Jump to content
XPEnology Community

DVA3221 loader development thread


Recommended Posts

4 hours ago, dimakv2014 said:

Basically the problem with "no GPU" on Quadro T400, is:

"Found VGA Controller: pciid 10ded00001fb2

Required Extension: No matching extention"

Pocopico, can you please somehow add driver extension? Thanks in advance) or tell me how to compile latest driver to work in this kernel)

 

P.s maybe its possible to just add the name of gpu to existing extension or fool it to be gtx1650 since its the same chip?

 

I'm not sure what you really expect :

01:00.0 VGA compatible controller: NVIDIA Corporation TU117 [GeForce GTX 1650] (rev a1) (prog-if 00 [VGA controller])
        Subsystem: Hewlett-Packard Company Device 8558
        Flags: bus master, fast devsel, latency 0, IRQ 135
        Memory at a2000000 (32-bit, non-prefetchable) [size=16M]
        Memory at 90000000 (64-bit, prefetchable) [size=256M]
        Memory at a0000000 (64-bit, prefetchable) [size=32M]
        I/O ports at 6000 [size=128]
        [virtual] Expansion ROM at a3000000 [disabled] [size=512K]
        Capabilities: [60] Power Management version 3
        Capabilities: [68] MSI: Enable+ Count=1/1 Maskable- 64bit+
        Capabilities: [78] Express Legacy Endpoint, MSI 00
        Capabilities: [100] Virtual Channel
        Capabilities: [250] Latency Tolerance Reporting
        Capabilities: [258] L1 PM Substates
        Capabilities: [128] Power Budgeting <?>
        Capabilities: [420] Advanced Error Reporting
        Capabilities: [600] Vendor Specific Information: ID=0001 Rev=1 Len=024 <?>
        Capabilities: [900] Secondary PCI Express
                LnkCtl3: LnkEquIntrruptEn-, PerformEqu-
                LaneErrStat: 0
        Capabilities: [bb0] Resizable BAR <?>
        Kernel driver in use: nvidia

 

The kernel driver in use is nvidia

# ll /lib/modules/nvidia*
-rw-r--r-- 1 root root 27439195 Oct  7  2021 /lib/modules/nvidia.ko
-rw-r--r-- 1 root root  1423643 Oct  7  2021 /lib/modules/nvidia-uvm.ko

# lsmod |grep nvidia
nvidia_uvm            886468  4
nvidia              20258115  608 nvidia_uvm

 

From dmesg :

...
[    0.292640] pci 0000:01:00.0: [10de:1f82] type 00 class 0x030000
[    0.293276] pci 0000:01:00.0: reg 0x10: [mem 0xa2000000-0xa2ffffff]
[    0.293924] pci 0000:01:00.0: reg 0x14: [mem 0x90000000-0x9fffffff 64bit pref]
[    0.294670] pci 0000:01:00.0: reg 0x1c: [mem 0xa0000000-0xa1ffffff 64bit pref]
[    0.295415] pci 0000:01:00.0: reg 0x24: [io  0x6000-0x607f]
[    0.295998] pci 0000:01:00.0: reg 0x30: [mem 0xa3000000-0xa307ffff pref]
[    0.296724] pci 0000:01:00.0: PME# supported from D0 D3hot
[    0.297317] pci 0000:01:00.0: System wakeup disabled by ACPI
[    0.297924] pci 0000:01:00.1: [10de:10fa] type 00 class 0x040300
[    0.298557] pci 0000:01:00.1: reg 0x10: [mem 0xa3080000-0xa3083fff]
...
[    0.345753] vgaarb: setting as boot device: PCI:0000:01:00.0
[    0.346334] vgaarb: device added: PCI:0000:01:00.0,decodes=io+mem,owns=io+mem,locks=none
[    0.347157] vgaarb: loaded
[    0.347442] vgaarb: bridge control possible 0000:01:00.0
...
[    0.627914] pci 0000:01:00.0: Video device with shadowed ROM
...
[    2.203613] DMAR: Hardware identity mapping for device 0000:01:00.0
[    2.204253] DMAR: Hardware identity mapping for device 0000:01:00.1
...
[    2.228063] iommu: Adding device 0000:01:00.0 to group 1
[    2.228864] iommu: Adding device 0000:01:00.1 to group 1
...
[    2.353045] pci 0000:01:00.0: Signaling PME through PCIe PME interrupt
[    2.353982] pci 0000:01:00.1: Signaling PME through PCIe PME interrupt
...
[   18.859215] systemd[1]: Created slice NVIDIARuntimeLibrary's slice.
[   18.865496] systemd[1]: Starting NVIDIARuntimeLibrary's slice.
...
[   21.757935] nvidia-nvlink: Nvlink Core is being initialized, major device number 247
[   21.767300] vgaarb: device changed decodes: PCI:0000:01:00.0,olddecodes=io+mem,decodes=none:owns=io+mem
[   21.820945] NVRM: loading NVIDIA UNIX x86_64 Kernel Module  440.44  Sun Dec  8 03:38:56 UTC 2019
[   21.855157] nvidia-uvm: Loaded the UVM driver, major device number 245.
...
[   22.308971] DMAR: 32bit 0000:01:00.0 uses non-identity mapping
...

 

Link to comment
Share on other sites

2 часа назад, Orphée сказал:

 

I'm not sure what you really expect :

01:00.0 VGA compatible controller: NVIDIA Corporation TU117 [GeForce GTX 1650] (rev a1) (prog-if 00 [VGA controller])
        Subsystem: Hewlett-Packard Company Device 8558
        Flags: bus master, fast devsel, latency 0, IRQ 135
        Memory at a2000000 (32-bit, non-prefetchable) [size=16M]
        Memory at 90000000 (64-bit, prefetchable) [size=256M]
        Memory at a0000000 (64-bit, prefetchable) [size=32M]
        I/O ports at 6000 [size=128]
        [virtual] Expansion ROM at a3000000 [disabled] [size=512K]
        Capabilities: [60] Power Management version 3
        Capabilities: [68] MSI: Enable+ Count=1/1 Maskable- 64bit+
        Capabilities: [78] Express Legacy Endpoint, MSI 00
        Capabilities: [100] Virtual Channel
        Capabilities: [250] Latency Tolerance Reporting
        Capabilities: [258] L1 PM Substates
        Capabilities: [128] Power Budgeting <?>
        Capabilities: [420] Advanced Error Reporting
        Capabilities: [600] Vendor Specific Information: ID=0001 Rev=1 Len=024 <?>
        Capabilities: [900] Secondary PCI Express
                LnkCtl3: LnkEquIntrruptEn-, PerformEqu-
                LaneErrStat: 0
        Capabilities: [bb0] Resizable BAR <?>
        Kernel driver in use: nvidia

 

The kernel driver in use is nvidia

# ll /lib/modules/nvidia*
-rw-r--r-- 1 root root 27439195 Oct  7  2021 /lib/modules/nvidia.ko
-rw-r--r-- 1 root root  1423643 Oct  7  2021 /lib/modules/nvidia-uvm.ko

# lsmod |grep nvidia
nvidia_uvm            886468  4
nvidia              20258115  608 nvidia_uvm

 

From dmesg :

...
[    0.292640] pci 0000:01:00.0: [10de:1f82] type 00 class 0x030000
[    0.293276] pci 0000:01:00.0: reg 0x10: [mem 0xa2000000-0xa2ffffff]
[    0.293924] pci 0000:01:00.0: reg 0x14: [mem 0x90000000-0x9fffffff 64bit pref]
[    0.294670] pci 0000:01:00.0: reg 0x1c: [mem 0xa0000000-0xa1ffffff 64bit pref]
[    0.295415] pci 0000:01:00.0: reg 0x24: [io  0x6000-0x607f]
[    0.295998] pci 0000:01:00.0: reg 0x30: [mem 0xa3000000-0xa307ffff pref]
[    0.296724] pci 0000:01:00.0: PME# supported from D0 D3hot
[    0.297317] pci 0000:01:00.0: System wakeup disabled by ACPI
[    0.297924] pci 0000:01:00.1: [10de:10fa] type 00 class 0x040300
[    0.298557] pci 0000:01:00.1: reg 0x10: [mem 0xa3080000-0xa3083fff]
...
[    0.345753] vgaarb: setting as boot device: PCI:0000:01:00.0
[    0.346334] vgaarb: device added: PCI:0000:01:00.0,decodes=io+mem,owns=io+mem,locks=none
[    0.347157] vgaarb: loaded
[    0.347442] vgaarb: bridge control possible 0000:01:00.0
...
[    0.627914] pci 0000:01:00.0: Video device with shadowed ROM
...
[    2.203613] DMAR: Hardware identity mapping for device 0000:01:00.0
[    2.204253] DMAR: Hardware identity mapping for device 0000:01:00.1
...
[    2.228063] iommu: Adding device 0000:01:00.0 to group 1
[    2.228864] iommu: Adding device 0000:01:00.1 to group 1
...
[    2.353045] pci 0000:01:00.0: Signaling PME through PCIe PME interrupt
[    2.353982] pci 0000:01:00.1: Signaling PME through PCIe PME interrupt
...
[   18.859215] systemd[1]: Created slice NVIDIARuntimeLibrary's slice.
[   18.865496] systemd[1]: Starting NVIDIARuntimeLibrary's slice.
...
[   21.757935] nvidia-nvlink: Nvlink Core is being initialized, major device number 247
[   21.767300] vgaarb: device changed decodes: PCI:0000:01:00.0,olddecodes=io+mem,decodes=none:owns=io+mem
[   21.820945] NVRM: loading NVIDIA UNIX x86_64 Kernel Module  440.44  Sun Dec  8 03:38:56 UTC 2019
[   21.855157] nvidia-uvm: Loaded the UVM driver, major device number 245.
...
[   22.308971] DMAR: 32bit 0000:01:00.0 uses non-identity mapping
...

 

You didnt read previous posts apparently, by that logic my gtx1660 should be incompatible same as someone's gtx1080, 1050ti and quadro p2000, at least my gtx1660 is confirmed by me to be fully working with AI, which is different gpu chip on same Turing, so its not caped to gtx1650 from hp only, as it doesnt require only 4gb of memory, since my card has 6gb. As for Quadro T400 it has same TU117 chip as your compatible 1650 gpu, so output doesn't mean its locked in the system only to specific gpu ) it may be locked to specific driver which has in its list few gpus or gddr5 memory. If you remember long time ago geforces were converted to quadro just by modifying driver, now its more complicated but doesn't mean its impossible.

 

440.44 version of driver gives a headache since it was released in Dec 19, it needs to be 5xx.xx+ since T400 was released in May 2021 which 1.5 years after. 

Same thing goes to GTX1650 with GDDR6, this revision of gpus was released in april 2020, thats why gtx1650 with gddr6 doesnt work, just because driver is old.

I guess p2000 have few revisions as well, the newer ones won't work with this outdated driver.

 

so the solution for those who tried their gpus and failed is new updated driver which needs to be implemented into the loader or synology: https://www.nvidia.com/Download/driverResults.aspx/191961/en-us/

Screenshot_2022-08-02-22-10-19-968_com.android.chrome.jpg

Edited by dimakv2014
Link to comment
Share on other sites

On 8/1/2022 at 2:20 PM, dimakv2014 said:

Well its obviously or driver needs to be compiled or has to do with synoconfig, since its some kind of software lock, Quadro T400 is much newer card than P2000, but still strange few people with p2000 comfirmed to be working, did you try to install modified nvidia library from dva3219 as someone suggested?

Not yet, wasn't sure if this could help. Thanks, will try when I can.

Link to comment
Share on other sites

Also it would help if you could tell what Brand and exact model of your P2000 if you have box or something, which revision is that, since Gigabyte for example marks some of P2000 as (Rev. 1.0)

 

Why this is important:

The first version which was released for download by Synology, Nvidia Runtime Library .spk 1.0.1-0014 had GPU driver update (see screenshot) and it has old outdated driver 440.44 (from December 2019) which didn't have many revisions and many current gpus in its list. Since then no new updates of the GPU drivers. Current driver version for Linux is 515.57 

 

P.S. on last screenshot its unpacked nvidia runtime lib.spk its clearly seen which driver it uses and I am currently trying to replace files with new one

the main problem that New versions of .spk files such as Nvidia Runtime Library.spk cannot be opened with 7zip and other archivators... https://www.synoforum.com/threads/how-to-open-spk-for-dsm-7-0.5261/ and old version of the Nv Lib doesn't want to install through Manual SPK

scr2.jpg

scr1.jpg

scr3.jpg

scr4.jpg

scr5.jpg

Edited by dimakv2014
p.s. added
Link to comment
Share on other sites

1 час назад, error403 сказал:

I'm afraid I got the card only (used) and there's no hint about the revision or anything else.

 

You can install HWiNFO if you have Windows and upload here screenshot of window with Chipset Revision info (as on my last screen)

 

First, download app and choose between the installer or the portable version of the app. Depending on the version you've chosen, you may download an executable file or an archive. If you have downloaded an archive, extract it somewhere on your computer and then run the HWiNFO executable file. Next, a UAC prompt is shown, and you need to press Yes. Next, you are shown the welcome screen to select whether you want to see only the sensors inside your PC and the information they provide, only the summary, or run the whole app. To find the information you need, select “Summary-only” and then press Run or press Run without selecting any option.

image.png.68cdb65703a57da557d5addc849681d9.png

image.png.12511608f84a40dd6859aa101abff3c8.png

If you did not select “Summary-only,” HWiNFO also displays another window with detailed information about the hardware inside your Windows computer. For example, to learn more about your graphics card, double click on Video Adapter and then on the name of your graphics card. Now, you get to see complete information about the graphics card, including its exact model name, manufacturer, memory, and so on. 

Can't say for sure if its enough info, usually "code name" of gpu is different with each revision but I dont know any better way to do it better unless drivers will be updated in Nvidia Runtime Library and everything will be working out of box. 

IMG_20220804_020812.png

Edited by dimakv2014
Link to comment
Share on other sites

17 hours ago, Orphée said:

Baremetal gtx1650.20220805_223033.jpg

 

You of course created a task for car & person detection in dedicated app in SS ?

Yes, i created a tasm for car and person, but it wont work and i dont know why :(

If i good remeber, on dva1622 it works good, but there is only 2 license for that - i need four.

quadro P620 works fine im think.

Link to comment
Share on other sites

18 минут назад, Bullseye сказал:

Yes, i created a tasm for car and person, but it wont work and i dont know why :(

If i good remeber, on dva1622 it works good, but there is only 2 license for that - i need four.

quadro P620 works fine im think.

What is your pci-e device id showing? 1CB6 or 1CBD? It should be something like this in nvidia-smi 

NVIDIA_DEV.1CB6 = "NVIDIA Quadro P620"
Device Id                         : 0x1CB610DE

Screenshot_2022-08-06-17-35-49-940_com.android.chrome.jpg

Edited by dimakv2014
Link to comment
Share on other sites

28 минут назад, Bullseye сказал:

It is in the system, im suppose :) 

But there is no any motion registered - i have car on the fron on garage, end even this is not recognized :(

 

 

image.thumb.png.99574b61c6772968170273e5faa2a908.png
image.thumb.png.2ca73acfaef72ed44cbe5d997fe49ec4.png

image.thumb.png.c94ff6f37991639a00237289038c602d.png

Well it looks like your problem is that all of your cameras are in Night mode, IR. AI works only in daytime mode with good lighting and camera must be on certain distance and angle from object, check recommendations: 

License Plate Recognition:

 https://global.download.synology.com/download/Document/Software/AdminGuide/Package/SurveillanceStation/All/enu/DVA_License_Plate_Recognition_AG_enu.pdf

Face Recognition: 

https://global.download.synology.com/download/Document/Software/AdminGuide/Package/SurveillanceStation/All/enu/Synology_Face_Recognition_AG_enu.pdf

 

P.S. also just in case where is your device id of gpu (see screenshot)

 

IMG_20220806_232351.png

Link to comment
Share on other sites

Look all kind of logs in /var/log/

You may find an error or something like it.

 

Edit :

image.thumb.png.066a81de76bf850b165dc416ec2db8c9.png

 

Look at GPU Memory Usage, it seems a bit more than your actual memory usage.

 

look logs in :

/var/log/surveillance/*

/var/log/synoscgi.log

/var/log/messages

 

Edit, and by the way, it works too with IR night vision :

image.thumb.png.969ce80aa0da276983edf96693c827f7.png

Edited by Orphée
Link to comment
Share on other sites

2 часа назад, Orphée сказал:

Look all kind of logs in /var/log/

You may find an error or something like it.

 

Edit :

image.thumb.png.066a81de76bf850b165dc416ec2db8c9.png

 

Look at GPU Memory Usage, it seems a bit more than your actual memory usage.

 

look logs in :

/var/log/surveillance/*

/var/log/synoscgi.log

/var/log/messages

 

Edit, and by the way, it works too with IR night vision :

image.thumb.png.969ce80aa0da276983edf96693c827f7.png

Not face recognition or the license plate recognition, this looks like people detection

Link to comment
Share on other sites

3 часа назад, Bullseye сказал:

This cameras work about one week, and in this one week they cant see any car on main route :( I know my id of GPU, and it shown in system :)

So maybe its camera issue? you can try downloading .mp4 file with cars or people walking on the street from youtube, load it as User Defined IP cam for Face Recognition and Lic Plate Recognition tests, and run nvidia-smi again to see gpu and memory usage. If it doesnt utilize then its definitely some problem with the gpu support. Also check the event detection settings on camera settings maybe recording is off on schedule or detection area is incorrect or sensitivity. As for cpu I would personally use J3455 because its intel and its faster cpu out of two, but since you are using AI, gpu is more important so I guess it doesnt really matter if amd works for you.

Edited by dimakv2014
Link to comment
Share on other sites

8 hours ago, dimakv2014 said:
Not face recognition or the license plate recognition, this looks like people detection

People, car, plate detection is part of the same task (Deep Video Analytics).
Face recognition is another task in dedicated app, like shown in a post above.
It's like you don't know what you are talking about, or you never tested it...
 
image.thumb.png.9d82f68c786dde5ba6344efa451361d0.png
 
Personnes = persons
Vehicules = car
3rd row is license plate when AI reads one.

This is the same and only Deep Video Analytics task I created.
It is able to detect body, car, licence plates.
 
for face recognition, it is in the other app "Face Recognition" with its dedicated task
image.thumb.png.8beb73f284197e4b4861b77f4eadb2d6.png
 
image.thumb.png.bf54e1a4ca2e5388bedad55e47bf4ada.png
 
Even in the night it catched a face (unknown of course as it still the night)...

Edited by Orphée
Link to comment
Share on other sites

2 часа назад, Orphée сказал:

People, car, plate detection is part of the same task (Deep Video Analytics).
Face recognition is another task in dedicated app, like shown in a post above.
It's like you don't know what you are talking about, or you never tested it...
 
image.thumb.png.9d82f68c786dde5ba6344efa451361d0.png
 
Personnes = persons
Vehicules = car
3rd row is license plate when AI reads one.

This is the same and only Deep Video Analytics task I created.
It is able to detect body, car, licence plates.
 
for face recognition, it is in the other app "Face Recognition" with its dedicated task
image.thumb.png.8beb73f284197e4b4861b77f4eadb2d6.png
 
image.thumb.png.bf54e1a4ca2e5388bedad55e47bf4ada.png
 
Even in the night it catched a face (unknown of course as it still the night)...

glad that you know everything, but I dont call this "Working" result since no face was recognized or license plate correctly in night time) but if it makes you happy good for you)

Edited by dimakv2014
Link to comment
Share on other sites

It still detects persons and vehicules.
Play with words if you want, the task does its job. Of course not as good as in daylight, but still, the configured task is working as good as it can do.
Feel free to consider it is not enough for your standards.
But yes AI advanced detection still works in IR night condition, no matter what you think.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...