Dope Tech Episode 2

<<@ahmetrefikeryilmaz4432 says : The arm sbc is the perfect home/small business router, firewall, NAS and DVR and it can do digital signage on three screens possibly simultaneously.>> <<@msn3wolf says : I have a question regarding the tests. The synthetic tests show that the Rock5 system is slightly faster than the N100 system, however the real application tests shows that the N100 system is almost twice as performant than the Rock5 one. Why is that? Could it be that the applications used are optimized for x86_64 systems? It would make sense since the availability of ARM desktop systems is fairly recent. I wonder how the applications comparison would be in a more ARM optimized environment like Android. Could you test Android-x86 vs Android-Arm applications on the on Rock5/N100 systems?>> <<@sgkonfetti says : You should consider having a look at the Minisforum UM890 Pro. https://store.minisforum.de/products/minisforum-um890-pro-mini-pc?utm_source=google&utm_medium=cpcg&gad_source=1&variant=43028691189943>> <<@jeromeprin5668 says : hI! Could you test both with emulators like ps2 emulators? IS it possible to run those machines with an android OS?>> <<@ankersman says : I'm looking to upgrade my system. Being on 12 volt solar my reason is saying ARM but my heart is saying Intel. I would have to try hard to justify nearly 2 amps but best case my guess it would be consuming on average more like 1.>> <<@andresteenberg-yk5is says : that donssent anser my qasting telt my fraes it n diffirent way how can n 32 bit cpu go furder than 4gb of ram if 4 gb is the standerd>> <<@geraldmansfield2631 says : Do you have an email address? I would like council on finding a computer to replace the origional on my 1999 Dodge ram Van. Dodge ram computers are insufficient.>> <<@andresteenberg-yk5is says : i dont understand how does n 86x base setem have 8 gb of ram an can only go to 4 gb ram of ram becuse it is n 32 bit cpu but can it gohigher than 4 gb ram>> <<@sprockkets says : I think the H4 Ultra in a mini itx case is now the next computer I would build around, esp when it has 3 video outputs.>> <<@pr0jectSkyneT says : The Rock 5 power supply must be really inefficient because the Orange Pi 5 SBC uses the same 3588 SoC but it only idles at around 2 watts.>> <<@protocetid says : If I follow the crystal rabbit will it take me to the Matrix?>> <<@hsew says : I wish AMD had a greater presence in this market.>> <<@yagoa says : For context Mac M3 is 2771.72 using 9W(average) in Web BaseMark 3.0>> <<@jpdj2715 says : Chris, having watched your RISC V videos, did you have a chance already to play with the cluster board and (seven) Sipeed Lichee Pi 4A modules on it? It seems like a nice proposition to have a mini-ITX multiprocessor computer. Question (1) is if there's any intelligence in the backplane. Then the next question (2), as it can run a version of Linux, is if it can behave like one multiprocessor computer. And finally, (3) as there are micro-kernel OS around that sort of removed the performance/throughput disadvantage relative to monoliths, is there anything micro-kernel you would run on these kind of systems? And (4) would a micro-kernel facilitate other "processor model" [a] software deployments? As to your definition of the "RISC" concept, while your videos to large extent may be an oratio pro domo, there will be some naives among the viewers. For these, the definition may send the wrong message. Thoughts go back to the 80s when Intel CPU fanboys shouted to the Motorola fanboys that their Intel processors ran at a higher clock speed and the Motorola folk would reply that theirs did more work per clock-cycle (as it had more complex instructions). It is easier to make a processor run fast at simple instructions, yes. And in the case of RISC that's what is done. It's a bit like sensing an SQRT calculation to an FPU and having to wait for the precise result to come back. It's faster to do it in 4 instructions on the ALU. Still, it needs 4 instructions to do the same thing, rather than 1. A very fundamental question I did not see answered in the RISC V videos anywhere, and comparisons with ARM, is if the low power consumption of the ARM, that it had from the first prototype out of the fab/foundry, is because ARM (company) made such an electronically efficient logic design. Or is it because the processor was RISC. And this triggers another idea for a future video maybe. Most people have zero idea how chips become what they are: logical design of electronic circuits, simulation of the design in software, conversion into a physical design, evaluation of the physical design (and going back to logical design if there are issues), printing of the physical design. The last step being done with printers that will cost about 1 billion $/£/€ today - with a few of those printers already being able to go as fine as steps of 3 nano-meter (at work in TSMC). [a] in software analysis, the "processor model" describes how software (components) is/are deployed across multiple physical processors.>> <<@SourojitBh says : Awesome video again! I think I would prefer the RK3588.>> <<@sayantantelecom says : Say something about RG kar case in India and compare with tech scams>> <<@deodotusalvika8991 says : Which better dor nextcloud and webserver?>> <<@xshadow-0 says : People say that the N100 completely crushes the RK3588 thanks to the software compatibility, but they forget there's a number of applications that require various embedded functionalities that the N100 doesn't have - such as GPIO support (with various communication buses), CSI camera support or even the built-in NPU. That's completely non-existent on the N100, where you're forced to use external MCUs for GPIO, external NPUs/TPUs/GPUs for AI acceleration and are limited to USB and HDMI when it comes to cameras and displays. N100 computers or SBCs are great for casual users, who simply want to have a low-power desktop or server, but for makers I believe they're out of the question. If you wanna tinker a little with simple robotics, internet of things or edge AI, the RK3588 wins. The only thing we need now is good software support, which, I admit, sucks very much (with the exception of Joshua's work).>> <<@ABouchez says : SInce such NICs are fine home servers too, a network Samba test could have help. I guess the RK3588 would have rocked here, because of its much better network interface.>> <<@johnlaurencepoole6408 says : What brand/model of power meter? And total price w/o screen, keyboard and mouse but with SSD, case and power?>> <<@slowlymakingsmoke says : That was a great example of real world vs synthetic tests. Just shows how software maturity and optimisation for a specific platform can have a massive impact.>> <<@rcdiekin says : nice video again chris. its kio on my other account. I'm not surprised the n100 beat the rk3588 a lower end x86 can compete and beat an arm based unit. the architectural designs of the chips are not equal in terms of performance in regards to ghz and power. x86 handles more complex computations and arm is simple. I've shown a few videos showing how real world emulation like ps2 on the radxa x2l kicked the crap out of pi5 and orange pi 5 in a linux environment in regards to emulation and other various tests. the geek bench scores and other measurable results mean nothing. great video again as always. I've been trying to get the word out to people that these sbcs by just hardware don't mean anything if its not paired with a great os to compliment it. these real world test results are the best way to decide whats what>> <<@mamdouh-Tawadros says : Thank you for the interesting video. But I wonder if we are comparing apples with oranges. Ubuntu is not the same, one for arm, the second for x86! Besides everything else is different.>> <<@voneschenbachmusic says : Amazing how far both platforms have come in terms of performance, efficiency and low cost. The 4K video playback in particular was amazing!>> <<@extremelydave says : You're going to make a me move to a mini ITX and Linux yet...I think it might be time to replace my 7 year old NUC computer that has served me extremely well, but 4K is making it chock these days....another smashing video Mr Barnett.>> <<@anon_y_mousse says : This pretty much matches my expectations. Choosing one architecture type over the other really depends on how impatient you are. RISC-based architectures are generally slower and partly because they use less power. In a lot of these graphics capabilities tests, more cores is an advantage, but clearly a faster chip will overpower that advantage as shown here. So it'd be nice to see a test with an older x86 chip that more closely matches the speed and core count of the RK3588 and see how that changes the result. However, given the low requirements of startup processes, the higher core count gives an advantage here. Although, once optimizations have caught up, maybe RISC-based architectures will beat x86, and that would be interesting.>> <<@LaurentCourtin says : I have the feeling that the Rock 5 ITX would make a quite good NAS/server motherboard. Not so sure how reliable it would be on the long term though.>> <<@ZoraMarlowe says : There is a way that nature speaks, that land speaks. Most of the time we are simply not patient enough, quiet enough, to pay attention to the story.>> <<@설카장-i8x says : How about using rk3588 for kubernates web server node?>> <<@ihateemael says : I think this is basically cpu speed vs more cores. Don't forget X86 runs 1.5 x faster.>> <<@WhiteoDead- says : yep OG has beat us all in tech since the 90s or 80s for sure mad respected❤‍🔥❤‍🔥❤‍🔥>> <<@PLAN50 says : 8gb with Linux is really not future proof and even now, pushing it.>> <<@williamhouse754 says : ❤❤❤❤❤❤>> <<@martinsmith5028 says : Interesting comparison of the two machines. There's not a lot between them so I think that the ultimate end use will determine which one to buy>> <<@jaxxx66 says : Have you tested with Cpu Governor set on "Performance"?>> <<@151mcx says : Well, that was quite boring. I figured he was going to use a snapdragon arm chip... Or at least compare and M1 arm chip with intel. Hmmm, I thought this was a tech channel. Not a pi vs intel. lol>> <<@SeeJayPlayGames says : ran glmark2 myself out of curiosity. I scored 8792 points. Core i5-4590 paired with GTX 1050 (non-Ti) 4GB. It's not mini-ITX, but it is small form factor. 10 year old platform, with nearly 8 year old GPU tech, as well.>> <<@nathanrussell2158 says : Just installed Ubuntu on one of my machines for first time in years, wow, I am impressed. Tempted to swap my mint installs over to it also.>> <<@hallo_welt1111 says : Was the same PSU used for the energy consumption test?>> <<@chromerims says : SATA on the ARM device are good for a little NAS. Lovely video 👍 Kindest regards, neighbours and friends.>> <<@shibotto says : x86/PC: take the iso, boot, install. ARM: one OS image for every ARM board, attach the disk to an already working computer (probably x86) and flash it... if hell exists, here it is.>> <<@StopTheDictator says : Watching this on rk3588 Orange Pi 5+ with armbian 6.11.0-rc3 kernel>> <<@mubeenmi4i547 says : Please check 4K60 and 1440p60 Playback also>> <<@richardwernst says : Excellent, thanks. Very comprehensive. Would have been interesting, if possible, to see boot times if somehow can eliminate the bios/mbd bootstrap delays. I think the N100 supposed NVME 4 channel, did I see that, I would have liked to see the speed comparison with the drive in that slot vs. the 2 channel. Lastly, would have found the speed difference for the cpus if the RK test had only specified 4 cpus, just as a comparison. But understand there are innumerable permutations and combinations of tests that always _could_ have been performed... :)>> <<@subbastionbastion2167 says : I don't understand the point of comparing different gpus when the title is about CPUs>> <<@MD_Builds says : Having seen those results I'm actually surprised. But I suspect something is wrong with your Rk3588 setup. As the Rk should have won on every bench mark over the n100. Your idle Wattage is also way too high. Something is wrong with that power management. Or some other device is using up that wattage, Its not simply the PSU efficiency. As it wont be 10watts even on the worst of PSUs. To hit 10Watts when a RK3588 at idle should be bellow 5Watts... around 2 watts. Also your max wattage was also a little off. Peaking at 14Watts total. But given you got something else burning at least 5-8 watts there, that means the Rk3588 isnt peaking out correctly either. As its max is around 15 watts. (and i think i saw it can be pushed to 20watts) All in all this shows something isnt right with the benchmark setup. Anyway interesting review. There's still some driver issues regarding the RK3588... Vulkan is still not fully supported. On paper the Rk3588 has the N100 beaten on every level. Just need to get that driver support up there.>> <<@Rcbeacon says : I wonder if the apparent slower performance of the RK3588 in real-life uses is due to GPU drivers and the way graphics are handled. The Intel CPU possibly having better drivers.>> <<@Kris-od3sj says : I have a request: can you please show us the output or `glxinfo | grep OpenGL` in any future reviews to 100% ensure that GPUs are really in use, to avoid an embarrassing situation like with the DC-Roma II video where it was rendering on the CPU (using llvmpipe), yet you claimed otherwise. `vainfo` would also be nice to see, to ensure that video decode acceleration is also supported.>> <<@richardokeefe7410 says : So the tests where the N100 shone were basically the ones that needed a good GPU, while the RK3588 won on CPU. There was no "build the kernel" test, no SQLite test, no MariaDB/MySQL test, no build-an-index-for-a-large-document-collection test, no format-a-500-page-document-using LaTeX test, no TensorFlow/Keras test, no BLAST test, no SuperCollider or Sonic Pi test, no firewall-under-heavy-attack test, ... ARM GPUs do not shine. This is not news.>> <<@mikebakkeyt says : Seems to me that the testing was more about GPU performance? Every graphical benchmark was clearly won by N100 whereas disk/compute seemed pretty much evenly matched with a slight lead for ARM. I think this all comes down to intended use-case. If I wanted a general purpose desktop it'd be N100 - for an always on server or media box it'd be ARM hands down based on up-front cost and vampiric power draw..>>