore the cooling demands, of our synthetically tortured Ryzen 9 9950X.
