A small leak of information from GeekBench 6, whose database highlights a result featuring a duo of Sierra Forest CPUs. The only problem is that they are currently unable to compete with AMD’s offering. For the time being, Intel CPUs are half as efficient as their rivals!
Sierra Forest: encouraging, but not yet up to scratch!
The score in question highlights a configuration featuring two Intel Sierra Forest CPUs. As a reminder, these processors are equipped with so-called “efficient” cores. These benefit from a simplified architecture that greatly reduces power consumption and size. This makes it possible to install a very large number of cores in a CPU. In this case, we’re talking about a dual-CPU configuration, each with 144 cores (288 in total)!
In terms of performance, the software boasts scores of :
- 855 pts single-core
- 7,770 pts multi-core
The first thing we notice is the low single-core score. This is quite logical, since E-Core-equipped CPUs are not the most violent in single-threaded mode. What’s more, in a server environment, this kind of usage doesn’t make sense.
The multi-core score, on the other hand, is far more interesting, as it easily rivals a configuration equipped with two Xeon Platinum 8480 processors with 56 cores and 112 threads running at up to 3.8 GHz boost. What’s more, we’re talking about Golden Cove P-Cores here. For your information, this configuration is rated at 6,500 pts to 7,500 pts in multi.
Nevertheless, it’s still not enough to compete with AMD’s Epyc 9754. This is a CPU with no fewer than 128 cores for 256 threads. However, with this processor, AMD follows the same logic as Intel, equipping it only with ” E-Cores ” (AMD sauce, of course), in this case Zen 4Cs. In its current state, AMD’s monster boasts 1,597 pts in single-core mode versus 16,455 pts in multi-core mode, i.e. more than double.
Now, let’s put that into perspective, since Sierra Forest is still in the development stage. As a reminder, these processors are due to see the light of day in the first half of 2024, so it’s not impossible that scores will evolve positively between now and then… But achieving 100% more performance seems hardly feasible.
The other explanation, as indicated by our colleagues, is that Geekbench 6 does not take advantage of the instruction/function sets of these CPUs. In other words, GeekBench represents a field of use that does not correspond to the CPU, and this is also a valid hypothesis.
