sfx2000 Posted October 4, 2018 Share Posted October 4, 2018 16 hours ago, tkaiser said: 100% meaningless numbers now since they do not provide any insights any more (you can not differentiate what's throttling and what's -- potentially inappropriate -- cpufreq scaling behavior). For two simple reasons: allowing very low cpufreq OPP trashes performance (on your Tinkerboard it's storage performance due to the clockspeeds not increasing fast enough) the way DVFS works at the lower end of the scale the differences in consumption and generated heat are negligible between silly ultra low MHz values and a reasonable lower limit (we in Armbian TEST for stuff like this with every new SoC family!) Closed your issue right now spending additional time to explain: https://github.com/ThomasKaiser/sbc-bench/issues/4 So you're suggesting your benchmark script is meaningless? I don't think so - it's very useful... and can be more useful in doing A/B testing - rather than set params, report them - and let the tester review and test as needed. the rk3288-tinker was armbian stock clocks - no tweaks there... it's just hopelessly impaired in a multiple thread environment - it gets heatsoaked, and no chance to recover with stock armbian - the numbers show that. The whole rk3288-tinker problem is something that is really driving me how to sort it - using what is on hand, and no code-impact to armbian - so going down the road of how to make the best use of resources available... Challenges at the moment with tinker is that armbian does do the lower limit, and still sorting out if the DT updates on mainstream have been pulled in as 4.14 has a lot of changes on the mainline. Hint on today's testing - I chose the Allwinner H3 device from FA - the NEO, as this is a well tested device (as you know) - just playing around with CPU schedulers... if you want the worksheet with backup on the testing - reach out to me directly... Yes - I know UnixBench is kind of relative - useless across different machines/platforms - but in relative testing, with the same board/kernel/compiler, it's very useful - as you can see the time to complete the actual test, each sequence is almost an hour under test - so fair chance that thermals would be the biggest problem... The NEO H3 - It doesn't throttle, but it's a fast race to minimum across performance/ondemand/schedutil... single thread shows some regression, but four threads bear interesting stuff when doing some fun with UnixBench... as you can see on the timeline - more than enough time for the H3 to get throttled, but it never does... did performance as a parity/sanity check - but the sample sizes across the benchs do suggest a high confidence interval on the results. Last run - I pushed the H3 into powersave to bring the clocks down, otherwise it would have been 4 cores running at max - 4 cores burning on performance, I suppose temps would have eventually come down - I let it run a bit - did a stint in powersave mode to get to a good entry point for temps at 40C on for a good entry point on ondemand and schedutil - they would always jump down to minclock there - perf wants to be fast in single thread - but these days... and with big.little, woe to be the kthread on the wrong core... NOTE - I've highlighted the performance regressions across things... and they deserve further study. With cpufreq schedutil, might improve... Looking forward - getting to a point where benchmarks can be useful, automated, and reproducible... ARM's LISA looks really good for the CI workflow - and it opens up the black boxes that most benchmarks do... https://github.com/ARM-software/lisa All told - upcoming items like affordable big.little boards, preempt-rt, and cpufreq coming into play is going to complicate things... it's the upcoming perfect storm for the team, so get ahead of it... I'm here to help - not trying to get into conflict - I've got more than a fair amount of time doing embedded stuff on arm - on QCOM, Marvell, Freescale, and TI - android and emdedded - I'm not crazy like the stuff over on xda, where things go strange. I'd rather work with you than against you - against you, I'll lose, that's true, but the community might also lose on the investment of my time and effort to contribute. 0 Quote Link to comment Share on other sites More sharing options...
NicoD Posted October 29, 2018 Share Posted October 29, 2018 I've just finished Benchmarking most of my sbc's. This is done to see how well benchmark tools perform. Conclussion is that I've found no perfect cpu-benchmark tool. With every tool there is a problem. 7-zip does good in single core tasks(100%) but with multicore tasks it doesn't use 100% of all cores. This percentage also differs from distro and from sbc to sbc. It also does a little bit better on 32-bit systems. With blender different versions perform different. Older versions are faster. It also performs a lot better on 64-bit. Sysbench does different tasks with different versions. It performs about 10x better on 64-bit systems. CPU-Miner only works on 64-bit systems. It performs best on Bionic. Gimp only uses 1 core. Performs better on 64-bit. GTKPerf tests desktop speed. It only uses 1 core. It doesn't give reliable information. Here all my results. I'll make a video about it by the end of the week. If anyone has suggestions. Please let me know. Reasons why benchmark tools can give different results ------------------------------------------------------ throttling 32-bit/64-bit Difference in cores A53/A7/A15/A72 distro (ubuntu/debian...) distro version kernel version driver versions compiler version software version/outdated repositories desktop Mate/Xfce/LXDE/... display resolution/headless background processes cpu clockspeed ram clockspeed/latency ram useage/swap/zram process sheduler optimizations for the system/distro crypto engine for encryption Undervoltage config settings Wifi dongle uses recources 7-zip works a bit better on 32-bit vs 64-bit, it doesn't use all cores at 100% in multi-core scores. The percentage differs with different distro's and boards. So it's not completely exact. Blender works a lot better on 64-bit than on 32-bit. It uses 100% of the cores. Sysbench works 10x better on 64-bit, different versions give different results. Version 1.xx does the test 10s and gives the amount of events. GIMP only uses 1 core, works better on 64-bit. GTKPerf tests desktop speed. It only uses 1 core. CPU Miner only works on 64-bit. Works better in Ubuntu Bionic than in Debian Stretch/Ubuntu Xenial. Blender : BMW render @ 1080p Gimp : BMW render result 1080p Filters -> Artistic -> Van Gogh -> ok Sysbench : sysbench --test=cpu --cpu-max-prime=20000 --num-threads="number of threads" run 7-zip : Numbers are average of 3 of decompressing only All tests are done with a fan when necessary so no throttling occurs. 64-bit SBC's NanoPC T3+ |SBC bench result |CPU Miner |7-zip s/c |7-zip b/c |7-zip multi avg. of 3 |Blender |GIMP |GTKPerf |Sysbench Armbian Bionic http://ix.io/1iRJ 10.99kH/s 1290 10254 1h10m25s 1m24s 10.11s 21692 Arbmian Stretch http://ix.io/1qiF 8.55kH/s 1275 10149 1h13m55s 1m32s 11.06s 3.2s NanoPi M4 |SBC bench result |CPU Miner |7-zip s/c |7-zip b/c |7-zip multi avg. of 3 |Blender |GIMP |GTKPerf |Sysbench Armbian bionic http://ix.io/1nLh 10.23kH/s 1335 2005 8352 1h13m50s 0m29s5 5.06s 26763 Armbian bionic nightly http://ix.io/1pDo 10.24kH/s 1329 1990 8292 1h13m28s 0m29s 5.12s 26733 Armbian stretch desktop http://ix.io/1odF 8.66kH/s 1350 1977 8400 1h14m12s 0m31s 5.24s 3.1s Armbian stretch dsk nightly //ix.io/1pM0 8.80kH/s 1359 1993 8500 1h15m04s 0m31s 5.32s 3.3s Armbian stretch core no fan //ix.io/1pKU 8.80-8.65kH/s 1353 1989 8461 Armbian stretch core //ix.io/1pL9 8.76kH/s 1354 1988 8456 Armbian stretch core nightly //ix.io/1pLf 8.82kH/s 1357 1994 8494 Lubuntu Bionic arm64 http://ix.io/1oGJ 9.24kH/s CPU Miner 1056 1551 6943 1h28m13s Lubuntu Bionic armhf http://ix.io/1pJ1 1111 1769 7705 2h02m54s 0m57s 6.97s 1666 32-bit Lubuntu Xenial armhf http://ix.io/1oCb 989 1507 6339 2h20m51s 0m59s 49.77s 49.7s 32-bit Khadas Vim2 Max |SBC bench result |CPU Miner |7-zip s/c |7-zip b/c |7-zip multi avg. of 3 |Blender |GIMP |GTKPerf |Sysbench Ubuntu Xenial http://ix.io/1qkA 6.86kH/s 823 1134 6682 1h14m39s 1m53s 16.26s 3.8s Armbian Bionic http://ix.io/1qnY 8.55kH/s 921 1272 7464 1h17m52s 1m32s 12.54s 19035 Odroid C2 |SBC bench result |CPU Miner |7-zip big core |7-zip multi avg. of 3 |Blender |GIMP |GTKPerf |Sysbench Armbian Stretch Core http://ix.io/1pZu 4.65kH/s 1390 5342 Armbian Stretch Core Nightly //ix.io/1pZJ 4.66kH/s 1391 5340 Armbian Stretch Desktop http://ix.io/1q1C 4.65kH/s 1394 5363 1m23s 11.66s 5.96s Armbian Stretch Desktop NGHT //ix.io/1p02 4.59kH/s 1394 5356 2h38m18s 1m23s 12s 6.0s Ubuntu Mate Bionic http://ix.io/1q2S clocked to 100Mhz 2h35m10s 1m17s 10.01s 12026 Ubuntu Mate Bionic OC Doesn't work/Clocked to 100Mhz 1607 5960 2h10m21s 1m09s 8.94s 13755 Rock64 |SBC bench result |CPU Miner |7-zip small core |7-zip multi avg. of 3 |Blender |GIMP |GTKPerf |Sysbench Armbian Stretch 1.5Ghz http://ix.io/1nCj 4.06kH/s 1406 5407 3h00n32s 1m39s 15.91s 7.0s OLD Armbian Stretch 1.3Ghz //ix.io/1iHB 3.80kH/s 1211 4904 Armbian Bionic 1.5Ghz core //ix.io/1qbK 5.00kH/s 1384 5379 10.0s Armbian Bionic 1.5Ghz dsk //ix.io/1qcb 4.94kH/s 1379 5326 2h55m56s 1m31s 15.00s 10172 32-bit SBC's Odroid XU4 |SBC bench result |7-zip s/c |7-zip b/c |7-zip multi avg. of 3 |Blender |GIMP |GTKPerf |Sysbench Debian Jessie http://ix.io/1q6X 950 1653 8823 1h12m19s 1m08s 18.53s 41.3s Ubuntu Bionic http://ix.io/1qbL 1219 2094 9395 1h44m19s 1m10s 14.36s 2200 Asus Tinker board |SBC bench result |7-zip big core|7-zip multi avg. of 3 |Blender |GIMP |GTKPerf |Sysbench Tinker OS 9.5 Stretch http://ix.io/1pRN 1983 7536 2h55m00s 1m19s 189.82s 63.7s Raspberry Pi 3B+ |SBC bench result |7-zip small core |7-zip multi avg. of 3 |Blender |GIMP |GTKPerf |Sysbench Raspbian Default no fan http://ix.io/1q10 1471 5027 2m09s 9.85s 88.2s Raspbian Default http://ix.io/1q1Q 1411 5371 5h47m31s 2m09s 10.04 79.5s Raspbian OC http://ix.io/1q5J 1591 6141 1m55s 8.81s 70.8s Ubuntu Mate Xenial http://ix.io/1q65 7-zip didn't work 2m17s 11.71s 90.5s Raspberry Pi 3B |SBC bench result |7-zip small core |7-zip multi avg. of 3 |Blender |GIMP |GTKPerf |Sysbench Raspbian Stretch : http://ix.io/1qob 1220 4681 2m31s 10.97 93s Software versions ----------------- GIMP Blender GTKPerf SysBench SBC-bench M4 : Lubuntu Xenial armhf 2.8.18 2.79b 0.40 0.4.12 0.6.1 Lubuntu Bionic armhf : 2.8.22 2.79b 0.40 1.0.11 LuaJIT 2.1.0-beta3 0.6.1 Armbian Stretch desktop 9.5 : 2.8.18 2.79b 0.40 0.4.12 0.6.1 Armbian Bionic : 2.8.22 2.79b 0.40 1.0.11 LuaJIT 2.1.0-beta3 0.6.1 Tinker : TinkerOS 9.5 Stretch : 2.8.18 2.79b 0.40 0.4.12 0.6.1 Odroid C2 : Armbian Stretch 9.5 : 2.8.18 2.79b 0.40 0.4.12 0.6.1 : Ubuntu Mate Bionic : 2.8.22 2.79b 0.40 1.0.11 LuqJIT 2.1.0-beta3 0.6.1 Doesn't work clocks to 100Mhz Rock64 : Armbian Stretch 9.5: 2.8.18 2.79b 0.40 0.4.12 0.6.1 : Armbian Bionic : 2.8.22 2.79b 0.40 1.0.11 0.6.2 RPi 3B : Raspbian Stretch 9.5 : 2.8.18 2.78a 0.40 0.4.12 0.6.2 RPi 3B+ : Raspbian Stretch 9.5 : 2.8.18 2.78a 0.40 0.4.12 0.6.1 Ubuntu Mate Xenial : 2.8.16 0.40 0.4.12 0.6.1 Odroid XU4 : Debian Jessie : 2.8.14 2.72b 0.40 0.4.12 0.6.1 7-zip doesn't work : Ubuntu Bionic : 2.8.22 2.79b 0.40 1.0.11 0.6.2 NanoPC T3+ : Armbian Bionic : 2.8.22 2.79b 0.40 1.0.11 0.4.6 Armbian Stretch : 2.8.18 2.79b 0.40 0.4.12 0.6.2 Khadas Vim2 Max : Ubuntu Xenial : 2.8.16 2.76b 0.40 0.4.12 0.6.2 Armbian Bionic : 2.8.22 2.79b 0.40 1.0.11 0.6.2 CPU Clocks ---------- NanoPi M4 : Armbian Bionic/Stretch : 2x2Ghz + 4X1.5Ghz 64-bit Lubuntu armhf/ARM64 : 2x1.8Ghz + 4X1.4Ghz armhf 32-bit / ARM64 64-bit Tinker Board : TinkerOS Stretch : 4x1.8Ghz 32-bit Odroid C2 : Armbian Stretch : 4x1.5Ghz 64-bit Ubuntu Mate Bionic : 4x1.5Ghz RAM 912Mhz 64-bit Ubuntu Mate Bionic OC : 4x1.75Ghz + RAM 1104Mhz 64-bit Rock64 : Armbian Stretch : 4x1.5Ghz 64-bit Armbian Bionic : 4x1.5Ghz 64-bit RPi 3B+ : Raspbian Stretch : 4x1.4Ghz no fan 4x1.2Ghz above 60°C 32-bit Raspbian Stretch OC : 4x1.570Ghz over_voltage=4 core_freq=500 sd_freq=510 32-bit Ubuntu Xenial : 4x1.4Ghz 32-bit Odroid XU4 : Debian Stretch : 4x1.4Ghz + 4x1.9Ghz 32-bit : Ubuntu Mate Bionic : 4x1.5Ghz + 4x2Ghz Underclocks when above 75°C 32-bit NanoPC T3+ : Armbian Bionic : 8x1.4Ghz 64-bit Some benchmark tools can give an estimate of the performance. But they are never an exact reflection. 0 Quote Link to comment Share on other sites More sharing options...
sfx2000 Posted October 30, 2018 Share Posted October 30, 2018 5 hours ago, NicoD said: I've just finished Benchmarking most of my sbc's. This is done to see how well benchmark tools perform. Conclussion is that I've found no perfect cpu-benchmark tool. With every tool there is a problem. Well - benchmarks are going to try to measure something specific... Toss in UnixBench across the boards -- https://github.com/sfx2000/byte-unixbench.git 0 Quote Link to comment Share on other sites More sharing options...
NicoD Posted October 30, 2018 Share Posted October 30, 2018 4 hours ago, sfx2000 said: Toss in UnixBench across the boards -- https://github.com/sfx2000/byte-unixbench.git Now too late. I`d need to prepare every image again. I can always add it later. I`m done for a while with benchmarking. It ain`t much fun. 0 Quote Link to comment Share on other sites More sharing options...
NicoD Posted November 1, 2018 Share Posted November 1, 2018 My video about benchmarking CPU`s of SBC`s. Here all my data gathered Reasons why benchmark tools can give different results ------------------------------------------------------ throttling 32-bit/64-bit Difference in cores A53/A7/A15/A72 distro (ubuntu/debian...) distro version kernel version driver versions compiler version software version/outdated repositories desktop Mate/Xfce/LXDE/... display resolution/headless background processes cpu clockspeed ram clockspeed/latency ram useage/swap/zram process sheduler optimizations for the system/distro crypto engine for encryption Undervoltage config settings Wifi dongle uses recources 7-zip works a bit better on 32-bit vs 64-bit, it doesn't use all cores at 100% in multi-core scores. The percentage differs with different distro's and boards. So it's not completely exact. Blender works a lot better on 64-bit than on 32-bit. Older versions are faster. It uses 100% of the cores. Sysbench works 10x better on 64-bit, different versions give different results. Version 1.xx does the test 10s and gives the amount of events. GIMP only uses 1 core, works better on 64-bit. GTKPerf tests desktop speed. It only uses 1 core. CPU Miner only works on 64-bit. Works better in Ubuntu Bionic than in Debian Stretch/Ubuntu Xenial. Blender : BMW render @ 1080p Gimp : BMW render result 1080p Filters -> Artistic -> Van Gogh -> ok Sysbench : sysbench --test=cpu --cpu-max-prime=20000 --num-threads="number of threads" run 7-zip : Numbers are average of 3 of decompressing only All tests are done with a fan when necessary so no throttling occurs. 64-bit SBC's NanoPC T3+ |SBC bench result |CPU Miner |7-zip s/c |7-zip b/c |7-zip multi avg. of 3 |Blender |GIMP |GTKPerf |Sysbench Armbian Bionic http://ix.io/1iRJ 10.99kH/s 1290 10254 1h10m25s 1m24s 10.11s 21692 Arbmian Stretch http://ix.io/1qiF 8.55kH/s 1275 10149 1h13m55s 1m32s 11.06s 3.2s NanoPi M4 |SBC bench result |CPU Miner |7-zip s/c |7-zip b/c |7-zip multi avg. of 3 |Blender |GIMP |GTKPerf |Sysbench Armbian bionic http://ix.io/1nLh 10.23kH/s 1335 2005 8352 1h13m50s 0m29s5 5.06s 26763 Armbian bionic nightly http://ix.io/1pDo 10.24kH/s 1329 1990 8292 1h13m28s 0m29s 5.12s 26733 Armbian stretch desktop http://ix.io/1odF 8.66kH/s 1350 1977 8400 1h14m12s 0m31s 5.24s 3.1s Armbian stretch dsk nightly //ix.io/1pM0 8.80kH/s 1359 1993 8500 1h15m04s 0m31s 5.32s 3.3s Armbian stretch core no fan //ix.io/1pKU 8.80-8.65kH/s 1353 1989 8461 Armbian stretch core //ix.io/1pL9 8.76kH/s 1354 1988 8456 Armbian stretch core nightly //ix.io/1pLf 8.82kH/s 1357 1994 8494 Lubuntu Bionic arm64 http://ix.io/1oGJ 9.24kH/s CPU Miner 1056 1551 6943 1h28m13s Lubuntu Bionic armhf http://ix.io/1pJ1 1111 1769 7705 2h02m54s 0m57s 6.97s 1666 32-bit Lubuntu Xenial armhf http://ix.io/1oCb 989 1507 6339 2h20m51s 0m59s 49.77s 49.7s 32-bit Khadas Vim2 Max |SBC bench result |CPU Miner |7-zip s/c |7-zip b/c |7-zip multi avg. of 3 |Blender |GIMP |GTKPerf |Sysbench Ubuntu Xenial http://ix.io/1qkA 6.86kH/s 823 1134 6682 1h14m39s 1m53s 16.26s 3.8s Armbian Bionic http://ix.io/1qnY 8.55kH/s 921 1272 7464 1h17m52s 1m32s 12.54s 19035 Odroid C2 |SBC bench result |CPU Miner |7-zip big core |7-zip multi avg. of 3 |Blender |GIMP |GTKPerf |Sysbench Armbian Stretch Core http://ix.io/1pZu 4.65kH/s 1390 5342 Armbian Stretch Core Nightly //ix.io/1pZJ 4.66kH/s 1391 5340 Armbian Stretch Desktop http://ix.io/1q1C 4.65kH/s 1394 5363 1m23s 11.66s 5.96s Armbian Stretch Desktop NGHT //ix.io/1p02 4.59kH/s 1394 5356 2h38m18s 1m23s 12s 6.0s Ubuntu Mate Bionic http://ix.io/1q2S clocked to 100Mhz 2h35m10s 1m17s 10.01s 12026 Ubuntu Mate Bionic OC Doesn't work/Clocked to 100Mhz 1607 5960 2h10m21s 1m09s 8.94s 13755 Rock64 |SBC bench result |CPU Miner |7-zip small core |7-zip multi avg. of 3 |Blender |GIMP |GTKPerf |Sysbench Armbian Stretch 1.5Ghz http://ix.io/1nCj 4.06kH/s 1406 5407 3h00n32s 1m39s 15.91s 7.0s OLD Armbian Stretch 1.3Ghz //ix.io/1iHB 3.80kH/s 1211 4904 Armbian Bionic 1.5Ghz core //ix.io/1qbK 5.00kH/s 1384 5379 10.0s Armbian Bionic 1.5Ghz dsk //ix.io/1qcb 4.94kH/s 1379 5326 2h55m56s 1m31s 15.00s 10172 32-bit SBC's Odroid XU4 |SBC bench result |7-zip s/c |7-zip b/c |7-zip multi avg. of 3 |Blender |GIMP |GTKPerf |Sysbench Debian Jessie http://ix.io/1q6X 950 1653 8823 1h12m19s 1m08s 18.53s 41.3s Ubuntu Bionic http://ix.io/1qbL 1219 2094 9395 1h44m19s 1m10s 14.36s 2200 Underclocks above 75°C Asus Tinker board |SBC bench result |7-zip big core|7-zip multi avg. of 3 |Blender |GIMP |GTKPerf |Sysbench Tinker OS 9.5 Stretch http://ix.io/1pRN 1983 7536 2h55m00s 1m19s 189.82s 63.7s Raspberry Pi 3B+ |SBC bench result |7-zip small core |7-zip multi avg. of 3 |Blender |GIMP |GTKPerf |Sysbench Raspbian Default no fan http://ix.io/1q10 1471 5027 2m09s 9.85s 88.2s Raspbian Default http://ix.io/1q1Q 1411 5371 5h47m31s 2m09s 10.04 79.5s Raspbian OC http://ix.io/1q5J 1591 6141 1m55s 8.81s 70.8s Ubuntu Mate Xenial http://ix.io/1q65 7-zip didn't work 2m17s 11.71s 90.5s Raspberry Pi 3B |SBC bench result |7-zip small core |7-zip multi avg. of 3 |Blender |GIMP |GTKPerf |Sysbench Raspbian Stretch : http://ix.io/1qob 1220 4681 2m31s 10.97s 93s Orange Pi+2 |SBC bench result |7-zip small core |7-zip multi avg. of 3 |Blender |GIMP |GTKPerf |Sysbench Armbian Bionic http://ix.io/1qtf 1050 4121 2m56s 19.07s 773 Software versions ----------------- GIMP Blender GTKPerf SysBench SBC-bench NanoPC T3+ : Armbian Bionic : 2.8.22 2.79b 0.40 1.0.11 0.4.6 Armbian Stretch : 2.8.18 2.79b 0.40 0.4.12 0.6.2 GIMP Blender GTKPerf SysBench SBC-bench M4 : Lubuntu Xenial armhf 2.8.18 2.79b 0.40 0.4.12 0.6.1 Lubuntu Bionic armhf : 2.8.22 2.79b 0.40 1.0.11 LuaJIT 2.1.0-beta3 0.6.1 Armbian Stretch desktop 9.5 : 2.8.18 2.79b 0.40 0.4.12 0.6.1 Armbian Bionic : 2.8.22 2.79b 0.40 1.0.11 LuaJIT 2.1.0-beta3 0.6.1 Khadas Vim2 Max : Ubuntu Xenial : 2.8.16 2.76b 0.40 0.4.12 0.6.2 Armbian Bionic : 2.8.22 2.79b 0.40 1.0.11 0.6.2 Odroid C2 : Armbian Stretch 9.5 : 2.8.18 2.79b 0.40 0.4.12 0.6.1 : Ubuntu Mate Bionic : 2.8.22 2.79b 0.40 1.0.11 LuqJIT 2.1.0-beta3 0.6.1 Doesn't work clocks to 100Mhz Rock64 : Armbian Stretch 9.5: 2.8.18 2.79b 0.40 0.4.12 0.6.1 : Armbian Bionic : 2.8.22 2.79b 0.40 1.0.11 0.6.2 Odroid XU4 : Debian Jessie : 2.8.14 2.72b 0.40 0.4.12 0.6.1 7-zip doesn't work : Ubuntu Bionic : 2.8.22 2.79b 0.40 1.0.11 0.6.2 Tinker : TinkerOS 9.5 Stretch : 2.8.18 2.79b 0.40 0.4.12 0.6.1 RPi 3B+ : Raspbian Stretch 9.5 : 2.8.18 2.78a 0.40 0.4.12 0.6.1 Ubuntu Mate Xenial : 2.8.16 0.40 0.4.12 0.6.1 RPi 3B : Raspbian Stretch 9.5 : 2.8.18 2.78a 0.40 0.4.12 0.6.2 Orange Pi+2 : Armbian Bionic : 2.8.22 0.40 1.0.11 0.6.2 CPU Clocks ---------- NanoPC T3+ : Armbian Bionic : 8x1.4Ghz 64-bit NanoPi M4 : Armbian Bionic/Stretch : 2x2Ghz + 4X1.5Ghz 64-bit Lubuntu armhf/ARM64 : 2x1.8Ghz + 4X1.4Ghz armhf 32-bit / ARM64 64-bit Khadas Vim 2 : Xenial/Bionic : 4x1Ghz + 4x1.5Ghz 64-bit Odroid C2 : Armbian Stretch : 4x1.5Ghz 64-bit Ubuntu Mate Bionic : 4x1.5Ghz RAM 912Mhz 64-bit Ubuntu Mate Bionic OC : 4x1.75Ghz + RAM 1104Mhz 64-bit Rock64 : Armbian Stretch : 4x1.5Ghz 64-bit Armbian Bionic : 4x1.5Ghz 64-bit Odroid XU4 : Debian Stretch : 4x1.4Ghz + 4x1.9Ghz 32-bit : Ubuntu Mate Bionic : 4x1.5Ghz + 4x2Ghz Underclocks when above 75°C 32-bit Tinker Board : TinkerOS Stretch : 4x1.8Ghz 32-bit RPi 3B+ : Raspbian Stretch : 4x1.4Ghz no fan 4x1.2Ghz above 60°C 32-bit Raspbian Stretch OC : 4x1.570Ghz over_voltage=4 core_freq=500 sd_freq=510 32-bit Ubuntu Xenial : 4x1.4Ghz 32-bit RPi 3B : Raspbian Stretch : 4x1.2Ghz 32-bit Orange Pi+2 : Armbian Bioic : 4x1.3Ghz 32-bit Some benchmark tools can give an estimate of the performance. But they are never an exact reflection. 0 Quote Link to comment Share on other sites More sharing options...
sfx2000 Posted January 5, 2019 Share Posted January 5, 2019 On 10/29/2018 at 10:33 PM, NicoD said: Now too late. I`d need to prepare every image again. I can always add it later. I`m done for a while with benchmarking. It ain`t much fun. that's ok, but what I have found is that UnixBench a good equalizer across different SoC's on these cheap little boards - even though UnixBench is still variable based on the compiler... It runs long enough (a full run can be 28 minutes) to show if thermals are a challenge (for TinkerBoard, absolutely with stock config compared to NanoPI NEO, where the Tinker has a much better architecture (A12) vs. NEO with A7, but the Neo can shed heat faster, and outperform the Tinker in the longer run - heck, even Pi3B+ runs UnixBench faster than Tinker... 1 Quote Link to comment Share on other sites More sharing options...
sfx2000 Posted January 6, 2019 Share Posted January 6, 2019 On 8/8/2018 at 2:05 AM, gprovost said: I get the following result for OpenSSL speed OpenSSL results: type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes aes-128-cbc 1280.56k 5053.40k 18249.13k 52605.27k 102288.04k 109390.51k aes-128-cbc 1285.51k 5030.68k 18256.13k 53001.90k 100128.09k 109188.44k aes-192-cbc 1276.82k 4959.19k 18082.22k 51421.53k 96897.71k 103093.59k aes-192-cbc 1290.35k 4961.09k 17777.24k 51629.74k 95647.06k 102596.61k aes-256-cbc 1292.07k 5037.99k 17762.90k 50542.25k 92782.59k 98298.54k aes-256-cbc 1281.35k 5050.94k 17874.77k 49915.90k 93164.89k 98822.83k In order to leverage on hw crypto engine, I had no choice but to use OpenSSL 1.1.1 lib (openssl-1.1.1-pre8) and I decided to use cryptodev-linux instead of AF_ALG since it gives me slightly better result (+5-10%). Something is wrong with your numbers... Armada 3720 without CESA - running the 3720 at 1GHz... all core - it's been my experience that the crypto accelerators are more of an offload rather than make things fast as arch's update and include instructions that support crypto... OpenSSL 1.1.0g 2 Nov 2017 $ openssl engine (dynamic) Dynamic engine loading support openssl speed -evp (crypto scheme) -elapsed type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes aes-128-cbc 100933.96k 305030.91k 580164.10k 767672.32k 848975.19k 862513.83k aes-256-cbc 94472.90k 248993.56k 411821.48k 501905.41k 536070.83k 534325.93k aes-128-gcm 66007.09k 179554.62k 332169.39k 424934.06k 460079.10k 460887.38k CESA makes a lot of sense with the ArmadaXP and Armada 38x as older ARM - ARMv8/aarch64 can change things quite a bit... 0 Quote Link to comment Share on other sites More sharing options...
hgabor47 Posted October 22 Share Posted October 22 I ran the sysbench on my Orange Pi Zero 3 1,5GB RAM DDR4 Allwinner H618 WiFi Bluetooth BLE Mini PC Orange Pi Zero3 Development Board Single Board Computer. sysbench --test=cpu --cpu-max-primt=20000 run Test: CPU 295.19 events per second (10sec = 2954) 1 thread = 295 EPS 0 Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.