epyc turin Benchmarks for a future article. 2 x AMD EPYC 9755 128-Core testing with a AMD VOLCANO (RVOT1001B BIOS) and ASPEED on Ubuntu 24.04 via the Phoronix Test Suite. a: Processor: 2 x AMD EPYC 9755 128-Core @ 2.70GHz (256 Cores / 512 Threads), Motherboard: AMD VOLCANO (RVOT1001B BIOS), Chipset: AMD Device 153a, Memory: 1520GB, Disk: 512GB SAMSUNG MZVL2512HCJQ-00B00 + 3201GB Micron_7450_MTFDKCB3T2TFS, Graphics: ASPEED, Network: Broadcom NetXtreme BCM5720 PCIe OS: Ubuntu 24.04, Kernel: 6.12.0-rc3-phx (x86_64), Compiler: GCC 13.2.0 + Clang 18.1.3, File-System: ext4, Screen Resolution: 1024x768 b: Processor: 2 x AMD EPYC 9755 128-Core @ 2.70GHz (256 Cores / 512 Threads), Motherboard: AMD VOLCANO (RVOT1001B BIOS), Chipset: AMD Device 153a, Memory: 1520GB, Disk: 512GB SAMSUNG MZVL2512HCJQ-00B00 + 3201GB Micron_7450_MTFDKCB3T2TFS, Graphics: ASPEED, Network: Broadcom NetXtreme BCM5720 PCIe OS: Ubuntu 24.04, Kernel: 6.12.0-rc3-phx (x86_64), Compiler: GCC 13.2.0 + Clang 18.1.3, File-System: ext4, Screen Resolution: 1024x768 LiteRT 2024-10-15 Model: SqueezeNet Microseconds < Lower Is Better a . 105380.0 |================================================================= b . 46189.4 |============================ WarpX 24.10 Input: Plasma Acceleration Seconds < Lower Is Better a . 23.47 |==================================================================== b . 23.43 |==================================================================== LiteRT 2024-10-15 Model: DeepLab V3 Microseconds < Lower Is Better a . 93820.7 |=========================================================== b . 102742.9 |================================================================= LiteRT 2024-10-15 Model: Inception V4 Microseconds < Lower Is Better a . 422782 |=================================================================== b . 380938 |============================================================ LiteRT 2024-10-15 Model: NASNet Mobile Microseconds < Lower Is Better a . 1831650 |================================================================== b . 1254561 |============================================= LiteRT 2024-10-15 Model: Mobilenet Float Microseconds < Lower Is Better a . 75229.3 |================================================================== b . 24645.9 |====================== LiteRT 2024-10-15 Model: Mobilenet Quant Microseconds < Lower Is Better a . 43601.8 |============================================================== b . 46489.2 |================================================================== LiteRT 2024-10-15 Model: Inception ResNet V2 Microseconds < Lower Is Better a . 824335 |=================================================================== b . 363512 |============================== LiteRT 2024-10-15 Model: Quantized COCO SSD MobileNet v1 Microseconds < Lower Is Better a . 63357.4 |========================================================== b . 71555.6 |================================================================== WarpX 24.10 Input: Uniform Plasma Seconds < Lower Is Better a . 20.38 |==================================================================== b . 20.38 |==================================================================== XNNPACK b7b048 Model: FP32MobileNetV1 us < Lower Is Better a . 126508 |=================================================================== b . 30213 |================ XNNPACK b7b048 Model: FP32MobileNetV2 us < Lower Is Better a . 349733 |=================================================================== b . 283027 |====================================================== XNNPACK b7b048 Model: FP32MobileNetV3Large us < Lower Is Better a . 322181 |=================================================================== b . 269833 |======================================================== XNNPACK b7b048 Model: FP32MobileNetV3Small us < Lower Is Better a . 306213 |=================================================================== b . 286163 |=============================================================== XNNPACK b7b048 Model: FP16MobileNetV1 us < Lower Is Better a . 121353 |=================================================================== b . 99495 |======================================================= XNNPACK b7b048 Model: FP16MobileNetV2 us < Lower Is Better a . 243557 |============================================================== b . 262462 |=================================================================== XNNPACK b7b048 Model: FP16MobileNetV3Large us < Lower Is Better a . 278473 |============================================================= b . 304488 |=================================================================== XNNPACK b7b048 Model: FP16MobileNetV3Small us < Lower Is Better a . 308464 |=================================================== b . 406223 |=================================================================== XNNPACK b7b048 Model: QS8MobileNetV2 us < Lower Is Better a . 202569 |========================================================= b . 238886 |=================================================================== Epoch 4.19.4 Epoch3D Deck: Cone Seconds < Lower Is Better a . 283.75 |=================================================================== b . 280.78 |================================================================== oneDNN 3.6 Harness: IP Shapes 1D - Engine: CPU ms < Lower Is Better a . 0.797341 |================================================================= b . 0.799482 |================================================================= oneDNN 3.6 Harness: IP Shapes 3D - Engine: CPU ms < Lower Is Better a . 0.652196 |================================================================= b . 0.652110 |================================================================= oneDNN 3.6 Harness: Convolution Batch Shapes Auto - Engine: CPU ms < Lower Is Better a . 0.298302 |================================================================ b . 0.301225 |================================================================= oneDNN 3.6 Harness: Deconvolution Batch shapes_1d - Engine: CPU ms < Lower Is Better a . 26.14 |=================================================================== b . 26.57 |==================================================================== oneDNN 3.6 Harness: Deconvolution Batch shapes_3d - Engine: CPU ms < Lower Is Better a . 0.423216 |================================================================= b . 0.421823 |================================================================= oneDNN 3.6 Harness: Recurrent Neural Network Training - Engine: CPU ms < Lower Is Better a . 761.16 |=================================================================== b . 758.56 |=================================================================== oneDNN 3.6 Harness: Recurrent Neural Network Inference - Engine: CPU ms < Lower Is Better a . 514.99 |=================================================================== b . 516.46 |===================================================================