zeroflow

As far as I can find, Plex does not support AMD iGPU for transcoding. Jellyfin will work, but support seems rather spotty. For other AI/ML work, it seems like ROCm is up and coming, but support - e.g. for Frigate object detection - is still a work in progress, especially for newer chips.

I often read that take. Yes, the J4125 was fine for a few easy / low effort transcodes, like 1080p to 720p for mobile streaming.

But I'm with you. The rest is fine, not great, but rather working well enough.

I totally agree with you.

The appeal for me was the "it just works" factor. It's a compact unit and setup was easy. Every self-built solution would either be rather large (factor for me) and more difficult to set up. And I think, that's what has kept Synology alive for so long. It allows entry level users to get into the selfhosting game with the bare minimum you need, especially if transcoding (Plex/Jellyfin) is mentioned.

As an anecdote, I've had exactly this problem when buying my last NAS some time ago. It was DS920+, DS923+ vs. QNAP TS-464. The arguments for QNAP were exactly what you write. Newer chip, 2.5G NICs, PCIe Slot, no NVMe vendor lock-in. So I bought the QNAP unit. And returned it 5 days later, because the UI was that much hot garbage and I did not want to continue using it.

Lately, the UGreen NAS series looks very promising. I'm hearing only good things about their own system AND (except for the smallest 2-bay solution) you can install TrueNAS. It mostly sounds too good to be true. Compact, (rather) powerful and flexible with support for the own OS.

As the next player, with mixed feelings about support, the Minisforum N5 Units also look promising / near perfect. 3x M.2 for Boot+OS, 5 HDD slots and a PCIe low-profile expansion slot.

I would say cases like the Fractal Node 304 or something like the HL4 / HL8 from 45homelab would be the best suited candidates.

Regarding mainboards - models from CWWK with lots of SATA ports have been trendy lately. But there are reports of problems. The other options are either using some obscure supermicro mainboards with lots of ports or using a HBA for expansion.

I want to mention a possible middle ground here: UGreen NAS Storage. All but the smallest model come the OS on a seperate M.2 drive. If you disable the watchdog in BIOS, you can use the models like a normal Server This would give you:

* 3x M.2 slots * 4, 6 or 8 SATA bays * N100 (4 bay), Pentium Gold 8505 (4 bay), i5-1235u (6 & 8 bay)

The M.2 slots are connected rather slow, but good enough for OS/app drives.

For example, my plan for the next NAS would be the 4-Slot N100 variant with TrueNAS. One M.2 SSD for boot, Two M.2 SSDs for Apps/Server duties in mirroring and the 4 drives in Raid-Z1.

It seems to be coming in multiple waves from multiple sides.

One of those is EUs ISA: First a display, now a warning and later actual interference with the driver.

And with the experiences with the current status are enough for me to be against those systems. The car doing an emergency stop because it saw a 30 sign an an adjacent road makes me not wanna purchase such a car. But there will be some time where no alternatives exist.

That "subscription" is one which I gladly pay due to multiple reasons:

1. It supports the developers(s) 2. The price can be directly attributed to cost for training 3. You can keep the models you trained during your subscription indefinately

That's pretty much the opposite to AgentDVR. I don't need hosted services for remote access or push notifications - I can do that myself. But if I want to abide the license terms, I need to purchase a monthly subscription for remote access over my own VPN.

The documentation is rather scarse on performance numbers, but it looks like the hierachy of price/performance is like Intel iGPU ("free"), Intel A310, Nvidia GPU.

I'm explicitly leaving out the Coral TPU, since it's been reported that the newer Intel CPUs (Core Ultra) seem to provide the same performance with it's iGPU.

I like the idea, but no.

They have a two-stage approach, first motion detection with - I think - OpenCV and then afterwards object detection of zones of interest with different object detection models, depending on your hardware.

It supports Coral TPU, Halio Accelerator and most GPUs. I think AMD is still the worst, since ROCm is not available on iGPUs.

Afterwards, they provide/support models like edgedet (Coral), YOLO-NAS, YOLO, D-Fine or RF-DETR.

They also offer paid access to a specially trained version of YOLO-NAS where you can also train your own images.

It's frustrating trying to work with ISPs. While there surely are business services available, as a normal customer, you're out of luck.

I count myself lucky, since I can at least get a public (dynamic) IPv4 address. The other option is DSLite with a single /64 via DHCP6. Well, I can't use that. So yeah, while it would be a fun journey trying out IPv6 - it's not feasable for me. And most likely, that holds true for others aswell.

There are options available - like the HE tunnelbroker - but those come with their own set of drawbacks.

A few years ago, I built a simpler version of that device myself.

As you said: A 90% solution is easy to hack together. I had mine displaying Skype for Business status so my colleagues would know if I was in a call or just listening to music.

The harder part or much rather the time consuming part is getting a good status message from the installed apps. It looks like there is plenty of software and APIs available.

I'm offering you a different viewpoint:

They made the analysis, how long the flash will live and saw, that it will make it out of the warranty period. Thus they did not opt for more durable and expensive flash and/or software change.

I've seen this myself before. One process step before release of the control module was a write cycle analysis to make sure the unit will live for at least 10 years (i think) before the guaranteed write cycles of the flash memory were consumed.

"That is left as an exercise for the reader"

That methods goal is to be tamper evident. You are referring to tamper proofing which opens up another whole can of worms.

My bigger question would be why do you need to verify boot firmware? If we know your goal, maybe there is a better way.

Tbh, I for myself would not care for physical intrusion. If someone (private or state sponsored) has the willingness to intrude into my home, them tampering with my PC is the least of my concerns. As someone else also mentioned: A $5 wrench will be more effective than any measures you can do by modifying your pc.

Regarding tamper evidence, there have been multiple Defcon / Blackhat talks about tamper evidence. One thing that comes into mind is vacuum sealing a notebook into a bag with colored beans and taking a photo. This way, it will be impossible to access the pc without disturbing the pattern of beans surrounding the PC. You just need the software to compare photos to know if the sealed bag has been tampered with.