EPYC 72F3 Retpolines Change In Linux 5.17 AMD EPYC 72F3 Retpoline benchmarks by Michael Larabel. New Default - Generic Retpolines: Processor: AMD EPYC 72F3 8-Core @ 3.70GHz (8 Cores / 16 Threads), Motherboard: Supermicro H12SSL-i v1.01 (2.0 BIOS), Chipset: AMD Starship/Matisse, Memory: 128GB, Disk: 3841GB Micron_9300_MTFDHAL3T8TDP + 1000GB Corsair Force MP600, Graphics: ASPEED, Monitor: VE228, Network: 2 x Broadcom NetXtreme BCM5720 2-port PCIe OS: Ubuntu 21.04, Kernel: 5.17.0-051700rc7daily20220310-generic (x86_64), Desktop: GNOME Shell 3.38.4, Display Server: X Server, Compiler: GCC 10.3.0, File-System: ext4, Screen Resolution: 1920x1080 AMD Retpolines Former Default: Processor: AMD EPYC 72F3 8-Core @ 3.70GHz (8 Cores / 16 Threads), Motherboard: Supermicro H12SSL-i v1.01 (2.0 BIOS), Chipset: AMD Starship/Matisse, Memory: 128GB, Disk: 3841GB Micron_9300_MTFDHAL3T8TDP + 1000GB Corsair Force MP600, Graphics: ASPEED, Monitor: VE228, Network: 2 x Broadcom NetXtreme BCM5720 2-port PCIe OS: Ubuntu 21.04, Kernel: 5.17.0-051700rc7daily20220310-generic (x86_64), Desktop: GNOME Shell 3.38.4, Display Server: X Server, Compiler: GCC 10.3.0, File-System: ext4, Screen Resolution: 1920x1080 LevelDB 1.22 Benchmark: Hot Read Microseconds Per Op < Lower Is Better New Default - Generic Retpolines . 10.54 |===================================== AMD Retpolines Former Default .... 10.50 |===================================== LevelDB 1.22 Benchmark: Fill Sync Microseconds Per Op < Lower Is Better New Default - Generic Retpolines . 1820.49 |=================================== AMD Retpolines Former Default .... 1806.76 |=================================== LevelDB 1.22 Benchmark: Overwrite MB/s > Higher Is Better New Default - Generic Retpolines . 17.1 |====================================== AMD Retpolines Former Default .... 16.9 |====================================== LevelDB 1.22 Benchmark: Overwrite Microseconds Per Op < Lower Is Better New Default - Generic Retpolines . 103.37 |==================================== AMD Retpolines Former Default .... 104.61 |==================================== LevelDB 1.22 Benchmark: Random Fill MB/s > Higher Is Better New Default - Generic Retpolines . 17.2 |====================================== AMD Retpolines Former Default .... 17.1 |====================================== LevelDB 1.22 Benchmark: Random Fill Microseconds Per Op < Lower Is Better New Default - Generic Retpolines . 103.22 |==================================== AMD Retpolines Former Default .... 103.68 |==================================== LevelDB 1.22 Benchmark: Random Read Microseconds Per Op < Lower Is Better New Default - Generic Retpolines . 10.58 |===================================== AMD Retpolines Former Default .... 10.55 |===================================== LevelDB 1.22 Benchmark: Seek Random Microseconds Per Op < Lower Is Better New Default - Generic Retpolines . 17.14 |===================================== AMD Retpolines Former Default .... 17.01 |===================================== LevelDB 1.22 Benchmark: Random Delete Microseconds Per Op < Lower Is Better New Default - Generic Retpolines . 99.88 |===================================== AMD Retpolines Former Default .... 98.95 |===================================== LevelDB 1.22 Benchmark: Sequential Fill MB/s > Higher Is Better New Default - Generic Retpolines . 16.9 |====================================== AMD Retpolines Former Default .... 16.5 |===================================== LevelDB 1.22 Benchmark: Sequential Fill Microseconds Per Op < Lower Is Better New Default - Generic Retpolines . 105.12 |=================================== AMD Retpolines Former Default .... 106.95 |==================================== Flexible IO Tester 3.29 Type: Random Read - Engine: IO_uring - Buffered: No - Direct: Yes - Block Size: 4KB - Disk Target: Default Test Directory IOPS > Higher Is Better New Default - Generic Retpolines . 239667 |==================================== AMD Retpolines Former Default .... 243000 |==================================== Flexible IO Tester 3.29 Type: Random Write - Engine: IO_uring - Buffered: No - Direct: Yes - Block Size: 4KB - Disk Target: Default Test Directory IOPS > Higher Is Better New Default - Generic Retpolines . 223933 |=================================== AMD Retpolines Former Default .... 229500 |==================================== PostMark 1.51 Disk Transaction Performance TPS > Higher Is Better New Default - Generic Retpolines . 6944 |==================================== AMD Retpolines Former Default .... 7282 |====================================== WireGuard + Linux Networking Stack Stress Test Seconds < Lower Is Better New Default - Generic Retpolines . 257.10 |==================================== AMD Retpolines Former Default .... 248.76 |=================================== Sockperf 3.7 Test: Throughput Messages Per Second > Higher Is Better New Default - Generic Retpolines . 539724 |================================= AMD Retpolines Former Default .... 587523 |==================================== Sockperf 3.7 Test: Latency Under Load usec < Lower Is Better New Default - Generic Retpolines . 3.666 |===================================== AMD Retpolines Former Default .... 3.641 |===================================== OSBench Test: Create Files us Per Event < Lower Is Better New Default - Generic Retpolines . 14.71 |===================================== AMD Retpolines Former Default .... 13.72 |=================================== OSBench Test: Create Threads us Per Event < Lower Is Better New Default - Generic Retpolines . 17.38 |==================================== AMD Retpolines Former Default .... 17.70 |===================================== OSBench Test: Launch Programs us Per Event < Lower Is Better New Default - Generic Retpolines . 49.36 |===================================== AMD Retpolines Former Default .... 48.63 |==================================== OSBench Test: Create Processes us Per Event < Lower Is Better New Default - Generic Retpolines . 43.14 |===================================== AMD Retpolines Former Default .... 43.36 |===================================== DaCapo Benchmark 9.12-MR1 Java Test: H2 msec < Lower Is Better New Default - Generic Retpolines . 4268 |====================================== AMD Retpolines Former Default .... 4228 |====================================== DaCapo Benchmark 9.12-MR1 Java Test: Jython msec < Lower Is Better New Default - Generic Retpolines . 3761 |====================================== AMD Retpolines Former Default .... 3697 |===================================== DaCapo Benchmark 9.12-MR1 Java Test: Tradesoap msec < Lower Is Better New Default - Generic Retpolines . 3660 |====================================== AMD Retpolines Former Default .... 3628 |====================================== DaCapo Benchmark 9.12-MR1 Java Test: Tradebeans msec < Lower Is Better New Default - Generic Retpolines . 5762 |====================================== AMD Retpolines Former Default .... 5634 |===================================== Renaissance 0.12 Test: Finagle HTTP Requests ms < Lower Is Better New Default - Generic Retpolines . 2378.7 |==================================== AMD Retpolines Former Default .... 2366.9 |==================================== Renaissance 0.12 Test: In-Memory Database Shootout ms < Lower Is Better New Default - Generic Retpolines . 4788.2 |=================================== AMD Retpolines Former Default .... 4878.7 |==================================== LZ4 Compression 1.9.3 Compression Level: 3 - Compression Speed MB/s > Higher Is Better New Default - Generic Retpolines . 61.24 |===================================== AMD Retpolines Former Default .... 61.36 |===================================== LZ4 Compression 1.9.3 Compression Level: 3 - Decompression Speed MB/s > Higher Is Better New Default - Generic Retpolines . 14455.1 |================================== AMD Retpolines Former Default .... 14817.3 |=================================== LZ4 Compression 1.9.3 Compression Level: 9 - Compression Speed MB/s > Higher Is Better New Default - Generic Retpolines . 59.57 |===================================== AMD Retpolines Former Default .... 59.69 |===================================== LZ4 Compression 1.9.3 Compression Level: 9 - Decompression Speed MB/s > Higher Is Better New Default - Generic Retpolines . 14586.2 |================================== AMD Retpolines Former Default .... 14807.4 |=================================== Zstd Compression 1.5.0 Compression Level: 8 - Compression Speed MB/s > Higher Is Better New Default - Generic Retpolines . 1343.6 |==================================== AMD Retpolines Former Default .... 1338.6 |==================================== Zstd Compression 1.5.0 Compression Level: 8 - Decompression Speed MB/s > Higher Is Better New Default - Generic Retpolines . 4255.1 |==================================== AMD Retpolines Former Default .... 4160.1 |=================================== Zstd Compression 1.5.0 Compression Level: 19 - Compression Speed MB/s > Higher Is Better New Default - Generic Retpolines . 51.6 |====================================== AMD Retpolines Former Default .... 51.9 |====================================== Zstd Compression 1.5.0 Compression Level: 19 - Decompression Speed MB/s > Higher Is Better New Default - Generic Retpolines . 3694.5 |==================================== AMD Retpolines Former Default .... 3668.5 |==================================== Node.js Express HTTP Load Test Requests Per Second > Higher Is Better New Default - Generic Retpolines . 6692 |===================================== AMD Retpolines Former Default .... 6910 |====================================== 7-Zip Compression 21.06 Test: Compression Rating MIPS > Higher Is Better New Default - Generic Retpolines . 86357 |===================================== AMD Retpolines Former Default .... 87210 |===================================== 7-Zip Compression 21.06 Test: Decompression Rating MIPS > Higher Is Better New Default - Generic Retpolines . 69393 |===================================== AMD Retpolines Former Default .... 69308 |===================================== ebizzy 0.3 Records/s > Higher Is Better New Default - Generic Retpolines . 619271 |==================================== AMD Retpolines Former Default .... 596554 |=================================== Timed GDB GNU Debugger Compilation 10.2 Time To Compile Seconds < Lower Is Better New Default - Generic Retpolines . 54.96 |===================================== AMD Retpolines Former Default .... 55.03 |===================================== Timed Godot Game Engine Compilation 3.2.3 Time To Compile Seconds < Lower Is Better New Default - Generic Retpolines . 111.51 |==================================== AMD Retpolines Former Default .... 111.58 |==================================== Timed Linux Kernel Compilation 5.16 Build: defconfig Seconds < Lower Is Better New Default - Generic Retpolines . 78.62 |===================================== AMD Retpolines Former Default .... 78.59 |===================================== Timed MPlayer Compilation 1.5 Time To Compile Seconds < Lower Is Better New Default - Generic Retpolines . 30.71 |===================================== AMD Retpolines Former Default .... 30.80 |===================================== Parallel BZIP2 Compression 1.1.13 FreeBSD-13.0-RELEASE-amd64-memstick.img Compression Seconds < Lower Is Better New Default - Generic Retpolines . 10.78 |===================================== AMD Retpolines Former Default .... 10.66 |===================================== Timed Eigen Compilation 3.3.9 Time To Compile Seconds < Lower Is Better New Default - Generic Retpolines . 61.69 |===================================== AMD Retpolines Former Default .... 61.10 |===================================== Hackbench Count: 8 - Type: Thread Seconds < Lower Is Better New Default - Generic Retpolines . 31.28 |===================================== AMD Retpolines Former Default .... 31.05 |===================================== Hackbench Count: 16 - Type: Thread Seconds < Lower Is Better New Default - Generic Retpolines . 62.20 |===================================== AMD Retpolines Former Default .... 62.03 |===================================== Hackbench Count: 8 - Type: Process Seconds < Lower Is Better New Default - Generic Retpolines . 29.95 |===================================== AMD Retpolines Former Default .... 28.83 |==================================== Hackbench Count: 16 - Type: Process Seconds < Lower Is Better New Default - Generic Retpolines . 58.78 |===================================== AMD Retpolines Former Default .... 58.36 |===================================== Hackbench Count: 32 - Type: Process Seconds < Lower Is Better New Default - Generic Retpolines . 119.83 |==================================== AMD Retpolines Former Default .... 117.98 |=================================== OpenSSL 3.0 Algorithm: SHA256 byte/s > Higher Is Better New Default - Generic Retpolines . 13238722860 |=============================== AMD Retpolines Former Default .... 13234833737 |=============================== OpenSSL 3.0 Algorithm: RSA4096 sign/s > Higher Is Better New Default - Generic Retpolines . 2372.3 |==================================== AMD Retpolines Former Default .... 2374.1 |==================================== OpenSSL 3.0 Algorithm: RSA4096 verify/s > Higher Is Better New Default - Generic Retpolines . 155003.1 |================================== AMD Retpolines Former Default .... 155031.8 |================================== Node.js V8 Web Tooling Benchmark runs/s > Higher Is Better New Default - Generic Retpolines . 11.26 |===================================== AMD Retpolines Former Default .... 11.31 |===================================== KeyDB 6.2.0 Ops/sec > Higher Is Better New Default - Generic Retpolines . 136793.59 |================================ AMD Retpolines Former Default .... 142573.34 |================================= PostgreSQL pgbench 14.0 Scaling Factor: 100 - Clients: 50 - Mode: Read Only TPS > Higher Is Better New Default - Generic Retpolines . 295179 |==================================== AMD Retpolines Former Default .... 295730 |==================================== PostgreSQL pgbench 14.0 Scaling Factor: 100 - Clients: 50 - Mode: Read Only - Average Latency ms < Lower Is Better New Default - Generic Retpolines . 0.170 |===================================== AMD Retpolines Former Default .... 0.169 |===================================== PostgreSQL pgbench 14.0 Scaling Factor: 100 - Clients: 100 - Mode: Read Only TPS > Higher Is Better New Default - Generic Retpolines . 302523 |==================================== AMD Retpolines Former Default .... 302399 |==================================== PostgreSQL pgbench 14.0 Scaling Factor: 100 - Clients: 100 - Mode: Read Only - Average Latency ms < Lower Is Better New Default - Generic Retpolines . 0.331 |===================================== AMD Retpolines Former Default .... 0.330 |===================================== PostgreSQL pgbench 14.0 Scaling Factor: 100 - Clients: 100 - Mode: Read Write TPS > Higher Is Better New Default - Generic Retpolines . 10383 |===================================== AMD Retpolines Former Default .... 10440 |===================================== PostgreSQL pgbench 14.0 Scaling Factor: 100 - Clients: 100 - Mode: Read Write - Average Latency ms < Lower Is Better New Default - Generic Retpolines . 9.641 |===================================== AMD Retpolines Former Default .... 9.590 |===================================== SQLite Speedtest 3.30 Timed Time - Size 1,000 Seconds < Lower Is Better New Default - Generic Retpolines . 50.48 |===================================== AMD Retpolines Former Default .... 49.01 |==================================== Google Draco 1.5.0 Model: Lion ms < Lower Is Better New Default - Generic Retpolines . 5088 |====================================== AMD Retpolines Former Default .... 5070 |====================================== Google Draco 1.5.0 Model: Church Facade ms < Lower Is Better New Default - Generic Retpolines . 6697 |====================================== AMD Retpolines Former Default .... 6649 |====================================== Memtier_benchmark 1.3 Protocol: Redis Ops/sec > Higher Is Better New Default - Generic Retpolines . 1324327.32 |================================ AMD Retpolines Former Default .... 1313054.11 |================================ Redis 6.0.9 Test: GET Requests Per Second > Higher Is Better New Default - Generic Retpolines . 2506720.25 |================================ AMD Retpolines Former Default .... 2438721.71 |=============================== Redis 6.0.9 Test: SET Requests Per Second > Higher Is Better New Default - Generic Retpolines . 1998825.92 |================================ AMD Retpolines Former Default .... 1995692.89 |================================ Stress-NG 0.13.02 Test: SENDFILE Bogo Ops/s > Higher Is Better New Default - Generic Retpolines . 200771.29 |================================= AMD Retpolines Former Default .... 193652.98 |================================ Stress-NG 0.13.02 Test: Semaphores Bogo Ops/s > Higher Is Better New Default - Generic Retpolines . 1326767.92 |================================ AMD Retpolines Former Default .... 1342170.84 |================================ Stress-NG 0.13.02 Test: Socket Activity Bogo Ops/s > Higher Is Better New Default - Generic Retpolines . 10751.13 |================================== AMD Retpolines Former Default .... 10567.09 |================================= Stress-NG 0.13.02 Test: Context Switching Bogo Ops/s > Higher Is Better New Default - Generic Retpolines . 4296558.49 |================================ AMD Retpolines Former Default .... 4326244.50 |================================ Stress-NG 0.13.02 Test: System V Message Passing Bogo Ops/s > Higher Is Better New Default - Generic Retpolines . 4521841.84 |================================ AMD Retpolines Former Default .... 4530010.92 |================================ Apache Cassandra 4.0 Test: Writes Op/s > Higher Is Better New Default - Generic Retpolines . 86420 |===================================== AMD Retpolines Former Default .... 85956 |===================================== Facebook RocksDB 6.22.1 Test: Random Fill Op/s > Higher Is Better New Default - Generic Retpolines . 605572 |=================================== AMD Retpolines Former Default .... 617250 |==================================== Facebook RocksDB 6.22.1 Test: Random Read Op/s > Higher Is Better New Default - Generic Retpolines . 61331721 |================================== AMD Retpolines Former Default .... 61586232 |================================== Facebook RocksDB 6.22.1 Test: Update Random Op/s > Higher Is Better New Default - Generic Retpolines . 393101 |==================================== AMD Retpolines Former Default .... 398526 |==================================== Facebook RocksDB 6.22.1 Test: Sequential Fill Op/s > Higher Is Better New Default - Generic Retpolines . 657523 |=================================== AMD Retpolines Former Default .... 670760 |==================================== Facebook RocksDB 6.22.1 Test: Read While Writing Op/s > Higher Is Better New Default - Generic Retpolines . 1734878 |=================================== AMD Retpolines Former Default .... 1746524 |=================================== Facebook RocksDB 6.22.1 Test: Read Random Write Random Op/s > Higher Is Better New Default - Generic Retpolines . 1391000 |=================================== AMD Retpolines Former Default .... 1386586 |=================================== nginx 1.21.1 Concurrent Requests: 20 Requests Per Second > Higher Is Better New Default - Generic Retpolines . 215287.79 |================================= AMD Retpolines Former Default .... 210605.86 |================================ nginx 1.21.1 Concurrent Requests: 100 Requests Per Second > Higher Is Better New Default - Generic Retpolines . 215190.56 |================================= AMD Retpolines Former Default .... 207615.65 |================================ nginx 1.21.1 Concurrent Requests: 200 Requests Per Second > Higher Is Better New Default - Generic Retpolines . 213684.40 |================================= AMD Retpolines Former Default .... 206147.39 |================================ Apache HTTP Server 2.4.48 Concurrent Requests: 100 Requests Per Second > Higher Is Better New Default - Generic Retpolines . 82500.75 |================================== AMD Retpolines Former Default .... 83397.57 |================================== Apache HTTP Server 2.4.48 Concurrent Requests: 200 Requests Per Second > Higher Is Better New Default - Generic Retpolines . 86362.18 |================================== AMD Retpolines Former Default .... 86408.48 |================================== Apache HTTP Server 2.4.48 Concurrent Requests: 500 Requests Per Second > Higher Is Better New Default - Generic Retpolines . 80298.15 |================================== AMD Retpolines Former Default .... 80970.15 |================================== InfluxDB 1.8.2 Concurrent Streams: 4 - Batch Size: 10000 - Tags: 2,5000,1 - Points Per Series: 10000 val/sec > Higher Is Better New Default - Generic Retpolines . 1326859.1 |================================= AMD Retpolines Former Default .... 1323460.2 |================================= InfluxDB 1.8.2 Concurrent Streams: 64 - Batch Size: 10000 - Tags: 2,5000,1 - Points Per Series: 10000 val/sec > Higher Is Better New Default - Generic Retpolines . 1638195.7 |================================= AMD Retpolines Former Default .... 1639910.0 |=================================