Just last week Intel revealed its Cryo Cooling Technology with EKWB, which takes thermo-electric heat transfer tech and turns it into a real-world product. Shortly after Cooler Master followed suit with an AIO-type solution, but thus far, the products have only had support for Intel sockets. This hasn't changed, but many of us were wondering what would happen if you installed this tech onto an AMD chip. So, you can guess who found his call to action: the one and only Der8auer. He tested the cooler on AMD's latest and greatest, the 16-core Ryzen 9 5950X.
Of course, installing the EKWB EK-QuantumX Delta TEC onto the AMD chip wasn't quite straight forward for two reasons: socket compatibility and software control. Der8auer gathered some bits and pieces from other coolers in an attempt to improvise a mounting mechanism that adapts the AM4 AMD socket to accommodate the Intel-socketed cooler, which for the most part resulted in a relatively neat fitment.
Then, because the Intel control software won't run on the AMD system, der8auer set up a second rig using an Intel 10900K-based system to control the software of the Intel Cryo cooling tech.
Software control is required for this device for a couple of reasons, but mainly to ensure condensation doesn't become a problem. When the system is idle, the TEC unit isn't fed as much power, but when the CPU is running at full blast, the software tells the TEC unit to cool as hard as it can in an attempt to keep the chip at sub-ambient temperatures. Fortunately, the unit has an 'unregulated mode' that allows it to override all the safeties.
5950X Cooled by Intel, Overclocking Results
Under single-core loads, with the TEC unit disabled the 5950X saw CCD temperatures of up to 90 °C under single-core Cinebench R15 loads. Enable the TEC, and thermals dropped to about 50 °C.
In-game, the 5950X pulled speeds of up to 5.050 Ghz for single-cores on its automatic boost profile, with the entire chip often boosting to 4.8 - 4.9 GHz on all cores at once (with variances between the CCDs).
Unregulated, with no load the TEC unit dropped the chip's thermals to as low as 2 °C. Without a load, Der8auer was able to push the chip to a staggering 5323 MHz on all cores, when the system promtly crashed.
For comparison, Der8auer used the Cryo Cooler to push the i9-10900K to 5.8 GHz for validation and about 5.6 GHz for single-threaded Cinebench workloads, which are higher clocks. But, that chip has six fewer cores and lower IPC compared to AMD's chip, so the Ryzen 9 5950X is for all intents and purposes faster than the Intel chip, also when using Intel cooling tech.
And an i9-10900K's is only good enough to be used as a software controller for Zen3 cpus ;)
However few apps are highly parallel and few cpus are used to run just one app at a time. So the benchmark of most importance is an Algorithm test. In other words a simulation of real-world processing.
And, believe it or not, the results aren't dependent on thread count pr even IPC or frequency. It's all about cpu architecture. I have seen this myself when a 16c/32t cpu matched a 16c/128t cpu although the 128t cpu had faster single and multi thread load performance.
The 32t thread cpu would be regarded as faster as it uses less resources to get the job done.
So using IPC and Hz etc as done in the gaming world as proof is a slippery slope.
The gaming benchmarks I have seen show Zen3 is top at 1080p but as the resolution increases it's intel that has the edge. Could be wrong.
My bet is that the architecture improvements with 11th gen will see it widen the gap.
But I just laugh at the debates about which cpu is faster. It's which cpu is better for you that's important.
Lastly, do not rely on any benchmark that does not involve the entire pc.
This really shows desperation. Intel as of now has nothing to compete with AMD, so they just are getting desperate.