Six SBC Benchmark: ODROID XU4, ROCKPro64 & More!

[@HongjianWang]

I believe the reported sysbench result is misinterpreted. For older version of sysbench, they run 10000 events and automatically stop. Meanwhile, the newer sysbench run 10s then stop. Therefore, the Rock64 processed 10k events within 7 seconds, while the RockPro64 processed 25k events in 10 seconds. The RockPro64 is definitely faster (given 50% more core counts).

[@ExplainingComputers]

Very helpful, thanks. Your comment led me to find this in a helpfile "While older versions of sysbench limit the number of events, sysbench >=1.0 limits the total execution time by default, in which case the total number of events or the number of events per second should be used as a performance indicator instead of the total execution time". So due to the different version of Sysbench, the comparative result for the RockPro64 is invalid -- and the RockPro64 is about 75% faster than the Rock64 and hence the winner in that test. Though I do find it bizarre that they have changed the test! A most strange thing to do.

[@leatheljamie]

@@ExplainingComputers makes sense 😂 love your channel btw, just found it and binge watching your videos :D

[@Gavinconaghty]

Yes, but the different version of tests are still comparable by dividing total events by total seconds. Results would be... Rasp pi 3 125 Rock64 1412 rock pro 2503 Odroid 258 Tinker 154 where higher is better. The reason Pi 3 is ~10 slower then Rock64 with same CPU must be the 32bit OS ? maybe? Great review. Thanks.

[@pdjames1729]

Yeah, the Rpi figures are anomalous - I know Raspian is ropey (Rpi-org are a bit lazy with 64bit refinements?) but theres no way 4x1.4 is one 20th the speed of 4x1.5 plus 2x2.0 - Or one 10th of 4x1.5 plus 4x1.5 - from base clock you'd expect to see about half, plus the architectural tweaks ... so maybe 3x the speed of a Pi3+. You notice all the 32bit distro's are flunking this in a big way... Again, that's likely to not be legit either. Regardless it does show that a #Rock64 is a bargain Mate' box... but I'd be getting the Pro with an Android distro xx I did a multi-boot noobs install on my first Pi - the difference between linux distros is quite substantial. Some could barely run VNC, others flew on it. So this is a useful overview but will often come down to how exactly the sbc is being used and which apps are best optimised for that hardware. xxRJ

[@movax20h]

10 second benchmark is semi useless (it is fine, if you know what you are looking for, and know how to properly interpret results) anyway, as the system will not reach thermal stability. You need ~100 seconds, to make sure you can sustain the performance. Also reporting any kind of benchmark results as time is very poor methodology. You should always convert to throughput if possible. (Only other case being explicit focus on latency, i.e. networking, server processing, etc). So do MB/s, or items/s, not seconds to finish X. The main reason is that to get reasonable results you need to gather big enough sample to remove a lot of semi random variations (I would additionally recommend to run it 3 times, and take the fastest result, not average). It also makes it easier to interpret results.

[@kennyluba4137]

Thank you for not abandoning the project. Straightforward tests, but nothing like this in the net! Thanks Christopher.

[@kennyluba4137]

Oh happy day! Prof Barnatt liked my comment. I’m a recent subscriber and watched all your videos. They are excellent and genuinely entertaining, please do not stop! Great work.

[@ExplainingComputers]

Thanks for the sub -- welcome aboard! :)

[@perrymcclusky4695]

Glad you didn’t have to abandon this video! I find your SBC videos always fascinating.

[@cokeforever]

I'm really glad I've found your channel. You seem to be the perfect speaker on behalf of computing technology, and thou i hardly learn new things, it is still very very comforting to be in your virtual presence, a well educated guy with great speech skill and actually talking sense on our beloved computers ) Loving your channel more and more each day) p.s. your British is wonderful, i mean they spoil the language a lot these days, and yours is very very reference one for me as non-native. Thank you!

[@RoboNuggie]

Thank you Christopher - you're on top form....watching this whilst drinking some green tea - can't be beat!

[@RoboNuggie]

I got beat ;-( Very nice Stephen....and some snacks?

[@RoboNuggie]

Even better.... man, you are making me hungry...roast potatoes, gravy....Yorkshire Puddings....

[@anderskirchenbauer3723]

Please remember that ARM bigLITTLE architecture is unusual in the computing world. In most threading applications the workload will be divided and passed equally to multiple threads. In ARM bigLITTLE the system is supposed to run little, and then when there is a workload toss it at the big cores to do the work and then power down again. What you are probably seeing here on all of the bigLITTLE chips is that an equal workload is being passed to the big and little cores, and it is spending most of its time waiting for the thread that were run on the little cores to return. You may get results more consistent with your expectations if you set your threads equal to the number of big cores instead of to the total core count so that it doesn't schedule on the big and the little cores at once and only runs on the big cores (mileage may vary based on kernel configuration). Additionally I respect your use of Raspbian on the Pi3, since the RPi foundation continues to keep it ARMv6 to be compatible with the original Pi and the Pi0, and it is what most people will use. If you want to benchmark Pi3 with 64-bit though, the aarch64 Fedora image will boot on a Pi3 out-of-the box.

[@ExplainingComputers]

Thanks for this. I must, as you note, test a 64 bit OS on the Pi 3B+ sometime. :)

[@mtbevins]

Thank you for your hard work. It is very challenging to benchmark so many dissimilar boards against each other so good job. I do think the real world program tests cary the most weight. Imagine a consumer tying to decide which computer to buy an Apple or A PC. One of the big determining factors for the consumer after comparing software library would be how well each ran the software. As always thanks for the great content and keep up the good work.

[@pdjames1729]

They are all Debian, to start with at least. As apples vs pears vs bananas comparisons go it's very useful. I'm not surprised Chris nearly gave up xx

[@darekvrnak3887]

"In the ideal world you're right." My top favorite quote from this channel

[@jonglass]

I think what would be interesting would be real-life-type tests--opening documents, doing something in LibreOffice (search and replace in Write or some kind of spreadsheet recalculating), or compiling a small project (like WordGrinder, for instance), and launch times for various apps. If someone wants to use one of these as a backup desktop replacement, I think these sorts of tests would be useful. :-) Oh, and thanks for sticking with it to the end. It was very much appreciated.

[@TomMannCenturia]

Great stuff as always, and the Pi Zero W still a brilliant bit of kit for lots of mini projects and it's inclusion was well worth it, as you said, to highlight how far we've come.

[@djk8541]

It's always a fantastic morning when ExplainingComputers and Techmoan both release a new video

[@evanj571]

Hi Mr Barnatt! I want to thank you for all your time and hard work. Know that you are one of my best and favorite youtubers. I always tune in to your channel eagerly to learn what you've got to offer and teach. Thank you and God bless

[@ExplainingComputers]

Thanks. :)

[@resrussia]

Thanks for another interesting video. By doing this type of video, you are contributing greatly to the knowledge of individuals using SBCs. As always, keep up the excellent you are doing for us!

[@stegoboard7495]

Paragraph says Raspberry Pi 2 B+, however it is the 3 B+. Very excellent work Chris. As always your videos are detailed, well filmed and straight forward.

[@ExplainingComputers]

Thanks -- now sorted. :)

[@fabiStgt]

So glad you didn't abandon this video! How about even doing a follow-up video with other benchmarks - you might want to take a look at the phoronix test suite for that! Anyway, keep up the good work, excellent video, as always! :)

[@ExplainingComputers]

I am considering an SBC media player head-to-head, and also working on a power consumption one. :)

[@weerobot]

No mini boards were harmed in this video... poor Rasp Pi Zero W

[@gameofcharging]

Raspberry pi zero or Zero w Poor because is a low cost board

[@imabeapirate]

Well done! I was really hoping for a comparison of power usage, since a lot of these have applications that run 24/7, but it was good to get an overview!

[@ExplainingComputers]

I am already working on a video about SBC power consumption! :) Featuring even more boards, but just power consumption tests.

[@sagarmakhija1994]

Clearly its a lots of hard work for creating this video's. Thanks for bringing such a great quality work to us.

[@IcicleFurry]

I was struggleing wich sbc i should take. Now I have a result and I am going to go the of the Rock64. Thank you, Christopher!

[@ExplainingComputers]

Excellent. The Rock64 is a nice board.

[@drapakdave]

Thank you so much for doing this - this is the result of a heck of a lot of work. It has helped me figure out which one I am going to buy...

[@quantuminfinity4260]

Thank you so much! I have been waiting for this video! I was just thinking I wonder if he would make that video about 2 days ago! But your video delivered all of the information in an even more comprehensive and overall nicer format then I had hoped!

[@jasongooden917]

I think running the tests with the OS most associated with the board is fine. Windows and Mac OS have been compared since the beginning, even PlayStation vs Xbox

[@kennethconnors5316]

all these boards have specific uses the ODROID fits my uses perfectly and for me best bang for the buck ,always like your excellent videos

[@manuel_youtube_ttt]

Good comparison, I didn't know those Rock boards, thank you for the detailed tables.

[@johnc3403]

you done well. I'm glad you stuck at it. The Rockpro64 is my next adventure off the back of your videos.. thank you

[@stephenrichards5386]

Thanks for both replies. I saw the LatteP review hoping to see your NUC review. Thanks again for your effort

[@ericartman0]

Thanks for all the info and the obvious effort it took compiling it. Wonderful stuff.

[@paullalonde5781]

Very good video, it help me in my choices. The material and the operating systems differences between the SBC tested are important and the services who are started at boot time and run in background on each machine can make a difference too. Thank-you M. Barnatt.

[@vishalbindel7660]

Excellent job done by ExplainingComputrs...... Head's off to you

[@larrywebber2971]

Thanks Chris. Good job! Also thanks to the commentators for pointing out the differences in the sysbench suite. I have an Odroid- XU4 and was not happy with the performance of the graphics so I created a non-GUI base on another SD card and installed XFCE. The results for me were markedly improved video performance, boot times, and overall stability when accessing via VNC. That said, you're right on for installing the most common OS builds.

[@coganng]

Great video about SBC. I find your videos about SBCs enjoyable. Time to solar power an SBC.

[@ExplainingComputers]

Ah, now that is a great idea! (Turns to a Pi Zero).

[@coganng]

By the way, will you ever run Windows on the Raspberry Pi 3? It is possible.

[@ExplainingComputers]

It is a real hack to run ARM Windows on a Pi 3, apparently . . .

[@coganng]

Will you ever give it a try and show us how it works?

[@bestbattle]

Looking forward to comparison with more SBCs! Thank you!

[@mindprobesven]

Great work on this video! Thanks for taking the time to do it.

[@B.I.T.E.]

Thank you so much Christopher. This question of performance to project and what board to use has stumped me. Now, I think the PI Zero will work just fine for some of them. Then the RockPro64 might have a spot for my DVR Motorbike capture plus more. Great video. I am so glad you didnt give up. I have learned alot from your videos and so appreciate your time..

[@factChecker01]

You do excellent work and made excellent decisions regarding the OS, boot disk, and tests to use. I really appreciate your efforts. Those are the results and decisions that are relevant to me.

[@SusanAmberBruce]

Always very interesting to see these benchmarks, Thank you for your hard work.

[@ritikbhambhani5656]

We glad to see your University Sir chirstopher ☺ we want learn much from you.

[@PaulFeakins]

I always remind my developers it's depreCated not depreCIated, I'm sure they love having such a pedantic boss! Great video as usual :)

[@mrmotofy]

Semantics are extremely important in life business or well...anything. Ignorant people whine and complain, but miscommunication creates problems. A chord is not a cord nor is a brake a break. Use the correct words to convey your thoughts.

[@ErnestGWilsonII]

I really like your videos! In this particular case I enjoyed your choice to run each board in the likely operating system flavor most folks would run. Although the results seemed a little strange, it did indeed point out the topic we all struggle with, price vs performance. Depending on a specific use case, saving money might make sense, while in a different scenario higher-performance might be needed. I thoroughly enjoyed this video as it gave me food for thought. As always, I am subscribed with notifications turned on and thumbs up!

[@ExplainingComputers]

Thanks for subscribing.

[@MarkTheMorose]

In terms of who was first to watch this video, I'm like the RasPi Zero: bringing up the rear. The latest RasPi continues to be the best all-rounder for its price, considering also the size of the user base and support. Nice to see that Mr Scissors and Stanley the knife got a rest this week. Small correction: video description says 'RasPi 2' instead of 3.

[@ExplainingComputers]

Oh dear, I am having typo problems this week . . .

[@BenQuigley]

I love you and your channel, this really reminds me of the BBC micro era style TV shows that were around back in the day.

[@walts555]

Thx Chris. Good comparison. Interesting comment details about sysbench. Changing a "benchmark test" does seem counter to the notion of "benchmark" and lands it into the category of jumbo shrimp and approximately accurate ;))

[@ExplainingComputers]

Very well put sir!

[@amancalledoss38]

Fantastic video, I loved every minute of this now let's overclock them all and do it again (only kidding). While I do love all of these boards I think the RPi still to this day is the best Pound/Dollar to performance single board out there or it could just be me being bias, all of them are exceptional value for money but its the huge community you are also buying into with the Pi and that part is priceless. On another note, Chris why haven't you made a Patreon account? I'm sure there are many of us out here who wait from Sunday to Sunday to watch another top class video from EC each week would like to be able to contribute towards the channel ?

[@ExplainingComputers]

Thanks for this -- the Pi certainly wins on value (and support). I keep pondering Patreon . . .

[@augurseer]

I'm a huge fan of the Orange PI as well. Well optimized. Neat variety of hardware

[@ThanassisTsiodras]

In case anyone is interested, on my Orange PI Zero (512MB, 4 cores, costs 9 euros in AliExpress) the same sysbench that Chris runs finishes in 93 seconds under Armbian. Not bad, given the price - almost the same performance as an RPI3 at 1/4th the price...

[@SergiuszRoszczyk]

Difference in sysbench may be related to instruction set. 32 bit ARM uses bloated, backwards compatible set, while 64 bit ARM uses clean, optimized instructions. It is not like in x86/x64 where instruction sets are compatible. ARM can switch between the two, so on 32-bit OS even 64-bit CPU work on bloated set. When working on 64-bit kernel it works on the optimized set.

[@thingsiplay]

The Raspberry Pi Zero have the big advantage to be much smaller and draining less power, which can be important in some scenarios. And I wish you would have done a Retro emulation gaming test, but this would be much more complex with more stuff going on (latency test, 2D and 3D performance).

[@merefield2585]

Thanks for all the great coverage of SBCs, Chris. Please cover the RISC V open-source initiative :)

[@jrherita]

Very awesome video EC! I also think these boards are a great selection of boards that have decent support from their OEMs. There are a ton of other interesting boards out there (InForce computing, etc) that often lack real support from the OEM or even community to make them 'easily usable' for projects or other purposes.

[@xybersurfer]

nice work. very interesting results. i'm guessing the operating system played a big role, although i didn't expect this Rock64 came through very well

[@allluckyseven]

Interesting results! I am still looking forward for that one killer application in order to get one of these, but it's great to see how they have evolved.

[@zbeekerm]

Hi Tom, I work for a *very* small company that does performance engineering for HPC (high performance computing, aka “super computers”). The other commenter could be correct that the results are due to different versions of sysbemch. If that is not the reason, then it could be possible that the benchmark software doesn’t load balance adequately for the asymmetric ROCKpro64 cores. But I would be inclined to believe that the result is in fact due to the version differences.

[@ExplainingComputers]

See pinned comment -- Sysbench has indeed changed its test!

[@DrNPCabd]

Awesome video, simple, straight, clear, fair. Sir, I salute you.

[@Freddie1980]

I've seen a few of your video's recently and I remembered I saw some of your earlier work from a few years ago and I have to say the quality of your video's has improved a lot (quality is maybe the wrong term perhaps presentation would be a better way of putting it). I found your video on the Odriod XU4 after watching a video about Sega Saturn emulation, apparently someone has written a version of Yabause (Sega Saturn emulator) that can run of a Obriod XU4 which is a remarkable feat of coding skill given the hardware limitations and the difficulty of emulating the Saturn. The take away from this video for me would be before buying any of these boards is given that this is hobby industry in many ways to get the best out of these devices you will want to buy a SBC that has excellent software support as that's as important if not more important then what hardware the device is running (the ODRIOD XU4 from the review is a prime example, on paper that SBC should have stomped all over it's rivals but it struggled to keep up with the Raspberry Pi 3 b+). Going back to my Saturn example I do wonder if something like the Rock64 can handle Yabause (apparently the Raspberry Pi 3 can't at the moment). Maybe this is avenue you could explorer in future video's there a lot of people out that who would to build an inexpensive emulating machine and a home media centre.

[@thewassock]

I've obtained similar Sysbench to the Rock64 figures on a Pi 3 B, by running a custom 64-bit Debian based installation. I gather the performance increase is due to 64-bit divides being performed in the processor instead of using a library routine to produce the same division. Performance hikes can also be achieved with a Pi 3 B on standard Raspbian by recompiling the sysbench binary for ARMv7 - standard Raspbian binaries are compiled for ARMv6 to ensure that they run on all Pi models.

[@ExplainingComputers]

Very interesting indeed. Thanks for sharing your 64 bit Pi test info.

[@davidhardaker192]

Thanks

[@christophejournoud2773]

You are right. Have a look on Libre Computer website, they talk about that in a comparative bench with Le Potato and Pi3

[@ErtugrulOzdemir-mf1gl]

finally i was waiting for this thanks

[@ha13151]

Five stars go to The Explaining Computers! ✯✯✯✯✯

[@charlesdeens8927]

Fantastic job. Fair and proper review.

[@AyadrixBF]

thank you so much for putting the time and effort in making this video!

[@nutsnproud6932]

Thanks for the video Chris. I use a cheap Chinese webcam and a Pi ZeroW to do a timelapse of the weather and it uploads to my web site once every 10 minutes. I have lots of Pi SBCs, one is a games emulator and one is running a NAS of 8 attached USB 2 drives via a hub just fine. Just an idea, if you would like viewers to run tests for you on certain SBC's you did not get time to do then I'm sure we would. Best wishes.

[@ExplainingComputers]

I very much like the Pi Zero W -- an amazing, small and good value board. I was working with one on a robot project most of today! :) Great idea about viewer benchmarking. Has got me thinking!

[@joshman5217]

Excellent video sir! Keep up the great work!

[@CognosSquare]

Great video. A computer + software is a complex thing to test indeed.

[@Battlenut1]

Great job, as always. I always look fwd to your upcoming videos. This is my favorite by far. I own 5 of 6 of these boards. Great to know which is best.

[@LiamGunes]

I like you’re methodological solution to the o/s availability issue. Keeping the difference consistent around a practical theme definitely helps when thinking about the entire vendor platform. Have you looked at running something like Kali Linux for a hardware to hardware comparison? Kali has ARM distros for a number of platforms, including a generic ARM version. It wouldn’t provide a purely comparable test, but it might provide a control mechanism to deal with optimization that occurs when hardware and o/s come from the same manufacturer. Full disclosure: I use Kali for some classes and presentations, so seeing it on other hardware platforms is interesting to me all by itself. Great video as always.

[@ExplainingComputers]

I will look into Kali, but suspect a version for the RockPro64 may be some time away . . .

[@Giblet535]

A *very* interesting benchmark, sir! Given the Sysbench difference, it might be interesting to see how quickly OpenCV can search a set of images for a shape on each board. I'm going to guess that Rock64 will win that as well, indicating either faster RAM or a faster path to RAM.

[@klontjespap]

i'm intrigued... i'm looking to build my own segway-style robot, with a lot of over-the-top features i'm pretty gash at electronics but i think my programming skills are decent enough to string something together, already picked out a ton of ICs from ali express that require the least amount of soldering, those which i can control through simple i2c or serial comms, and not requiring much else in the way of latches and analog circuitry , although i'm fine with having a couple of shift registers if need be. using a full-fledged linux computer with ansi C, and usb (and therefor support for regular usb webcams, sound cards, and a much wider variety of everything really) would take a load off my shoulder, although i''m got pretty decent at arduino C, becuase i like working with very limited hardware for the sake of doing things right. (yeah i'm one of those optimizing too early types, but hey it did teach me to get things like fast polling right, long before i even knew how to work interrupts lol) it may turn out to be an arduino nano knockoff (or 2 or 3) and one of these bad boys, if opencv performance is really worth doing anything real-time with (a lenient real-time, meaning 2-5 seconds) there sure seems something about the rockpro numbers in gimp that would indicate a faster link despite only being 4x2 instaed of 4x4 cpu cores, i figured, it was badly optimized to use all cores, it would tank instead of excel on the rockpro, because it only has 2 of the faster ones.

[@NetRolller3D]

Keep in mind that Raspbian is compiled for ARMv6, so it's essentially putting the 3B+'s CPU in an emulated mode.

[@dhgodzilla1]

Thank you for the video it was helpful for me as I do want to buy a few mini comps in the near future.

[@JB52520]

Hoorah! Thanks for finishing this video. I thought the Odroid would do better than it did, so this was educational. I still love mine. Even though its graphics performance is rather poor, it makes a great git server.

[@NomadicSage]

Another awesome video sir, it's really great learning from you. Please make more programming videos.

[@gusgone4527]

Excellent video, had me on the edge of my seat awaiting the head to head results. Or was it because my coffee was just out of reach? Either way, great fun for fellow SBC enthusiasts. In the next one, please bring back Mr Scissors. (Having first completed the necessary risk assessment. Clearly displayed in the work space.)

[@jazzochannel]

Correction! for the sysbench command the syntax will be identical while the parameters will change.

[@freesaxon6835]

Chris, you put a great deal of work into that one! The ' progress' in the evolution of these small factor boards is slow, of course they a specialist item in a way. Often wonder why a mainstream motherboard company hasn't pushed the frontier of micro ATX boards a little harder. Perhaps going the extra mile and producing ultra micro ATX board. They have the basic architecture worked out, but no, they don't go for that market, which could be a dual market place

[@ExplainingComputers]

Thanks. :) Ultra micro-ITX is Pico-ITX. Not very mainstream though . . .

[@freesaxon6835]

@@ExplainingComputers ah yes I meant ITX, ...... yes that's the point, mainstream motherboard manufacturers seem to 'blank' a possible market.

[@SouravTechLabs]

Rock64 and Tinker Board S clearly contradict each other by Synbench time and Gimp time!

[@ExplainingComputers]

:)

[@hansfehlow9657]

I think the the bottom line, for this scale of computer, what do you want to pay for the features. I have purchased many versions of the Pi, mostly because they are less expensive for tinkering. Lots of fun per buck. In my opinion the more expensive boards are not powerful enough for real day to day computing and therefore too expensive for tinkering. Thanks for the video, I do enjoy everyone!

[@mendoxx097]

It would be nice to watch the rockpro ones and the odroid UX4 actually on this same comparison with actual boards so we can see the improvements with time going on and how well or bad they get old aside the new odroid c2 or raspberry pi4 B for example

[@zetaconvex1987]

Thanks Chris. You obviously put a lot of work into this video. Judging by the results, the Pi3 really does hit the sweet spot between price and performance (IMO). The other ones are too pricey, and I'm sure if, ultimately, they're positioned well in the market to be sustainable. Another intangible benefit of the Pi3 is support, a much underrated factor when choosing a board. In a way, it's a pity that Linux is the only OS available for these boards. It's quite heavyweight, and it would be nice to see the emergence of an alternative. Maybe some unikernel design for the Pi0, perhaps? Or something lightweight like an AmigaOS. Or maybe they could just modernise RiscOS. Finally, I'd be interested in seeing a video on RISC-V, as I've been hearing it mentioned with increasing frequency.

[@ElmerFuddGun]

I think these tests show how one can cherry pick tests to suit the product they are selling. Just like automobile marketing or anything else you pick the tests that work best for you and worst for your competitors. But in the real world it depends on the end user's uses and comparing those exact uses that will determine the best product to use. It will be interesting to see how these test results change in the future when the same OS and version can be tested on all the boards. It's great having all this competition in this market that I wouldn't have thought possible only a few years ago.

[@VIPINSAINI_20]

Need More SBC Videos 🙍🙍🙍🙍🙍🙍🙍🙍 Love Your Work.😄😄😄

[@ExplainingComputers]

Thanks. I've some interesting reviews of new SBCs coming up -- including the LattePanda Alpha! :)

[@VIPINSAINI_20]

+ExplainingComputers 😄😄😄😄😄😄

[@glenswada]

I tried same tests on UDOO X86 (my desktop pc) and got 14sec and 9sec compared to RockPro64 10sec and 7.5 sec. Impressed.

[@fastjp72]

Why do you think the XU4 is so slow compared to the other faster SBC's?

[@ExplainingComputers]

It will be the heaviness of the operating system, and optimization (or lack thereof) for the board holding it back. The hardware itself is fast. In contrast, the Raspberry Pi 2 B+ has much slower hardware, but very well optimized software.

[@elgrangatopardo]

In my experience, depending on kernel and version of sysbench, new versions of sysbench always run the tests for 10 seconds and the performance is given for the number of events executed during that 10 seconds, so results are not comparable

[@ExplainingComputers]

I find this very hard to believe -- the test factors prime numbers to the value specified (here 20,000). The result in terms of the final time figure is the time taken to complete this task.

[@elgrangatopardo]

@@ExplainingComputers I runned before sysbench on ubuntu 16.04, and in 18.04 things changed and the test always run for 10 seconds, for any machine (you can try it too), I think this is what is happening with rock pro 64. Best regards

[@ExplainingComputers]

Sysbench is not a test. It is a utility that offers a range of tests of user defined parameters. The test I am using here requires the user to define the highest prime to factor, and then runs for the length of time that takes. I've no idea which sysbench test you have been running, and with which parameters. But the test I am using here I have used for many years on scores of OS and different computers, and it does not always run for 10 seconds :)

[@ExplainingComputers]

My apologies here -- I now stand corrected. The issue is not Ubuntu though, but the fact that Sysbench have changed their own test. Which I really cannot believe. To cite a helpfile: "While older versions of sysbench limit the number of events, sysbench >=1.0 limits the total execution time by default, in which case the total number of events or the number of events per second should be used as a performance indicator instead of the total execution time". So you are right that Sysbench by default now runs 10 seconds only, making comparisons between versions impossible. Sorry again.

[@cdl0]

Benchmark results for a cheap HP notebook with 1.35-GHz AMD E1-6010 APU (2 cores) AMD Radeon R2 Graphics and 4 Gb ram running 64-bit Ubuntu 16.04 are as follows: sysbench = 21.2 s; Gimp = 10 s. The machine is nearly four years old (i.e. October 2014) and the original price when new was about 170 GBP including 20% VAT and delivery. It seems to take ages to boot, but does run loads of stuff, and only has a slow hard disc, which is probably the limiting factor.

[@prashanthb6521]

You are awesome. Thanks for all your videos.

[@majorhavik6637]

Yes, running the tests with the OS most associated with the board makes sense. Really what choice do you have? You have to install the distro that the SBC maker has for it. But, why not have the two Rock64 boards in the test be the versions with the same amount of RAM? The "SBC Tested System" table showed the Rock64 had 2Gb and the Pro had 4Gb.

[@ExplainingComputers]

The Rock boards were the ones I had -- I bought the Rock64 earlier in the year, and was kindly supplied with the RockPro64 as a review sample. Practicalities! :)

[@majorhavik6637]

Ah, I see. I'm sure we can safely assume the results for the Rock64 "Boot" and "Gimp" tests would have been more like the Rock64 Pro results if both boards had 4gb anyways. Thanks for doing this kind of test. I'm sure it wasn't easy.

[@Henriquems856]

What a lovely SBC video!!

[@cheddercheese4910]

Hi Wich sbc is my best choice for gaming. Please let me know.

[@ExplainingComputers]

The Pi 3B+ and TinkerBoard are probably best supported for RetroPie and retro gaming.

[@andrewjameslyon]

Could you do a video testing and comparing the various video outputs on the RockPro64? I've found it nearly impossible to find out exactly WHAT screens are compatible with the DSI, eDP, etc. connections on various SBCs. Can I use salvaged mobile-phone screens via MIPI DSI? Laptop screens via eDP? Are there software solutions, or just expensive controller boards from eBay and AliExpress? Am I limited to 'official' proprietary screens for these SBCs?

[@beauslim]

good choice of boards to test. I have 3 and I'm wondering about the other 3. To be fair, I think most people run the Zero headless with Raspbian lite.

[@rtpman1953]

Thanks. Very interesting results indeed. Shifting the topic, many laptops come with a hard drive and m.2 slot where both can be used. I'd love to see a walkthrough of safely moving the OS from the HD to the m.2 drive, leaving the slower HD for storage. Love your videos.

[@ExplainingComputers]

Note that much such laptops will have a SATA and not NVMe M.2 port, though a speed improvement could still be had. An interesting idea. Noted! :)

[@CaptainSkurvy]

Thank you very much for yet another excellent informative video I also think the boot test using the best OS and Gimp show people the best view of the boards. ♿️🖖🇦🇺

[@emulatorretro]

The Rock64 Results make sense. I have installed emulators on my Rock64 4gb model running 64gb EMMC, and it can speed through Dreamcast and N64 games just fine. Not to mention that it can pop along through web pages and run my image editing programs from google store with no issue. I have yet to buy a Rock64 PRO, that will become my mobile desktop.

[@arthurdent8091]

Hi Chris, It's me Tom/ My prior question is this. I was trying to image my laptop with a mirrored 2TB external SATA hard drives. I don't leave it connected to minimize any chance of infection. I had neglected to delete an older image file and the unit ran out of room. It stopped dead in the middle of creating an image. I was nosing around through the Acronis software ( The Acronis rep swore that it positively was NOT anything having to do with Acronis ) when a message box came up offering to "disconnect" the device ( I was running through my Cat 5 port on the Buffalo enclosure and laptop) and since I was trying to restart the application I thought I could just reconnect the enclosure and get on with it. Despite my best efforts, I cannot find any reconnect option. I tried the Windows CMD line, Acronis application and power cycled everything for several days. No Luck. For a reason I cannot figure out I simply cannot get Win 10 to "see" / mount the drive(s). Since I suspect that Acronis is running a Linux variant under a Windows skin I was hoping that a Linux command might be able to get it to remount ergo my earlier email to you. If you have any suggestions/hints / ideas I am all ears. Thank you for your time.

[@ExplainingComputers]

I've been in a similar situation with an SSD not that long ago following a power failure on a Linux system. I eventually got out of it as follows. Though please do this with extreme care! First, connect the drive in Windows, and open a command prompt (right-clicking to "run as administrator"). Then type "diskpart" (do not enter the quotes!), and when you get a disk part prompt, enter "list disk", and you will get a list of drives. Then enter "select disk #", where # is the number of the drive you have issues with. Be very, very careful to get this right! Then enter "clean". This will remove absolutely everything from the disk (including all data and partitions). The disk will now need to be initialized as if it was new from the factory, but will now appear in Windows disk management, where you can set it up again (right click computer, manage, and disk management). Good luck -- and be careful not to wipe your boot drive! :)

[@arthurdent8091]

Thanks Chris. I will try it in the morning when I can approach it with a clear head. I learned that while my desire is great I should only do those this that can't be undone in the am. A long story. Again thanks.

[@rondlh20]

Amazing job again, very interesting. The huge differences in the results are surprising, perhaps the results should be compared, but I don't know if that is possible

[@willgilliam9053]

Rock64Pro is keeping up with an AMD A4 6300U using both cores doing the prime test. Quite impressive. Id love to see the results of cat /proc/cpuinfo for the two types of cores that its using. I love the idea of HMP computing with 4 PCI-E lanes. Can it used to power an nvidia graphics card? Some of them have to have arm drivers as well as 4x mode

[@shubhendragautam]

woowo...these results are awesome and very helpful in deciding to buy/use..thanks for this video...

[@ehs03y3ol]

I understand that the GIMP needs to use memory-CPU bandwitich, meanwhile, sysbench only uses CPU instructions to measure. That's why in real world application 64 bits arquitecture are being faster even due relative slower CPU, they do have double memory bandwich...

[@impatientmaker8469]

Thanks for this video! Very informative - much to ponder...

[@lmarloe]

Thank you. It became clear to me that coming PINEBOOK Pro will be not usable as desktop

[@pascalsampatsing5528]

Dear C. Barnatt, I'm a autodidact in learning mechatronica (Pi3 and arduino) via python. My question; Could you explain what Julia is, and why there is so much interest

[@C-MAGs]

I've always wanted a rock board. At the time of looking, there wasn't very many around. Anyway, I have the Pi 3 already and for twice the price, I'm not so sure on the Rock boards now. Hmmm.

[@ExplainingComputers]

The Pi 3 remains hard to beat, esp. when it comes to software. Though the Rock64 boards are nice with their USB 3.0 interface for some applications (eg NAS).

[@CarcharothQuijadasdelased]

The Rasp Pi 3B+ is behind in connectivity, in performance is really good. The best option for the price looks like is the ROCK64 2Gb, same price than Rasp 3, more connectivity, 64 bits, more memory. Rasp Zero still is a good option for low power, simple tasks or of you don't mind waiting more time.

[@CarcharothQuijadasdelased]

Yep, DDR2 is a bit obsolete, but at the end of the day the "RAW" specifications and synthetic benchmarks are "nothing" against "real world" performance (this test is very short as our boi Christopher said). ***What we really need and want is real optimization, thats why Pi3 made a nice job in GIMP (just compare launch titles and "end of life" games for any game console).