AMD Ryzen 5 3600XT / Ryzen 7 3800XT / Ryzen 9 3900XT Linux Performance In 130+ Benchmarks

Written by Michael Larabel in Processors on 7 July 2020 at 09:00 AM EDT. Page 1 of 9. 23 Comments.

After the AMD Ryzen 3000XT series was announced last month, these new higher-clocked Zen 2 desktop processors are shipping today. Here are 130+ benchmarks on each of the Ryzen 5 3600XT, Ryzen 7 3800XT, and Ryzen 9 3900XT parts compared to various Intel and AMD CPUs. Tests under Ubuntu Linux and also complemented by performance-per-Watt / power and performance-per-dollar data points.

These "XT" processors were announced in mid-June as still being Zen 2 based like the rest of the Ryzen 3000 desktop line-up but with slight increases to the base and boost clock frequencies to ratchet up the competition on Intel's new Comet Lake processors.

AMD advertises the Ryzen 9 3900XT as offering around 4% better single-threaded performance with a 100MHz increase to the boost clock frequency. The Ryzen 9 3900XT is priced at $499 putting it right in line with the Core i9 10900K.

The Ryzen 7 3800XT meanwhile is the new $399 part with a 200Mhz boost clock frequency increase over the 3800X while still being 8 cores / 16 threads.

The Ryzen 5 3600XT meanwhile comes in at $249 USD and has a 100MHz boost clock bump over the Ryzen 5 3600X for this 6c/12t processor. The Ryzen 5 3600XT is viewed as the direct competitor to the Intel Core i5 10600K.

These "Matisse Refresh" processors otherwise are quite similar to the existing Ryzen 3000 parts. The new XT processors should be available in retail channels beginning today.

As AMD did supply these new 3600XT/3800XT/3900XT processors timely, we do have many benchmarks to share. In fact, 130+ benchmarks were run on each of the CPUs being tested in time for launch day. The CPUs benchmarked for today's article included:

- Ryzen 5 3600X
- Ryzen 7 3700X
- Ryzen 9 3900X
- Ryzen 9 3950X
- Ryzen 5 3600XT
- Ryzen 7 3800XT
- Ryzen 9 3900XT
- Core i5 10600K
- Core i9 10900K

On all of those processors, the different distinct workloads tested included:

- ACES DGEMM
- Appleseed
- BRL-CAD
- Basis Universal
- Blender
- Build2
- C-Blosc
- C-Ray
- Chaos Group V-RAY
- CloverLeaf
- Coremark
- Crafty
- Cryptsetup
- DaCapo Benchmark
- Darktable
- Darmstadt Automotive Parallel Heterogeneous Suite
- DeepSpeech
- Embree
- FFTW
- G'MIC
- GEGL
- GIMP
- GNU Octave Benchmark
- GROMACS
- Git
- Himeno Benchmark
- IndigoBench
- Inkscape
- Intel Open Image Denoise
- Java Gradle Build
- John The Ripper
- LAMMPS Molecular Dynamics Simulator
- LibreOffice
- N-Queens
- NAMD
- Nebular Empirical Analysis Tool
- Nettle
- Numenta Anomaly Benchmark
- Numpy Benchmark
- OSPray
- OpenVKL
- POV-Ray
- PyBench
- RawTherapee
- Rodinia
- SVT-AV1
- SVT-HEVC
- SVT-VP9
- Scikit-Learn
- Selenium
- Stockfish
- TTSIOD 3D Renderer
- Tachyon
- Timed FFmpeg Compilation
- Timed GCC Compilation
- Timed HMMer Search
- Timed ImageMagick Compilation
- Timed LLVM Compilation
- Timed Linux Kernel Compilation
- Timed MPlayer Compilation
- Timed PHP Compilation
- Tungsten Renderer
- WireGuard + Linux Networking Stack Stress Test
- XZ Compression
- YafaRay
- Zstd Compression
- asmFish
- ctx_clock
- dav1d
- libavif avifenc
- m-queens
- oneDNN
- rays1bench
- x264

With varying test combinations, this in turn led to 134 benchmarks being run each time. All of the testing, of course, was made possible via the Phoronix Test Suite. Along with the raw performance metrics are many additional data points.


Related Articles