3600x 2024 AMD Ryzen 9 3900XT 12-Core testing with a MSI MEG X570 GODLIKE (MS-7C34) v1.0 (1.B3 BIOS) and AMD Radeon RX 56/64 8GB on Ubuntu 22.04 via the Phoronix Test Suite. a: Processor: AMD Ryzen 9 3900XT 12-Core @ 3.80GHz (12 Cores / 24 Threads), Motherboard: MSI MEG X570 GODLIKE (MS-7C34) v1.0 (1.B3 BIOS), Chipset: AMD Starship/Matisse, Memory: 2 x 8GB DDR4-4133MT/s F4-4133C19-8GTZKW, Disk: 500GB Seagate FireCuda 520 SSD ZP500GM30002, Graphics: AMD Radeon RX 56/64 8GB, Audio: AMD Vega 10 HDMI Audio, Monitor: DELL U2723QE, Network: Realtek Device 2600 + Realtek Killer E3000 2.5GbE + Intel Wi-Fi 6 AX200 OS: Ubuntu 22.04, Kernel: 6.8.0-47-generic (x86_64), Desktop: GNOME Shell 42.9, Display Server: X Server + Wayland, OpenGL: 4.6 Mesa 22.0.1 (LLVM 13.0.1 DRM 3.57), Vulkan: 1.3.204, Compiler: GCC 11.4.0, File-System: ext4, Screen Resolution: 3840x2160 b: Processor: AMD Ryzen 9 3900XT 12-Core @ 3.80GHz (12 Cores / 24 Threads), Motherboard: MSI MEG X570 GODLIKE (MS-7C34) v1.0 (1.B3 BIOS), Chipset: AMD Starship/Matisse, Memory: 2 x 8GB DDR4-4133MT/s F4-4133C19-8GTZKW, Disk: 500GB Seagate FireCuda 520 SSD ZP500GM30002, Graphics: AMD Radeon RX 56/64 8GB, Audio: AMD Vega 10 HDMI Audio, Monitor: DELL U2723QE, Network: Realtek Device 2600 + Realtek Killer E3000 2.5GbE + Intel Wi-Fi 6 AX200 OS: Ubuntu 22.04, Kernel: 6.8.0-47-generic (x86_64), Desktop: GNOME Shell 42.9, Display Server: X Server + Wayland, OpenGL: 4.6 Mesa 22.0.1 (LLVM 13.0.1 DRM 3.57), Vulkan: 1.3.204, Compiler: GCC 11.4.0, File-System: ext4, Screen Resolution: 3840x2160 c: Processor: AMD Ryzen 9 3900XT 12-Core @ 3.80GHz (12 Cores / 24 Threads), Motherboard: MSI MEG X570 GODLIKE (MS-7C34) v1.0 (1.B3 BIOS), Chipset: AMD Starship/Matisse, Memory: 2 x 8GB DDR4-4133MT/s F4-4133C19-8GTZKW, Disk: 500GB Seagate FireCuda 520 SSD ZP500GM30002, Graphics: AMD Radeon RX 56/64 8GB, Audio: AMD Vega 10 HDMI Audio, Monitor: DELL U2723QE, Network: Realtek Device 2600 + Realtek Killer E3000 2.5GbE + Intel Wi-Fi 6 AX200 OS: Ubuntu 22.04, Kernel: 6.8.0-47-generic (x86_64), Desktop: GNOME Shell 42.9, Display Server: X Server + Wayland, OpenGL: 4.6 Mesa 22.0.1 (LLVM 13.0.1 DRM 3.57), Vulkan: 1.3.204, Compiler: GCC 11.4.0, File-System: ext4, Screen Resolution: 3840x2160 Glibc Benchmarks 2.39 Benchmark: cos ns < Lower Is Better a . 81.31 |==================================================================== b . 78.44 |================================================================== c . 81.37 |==================================================================== Glibc Benchmarks 2.39 Benchmark: exp ns < Lower Is Better a . 16.77 |=========================================================== b . 16.99 |============================================================ c . 19.23 |==================================================================== Glibc Benchmarks 2.39 Benchmark: ffs ns < Lower Is Better a . 6.03612 |================================================================= b . 6.08496 |================================================================== c . 6.07068 |================================================================== Glibc Benchmarks 2.39 Benchmark: pow ns < Lower Is Better a . 42.70 |================================================================= b . 44.64 |==================================================================== c . 44.64 |==================================================================== Glibc Benchmarks 2.39 Benchmark: sin ns < Lower Is Better a . 71.89 |==================================================================== b . 71.04 |=================================================================== c . 71.02 |=================================================================== Glibc Benchmarks 2.39 Benchmark: log2 ns < Lower Is Better a . 13.46 |==================================================================== b . 13.50 |==================================================================== c . 12.92 |================================================================= FurMark 2.1 Resolution: 1920 x 1080 - Demo: Furmark Knot Vulkan - MSAA: Off FPS > Higher Is Better a . 15.96 |==================================================================== FurMark 2.1 Resolution: 1920 x 1080 - Demo: FurMark Knot OpenGL - MSAA: Off FPS > Higher Is Better a . 11.25 |==================================================================== Glibc Benchmarks 2.39 Benchmark: modf ns < Lower Is Better a . 6.69978 |================================================================ b . 6.55046 |=============================================================== c . 6.89904 |================================================================== Glibc Benchmarks 2.39 Benchmark: sinh ns < Lower Is Better a . 29.16 |=================================================================== b . 28.16 |================================================================= c . 29.45 |==================================================================== Glibc Benchmarks 2.39 Benchmark: sqrt ns < Lower Is Better a . 10.38790 |================================================================= b . 7.84162 |================================================= c . 10.32780 |================================================================= Glibc Benchmarks 2.39 Benchmark: tanh ns < Lower Is Better a . 39.62 |=================================================================== b . 40.09 |==================================================================== c . 39.61 |=================================================================== Glibc Benchmarks 2.39 Benchmark: asinh ns < Lower Is Better a . 31.79 |==================================================================== b . 31.53 |=================================================================== c . 31.79 |==================================================================== Glibc Benchmarks 2.39 Benchmark: atanh ns < Lower Is Better a . 36.99 |==================================================================== b . 36.99 |==================================================================== c . 35.79 |================================================================== Glibc Benchmarks 2.39 Benchmark: ffsll ns < Lower Is Better a . 6.06866 |================================================================== b . 6.03505 |================================================================== c . 6.06880 |================================================================== Glibc Benchmarks 2.39 Benchmark: sincos ns < Lower Is Better a . 41.18 |=================================================================== b . 41.98 |==================================================================== c . 41.20 |=================================================================== Glibc Benchmarks 2.39 Benchmark: pthread_once ns < Lower Is Better a . 6.03187 |================================================================== b . 6.03169 |================================================================== c . 5.78067 |=============================================================== LeelaChessZero 0.31.1 Backend: BLAS Nodes Per Second > Higher Is Better a . 63 |======================================================================= XNNPACK b7b048 Model: FP32MobileNetV1 us < Lower Is Better a . 2454 |==================================================================== b . 2502 |===================================================================== c . 2430 |=================================================================== XNNPACK b7b048 Model: FP32MobileNetV2 us < Lower Is Better a . 2263 |===================================================================== b . 2220 |==================================================================== c . 2206 |=================================================================== XNNPACK b7b048 Model: FP32MobileNetV3Large us < Lower Is Better a . 2505 |===================================================================== b . 2512 |===================================================================== c . 2479 |==================================================================== XNNPACK b7b048 Model: FP32MobileNetV3Small us < Lower Is Better a . 1183 |==================================================================== b . 1194 |===================================================================== c . 1189 |===================================================================== XNNPACK b7b048 Model: FP16MobileNetV1 us < Lower Is Better a . 2493 |===================================================================== b . 2485 |===================================================================== c . 2491 |===================================================================== XNNPACK b7b048 Model: FP16MobileNetV2 us < Lower Is Better a . 1915 |===================================================================== b . 1911 |===================================================================== c . 1912 |===================================================================== XNNPACK b7b048 Model: FP16MobileNetV3Large us < Lower Is Better a . 2178 |===================================================================== b . 2174 |===================================================================== c . 2170 |===================================================================== XNNPACK b7b048 Model: FP16MobileNetV3Small us < Lower Is Better a . 1084 |===================================================================== b . 1082 |===================================================================== c . 1082 |===================================================================== XNNPACK b7b048 Model: QS8MobileNetV2 us < Lower Is Better a . 1377 |===================================================================== b . 1371 |===================================================================== c . 1371 |===================================================================== LiteRT 2024-10-15 Model: Inception ResNet V2 Microseconds < Lower Is Better a . 32634.7 |================================================================= b . 32941.0 |================================================================== c . 32820.2 |================================================================== LiteRT 2024-10-15 Model: Inception V4 Microseconds < Lower Is Better a . 38445.9 |================================================================= b . 38646.1 |================================================================= c . 39144.4 |================================================================== LiteRT 2024-10-15 Model: Mobilenet Float Microseconds < Lower Is Better a . 2282.72 |================================================================== b . 2239.60 |================================================================= c . 2259.66 |================================================================= FurMark 2.1 Resolution: 1920 x 1080 - Demo: FurMark Vulkan - MSAA: Off FPS > Higher Is Better a . 38.17 |==================================================================== NAMD 3.0 Input: ATPase with 327,506 Atoms ns/day > Higher Is Better a . 0.77506 |================================================================= b . 0.78593 |================================================================== c . 0.78383 |================================================================== LiteRT 2024-10-15 Model: DeepLab V3 Microseconds < Lower Is Better a . 3688.37 |================================================================== b . 3653.79 |================================================================= c . 3629.77 |================================================================= oneDNN 3.6 Harness: IP Shapes 1D - Engine: CPU ms < Lower Is Better a . 2.55382 |=============================================================== b . 2.66658 |================================================================== c . 2.62929 |================================================================= oneDNN 3.6 Harness: Deconvolution Batch shapes_3d - Engine: CPU ms < Lower Is Better a . 5.33741 |================================================================== b . 5.36237 |================================================================== c . 5.35165 |================================================================== oneDNN 3.6 Harness: Recurrent Neural Network Inference - Engine: CPU ms < Lower Is Better a . 1370.26 |================================================================== b . 1364.38 |================================================================== c . 1364.68 |================================================================== C-Blosc 2.11 Test: blosclz shuffle - Buffer Size: 8MB MB/s > Higher Is Better a . 11575.0 |================================================================== b . 11536.1 |================================================================== c . 11542.8 |================================================================== C-Blosc 2.11 Test: blosclz shuffle - Buffer Size: 16MB MB/s > Higher Is Better a . 11523.1 |================================================================== b . 11253.3 |================================================================ c . 11379.4 |================================================================= C-Blosc 2.11 Test: blosclz shuffle - Buffer Size: 32MB MB/s > Higher Is Better a . 10369.6 |================================================================= b . 10252.8 |================================================================ c . 10512.1 |================================================================== C-Blosc 2.11 Test: blosclz shuffle - Buffer Size: 64MB MB/s > Higher Is Better a . 8550.6 |=================================================================== b . 8530.4 |=================================================================== c . 8536.4 |=================================================================== C-Blosc 2.11 Test: blosclz noshuffle - Buffer Size: 8MB MB/s > Higher Is Better a . 9500.0 |================================================================= b . 9622.5 |================================================================== c . 9750.8 |=================================================================== C-Blosc 2.11 Test: blosclz shuffle - Buffer Size: 128MB MB/s > Higher Is Better a . 6896.5 |=================================================================== b . 6893.0 |=================================================================== c . 6900.4 |=================================================================== C-Blosc 2.11 Test: blosclz bitshuffle - Buffer Size: 8MB MB/s > Higher Is Better a . 10444.6 |================================================================== b . 10497.1 |================================================================== c . 10361.2 |================================================================= LiteRT 2024-10-15 Model: SqueezeNet Microseconds < Lower Is Better a . 2548.21 |================================================================== b . 2553.50 |================================================================== c . 2546.22 |================================================================== oneDNN 3.6 Harness: IP Shapes 3D - Engine: CPU ms < Lower Is Better a . 15.94 |==================================================================== b . 12.07 |=================================================== c . 12.23 |==================================================== LiteRT 2024-10-15 Model: NASNet Mobile Microseconds < Lower Is Better a . 12847.8 |================================================================== c . 12877.1 |================================================================== oneDNN 3.6 Harness: Convolution Batch Shapes Auto - Engine: CPU ms < Lower Is Better a . 19.00 |==================================================================== b . 18.91 |==================================================================== c . 18.85 |=================================================================== LiteRT 2024-10-15 Model: Mobilenet Quant Microseconds < Lower Is Better a . 1698.80 |================================================================== b . 1689.17 |================================================================== c . 1686.52 |================================================================== oneDNN 3.6 Harness: Deconvolution Batch shapes_1d - Engine: CPU ms < Lower Is Better a . 7.03818 |================================================================== b . 5.96129 |======================================================== c . 6.03743 |========================================================= LiteRT 2024-10-15 Model: Quantized COCO SSD MobileNet v1 Microseconds < Lower Is Better a . 2646.99 |================================================================== b . 2652.69 |================================================================== c . 2648.53 |================================================================== oneDNN 3.6 Harness: Recurrent Neural Network Training - Engine: CPU ms < Lower Is Better a . 2654.70 |================================================================== b . 2666.16 |================================================================== c . 2651.48 |================================================================== C-Blosc 2.11 Test: blosclz noshuffle - Buffer Size: 16MB MB/s > Higher Is Better a . 8745.0 |================================================================= b . 8834.7 |================================================================== c . 8993.3 |=================================================================== C-Blosc 2.11 Test: blosclz noshuffle - Buffer Size: 32MB MB/s > Higher Is Better a . 6738.3 |================================================================== b . 6814.6 |=================================================================== c . 6865.3 |=================================================================== C-Blosc 2.11 Test: blosclz noshuffle - Buffer Size: 64MB MB/s > Higher Is Better a . 5462.1 |=================================================================== b . 5424.4 |=================================================================== c . 5439.3 |=================================================================== C-Blosc 2.11 Test: blosclz bitshuffle - Buffer Size: 16MB MB/s > Higher Is Better a . 10404.9 |================================================================= b . 10549.8 |================================================================== c . 10414.3 |================================================================= C-Blosc 2.11 Test: blosclz bitshuffle - Buffer Size: 32MB MB/s > Higher Is Better a . 9784.0 |=================================================================== b . 9540.4 |================================================================= c . 9643.8 |================================================================== FurMark 2.1 Resolution: 1920 x 1080 - Demo: FurMark OpenGL - MSAA: Off FPS > Higher Is Better a . 42.12 |==================================================================== b . 41.92 |==================================================================== C-Blosc 2.11 Test: blosclz bitshuffle - Buffer Size: 64MB MB/s > Higher Is Better a . 8054.3 |=============================================================== b . 8517.2 |=================================================================== c . 8352.2 |================================================================== C-Blosc 2.11 Test: blosclz noshuffle - Buffer Size: 128MB MB/s > Higher Is Better a . 4695.4 |=================================================================== b . 4696.9 |=================================================================== c . 4715.0 |=================================================================== C-Blosc 2.11 Test: blosclz bitshuffle - Buffer Size: 128MB MB/s > Higher Is Better a . 6703.8 |=================================================================== b . 6587.4 |================================================================== c . 6615.2 |==================================================================