Raspberry Pi 5 GPU Benchmarks

Raspberry Pi 5 vs. Raspberry Pi 4 graphics benchmarks by Michael Larabel for a future article. ARMv8 Cortex-A72 testing with a BCM2835 Raspberry Pi 400 Rev 1.0 and vc4drmfb on Debian 11 via the Phoronix Test Suite.

Compare your own system(s) to this result file with the Phoronix Test Suite by running the command: phoronix-test-suite benchmark 2309287-NE-2309280NE47
Jump To Table - Results

View

Do Not Show Noisy Results
Do Not Show Results With Incomplete Data
Do Not Show Results With Little Change/Spread
List Notable Results
Show Result Confidence Charts

Statistics

Show Overall Harmonic Mean(s)
Show Overall Geometric Mean
Show Wins / Losses Counts (Pie Chart)
Normalize Results
Remove Outliers Before Calculating Averages

Graph Settings

Force Line Graphs Where Applicable
Convert To Scalar Where Applicable
Prefer Vertical Bar Graphs

Additional Graphs

Show Perf Per Clock Calculation Graphs Where Applicable

Multi-Way Comparison

Condense Multi-Option Tests Into Single Result Graphs

Table

Show Detailed System Result Table

Run Management

Highlight
Result
Hide
Result
Result
Identifier
View Logs
Performance Per
Dollar
Date
Run
  Test
  Duration
Raspberry Pi 5
September 28 2023
  5 Hours, 15 Minutes
Raspberry Pi 4
September 28 2023
  18 Hours, 41 Minutes
Invert Hiding All Results Option
  11 Hours, 58 Minutes
Only show results matching title/arguments (delimit multiple options with a comma):
Do not show results matching title/arguments (delimit multiple options with a comma):


Raspberry Pi 5 GPU BenchmarksProcessorMotherboardChipsetMemoryDiskGraphicsMonitorNetworkOSKernelDesktopDisplay ServerOpenGLCompilerFile-SystemScreen ResolutionRaspberry Pi 4Raspberry Pi 5ARMv8 Cortex-A72 @ 1.80GHz (4 Cores)BCM2835 Raspberry Pi 400 Rev 1.0Broadcom BCM27114096MB128GB SF128vc4drmfbDELL P2210HDebian 115.15.84-v8+ (aarch64)LXDEX Server 1.20.112.1 Mesa 20.3.5GCC 10.2.1 20210110ext41920x1080ARMv8 Cortex-A76 @ 2.40GHz (4 Cores)BCM2835 Raspberry Pi 5 Model B Rev 1.08GB32GB SL32GV3D 7.1 8GBDELL P2415QDevice 1de4:0001Debian 126.1.0-rpi3-rpi-2712 (aarch64)X Server + Wayland3.1 Mesa 23.2.0~rc3-1+rpt3GCC 12.2.03840x2160OpenBenchmarking.orgKernel Details- Raspberry Pi 4: snd_bcm2835.enable_compat_alsa=0 snd_bcm2835.enable_hdmi=1- Raspberry Pi 5: cfg80211.ieee80211_regdom=USCompiler Details- Raspberry Pi 4: --build=aarch64-linux-gnu --disable-libquadmath --disable-libquadmath-support --disable-werror --enable-bootstrap --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-fix-cortex-a53-843419 --enable-gnu-unique-object --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-link-mutex --enable-multiarch --enable-nls --enable-objc-gc=auto --enable-plugin --enable-shared --enable-threads=posix --host=aarch64-linux-gnu --program-prefix=aarch64-linux-gnu- --target=aarch64-linux-gnu --with-build-config=bootstrap-lto-lean --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-target-system-zlib=auto -v - Raspberry Pi 5: --build=aarch64-linux-gnu --disable-libquadmath --disable-libquadmath-support --disable-werror --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-fix-cortex-a53-843419 --enable-gnu-unique-object --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-multiarch --enable-nls --enable-objc-gc=auto --enable-plugin --enable-shared --enable-threads=posix --host=aarch64-linux-gnu --program-prefix=aarch64-linux-gnu- --target=aarch64-linux-gnu --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-target-system-zlib=auto -v Processor Details- Scaling Governor: cpufreq-dt ondemandPython Details- Raspberry Pi 4: Python 3.9.2- Raspberry Pi 5: Python 3.11.2Security Details- Raspberry Pi 4: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + mmio_stale_data: Not affected + retbleed: Not affected + spec_store_bypass: Vulnerable + spectre_v1: Mitigation of __user pointer sanitization + spectre_v2: Vulnerable + srbds: Not affected + tsx_async_abort: Not affected - Raspberry Pi 5: gather_data_sampling: Not affected + itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + mmio_stale_data: Not affected + retbleed: Not affected + spec_rstack_overflow: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl + spectre_v1: Mitigation of __user pointer sanitization + spectre_v2: Mitigation of CSV2 BHB + srbds: Not affected + tsx_async_abort: Not affected

Raspberry Pi 4 vs. Raspberry Pi 5 ComparisonPhoronix Test SuiteBaseline+154.6%+154.6%+309.2%+309.2%+463.8%+463.8%618.4%570.4%505.3%465.6%436.8%419.5%412.9%404.1%374.7%367.3%362%346%330.6%297.5%277.1%258.1%250.6%250%232.8%215.4%209.4%197%171%167.3%165.2%163.2%Vulkan GPU - shufflenet-v2Vulkan GPU - FastestDetVulkan GPU - mnasnetVulkan GPU - blazefaceVulkan GPU - regnety_400mVulkan GPU - resnet18Vulkan GPU-v3-v3 - mobilenet-v3Vulkan GPU - efficientnet-b0Vulkan GPU-v2-v2 - mobilenet-v2OpenGL ES 3.x - Off - Off - 1920 x 1080Vulkan GPU - resnet50OpenGL ES 3.x - On - Off - 1920 x 10801920 x 1080Vulkan GPU - mobilenetVulkan GPU - yolov4-tinyVulkan GPU - googlenetOpenGL ES 3.x - Off - On - 1920 x 1080OpenGL ES 3.x - On - On - 1920 x 1080Vulkan GPU - vgg16Vulkan GPU - squeezenet_ssdVulkan GPU - vision_transformerVulkan GPU - alexnetOpenGL 3.x - Off - On - 1920 x 1080OpenGL 3.x - On - Off - 1920 x 1080OpenGL 3.x - Off - Off - 1920 x 1080OpenGL 3.x - On - On - 1920 x 1080NCNNNCNNNCNNNCNNNCNNNCNNNCNNNCNNNCNNyquake2NCNNyquake2GLmark2NCNNNCNNNCNNyquake2yquake2NCNNNCNNNCNNNCNNyquake2yquake2yquake2yquake2Raspberry Pi 4Raspberry Pi 5

Raspberry Pi 5 GPU Benchmarksncnn: Vulkan GPU - mobilenetncnn: Vulkan GPU-v2-v2 - mobilenet-v2yquake2: OpenGL ES 3.x - Off - Off - 1920 x 1080glmark2: 1920 x 1080yquake2: OpenGL ES 3.x - Off - On - 1920 x 1080yquake2: OpenGL ES 3.x - On - Off - 1920 x 1080yquake2: OpenGL ES 3.x - On - On - 1920 x 1080yquake2: OpenGL 3.x - Off - Off - 1920 x 1080yquake2: OpenGL 3.x - On - Off - 1920 x 1080ncnn: Vulkan GPU-v3-v3 - mobilenet-v3ncnn: Vulkan GPU - shufflenet-v2ncnn: Vulkan GPU - mnasnetncnn: Vulkan GPU - efficientnet-b0ncnn: Vulkan GPU - blazefacencnn: Vulkan GPU - googlenetncnn: Vulkan GPU - vgg16ncnn: Vulkan GPU - resnet18ncnn: Vulkan GPU - alexnetncnn: Vulkan GPU - resnet50ncnn: Vulkan GPU - yolov4-tinyncnn: Vulkan GPU - squeezenet_ssdncnn: Vulkan GPU - regnety_400myquake2: OpenGL 3.x - Off - On - 1920 x 1080ncnn: Vulkan GPU - vision_transformerncnn: Vulkan GPU - FastestDetyquake2: OpenGL 3.x - On - On - 1920 x 1080Raspberry Pi 4Raspberry Pi 5198.5466.1716.5498.916.18.689.688.349.3926.1554.9679.399.05123.26564.78131.1780.16273.48223.68137.5967.0588.22008.6240.8387.449.9513.9477.121131.271.830.1237.6236.09.633.649.0815.751.6034.42169.7025.2526.9959.1959.3143.6212.49239.0649.126.09230.0OpenBenchmarking.org

NCNN

NCNN is a high performance neural network inference framework optimized for mobile and other platforms developed by Tencent. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: mobilenetRaspberry Pi 5Raspberry Pi 44080120160200SE +/- 0.10, N = 3SE +/- 1.94, N = 349.95198.54-fopenmp - MIN: 49.19 / MAX: 53.8-O3 -lgomp -lpthread -pthread - MIN: 194.17 / MAX: 219.311. (CXX) g++ options: -rdynamic

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU-v2-v2 - Model: mobilenet-v2Raspberry Pi 5Raspberry Pi 41530456075SE +/- 0.05, N = 3SE +/- 1.02, N = 313.9466.17-fopenmp - MIN: 13.53 / MAX: 14.41-O3 -lgomp -lpthread -pthread - MIN: 64.34 / MAX: 97.371. (CXX) g++ options: -rdynamic

yquake2

This is a test of Yamagi Quake II. Yamagi Quake II is an enhanced client for id Software's Quake II with focus on offline and coop gameplay. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is Betteryquake2 8.10Renderer: OpenGL ES 3.x - AF: Off - MSAA: Off - Resolution: 1920 x 1080Raspberry Pi 5Raspberry Pi 420406080100SE +/- 0.07, N = 3SE +/- 0.09, N = 377.116.51. (CC) gcc options: -shared -lm -ldl -rdynamic -lSDL2 -O2 -pipe -fomit-frame-pointer -std=gnu99 -fno-strict-aliasing -fwrapv -fvisibility=hidden -MMD -fPIC

GLmark2

This is a test of GLmark2, a basic OpenGL and OpenGL ES 2.0 benchmark supporting various windowing/display back-ends. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgScore, More Is BetterGLmark2 2023.01Resolution: 1920 x 1080Raspberry Pi 5Raspberry Pi 45010015020025021149

yquake2

This is a test of Yamagi Quake II. Yamagi Quake II is an enhanced client for id Software's Quake II with focus on offline and coop gameplay. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is Betteryquake2 8.10Renderer: OpenGL ES 3.x - AF: Off - MSAA: On - Resolution: 1920 x 1080Raspberry Pi 5Raspberry Pi 4714212835SE +/- 0.03, N = 3SE +/- 0.03, N = 331.28.91. (CC) gcc options: -shared -lm -ldl -rdynamic -lSDL2 -O2 -pipe -fomit-frame-pointer -std=gnu99 -fno-strict-aliasing -fwrapv -fvisibility=hidden -MMD -fPIC

OpenBenchmarking.orgFrames Per Second, More Is Betteryquake2 8.10Renderer: OpenGL ES 3.x - AF: On - MSAA: Off - Resolution: 1920 x 1080Raspberry Pi 5Raspberry Pi 41632486480SE +/- 0.00, N = 3SE +/- 0.12, N = 371.816.11. (CC) gcc options: -shared -lm -ldl -rdynamic -lSDL2 -O2 -pipe -fomit-frame-pointer -std=gnu99 -fno-strict-aliasing -fwrapv -fvisibility=hidden -MMD -fPIC

OpenBenchmarking.orgFrames Per Second, More Is Betteryquake2 8.10Renderer: OpenGL ES 3.x - AF: On - MSAA: On - Resolution: 1920 x 1080Raspberry Pi 5Raspberry Pi 4714212835SE +/- 0.00, N = 3SE +/- 0.03, N = 330.18.61. (CC) gcc options: -shared -lm -ldl -rdynamic -lSDL2 -O2 -pipe -fomit-frame-pointer -std=gnu99 -fno-strict-aliasing -fwrapv -fvisibility=hidden -MMD -fPIC

OpenBenchmarking.orgFrames Per Second, More Is Betteryquake2 8.10Renderer: OpenGL 3.x - AF: Off - MSAA: Off - Resolution: 1920 x 1080Raspberry Pi 5Raspberry Pi 450100150200250SE +/- 0.99, N = 3SE +/- 0.35, N = 3237.689.61. (CC) gcc options: -shared -lm -ldl -rdynamic -lSDL2 -O2 -pipe -fomit-frame-pointer -std=gnu99 -fno-strict-aliasing -fwrapv -fvisibility=hidden -MMD -fPIC

OpenBenchmarking.orgFrames Per Second, More Is Betteryquake2 8.10Renderer: OpenGL 3.x - AF: On - MSAA: Off - Resolution: 1920 x 1080Raspberry Pi 5Raspberry Pi 450100150200250SE +/- 1.30, N = 3SE +/- 0.32, N = 3236.088.31. (CC) gcc options: -shared -lm -ldl -rdynamic -lSDL2 -O2 -pipe -fomit-frame-pointer -std=gnu99 -fno-strict-aliasing -fwrapv -fvisibility=hidden -MMD -fPIC

NCNN

NCNN is a high performance neural network inference framework optimized for mobile and other platforms developed by Tencent. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU-v3-v3 - Model: mobilenet-v3Raspberry Pi 5Raspberry Pi 41122334455SE +/- 0.04, N = 3SE +/- 0.86, N = 39.6349.39-fopenmp - MIN: 9.39 / MAX: 10.03-O3 -lgomp -lpthread -pthread - MIN: 47.7 / MAX: 57.961. (CXX) g++ options: -rdynamic

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: shufflenet-v2Raspberry Pi 5Raspberry Pi 4612182430SE +/- 0.04, N = 3SE +/- 0.55, N = 33.6426.15-fopenmp - MIN: 3.51 / MAX: 3.91-O3 -lgomp -lpthread -pthread - MIN: 25.17 / MAX: 38.681. (CXX) g++ options: -rdynamic

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: mnasnetRaspberry Pi 5Raspberry Pi 41224364860SE +/- 0.08, N = 3SE +/- 0.72, N = 39.0854.96-fopenmp - MIN: 8.59 / MAX: 9.49-O3 -lgomp -lpthread -pthread - MIN: 53.41 / MAX: 82.611. (CXX) g++ options: -rdynamic

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: efficientnet-b0Raspberry Pi 5Raspberry Pi 420406080100SE +/- 0.12, N = 3SE +/- 1.00, N = 315.7579.39-fopenmp - MIN: 15.31 / MAX: 16.35-O3 -lgomp -lpthread -pthread - MIN: 77.6 / MAX: 88.491. (CXX) g++ options: -rdynamic

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: blazefaceRaspberry Pi 5Raspberry Pi 43691215SE +/- 0.00, N = 3SE +/- 0.18, N = 31.609.05-fopenmp - MIN: 1.54 / MAX: 1.78-O3 -lgomp -lpthread -pthread - MIN: 8.67 / MAX: 11.181. (CXX) g++ options: -rdynamic

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: googlenetRaspberry Pi 5Raspberry Pi 4306090120150SE +/- 0.03, N = 3SE +/- 2.18, N = 334.42123.26-fopenmp - MIN: 33.91 / MAX: 34.89-O3 -lgomp -lpthread -pthread - MIN: 120.08 / MAX: 141.421. (CXX) g++ options: -rdynamic

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: vgg16Raspberry Pi 5Raspberry Pi 4120240360480600SE +/- 0.20, N = 3SE +/- 2.50, N = 3169.70564.78-fopenmp - MIN: 166.31 / MAX: 177.27-O3 -lgomp -lpthread -pthread - MIN: 549.85 / MAX: 609.281. (CXX) g++ options: -rdynamic

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: resnet18Raspberry Pi 5Raspberry Pi 4306090120150SE +/- 0.10, N = 3SE +/- 2.68, N = 325.25131.17-fopenmp - MIN: 24.66 / MAX: 39.94-O3 -lgomp -lpthread -pthread - MIN: 127.45 / MAX: 147.51. (CXX) g++ options: -rdynamic

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: alexnetRaspberry Pi 5Raspberry Pi 420406080100SE +/- 0.06, N = 3SE +/- 1.04, N = 326.9980.16-fopenmp - MIN: 26.46 / MAX: 48.25-O3 -lgomp -lpthread -pthread - MIN: 77.99 / MAX: 94.431. (CXX) g++ options: -rdynamic

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: resnet50Raspberry Pi 5Raspberry Pi 460120180240300SE +/- 0.11, N = 3SE +/- 0.19, N = 359.19273.48-fopenmp - MIN: 58.06 / MAX: 68.62-O3 -lgomp -lpthread -pthread - MIN: 271.84 / MAX: 295.41. (CXX) g++ options: -rdynamic

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: yolov4-tinyRaspberry Pi 5Raspberry Pi 450100150200250SE +/- 0.16, N = 3SE +/- 0.29, N = 359.31223.68-fopenmp - MIN: 58.45 / MAX: 104.55-O3 -lgomp -lpthread -pthread - MIN: 220.49 / MAX: 246.881. (CXX) g++ options: -rdynamic

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: squeezenet_ssdRaspberry Pi 5Raspberry Pi 4306090120150SE +/- 0.02, N = 3SE +/- 0.21, N = 343.62137.59-fopenmp - MIN: 42.99 / MAX: 44.47-O3 -lgomp -lpthread -pthread - MIN: 136.58 / MAX: 156.341. (CXX) g++ options: -rdynamic

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: regnety_400mRaspberry Pi 5Raspberry Pi 41530456075SE +/- 0.08, N = 3SE +/- 0.96, N = 312.4967.05-fopenmp - MIN: 12.19 / MAX: 13.12-O3 -lgomp -lpthread -pthread - MIN: 65.14 / MAX: 85.911. (CXX) g++ options: -rdynamic

yquake2

This is a test of Yamagi Quake II. Yamagi Quake II is an enhanced client for id Software's Quake II with focus on offline and coop gameplay. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is Betteryquake2 8.10Renderer: OpenGL 3.x - AF: Off - MSAA: On - Resolution: 1920 x 1080Raspberry Pi 5Raspberry Pi 450100150200250SE +/- 2.81, N = 4SE +/- 0.46, N = 3239.088.21. (CC) gcc options: -shared -lm -ldl -rdynamic -lSDL2 -O2 -pipe -fomit-frame-pointer -std=gnu99 -fno-strict-aliasing -fwrapv -fvisibility=hidden -MMD -fPIC

NCNN

NCNN is a high performance neural network inference framework optimized for mobile and other platforms developed by Tencent. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: vision_transformerRaspberry Pi 5Raspberry Pi 4400800120016002000SE +/- 0.51, N = 3SE +/- 15.11, N = 3649.122008.62-fopenmp - MIN: 626.69 / MAX: 668.91-O3 -lgomp -lpthread -pthread - MIN: 1900.59 / MAX: 2112.981. (CXX) g++ options: -rdynamic

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: FastestDetRaspberry Pi 5Raspberry Pi 4918273645SE +/- 0.08, N = 3SE +/- 0.17, N = 36.0940.83-fopenmp - MIN: 5.85 / MAX: 6.7-O3 -lgomp -lpthread -pthread - MIN: 40.04 / MAX: 62.821. (CXX) g++ options: -rdynamic

yquake2

This is a test of Yamagi Quake II. Yamagi Quake II is an enhanced client for id Software's Quake II with focus on offline and coop gameplay. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is Betteryquake2 8.10Renderer: OpenGL 3.x - AF: On - MSAA: On - Resolution: 1920 x 1080Raspberry Pi 5Raspberry Pi 450100150200250SE +/- 1.86, N = 3SE +/- 0.66, N = 3230.087.41. (CC) gcc options: -shared -lm -ldl -rdynamic -lSDL2 -O2 -pipe -fomit-frame-pointer -std=gnu99 -fno-strict-aliasing -fwrapv -fvisibility=hidden -MMD -fPIC