Currently playing with this 12500h mITX: And added some jank for the P1000: CPU could max out at 100c during Cinebench, but normally mid 30c during idle. Up to 60c for some P core action (eg. Frigate 1 camera CPU detector). GPU max out at mid 80c during tensorRT compile for Frigate. Hovers around 70c with 3 cameras. Not great, not bad.
I saw some decent temperature improvements on the 12450H by re-pasting the interface between that heatspreader and the mobile CPU. Mine is a 10729 board and yours is Erying, but I think the heatspreader is still just screwed from the back of the board.
And that's before you factor in the cost of shipping it somewhere that's not NA... 'built like a tank' = heavy = expensive to ship. Even if you can flatpak it like those Parvum cases. The lack of drive cooling, the use of a flex ATX psu for no appreciable reason [I'm sure you could have accommodated a sfx in the the same footprint or a negligble increase if they'd wanted to] 'But it's priced competitively against NAS boxes' sure... but what it's not priced competitively against is the likes of Silverstone's NAS cases... and the HL8 is competing against those as much as whatever commodity boxen QNAP, Synology or whoever are selling. Take the DS380, also 8 bay, also itx... No it isn't toolless and yes the backplane might be a bit janky compared to 45drives'... ...it's also 1/3 the price. And doesn't require a weird-ass hard to source format of PSU. The HL15 has/had fewer competing options
TBH the CPU thermal seems fine. Fitting a fan to the cooler as it was designed would drop temperature to more than manageable level. HWinfo says it only draws 80 odd watts during Cinebench. Running it without a fan works okay at the moment. Even with a little constant load as mentioned. Yes, heat spreader is screwed in from the back. It’s well done because the back plate seems same size as 1700 backplate. Easily fits coolers. I’m going to revisit the GPU cooling mentioned in the other thread.
I have been really tempted to pick up the MS-A1 and an oculink dock to play with. I certainly don’t need it, but the want is strong!
My want was stronger than yours, apparently. I ordered the DEG1 Oculink dock as well but it was back-ordered and should be shipped this week. I justified it by resolving to sell off all my AM4 stuff and, maybe, my 3090 FE but that's going nowhere until I've a had chance to throw it on the DEG1; you know, for science.
Well.... it ain't perfect. The mounting screws for heatsink that covers the RAM and two of the four m.2s actually secure the m.2s; so you kinda fart about placing it on top and hope they align correctly before tightening. It's fiddly and less than perfect. The main issue, even though I will eventually be running Bazzite or some form of Linux OS, is that Windows 11 won't install because it cannot detect the TPM 2.0 module, which is present and enabled in the BIOS. I know there are ways around this, but it's hardly the point. This is a brand new PC in 2024. This was a barebones kit, but I know the retail full PC runs Win 11 out of the box, so I'm hoping this is just a firmware/BIOS thing that needs addressing. I have seen this reported elsewhere and one guy claims they are sending him an update to fix it. Anyhoo, I've emailed support and I'm awaiting their response. Other than that - I just dropped in an m.2 with a Win 10 install from my old 5800X3d rig and it was up and running without drama. Not done much in the way of gaming yet - I spent far too much time on the TPM issue, but I'll be running a few games this weekend.
It's not an enclosure. It's basically a deck on which you mount the PSU and GPU. It's the Oculink interface that intrigues me. I've run TB3 eGPUs, which have a theoretical max bandwidth of 40gbps, but usually produce around 30gbps. Oculink taps out at 63gbps but typically achieves low 50s. It'll definitely bottleneck the 3090 but the 4060 might be just fine. ETA prime tried out a 4090 on it and only lost about 10 to 15% on synthetic benchmark scores, despite the massive disparity in bandwidth.
Nice. I did think that it didn't have enough sides to be an enclosure but wasn't sure if they were just underneath or not
Small PCB. See, now I'm thinking: 3D printed smaller enclosure (PETG), based around a HD Plex GaN PSU... ...no, wait, I have a spare 400W DC-ATX converter and a 330W Dell power brick... Just gunna peek over the edge and see how deep this rabbit hole might beeeeeeeeee...
Ok, what I should be doing right now is packing for a two week cruise which sails tomorrow afternoon. What I'm actually doing is playing around with a Gen 7 Microserver chassis... and sizing up how I can make a standard Mini-ITX board fit in there. Why? Well, I have one of these in the loft: 6x2.5" bays in a single 5.25" slot. I can make the replacement board's x16 slot line up with the outermost pci slot on the case to receive a low profile single slot quadro card so, providing I don't fsck up cutting the back to accommodate a standard IO shield, I can drop in my i5-12450H MoDT ITX board with a HBA / SATA card and make the perfect SFF home server, with ten externally accessible drive bays and three onboard m.2 slots; which will let me add my m.2 to 10GbE NIC as well as an OS drive and a NVME cache. I have 2 x 8TB SATA SSDs, 4 x 4TB SATA SSDs and 4 x 8TB SATA HDDs, which will all fit in this rig, if I can pull it off. The only real obstacles are cutting the back IO accurately and if I can make the CPU cooler fit (which I think is doable). I have to stop now, because I can hear exaggerated sighing and occasional tutting from the other end of the lounge.
