TensorFlow Lite Kernel Regression

AMD EPYC 7702 64-Core testing with a ASRockRack EPYCD8 (P2.10 BIOS) and ASPEED on Ubuntu 20.04 via the Phoronix Test Suite.

HTML result view exported from: https://openbenchmarking.org/result/2009103-FI-TENSORFLO07.

TensorFlow Lite Kernel RegressionProcessorMotherboardChipsetMemoryDiskGraphicsAudioNetworkOSKernelDesktopDisplay ServerDisplay DriverCompilerFile-SystemScreen Resolutionv5.5v5.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 EPYC 7702 64-Core @ 2.00GHz (64 Cores / 128 Threads)ASRockRack EPYCD8 (P2.10 BIOS)AMD Starship/Matisse126GB3841GB Micron_9300_MTFDHAL3T8TDP + 32GB Flash DiskASPEEDAMD Starship/Matisse2 x Intel I350Ubuntu 20.045.5.0-pts (x86_64)GNOME Shell 3.36.4X Server 1.20.8modesetting 1.20.8GCC 9.3.0ext41024x7685.4.0-pts (x86_64)5.4.0-rc7-pts (x86_64) 202009105.4.0-pts (x86_64)5.4.0-rc8-pts (x86_64) 202009105.4.0-rc1-pts (x86_64) 20200910OpenBenchmarking.orgProcessor Details- Scaling Governor: acpi-cpufreq ondemand - CPU Microcode: 0x830101cSecurity Details- v5.5: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional IBRS_FW STIBP: always-on RSB filling + tsx_async_abort: Not affected- v5.4: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional IBRS_FW STIBP: always-on RSB filling + tsx_async_abort: Not affected- 8c39f71ee2019e77ee14f88b1321b2348db51820: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional IBRS_FW STIBP: always-on RSB filling + tsx_async_abort: Not affected- 3b397c7ccafe0624018cb09fc96729f8f6165573: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional IBRS_FW STIBP: always-on RSB filling + tsx_async_abort: Not affected- 89d57dddd7d319ded00415790a0bb3c954b7e386: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional IBRS_FW STIBP: always-on RSB filling + tsx_async_abort: Not affected- 9e7a03233e02afd3ee061e373355f34d7254f1e6: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional IBRS_FW STIBP: always-on RSB filling + tsx_async_abort: Not affected- df28204bb0f29cc475c0a8893c99b46a11a4903f: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional IBRS_FW STIBP: always-on RSB filling + tsx_async_abort: Not affected- 0e3149f86b99ddabde8c5029eea0a9267e34f1a0: l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional IBRS_FW STIBP: always-on RSB filling- 1ae78780eda54023a0fb49ee743dbba39da148e0: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional IBRS_FW STIBP: always-on RSB filling + tsx_async_abort: Not affected- 77a05940eee7e9891cd6add7a690a3e762ee21b0: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional IBRS_FW STIBP: always-on RSB filling + tsx_async_abort: Not affected- b21feab0b865c36b24d7a60b55a10c7033b03159: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional IBRS_FW STIBP: always-on RSB filling + tsx_async_abort: Not affected- 74c578759f15cb5a0d0107759bdad671d7b52ab9: l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional IBRS_FW STIBP: always-on RSB filling- 2ab4092fc82d6001fdd9d51dbba27d04dec967e0: l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional IBRS_FW STIBP: always-on RSB filling- a34983470301018324f0110791da452fee1318c2: l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional IBRS_FW STIBP: always-on RSB filling- 0b0695f2b34a4afa3f6e9aa1ff0e5336d8dad912: l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional IBRS_FW STIBP: always-on RSB filling- fcf0553db6f4c79387864f6e4ab4a891601f395e: l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional IBRS_FW STIBP: always-on RSB filling

TensorFlow Lite Kernel Regressiontensorflow-lite: Inception V4tensorflow-lite: Mobilenet Quanttensorflow-lite: Inception ResNet V2v5.5v5.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.587633285164739820.5717903100043643055.187921483441339191.7713821100515743430.788403999324843193.188974482425440402.175199884135640473.772055999683743319.588035899276043311.587786898746043226.688038598854143170.787851197686142829.178640286822739813.875580496853343036.680486195454643282.4859121OpenBenchmarking.org

TensorFlow Lite

Model: Inception V4

OpenBenchmarking.orgMicroseconds, Fewer Is BetterTensorFlow Lite 2020-08-23Model: Inception V4v5.5v5.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 +/- 5786.28, N = 3SE +/- 10230.46, N = 15SE +/- 2076.06, N = 3SE +/- 8756.66, N = 15SE +/- 3153.29, N = 3SE +/- 1129.67, N = 3SE +/- 6623.24, N = 13SE +/- 7882.56, N = 3SE +/- 3016.30, N = 3SE +/- 10918.27, N = 3SE +/- 2281.23, N = 3SE +/- 1203.18, N = 3SE +/- 1426.28, N = 3SE +/- 13759.34, N = 3SE +/- 7633.77, N = 3SE +/- 917.91, N = 399550485164710004368344131005157993248824254841356996837992760987460988541976861868227968533954546

TensorFlow Lite

Model: Mobilenet Quant

OpenBenchmarking.orgMicroseconds, Fewer Is BetterTensorFlow Lite 2020-08-23Model: Mobilenet Quantv5.5v5.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 +/- 88.14, N = 3SE +/- 436.64, N = 15SE +/- 109.54, N = 3SE +/- 601.92, N = 3SE +/- 131.82, N = 3SE +/- 282.29, N = 3SE +/- 362.35, N = 15SE +/- 389.69, N = 15SE +/- 51.34, N = 3SE +/- 139.06, N = 3SE +/- 73.64, N = 3SE +/- 64.53, N = 3SE +/- 116.94, N = 3SE +/- 550.88, N = 3SE +/- 136.90, N = 3SE +/- 40.06, N = 343291.539820.543055.139191.743430.743193.140402.140473.743319.543311.543226.643170.742829.139813.843036.643282.4

TensorFlow Lite

Model: Inception ResNet V2

OpenBenchmarking.orgMicroseconds, Fewer Is BetterTensorFlow Lite 2020-08-23Model: Inception ResNet V2v5.5v5.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 +/- 7137.69, N = 3SE +/- 4913.76, N = 3SE +/- 7425.32, N = 3SE +/- 2546.21, N = 3SE +/- 3550.66, N = 3SE +/- 1311.25, N = 3SE +/- 4037.36, N = 3SE +/- 10502.35, N = 3SE +/- 3051.65, N = 3SE +/- 1416.31, N = 3SE +/- 873.96, N = 3SE +/- 1369.49, N = 3SE +/- 2377.85, N = 3SE +/- 10351.22, N = 4SE +/- 8431.87, N = 15SE +/- 2645.52, N = 3876332717903879214713821884039889744751998720559880358877868880385878511786402755804804861859121

Phoronix Test Suite v10.8.4