Jump to content
XPEnology Community

dimakv2014

Member
  • Posts

    305
  • Joined

  • Last visited

  • Days Won

    5

Everything posted by dimakv2014

  1. Grats then you can ignore my previous post, so if you say 4x sata adapter works can you please make pictures and share in details minipc model? Also you can make more than one FaceRec tasks on this cpu I think 3, becuase on j4125 two facerec tasks loads cpu to 86-90% so for J5040 3 tasks will be max with little exploit of using ActionRules in Surveillance Station menu
  2. You should try Pocopico's Loader https://github.com/pocopico/tinycore-redpill it should work, I did a reaseach and AI works with J5040. Also you can try Peter's Su M Shell loader it can work as well, most problems people have with AI not working on APRL for some reason Arc is basically same as APRL. And as Orphee suggested you don't need a Patch for this to work on this cpu. Just browse this forum topic there is an instruction how to install Pocopico's TCRP Loader.
  3. Sorry mate, just tried to prevent people from buying non-working components, which definitely won't help to build a loader with better experience, I didn't realize it was something bad. But I don't agree about going too far from Develop and Refine. Develop - means to make something new from scratch. Refine - typically means to make something better or more precise. For example improving hardware for DVA1622 to make better use of it can be defined as Refine. So discussions about hardware which is part of the loader because without it you cannot obviously build a loader is very important and part of testing. If you don't agree you can ask Developer to judge, I don't mind 😉
  4. Buy it, test it, make a review but I doubt that thing on image will work😁 but you can try😉 But what will work however is this, take a look at connector pins and find a difference between this and what you shown. But it will work only if you have SATA m.2 slot not NVME which most cheap Chinese MiniPCs have.
  5. Thats mSATA adapter on image and not m.2. Moreover if miniPC has m.2 slot it must clearly state whether it supports m.2 sata or only pci-e nvme drives or both. You can't put m.2 sata (M+B Key) ssd into m.2 pci-e slot it will not work... Its basically same thing as trying to put DDR3 memory into DDR4 slot or vise versa
  6. Nvme = pci-e, minipci-e. So like I said only 2 sata (regular or msata) means 2 drives in DVA1622
  7. Not really, you will still have 2 drives since DVA1622 doesn't support nvme drives as a Storage, only for Cache. And its not loader problem its Synology DSM. There are however other than DVA models that do support nvme drives for storage but only as secondary drive so DSM cannot be installed on them.
  8. I dont know abot J6426 noone reported it working on it, but there were discussions on forum about N100, I believe this patch is needed, but you ll have to search it yourself or risk it and test it out. Just in case if you going to buy one of these make sure you can return it back to the seller without any problems in case it won't work.
  9. On dva1622 it will work even without gtx 1050ti since it doesn't use dedicated gpu instead it uses integrated gpu in Intel's cpu. In order to use gtx 1050ti in AI to get more tasks which is 4 times more than dva1622 by default you should look into DVA3219. Your cpu is fine for both DVA1622 (without GTX 1050ti) and DVA3219 (with GTX 1050Ti) if you want to use AI.
  10. Not bad choice its faster than J4125 on igpu side which means AI will be more reliable, but a little slower on cpu speed side but I doubt that you will notice a difference in perfomance in dva1622. But 6watt vs 10watt is significant difference in terms of cooling. For example J4125 even with SoC just passive cooling is not enough when it comes to AI, while Pentimum N5000 with it's 6watts may get away with passive cooling to make the system completely noiseless. Also remember to use sata SSD and to boost performance a little bit you can use small capacity nvme m.2 ssd for Cache, unfortunately its not possible yet to use nvme as a volume in DVA1622.
  11. There is a solution, try N4xxx or J4xxx cpu as I told you before. The reason for AI not working in your case might be because DVA1622 hardcoded with specific igpus Intel UHD and your cpu has HD (without Ultra) all cpus that I have tested AI with had UHD 530 to UHD 630 and it worked fine. Celerons J4105, J4115, J4125, also I believe N4000 has UHD600 series igpu. Core i3 6100 has UHD530 works fine, i7 8700k has UHD 630 AI works fine. Look into this way.
  12. You can try if it says its supported by this patch but out of box it won't work. I personally haven't tried this patch
  13. Ok let me rephrase it its Celeron 4xxx cpu Series minimum required for AI to work in DVA1622 and Gen 4 minimum for Core Series
  14. Nope, you have 3rd Gen and it clearly says minimum is 4gen lets say Celeron N4xxx J4xxx etc. But since you have AVX2/Movbe cpu instruction supported you can have AI on DVA3221 with Nvidia GTX 1650/1660 gpu. But not on igpu with DVA1622
  15. So basically those .git files are archives and directories. So lets say if we unpack them create following folders as well as empty "auxfiles" folder then we will not need "prepare-offline.bat" file at all? Actually 3 folders because I noticed auxfiles folder is there by default it just confused me because in .bat file it makes dir again ))
  16. The last piece of puzzle is prepare-offline.bat which is still requires online access to github since it redirects to 3 .git files. Basically those 3 files (addons.git, modules.git, load.git) must be in dir "auxfiles" if I am not mistaken: "git clone -c core.autocrlf=false https://github.com/PeterSuh-Q3/tcrp-modules.git git clone -c core.autocrlf=false https://github.com/PeterSuh-Q3/tcrp-addons.git git clone -c core.autocrlf=false https://github.com/PeterSuh-Q3/redpill-load.git mkdir auxfiles"
  17. Wow great job 😀 So as far as understand from Linux its possible to do step "1." GParted (alternative to MiniTool) and then steps 2-3 can be skipped. Step "5." Prepare offline.bat can be converted to .sh and can be executed in local path instead of having to download it from GitHub (in case something happens to repository)?
  18. Too bad that someone tries to close such projects, I hope everything will be solved and back to normal. This particular situation shows that other projects and other developers could be at risk as well. Also since many things within loaders require online access it would be impossible to install it offline if account will be banned. Correct me if I am wrong. Big thanks to wjz304 for hard work and wish everything will get back to normal soon enough without any legal proceedings
  19. The problem could be because of a wrong Package Name. Currently with custom driver: And it should be:
  20. In order for someone to help you out, you need to provide more info: 1. You use Vmware ESXi 8 with igpu Passthrough, do you have second gpu in the system? You cannot passthrough only available gpu. 2. Most people here prefer Proxmox over VMware since it has much larger user community and more features in virtualization, its just more popular. 3. You said you have created dumps and it would be helpful if you would post them in the message. 4. Explain step-by-step what you did, version of loader, which loader you have used, vm settings and also about i915.ko patch how you applied from which developer etc...
  21. No problem) basically what this trick does is just Enables all Disabled tasks upon motion detection on any of your cameras (and you can create theoretically unlimited number of Disabled tasks). The only limit is GPU, put GTX1660 and its possible to run about 24 tasks. Its good if you want to have 8 cams with Face Rec, Licence Plate Rec, Object detection and so on, on each of them simultaneously.
  22. the point is to use GPU to the max, lets say DVA3221 allows you only 6 tasks on Face Rec and you have more than 6 cameras or want to use additional DVA functions on those cameras along with Fac Rec. In my case I have 2 Face Recognition tasks working at once in DVA1622 while its not possible with default settings. Cpu load is up to 95% on Celeron J4105, but in case of DVA3221 with Nvidia gpu it utilizes about 50-60% on 6 Face Rec tasks so you can add more. Go to Surveillance Station, first create tasks (Face Rec, Lic Plate Rec etc...) but disable them and then click on Main Menu, choose Action Rule (Please ask if you have any questions): 1) create motion detection zones and conditions for their detection on cameras, while choosing "Surveillance station" as the detection algorithm and not camera algorithms; 2) create the necessary AI and face recognition tasks for each camera but do not enable them (you still cannot enable more than 2 tasks if on dva1622); 3) in the surveillance station, open the "Action rules" package and create the following rule in it: a) in the conditions select "started" as the rule type, and "Interruptible" as the action type (these values are specified by default); b) in the "events" tab, select "camera" as the source, select cameras as the "device" (for example, I selected all cameras), select "motion detection" as the event; c) in the "actions" tab, as the "action device", select "Deep Video Analytics" or "Face recognition", as the "task" select one of the AI or face recognition tasks that you created in step 2 of this guide, as " actions" select "enable task", d) then on the same tab at the top left click the "add" button and create by analogy "Action 2", "Action 3", "Action 4" and "Action 5" (in just one rule, you can set up to 5 actions, if there is a need for more actions, then create a new rule by analogy); e) in the tab "schedule" you set "active always".
  23. Just read from #800 instruction on Pocopico's loader with web interface there is a tutorial, nothing changed since then applies to all builds DVA1622 and DVA3221. As for getting more tasks using Action Rule in SS, Orphee probably knows this trick)
×
×
  • Create New...