CPU performance tests
For a direct comparison of pure CPU performance, we added five other candidates, as well as running the FX-8370E at the deliberate three clock rates. This results in 8 different constellations, even though the Intel Cire i7 4790K naturally plays in a different league in terms of price. But we also wanted to be fair enough to compare 8 threads Intel with 8 threads AMD. Let’s now start with what AMD doesn’t really like: the IPC. What happens if you can or want to use only a single thread? Cinebench gives us a pretty good overview:
We can see very clearly that the CPU really only clocks at a reduced rate and that the performance scales pretty much exactly with what AMD already sold as FX. At 4.2 GHz, everything is then pretty much the same again and the FX-8370 without E is almost caught up. However, the difference to Intel’s Haswell CPUs is already frightening, so it can actually only go upwards from now on. The APU also shows the FX-3870E the rear lights, even if only just.
If, on the other hand, we take all eight threads as a guide, the world suddenly looks a lot rosier. The direct comparison between the Core i7-4790K and the FX-3870E in all three clock speeds does not improve the ratio, but the candidates with only four threads are clearly at a disadvantage:
Of course, such very well parallelizable tasks really invite us to watch the CPUs with more than four threads, so that we also grant the FX as such a success as a number cruncher as a temporary highlight:
. However, the euphoria quickly drops a bit when it comes to real-world tasks. At least the FX-8370 and the variant overclocked to 4.2 GHz can be attested to not having lost sight of the Core i5-4690K. Thus, an FX is still comparatively well suited for such purposes.
Of course, we could add more synthetic and application benchmarks, but these won’t change the core statement, especially since we have written our fingers to the bone about the FX for years and the architecture is still the same. The FX-8370E is only worthwhile in CPU-intensive applications when there are many tasks that can be parallelized well and the software can also use the advantage of as many threads as possible sensibly.
However, it is never overtaxed in the daily application environment as long as it does not depend on computationally intensive tasks that are reduced to one thread. Then – at least in comparison to the competitors – things can get tight.
18 Antworten
Kommentar
Lade neue Kommentare
Veteran
Urgestein
Urgestein
Urgestein
Urgestein
Mitglied
Urgestein
Urgestein
Urgestein
Mitglied
Urgestein
Neuling
Mitglied
Mitglied
1
Mitglied
Mitglied
Mitglied
Alle Kommentare lesen unter igor´sLAB Community →