I built one of these. It is not possible to do so for 60$ anymore since the board itself goes for 150+. On top of that you still need a PSU, NVMe, a high pressure fan, DP to HDMI adapter, possibly BT and Wifi adapters, and a 3d-printed or DIY case.
It is also a very hacky build. You need to flash the BIOS which enables a few interesting unlocks: from 24 to 40 GPU compute units and 6 to 8 CPU cores. It's a bit of a silicon lottery and will depend on your specific board and you need to test everything thoroughly yourself, honing on a config that works for you.
The software side of things is Linux, so if you're familiar with setting it up it shouldn't be difficult to go through the different steps. There are a few distros that work well. I use Bazzite and boot directly to Steam big picture, the UX is great. You can exit to desktop and you have a fully usable Linux box that you can play around with from the comfort of your sofa.
One of the downsides is that it uses a lot of power when idling, about 80W. I just turn it off after using it. I'm working on a QOL feature: an ESP32-based power switch that starts the board when a specific BT device is detected (usually a controller). Once everything is setup the board works very nicely. It can be a bit loud depending on your fan/case combination. I'm currently printing this one[0]
It delivers about the same punch as a standard PS5. The hardware is remarkably similar, and much better than mini PCs set up with AMD APUs, unless you go Strix halo but that is another price league (10x). You can play Cyberpunk 2077 decently on it on settings similar to an AMD RX 6700XT or slightly below RTX 3070.
A pretty impressive number is the memory bandwidth at ~410 GB/s, which should run local LLMs like Gemma 4 or Qwen3.8 27B at decent speed but with limited context size as its just 16Gb
Would I build it again? It was a lot of fun, but financially is not a crazy deal if you need to buy all the materials and factor your time in. Still much cheaper than a Steam Box!
Was it possible to build it for $60 when you built it though?
As for the idle dray, a lot of PC's have Wake On Lan so you can power it on via a special IP packet. I could't get it to work on mine, so i pus an esp32 based power monitoring plug in front of it, and configured the PC to power on directly when powered. When the power monitoring sees the PC sleeping for some time, it cuts the power. I control the plug via home assistant, so i can switch it on remotely
No. I bought it for about 150 shipping included from Canada to Spain. I reckon I spent about 300$ in total, the most expensive thing being the 1Tb NVMe.
I don't have an ethernet cable where the box is, so WoL is not an option for me either. I could have used something similar to your approach (very smart, by the way!) but decided to go ESP aiming for the most console-like experience of just grabbing the controller and playing your game
WoWLAN is a thing. On my motherboard it works so well that it wakes up whenever my phone, laptop or steam deck connect to the WiFi, not just when I actually send the magic packets!
First time I hear about WoWLAN, thanks! I will check if it can work with my cheap usb Wifi adapter.
Yours seems a bit too eager though! I wouldn't want the thing to start up every time I get home as it turns the projector on, which lowers the projector screen, which blocks the (small) apartment's main entrance when lowered and would make me enter home on my fours ¯\_(ツ)_/¯
Yeah it's not actually intentional; I have sunk a lot of time into trying to stop it happening. The firmware on my motherboard sucks; I couldn't even suspend it before the last bios update. Relatively new board.
Hello fellow Spaniard! I just checked the prices and they don't make any sense anymore. A few weeks back you could find the board in Aliexpress for 150eur, now it's twice that price or even more. I got mine fom eBay, didn't think of checking Aliexpress at the time. As per the other parts, I got a quite fancy Metalfish PSU [0] (was on sale when I got it), a power button [1] from Ali (not necessarily needed) and an ESP32 [2] for the remote power on. Then a high pressure fan, I went with Arctic P12 Pro [3]. Wifi and BT adapters whatever works on Linux will work.
I got one for $130 ~8-10 weeks ago. Whew. Lucky. The Steam Machine I think is part of why these things are now suddenly super in demand but also, yeah, it was a riotously good deal for a long time; thank you for this one good thing crypto.
Just to note there are already over a half dozen esp32 based power controllers. The BT detect is cool. Reddit is at the "Yet Another esp32 controller" phase, has been for a while. (I'm not sure if yours is the first or not to do that, I think I've seen similar but maybe it was just yours I saw.) Bless you esp32. Thank you for actually having the obvious good thing, wireless, available for the world (and why did no one else deign do so?)! https://www.reddit.com/r/BC250Gaming/comments/1uylmj0/yet_an...
I don’t see how you could run Qwen3.8 27B on 16GB of memory that’s shared with Linux. Are people running models at 2bit quants? Are they even worth bothering with? I had assumed you go down to 4bit and if you need to go smaller you have to lose parameters.
You can, it's kind of cool to have this capability on something gaming at such a low price tag. Its not optimal for your time but beats nothing by a LOT. And that model is pretty reliable.
It was all the rage towards getting 4K 120 Hz VRR on Linux before AMD came along to contribute proper HDMI 2.1 support to the kernel. There is also mention in that thread of a CableMatters adapter that could be flashed with an alternate firmware but it wasn't as stable.
In my particular case the HDMI goes directly into a sound bar which does not have a DP input. I assume most people building a box like this will connect it to a TV, which is most commonly HDMI-only. So technically not a strong requirement but practically almost
Not sure why you got downvoted for this - all but the cheapest monitors include DisplayPort because DisplayPort is both technically superior to HDMI and royalty-free.
The only thing HDMI really has going for it TV's/home theater equipment. (And that's partially because most of the manufacturers are part of the HDMI forum, so they're the ones receiving the royalties that everyone else pays to use HDMI.)
This so much! What great games, probably my favorites over the early FPSs of the time, but in close contention with Day of the Tentacle and Monkey Island
Spain is at 96.79% (June 2025), up from 94.79% a year before. See [0]
I think France and Spain are the best of the big European countries regarding coverage, but also adoption: If we look at the adoption ranking globally, Spain sits fourth at 90%, behind the UAE, China and South Korea and ahead of France at 83.6%.
Very few European countries convert coverage into subscriptions that efficiently, and I think it's driven by competition. You can get 500Mb symmetric connection for as little 10 euro [1], 20 for 2Gbps. DIGI has just gone crazy and in their own network they are offering symmetric 10Gbps for 25. Most people I know have 1Gbps because when is that not enough?
In Italy we are at 78% of addresses covered. That doesn't mean automatically that one gets fiber. First at all, if someone doesn't need to download heavy stuff and has a good coverage a 5G SIM with unlimited data (something you can get for 10/15 euros a month) is cheaper than a fiber connection (that starts from 25 euros).
Second a lot of people has still FTTC even if it has FTTH, mostly because FTTC offers a speed that is enough for most people (100Mbit/s) and upgrading may have costs (the installation of the fiber is at the expense of the ISP, but if you need to make modifications, such as put a pipe from your house to the street in which they will put the fiber, it's at your expense). In some cases even if fiber is present it's difficult/not possible to get it into your house (for example you live in a condo and the other tenants doesn't want to spend money to make the modifications that are needed to pass the wires).
In any case they have started to switch off copper (since it costs money to maintain an infrastructure that has decades), so either way customers will be forced to move to fiber or 5G network in the upcoming years. In any case a fault on a copper line is likely not to be repaired...
You can use any one of many MVNOs to get unlimited data for around that price, say less adjusted for household income, in the USA.
A favorite of mine is https://www.usmobile.com/ which is $25/mo list price (and less if you pay for the year, or if they're having a sale) which will give you Unlimited Data and 1GB a month of global roaming data
Only some places. Netherlands is €45-60 for 1 gbps (we don't make 4x-6x the money Spaniards do), unless you live in an area that actually sees multi-line competition.
We do have an interesting sidegrade: 5G Klik & Klaar from Odido (formerly T-Mobile / Deutsche Telekom) for €26. If you point it the right way out your window, you can get 700-1000 mbps down and 150 (capped) mbps up. Ping is around 35ms. Horrid modem firmware though, you get ping spikes up to 400-850ms (!!) under load due to bufferbloat, but you can of course solve that with your own router and SQM :+)
In the U.S. I perceive MVNOs to be a scam. Like Tracfone was OK in 2011 but somewhere afterwards it became normal to go around rural upstate NY and no service and go around NYC and no service and go to LA and no service.
This year I got a real phone with a real plan and I have coverage in many places now me I can finally see why people get excited about apps.
That's kind of a broad brush, Google Fi is great, especially the fact that they work mostly seamlessly for the same price in like every country, and they're an MVNO.
It proves that you can’t trust the claim that “it is the same network”, I mean it is the same network but you may find yourself riding the back of the bus and there is no independent way to check what service you are actually getting.
Well I never expected it to be treated exactly the same as their first party customers, did you? Especially if I’m paying a much lower rate. Just have to evaluate the actual results, and hop if you’re unhappy. I’ve been extremely happy with mine, so I’ve stuck with it. It’s a bad choice if you’re out and about a lot, since it’s not cheap when you go unlimited data, but working from home on WiFi, the metered option, I end up paying like $30-$40/mo, and there’s no need to go get a new sim when I travel. For a different lifestyle shape, I’d be shopping for one that fit that better.
I have both T-Mobile via a family plan and US Mobile (AT&T MVNO) via a yearly plan.
US Mobile serves as my data eSIM, where it gives me unlimited hotspot 20GB of roaming/month. Previously I bought eSIMs when traveling since T-Mobile is only 5GB/mo hotspot/roaming and earlier at 256 Kbps.
T-Mobile is better for calls and international texts, due to it being unlimited. I also prefer keeping a non-MVNO as my call/text number so I can go to a T-Mobile store and activate a spare phone if I lose mine (which I never did), or my phone fails. With US Mobile, I'm not sure how I'd recover a lost phone.
Since the Pixel 10 Pro lacks a SIM card slot, this is important. But hey, I used my Pixel 3 on Sprint with eSIM; I went T-Mobile 3 months before they merged.
US Mobile has done well with me, although anything not-T-Mobile is really slow nowadays.
I do use US Mobile, and I recently switched to it because T-mobile made it so difficult to pay a bill, it was nearly impossible to get them to accept money.
And a ~decade ago in SF I did have Sonic 1Gbps symmetric fiber for something like $40 or $50/month, but that too is twice the cost of Italy.
Can't speak for Italy, but there have been some consequences in many EU jurisdictions. The EU has been pushing HARD for competition, etc, but it has left many European telecoms in debt-heavy financial situations that has caused issues around future investment and in some countries the (mobile) coverage is good on paper, but the networks are themselves actually terrible (looking at you Germany).
But yeah, it's otherwise a problem with competition enforcement.
Because deploying fiber is so hard in NYC, I use Calyx Institute (https://calyxinstitute.org/membership) on a MikroTik Chateau 5G R16. Calyx uses T-Mobile towers. I use the internet via L2TP to a VPS to give me a static IPv4 and IPv6 /56 block. This is only for upload speeds. I get 45-80 Mbps uploads.
I used Spectrum for four months but because of my L2TP setup, DOCSIS ACK suppression doesn't work; it's TCP inside L2TP UDP and then ACKs are then sent over the wire. This meant sluggish internet under load. Spectrum did install a high-split amp but it's still not turned on afaict.
Row townhouses in Brooklyn and Queens have Fios, but not in Manhattan. I wish I could just do a NYT ad but that'd force me to liquidate a decent chunk of my E*Trade and J.P. Morgan accounts.
I am right now visiting Belgium and I saw multiple towns, including historic neighborhoods with Proximus fiber splitters: Brussels, Antwerpe, Ghent and Bruges. The fiber and Telenet Coax is so well organized when compared to NYC's coax mess. However, where Fios is available, Verizon does an okay job with organizing fiber.
Verizon said killing line sharing would bring more fiber. Yet Proximus built fiber in historic cities more than Verizon did. I don't believe Orange France legally has to share fiber, but in return Orange can't block Free/SFR with payola deals.
It’s not uncommon for cities to be worse than rural areas now. In Ireland, nearly all rural premises have 10gbit-capable fibre available (will be literally all by early next year) but parts of central Dublin are stuck with VDSL. Retrofitting in cities can be quite slow and expensive.
The main issue with copper comes from the active infrastructure rather than the physical lines.
I work at an ISP that gets DSLAM equipment from Huawei, we run one of the few G.fast VDSL deployments in the world using this particular hardware. Huawei has stopped manufacturing this product line (the Swiss market is a rounding error ;), which will force us to fully migrate some neighborhoods, only for backup inventory.
Germany being 15 years late, I assume you can still buy similar hardware from AVM though.
I am in Italy right now, in fact, spent a lot of time here. I can safely say, the 5G sim coverage is extremely spotty, and when you can get internet from a physical line, it's generally pretty poor speeds. I am pretty sure a lot of that 78% includes 1 bar of LTE coverage, so it's "technically" true.
Several of the issues you point have been solved in France via regulations (on the ISP and on the tenants of condos). It tools several years, even decades, but we ended up with fiber regulations that are good (could have been better and faster).
I spent time in the interior of Portugal that had fibre internet, it was very impressive and incredibly helpful to the village. Better payment services and banking for example.
UK is at 85% FTTH or 91% gigabit (including DOCSIS): https://labs.thinkbroadband.com/local/. Given VM are a fair way through their XGS-PON upgrade it'll probably be 90%+ FTTH imminently.
I pay £6 per month for a sim only contract, which gives me 4GB of data. In the USA I feel like I was paying closer to $40-50 for a google fi sim? Which admittedly did have better roaming. Now I just get an e-sim for travel and it works out to be £3ish per day
I’m in the same camp as you. I’ve just decided that the M5 Ultra is gonna be my next machine. Plus an Air for the few times when I need a computer on the go. I’m invested in Tailscale already so an ubiquitous Studio Ultra seems like a sweet idea
I checked eBay and they are now around 200 usd and up. All the sellers located in the US and Canada, so a bit more involved for hackers looking into building one from Europe. I still pulled the trigger on one, not sure I will use it much but I'm definitely gonna have a lot of fun building it!
This is what I have been waiting for in order to process my massive action camera library, I will definitely give it a try!
I will start with older standard GoPro videos, and hope that by the time I make it to the Insta360 videos the editor supports 360 video (unless it does already!)
It is 17% more token-efficient than 3.5 and performs significantly better in coding and tool usage benchmarks.
It is also cheaper than 3.5:
> This enhanced efficiency is also combined with a lower price than 3.5 Flash. At $1.50/1M input tokens and $7.50/1M output tokens, 3.6 Flash reduces the overall cost per agentic task, making agents more cost-effective to build and run.
This feels a lot like free spins in slot machines. I have a friend obsessed with using up all the Fable usage which supposedly ends today. He's doing a lot of stuff of questionable usefulness just because it's a resource perceived as scarce.
Yeah, it's the gacha of software development. I too redline my subscription every single week just because I want to get the most value out of the money I spent. Literally 100% weekly usage, every single week.
I am using the resulting software and documents though. I'm daily driving the projects I wrote with AI. Started new ones for fun, such as decompilation projects for my cherished childhood games. Hardened my router and my LAN. I just keep finding an unending amount of things to do.
It took over a month to complete a full Fable code review of my lone lisp project.
That's what half of people's AI developed projects feels like to me. I have colleagues using Openclaw, and having an entire team of agents managing their home infrastructure. And I'm just thinking "why?". They're not developing anything that useful, their homes are already automated via home assistant. It's just agents for agents sake. Outside of my work life, I find very little need to be using AI.
AI in general is a slot machine and it wouldn't surprise me if that was a large part of the psychosis. You prompt the lever, watch the display spin, and sometimes, just sometimes, a bunch of coins come out.
There's no psychosis in noting that something consistently produces good work, or in wanting to get the most out what is effectively an employee - hired for the price of the subscription. Seriously, if that's 'psychosis', then every employer on the planet eager to wring out productivity from their employees is afflicted with the same psychosis.
Isn't it rather a psychosis to insist that people are imagining the results they see, just because you don't like the tech and can't stomach the thought there might be something in it?
"Beginning July 20, Claude Fable 5 will be included in all Max and Team Premium plans, at 50% of limits.
Pro and Team Standard users will continue to have access to Fable via usage credits, and will receive a one-time $100 credit.
Demand for Fable has been challenging to predict, which is why we rolled it out to subscription plans in stages, extending access several times as we secured additional capacity." [0]
I am assuming he means that Fable will remain part of the subscription plans going forward, but only for Max and Team Premium plans. For other plans it will be available via credits only.
Bring it on! Hoping that they release smaller sizes of Qwen3.8. I use the 35B MoE and 27B dense models locally and most of the time I don’t need to reach out to Claude. Extremely useful specially when requests include sensitive and/or personal data
It would be great if they'd release an MoE model somewhere between the 35B size of 3.6 and the 122B version of 3.5 - it could be a great balance of speed and ability for people with reasonably powerful but not insane home computers.
Yeah, this is what I'm holding out for, the NVFP4 variant of 3.5 122B is blazing fast with reasonable quality and even with max context fits perfectly within 96GB.
Yeah, Qwen3.5-122B-A10B-NVFP4 produces better responses than Qwen3.6-27B-NVFP4 (both from unsloth), but I'm mostly using them for programming in various ways, mostly Rust, Clojure, Python and JavaScript, and some translations tasks, but not much more than that, so YMMV.
Edit: as a concrete example, I'm working on a "optimization framework via agent harness" right now, Qwen3.6-27B-NVFP4 is often unable to actually complete the optimization within 100 turns, while Qwen3.5-122B-A10B-NVFP4 has no issues finishing within ~50 turns or so.
> Yeah, Qwen3.5-122B-A10B-NVFP4 produces better responses than Qwen3.6-27B-NVFP4 (both from unsloth), but I'm mostly using them for programming in various ways, mostly Rust, Clojure, Python and JavaScript, and some translations tasks, but not much more than that, so YMMV.
>
> Edit: as a concrete example, I'm working on a "optimization framework via agent harness" right now, Qwen3.6-27B-NVFP4 is often unable to actually complete the optimization within 100 turns, while Qwen3.5-122B-A10B-NVFP4 has no issues finishing within ~50 turns or so.
But why are you using Qwen3.6-27B-NVFP4 compared to the FP8 or full version? In my experience the Q8 of 27B is on par sometimes better than 122B. I am experiemnting witb higher quants for 122B to fit on my Strix Halo, but still, the difference honestly for my workflow is not that much. I just wish they released 3.6-122B version.
This seems more of a battle for frontier AI supremacy. I'm afraid that small capable models have been left in the dust. Big labs don't really want to hand over the golden eggs goose to the end user. Possibly the hardware vendors(e.g. Nvidia) may want to play in that area as well, to pull money from all parties.
> I'm afraid that small capable models have been left in the dust
I wholly disagree. Rather than going the "everything is a claude code skill" route, I've been hacking together purpose-built harnesses for all sorts of tasks, and in that environment a wee little baby model can do some really useful things. You end up burning lots of tokens making the thing, but then all that investment comes back when the resulting tool works perfectly fine on a dinky little model that fits on my 3060 Ti.
Google makes Gemma 4 31B QAT; that's not a small lab. It's one of the better models out there for consumer hardware. Allows me to run it on a 7900XTX with 64k context.
someone will keep putting out consumer level models. Once you have the larger models, you can derive the smaller onces.
Europe will definitely be interested in democratizing these things if China starts losing interests; from there, there'll be more countries looking to keep their citizens entrained in their own Country's infrastructure.
It'll especially be true if the memory cartel keeps prices high and NVIDIA tries to gouge higher memory models.
It's an arms race everyone can join because PC hardware was mostly democratized in the last decade.
nvidia and amd don't give a flying fuck about end users right now while they can milk triple digit markups from infinite money VCs via data center GPUs.
there's been a lot of research about reducing models by taking out layers; there's also using it to train smaller models by optimizing parameters.
I dont see most model building as anything more than a pig at a slop troth, despite the level of sophistication; they're still rarely pruning the input beyond random sampling.
I run a 2x 3090 rig, but a single 3090 already provides a great experience at a reasonable quant and context size. On a single card I used to run Qwen_Qwen3.6-27B-Q4_K_M or similarly quantized 35B MoE at 65536 context size
It is also a very hacky build. You need to flash the BIOS which enables a few interesting unlocks: from 24 to 40 GPU compute units and 6 to 8 CPU cores. It's a bit of a silicon lottery and will depend on your specific board and you need to test everything thoroughly yourself, honing on a config that works for you.
The software side of things is Linux, so if you're familiar with setting it up it shouldn't be difficult to go through the different steps. There are a few distros that work well. I use Bazzite and boot directly to Steam big picture, the UX is great. You can exit to desktop and you have a fully usable Linux box that you can play around with from the comfort of your sofa.
One of the downsides is that it uses a lot of power when idling, about 80W. I just turn it off after using it. I'm working on a QOL feature: an ESP32-based power switch that starts the board when a specific BT device is detected (usually a controller). Once everything is setup the board works very nicely. It can be a bit loud depending on your fan/case combination. I'm currently printing this one[0]
It delivers about the same punch as a standard PS5. The hardware is remarkably similar, and much better than mini PCs set up with AMD APUs, unless you go Strix halo but that is another price league (10x). You can play Cyberpunk 2077 decently on it on settings similar to an AMD RX 6700XT or slightly below RTX 3070.
A pretty impressive number is the memory bandwidth at ~410 GB/s, which should run local LLMs like Gemma 4 or Qwen3.8 27B at decent speed but with limited context size as its just 16Gb
Would I build it again? It was a lot of fun, but financially is not a crazy deal if you need to buy all the materials and factor your time in. Still much cheaper than a Steam Box!
- [0] https://makerworld.com/en/models/3021754-bc250-case-mkuu-fsp...
Edit: GPU performance comparison
reply