PCCooler CPS RZ820: Against the elite, weighing over 1.8 kg

PCCooler CPS RZ820 in detail

They were so emphatic about their cooler being optimized for Intel LGA 1700 processors that we had to try it out. And given that the CPS RZ820 is a “hefty chunk” of material, the reference sample selection included the kind of competitive coolers that are at the top of the tower model database. These should give the PCCooler CPS RZ820 a good run for its money. At the same time, this cooler should not be more expensive.

Before we turn to the new methodology for testing coolers built on the Intel LGA 1851 platform, let’s do some more testing on Intel LGA 1700, specifically on the Intel Core i9-13900K processor. The shape of its IHS and the layout of hotspots are different compared to processors that do not combine small and large cores (big.LITTLE).

And coolers optimized for specific conditions can thus perform better than in cases where the manufacturer has not tried to achieve the best possible contact. This is typically on platforms that are supported (and the cooler can be installed on them), but the base of the cooler doesn’t “fit” as well on the CPU heatspreader. Behind this you can imagine various surface treatments that are supposed to lead to absorbing as much heat as possible, resulting in, of course, lower CPU temperatures than with “inferior” coolers.

The PCCooler CPS RZ820 Black is supposed to be one of such platform-optimized (LGA 1700) coolers. Its manufacturer vehemently pointed this out with the lower-end CPS RZ620 as well. The latter achieved very good results already on the Core i9-11900K (LGA 1200) on which this cooler was not optimized, especially in relation to the relatively lower price (compared to competing models with comparable cooling performance). The development of the even more robust RZ820-BK cooler was similarly approached in this regard.

We pitted the PCCooler’s top tower cooler against the BeQuiet! Dark Rock Elite, Cooler Master MasterAir MA824 Stealth, DeepCool Assassin IV and Noctua NH-D15 (G1). The first three of the compared coolers took the top spots in our tests. The older Noctua model ends up in “fourth place”, but again, it is the most widely used among users. We have already covered all the coolers in the comparison in separate articles. The quickest way to get to them is to click on their respective links in the first sentence of this paragraph.

We’ll be covering the Noctua NH-D15 G2 HBC cooler in tests later on, but it’s possible that it will already be on the Intel LGA 1851 platform. But it will be in comparison with the PCCooler CPS RZ820 cooler and probably with the other models from this selection as well. At the moment, we’d rather not promise anything due to the large time commitment of the new, permanent test methodology built on Intel LGA 1851 and AMD AM5 platforms. We’ll see…

PCCooler CPS RZ820 in detail

First of all, the very high weight, the first thing to highlight with the RZ820 cooler. While the Cooler Master MasterAir MA824 as the second heaviest of the tested coolers has 1460 grams, the RZ820 has almost 400 grams more (about 1840 grams in total). The Noctua NH-D15 G1 is “only” about 1320 grams, and the old heads will also remember the Thermalright Silver Arrow SB-E (Extreme), which was a top-of-the-line dual-tower cooler that fit into one kilogram even with fans.

The RZ820 cooler is almost twice as heavy. That’s just to illustrate what you’re dealing with. The backplate is otherwise metal though, with molded edges, so prevention against flexing of the motherboard or socket under the heft is good here.

The total weight is also made up of elements that are not directly involved in cooling. For example, just the aesthetic roof with illumination is one hundred grams. In the upper left corner (that is, if you have the cooler installed vertically) is the CPS logo. This has a fixed orange color that is strictly given, and you can only adjust the color or effect of the ARGB LED behind the triangular light pipe (with a mosaic of small squares) in the bottom left corner.

But the structure consists mainly of two towers with massive bundles of aluminum fins. In height, the cooler stands at 165 mm, which is still acceptable in terms of compatibility with cases, and stretches more in width. In this axis it is up to 160 mm.

Compatibility with higher RAM modules is solved partly by removing part of the fins that are above the first two DIMM slots and partly by the possibility of offsetting the front fan. Memory modules up to 45 mm in height can fit underneath it in the default position. In two steps, one centimeter each, you can also raise it, but this changes the overall height to 175 mm and 185 mm respectively, which many cases will not be able to cope with anymore. Anyway, there is also the option to install possibly 70 mm memory under the fan. And actually have such tall modules in all four slots.

In terms of cooling performance, the RZ820 should benefit from use on the Intel LGA 1700 platform, for which the cooler base is optimized. What exactly this means we can’t tell you, because the manufacturer tries to keep some “know-how” to themselves, but if there is really something to it, it is clear that it is supposed to provide better contact with the surface of the processor IHS and apparently also in the places where there are the greatest requirements for heat flux. Thus, faster heat exchange could be ensured in these locations, which may eventually expand the bottleneck a bit. Theoretically.

Measuring anything in terms of flatness, what is convex, where and how much convex, doesn’t make much sense, we can’t draw any conclusions from it, but a thermal paste imprint might be useful to someone for research purposes.

Then there is the separate heatsink, which is on the other side of the base, on the top, which is remarkable in connection with the base. This used to be seen quite often on older coolers, but it has gradually disappeared. PCCooler considers it relevant, though. And BeQuiet! too (ion the Dark Pro Elite cooler).

The fins of the heatsink are straight in some places, wavy in others. For example, they are straight on the inside, around the 140mm fan in the middle, and wavy on the outside. At the back and on the other side, opposite the 120mm inlet fan. Whether some psychoacoustic optimization (by breaking up the tonal peaks) is being pursued with this, or whether it’s mainly a matter of appearance, it’s hard to say. However, it is possible that it is a combination of both.

You know the front fan from the analysis of the F5 R120 BK, it’s the same model with a speed range of 500–2200 rpm. The only difference is the atypical mounting (via rails) and therefore more complicated replacement in case of its failure. Otherwise, though, it will be the strength of the cooler, we know this from the top-notch results in separate tests.

Atypical is also the frame of the middle 140 mm fan and actually its (de)mounting as well. It can be accessed by removing the roof. It is held on magnets at different points so that the mounting can be described as stable. The roof would probably stay in place even during more aggressive transport of the computer build, which, of course, should not be installed on the motherboard at all in such conditions.

Once the roof is removed, the 140-millimeter fan is removed by pulling its recessed brackets from the top of the frame. It’s going to be a bit stiff, but there’s no need to worry about it. The fan intentionally puts up more resistance when removed, as the process is hindered by the larger rubber inserts on the sides. These are there to keep the fan well in place since it is not fixed in place, as is typically the case with traditional clips.

Testing methodology

We test the coolers in seven modes, six of which are normalized to comparable noise levels (31, 33, 36, 39, 42 and 45 dBA) with a close-up measurement technique for the highest possible resolution, in a manner familiar from our graphics card test methodology. The noise meter sensor (Reed R8080) is approximately 20 centimeters from the cooler, to the side (at a 30-degree angle). Finally, we also test how the coolers handle the Intel Core i9-13900K processor at maximum fan speed. In this mode, the noise level is naturally different for each cooler.

The test processor (on the Asus ROG Strix Z790-E Gaming WiFi motherboard, by the way) is tuned to always achieve the same power consumption (about 250W, 5.0GHz on the P cores and with 4.0GHz on the E cores) under the specified load. Consumption stability is checked during tests by current clamps directly on the power cables (EPS). It does not change even in quiet modes where the available cooling performance is lower. The thermal paste used is always the one that comes with the accessories, i.e. the one with which the particular cooler is shipped. In the case of the RZ820, it’s the PCCooler EX90 paste. The application of the thermal paste is a “cross”. This technique has also proven successful for coolers with coldplates in tests of various thermal paste application techniques. In these, we also tested on a processor with a relatively larger IHS, as it is also on the Intel LGA 1700 test platform.

Cinebench R23 is used to load the processor, that is, a 10-minute test. The measurements take place in a wind tunnel with four Noctua NF-S12A PWM fans in an equilibrium ratio of two (input) to two (output) at about 500 rpm. The air temperature is constant, controlled at the entrance to the tunnel. It can never exceed a narrow range of 21.0–21.1 °C.

In addition to the average of the maximum temperatures of the P and E cores, we also record the temperature around the socket in the VRM region. One sensor is labeled “VRM”, the other “VR VCC”. The CPU cooler also has an impact on cooling these components, and better results in these areas can be a good added value. In critical cases, typically on cheaper motherboards, CPU performance is shaped exactly by the heating or rather overheating of the VRM. When the temperature limit is exceeded, that’s when the multiplier (CPU) drops down massively.

• Contents
• PCCooler CPS RZ820 in detail
• Results: Maximum performance
• Results: Higher performance (45 dBA)
• Results: Medium performance (42 dBA)
• Results: Lower noise level (39 dBA)
• Results: Low noise level (36 dBA)
• Results: Very low noise level (33 dBA)
• Results: Audibility threshold (31 dBA)
• Conclusion

Flattr this!