Test results:
This is the part that will interest us most, allowing us to evaluate the performance of this ultra-high-end configuration from AMD. As a reminder, our copy of the Ryzen Treadripper 7980X is equipped with 64 cores/128 threads, and the Treadripper 7970X with 32 cores/64 threads. We’ve gone back to our usual tests and added the performance of this processor at stock, i.e. without any overclocking.
Before getting started, it’s worth noting that the card takes an extremely long time to start up for the first time! For a moment, we even thought we’d made a bad assembly, but we just had to be patient. The same applies to CMOS clear. The loading time can take from 4 to 5 minutes. Once the BIOS parameters have been set, entry into the OS takes just ten seconds.
Reached frequencies and core activity?
The idea is to see if the frequencies announced by AMD are indeed reached, but also and above all if, during a multithreaded task, all the cores are in action. We can see that in a Cinebench R23 benchmark, the 64 cores of our 7980X are at 100% capacity.
As far as temperatures are concerned, they’re pretty good for a processor with so many cores. It remains to be seen whether we’ll be able to manage them when overclocking.
Cinebench multi-core in R20, R23 and R24 :
Let’s start with a CPU benchmark staple, the Cinebench series. We’ve opted for Cinebench R20, R23 and the latest, R24. Scores will be given in single-core and multi-core. By this we mean that the benchmark tests performance on a single processor core and then on all available cores.
To do this, we’re going to use Benchmate, which has just been updated to version 11.3.0 and can be downloaded here. The advantage of Benchmate is that it already contains a whole series of benchmarks, and what’s more, it’s recognized and certified when you want to encode your results on the Hwbot site.
Well, as you might have guessed, it hits high and performs very hard. With our Threadripper 7970X, the multi score is 25646 points and 37653 points with the Threadripper 7980X. By comparison, the 56-core Xeon W9-3495X scored 26826 points. So we’re well ahead of the pack, with the 7970X doing almost as well as Intel’s Xeon.
As regards Cinebench R23, the Threadripper 7970X scores 65710 points and the Threadripper 7980X 98427 points. In comparison, the 56-core Xeon W9-3495X scored 71115 points. If we make a rule of four to look at the number of points per core, the advantage goes to AMD!
Finally, we put the two Ryzen Threadrippers of the day through their paces on the latest Cinebench benchmark, the R24. Our 7970X scores 3710 points and the incredible 7980X a total score of 5862 points. We have no comparative scores with the Xeon, as the benchmark had not yet been released.
Geekbench multi core 3.4.4 and 6 :
This is a benchmark available in several revisions. Version 5 has been increasingly used in recent months, and is often used for CPU performance leaks. It provides two scores: one single-core and one multi-core. The version used for testing is 3.4.4 and 6, as we were unable to run Geekbench 5. Please note that to take full advantage of these three benchmarks, a license is required to run the 64-bit benchmark.
Let’s start with the Geekbench 3 score of our Threadripper 7970X, which scores 8902 points in single and 188923 in multi. For Geekbench 6, it’s 3068 in single and 25135 points in multi.
The Geekbench 3 score for the Threadripper 7980X is 8411 points single (6587 for the W9-3495X) and 288461 multi (206282 for the W9-3495X). On Geekbench 6, it’s 3023 in single mode (2238 for the W9-3495X) and 26272 in multi mode (16258 for the W9-3495X). In other words, AMD is crushing the competition on this type of processor.
Y-Cruncher benchmark :
A new benchmark we’ve just added, as it’s increasingly used by brands to showcase processor performance. To use it, simply download the latest version of CPU-Z, currently 2.0.2. Then go to the “Bench” tab to check single-thread and multi-thread performance. The benchmark version we used is 17.01.64. We also took the opportunity to test Y-cruncher.
We haven’t actually done any application testing, since unfortunately we don’t have any points of comparison. What’s more, some benchmarks require licenses that we don’t have. We plan to update our CPU test protocol at the beginning of 2024, and we’ll be adding a few more tests then.
