&gt; The main challenge is latency since you have to do much more frequent communication.Earlier this year I experimented with building a cluster to do tensor parallelism across large cache CPUs (AMD EPYC 7773X have 768mb of L3). My thought was to keep an entire model in SRAM and take advantage of the crazy memory bandwidth between CPU cores and their cache, and use Infiniband between nodes for the scatter&#x2F;gather operations.Turns out the sum of intra-core latency and PCIe latency absolutely dominate. The Infiniband fabric is damn fast once you get data to it, but getting it there quickly is a struggle. CXL would help but I didn&#x27;t have the budget for newer hardware. Perhaps modern Apple hardware is better for this than x86 stuff.

Exo-Labs is an open source project that allows this too, pipeline parallelism I mean not the latter, and it&#x27;s device agnostic meaning you can daisy-chain anything you have that has memory and the implementation will intelligently shard model layers across them, though its slow but scales linearly with concurrent requests.Exo-Labs: <a href="https:&#x2F;&#x2F;github.com&#x2F;exo-explore&#x2F;exo" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;exo-explore&#x2F;exo</a>

But that&#x27;s only for prefilling right? Or is it beneficial for decoding too (I guess you can do KV lookup on shards, not sure how much speed-up that will be though).

For a bit more context, those posts are using pipeline parallelism. For N machines put the first L&#x2F;N layers on machine 1, next L&#x2F;N layers on machine 2, etc. With pipeline parallelism you don&#x27;t get a speedup over one machine - it just buys you the ability to use larger models than you can fit on a single machine.The release in Tahoe 26.2 will enable us to do fast tensor parallelism in MLX. Each layer of the model is sharded across all machines. With this type of parallelism you can get close to N-times faster for N machines. The main challenge is latency since you have to do much more frequent communication.

Tensor Parallel test with RDMA last week
<a href="https:&#x2F;&#x2F;x.com&#x2F;anemll&#x2F;status&#x2F;1996349871260107102" rel="nofollow">https:&#x2F;&#x2F;x.com&#x2F;anemll&#x2F;status&#x2F;1996349871260107102</a>Note fast sync workaround

No way buying a bunch of minis could be as efficient as much denser GPU racks. You have to consider all the logistics and power draw, and high end nVidia stuff and probably even AMD stuff is faster than M series GPUs.What this does offer is a good alternative to GPUs for smaller scale use and research. At small scale it’s probably competitive.Apple wants to dominate the pro and serious amateur niches. Feels like they’re realizing that local LLMs and AI research is part of that, is the kind of thing end users would want big machines to do.

I haven’t looked yet but I might be a candidate for something like this, maybe. I’m RAM constrained and, to a lesser extent, CPU constrained. It would be nice to offload some of that. That said, I don’t think I would buy a cluster of Macs for that. I’d probably buy a machine that can take a GPU.

I think it’s going to be great for smaller shops that want on premise private cloud. I’m hoping this will be a win for in-memory analytics on macOS.

The lack of official Linux&#x2F;BSD support is enough to make it DOA for any serious large-scale deployment. Until Apple figures out what they&#x27;re doing on that front, you&#x27;ve got nothing to worry about.

I’m hoping this isn’t as attractive as it sounds for non-hobbyists because the performance won’t scale well to parallel workloads or even context processing, where parallelism can be better used.Hopefully this makes it really nice for people that want the experiment with LLMs and have a local model but means well funded companies won’t have any reason to grab them all vs GPUs.

Yeah, two Mac Studios is going to be ~400 W.

Almost the most impressive thing about that is the power consumption. ~50 watts for both of them? Am I reading it wrong?

I follow the MLX team on Twitter and they sometimes post about using MLX on two or more joined together Macs to run models that need more than 512GB of RAM.A couple of examples:Kimi K2 Thinking (1 trillion parameters): <a href="https:&#x2F;&#x2F;x.com&#x2F;awnihannun&#x2F;status&#x2F;1986601104130646266" rel="nofollow">https:&#x2F;&#x2F;x.com&#x2F;awnihannun&#x2F;status&#x2F;1986601104130646266</a>DeepSeek R1 (671B): <a href="https:&#x2F;&#x2F;x.com&#x2F;awnihannun&#x2F;status&#x2F;1881915166922863045" rel="nofollow">https:&#x2F;&#x2F;x.com&#x2F;awnihannun&#x2F;status&#x2F;1881915166922863045</a> - that one came with setup instructions in a Gist: <a href="https:&#x2F;&#x2F;gist.github.com&#x2F;awni&#x2F;ec071fd27940698edd14a4191855bba6" rel="nofollow">https:&#x2F;&#x2F;gist.github.com&#x2F;awni&#x2F;ec071fd27940698edd14a4191855bba...</a>

Outside of YouTube influencers, I doubt many home users are buying a 512G RAM Mac Studio.

it&#x27;s not like regular people can afford this kind of Apple machine anyway.

It’s gonna suck if all the good Macs get gobbled up by commercial users.

It already is depending on your needs.

It would be incredibly ironic if, with Apple&#x27;s relatively stable supply chain relative to the chaos of the RAM market these days (projected to last for years), Apple compute became known as a cost-effective way to build medium-sized clusters for inference.

Just keep going! 2TB of swap disk for 0.0000001 t&#x2F;sec

I did the same, then put in 14 3090&#x27;s. It&#x27;s a little bit power hungry but fairly impressive performance wise. The hardest parts are power distribution and riser cards but I found good solutions for both.

And if you get bored of that, you can flip the RAM for more than you spent on the whole system!

You can keep scaling down! I spent $2k on an old dual-socket xeon workstation with 768GB of RAM - I can run Deepseek-R1 at ~1-2 tokens&#x2F;sec.

There is no way to achieve a high throughput low latency connection between 25 Strix Halo systems. After accounting for storage and network, there are barely any PCIe lanes left to link two of them together.You might be able to use USB4 but unsure how the latency is for that.

You could use llama.cpp rpc mode over &quot;network&quot; via usb4&#x2F;thunderbolt connection

For $50K, you could buy 25 Framework desktop motherboards (128G VRAM each w&#x2F;Strix Halo, so over 3TB total) Not sure how you&#x27;ll cluster all of them but it might be fun to try. ;)

GPUs: 4x NVIDIA RTX 6000 Blackwell (96GB VRAM each)
• Cost: 4 × $9,000 = $36,000• CPU: AMD Ryzen Threadripper PRO 7995WX (96-Core)
• Cost: $10,000• Motherboard: WRX90 Chipset (supports 7x PCIe Gen5 slots)
• Cost: $1,200• RAM: 512GB DDR5 ECC Registered
• Cost: $2,000• Chassis &amp; Power: Supermicro or specialized Workstation case + 2x 1600W PSUs.
• Cost: $1,500• Total Cost: ~$50,700It’s a bit maximalist, but if you had to spend $50k it’s going to be about as fast as you can make it.

What&#x27;s the math on the $50k nvidia cluster? My understanding these things cost ~$8k and you can at least get 5 for $40k, that&#x27;s around half a tb.That being said, for inference mac still remain the best, and the M5 Ultra will even be a better value with its better PP.

Are you factoring in the above comment about as yet un-implemented parallel speed up in there? For on prem inference without any kind of asic this seems quite a bargain relatively speaking.

the GH&#x2F;GB compute has LPDDR5X - a single or dual GPU shares 480GB, depending if it&#x27;s GH or GB, in addition to the HBM memory, with NVLink C2C - it&#x27;s not bad!

Apple deploys LPDDR5X for the energy efficiency and cost (lower is better), whereas NVIDIA will always prefer GDDR and HBM for performance and cost (higher is better).

That thing will be extremely expensive I guess. And neither CPU nor GPU have that much memory. It&#x27;s also not a great workstation either - macOS is a lot more comfortable to use.

what about a GB300 workstation with 784GB unified mem?

Just wait for the M5-Ultra with a terabyte of RAM.

15 t&#x2F;s way too slow for anything but chatting, call and response, and you don&#x27;t need a 3T parameter model for thatWake me up when the situation improves

dang I wish I could share md tables.Here’s a text edition:
For $50k the inference hardware market forces a trade-off between capacity and throughput:* Apple M3 Ultra Cluster ($50k): Maximizes capacity (3TB). It is the only option in this price class capable of running 3T+ parameter models (e.g., Kimi k2), albeit at low speeds (~15 t&#x2F;s).* NVIDIA RTX 6000 Workstation ($50k): Maximizes throughput (&gt;80 t&#x2F;s). It is superior for training and inference but is hard-capped at 384GB VRAM, restricting model size to &lt;400B parameters.To achieve both high capacity (3TB) and high throughput (&gt;100 t&#x2F;s) requires a ~$270,000 NVIDIA GH200 cluster and data center infrastructure. The Apple cluster provides 87% of that capacity for 18% of the cost.

That screw hole is just the regular locking USB-C variant, is it not?See for example:<a href="https:&#x2F;&#x2F;www.startech.com&#x2F;en-jp&#x2F;cables&#x2F;usb31cctlkv50cm" rel="nofollow">https:&#x2F;&#x2F;www.startech.com&#x2F;en-jp&#x2F;cables&#x2F;usb31cctlkv50cm</a>

For #2, OWC puts a screw hole above their dock&#x27;s thunderbolt ports so that you can attach a stabilizer around the cord<a href="https:&#x2F;&#x2F;www.owc.com&#x2F;solutions&#x2F;thunderbolt-dock" rel="nofollow">https:&#x2F;&#x2F;www.owc.com&#x2F;solutions&#x2F;thunderbolt-dock</a>It&#x27;s a poor imitation of old ports that had screws on the cables, but should help reduce inadvertent port stress.The screw only works with limited devices (ie not the Mac Studio end of the cord) but it can also be adhesive mounted.<a href="https:&#x2F;&#x2F;eshop.macsales.com&#x2F;item&#x2F;OWC&#x2F;CLINGON1PK&#x2F;" rel="nofollow">https:&#x2F;&#x2F;eshop.macsales.com&#x2F;item&#x2F;OWC&#x2F;CLINGON1PK&#x2F;</a>

I have something similar from MyElectronics, and it works, but it&#x27;s a bit expensive, and still imprecise. At least the power button isn&#x27;t in the back corner underneath!

I have no experience with this, but for what it&#x27;s worth, looks like there&#x27;s a rack mounting enclosure available which mechanically extends the power switch: <a href="https:&#x2F;&#x2F;www.sonnetstore.com&#x2F;products&#x2F;rackmac-studio" rel="nofollow">https:&#x2F;&#x2F;www.sonnetstore.com&#x2F;products&#x2F;rackmac-studio</a>

This is a semi-solved problem e.g. <a href="https:&#x2F;&#x2F;www.sonnetstore.com&#x2F;products&#x2F;thunderlok-a" rel="nofollow">https:&#x2F;&#x2F;www.sonnetstore.com&#x2F;products&#x2F;thunderlok-a</a>Apple’s chassis do not support it. But conceptually that’s not a Thunderbolt problem, it’s an Apple problem. You could probably drill into the Mac Studio chassis to create mount points.

&quot;... Thunderbolt is great for peripherals, but as a semi-permanent interconnect, I have worries over the port&#x27;s physical stability ...&quot;Thunderbolt as a server interconnect displeases me aesthetically but my conclusion is the opposite of yours:If the systems are locked into place as servers in a rack the movements and stresses on the cable are much lower than when it is used as a peripheral interconnect for a desktop or laptop, yes ?

VNC over SSH tunneling always worked well for me before I had Apple Remote Desktop available, though I don&#x27;t recall if I ever initiated a connection attempt from anything other than macOS...erase-install can be run non-interactively when the correct arguments are used. I&#x27;ve only ever used it with an MDM in play so YMMV:<a href="https:&#x2F;&#x2F;github.com&#x2F;grahampugh&#x2F;erase-install" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;grahampugh&#x2F;erase-install</a>

With MDM solutions you can not only get software update management, but even full LOM for models that support this.
There are free and open source MDM out there.

They do still sell the Mac Pro in a rack mount configuration. But, it was never updated for M3 Ultra, and feels not long for this world.

&gt; To that last point, I&#x27;ve been trying to figure out a way to, for example, upgrade to macOS 26.2 from 26.1 remotely,I think you can do this if you install a MDM profile on the Macs and use some kind of management software like Jamf.

There are open source MDM projects, I&#x27;m not familiar but <a href="https:&#x2F;&#x2F;github.com&#x2F;micromdm&#x2F;nanohub" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;micromdm&#x2F;nanohub</a> might do the job for OS upgrades.

iOS build runner. Good luck developing cross-platform apps without a Mac!

It’s been terrible for years&#x2F;forever. Even Xserves didn’t really meet the needs of a professional data centre. And it’s got worse as a server OS because it’s not a core focus. Don’t understand why anyone tries to bother - apart from this MLX use case or as a ProRes render farm.

This implies you&#x27;d run more than one Mac Studio in a cluster, and I have a few concerns regarding Mac clustering (as someone who&#x27;s managed a number of tiny clusters, with various hardware):1. The power button is in an awkward location, meaning rackmounting them (either 10&quot; or 19&quot; rack) is a bit cumbersome (at best)2. Thunderbolt is great for peripherals, but as a semi-permanent interconnect, I have worries over the port&#x27;s physical stability... wish they made a Mac with QSFP :)3. Cabling will be important, as I&#x27;ve had tons of issues with TB4 and TB5 devices with anything but the most expensive Cable Matters and Apple cables I&#x27;ve tested (and even then...)4. macOS remote management is not nearly as efficient as Linux, at least if you&#x27;re using open source &#x2F; built-in toolingTo that last point, I&#x27;ve been trying to figure out a way to, for example, upgrade to macOS 26.2 from 26.1 remotely, without a GUI, but it looks like you _have_ to use something like Screen Sharing or an IP KVM to log into the UI, to click the right buttons to initiate the upgrade.Trying &quot;sudo softwareupdate -i -a&quot; will install minor updates, but not full OS upgrades, at least AFAICT.

That exists but it&#x27;s only for iUsers running Apple models. <a href="https:&#x2F;&#x2F;security.apple.com&#x2F;blog&#x2F;private-cloud-compute&#x2F;" rel="nofollow">https:&#x2F;&#x2F;security.apple.com&#x2F;blog&#x2F;private-cloud-compute&#x2F;</a>

The advantages of having a single big memory per gpu are not as big in a data center where you can just shard things between machines and use the very fast interconnect, saturating the much faster compute cores of a non Apple GPU from Nvidia or AMD

Apple should setup their own giant cloud of M chips with tons of vram, make Metal as good as possible for AI purposes, then market the cloud as allowing self-hosted models for companies and individuals that care about privacy. They would clean up in all kinds of sectors whose data can&#x27;t touch the big LLM companies.

Also see <a href="https:&#x2F;&#x2F;www.engadget.com&#x2F;ai&#x2F;you-can-turn-a-cluster-of-macs-into-an-ai-supercomputer-in-macos-tahoe-262-191500778.html" rel="nofollow">https:&#x2F;&#x2F;www.engadget.com&#x2F;ai&#x2F;you-can-turn-a-cluster-of-macs-i...</a>

That doesn&#x27;t answer the question, which was how to get a high-speed interconnect between a Mac and a DGX Spark. The most likely solution would be a Thunderbolt PCIe enclosure and a 100Gb+ NIC, and passive DAC cables. The tricky part would be macOS drivers for said NIC.

Sparks are built for this and actually have Connect-X 7 NICs built in! You just need to get the SFPs for them. This means you can natively cluster them at 200Gbps.

Is there any way to connect DGX Sparks to this via USB4? Right now only 10GbE can be used despite both Spark and MacStudio having vastly faster options.

I am waiting for M5 studio but due to current price of hardware I&#x27;m not sure it will be at a level that I would call affordable. Currently I&#x27;m watching for news and if there is any announcement prices will go up I&#x27;ll probably settle for an M4 Max.

I&#x27;ve been testing HPL and mpirun a little, not yet with this new RDMA capability (it seems like Ring is currently the supported method)... but it was a little rough around the edges.See: <a href="https:&#x2F;&#x2F;ml-explore.github.io&#x2F;mlx&#x2F;build&#x2F;html&#x2F;usage&#x2F;distributed.html#getting-started-with-mpi" rel="nofollow">https:&#x2F;&#x2F;ml-explore.github.io&#x2F;mlx&#x2F;build&#x2F;html&#x2F;usage&#x2F;distribute...</a>

Sure, there’s nothing about it that’s tied to ML. It’s faster interconnect , use it for many kinds of shared compute scenarios.

That’s great for AI people, but can we use this for other distributed workloads that aren’t ML?

Will Apple be able to ramp up M3 Ultra MacStudios if this becomes a big thing?Is this part of Apple’s plan of building out server side AI support using their own hardware?If so they would need more physical data centres.I’m guessing they too would be constrained by RAM.

Maybe Apple should rethink bringing back Mac Pro desktops with pluggable GPUs, like that one in the corner still playing with its Intel and AMD toys, instead of a big box full of air and pro audio cards only.

<a href="https:&#x2F;&#x2F;social.treehouse.systems&#x2F;@janne&#x2F;115509948515319437" rel="nofollow">https:&#x2F;&#x2F;social.treehouse.systems&#x2F;@janne&#x2F;115509948515319437</a> nvidia on a 2023 Mac Pro running linux :p

George Hotz made nvidia running on macs with his tinygrad via usb4<a href="https:&#x2F;&#x2F;x.com&#x2F;__tinygrad__&#x2F;status&#x2F;1980082660920918045" rel="nofollow">https:&#x2F;&#x2F;x.com&#x2F;__tinygrad__&#x2F;status&#x2F;1980082660920918045</a>

As someone not involved in this space at all, is this similar to the old MacOS Xgrid?<a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Xgrid" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Xgrid</a>

Do we think TB4 is on the table or is there a technical limitation?

For security, the feature requires setting a special option with the recovery mode command line:rdma_ctl enable

This sounds like a plug’n’play physical attack vector.

It&#x27;s faster and lower latency than standard Thunderbolt networking. Low latency makes AI clusters faster.

Can someone do an ELI5, and why this is important?

I found some useful clues here. Looks like it uses the regular InfiniBand RDMA APIs.<a href="https:&#x2F;&#x2F;github.com&#x2F;Anemll&#x2F;mlx-rdma&#x2F;commit&#x2F;a901dbd3f9eeefc62800da1142702bda8ea4f637" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;Anemll&#x2F;mlx-rdma&#x2F;commit&#x2F;a901dbd3f9eeefc628...</a>

Anyone found any APIs related to this?I&#x27;d have some other uses for RDMA between Macs.

at current memory prices today&#x27;s cheap is yesterday&#x27;s obscenely expensive - Apple&#x27;s current RAM upgrade prices are cheap

I imagine that M5 Ultra with Thunderbolt 5 could be a decent contender for building plug and play AI clusters. Not cheap, but neither is Nvidia.

You&#x27;ve been able to run inference on multiple Macs for around a year but now it&#x27;s much faster.

As someone that is not familiar with rdma, dos it mean I can connect multiple Macs and run inference? If so it’s great!

I hope so, I want to replace my M1 Pro with MacBook Pro with M5 Pro when they release it next year.

Apple tends to book its fab time &#x2F; supplier capacity years in advance

Hoping Apple has secured plentiful DDR5 to use in their machines so we can buy M5 chips with massive amounts of RAM soon.

The former will never happen and the latter is a certainty.

The end goal is that Macs become good local LLM inference machines and for AI devs to keep using Macs.

This is such a weird project. Like where is this running at scale? Where’s the realistic plan to ever run this at scale? What’s the end goal here?Don’t get me wrong... It’s super cool, but I fail to understand why money is being spent on this.

I don&#x27;t believe the standard supports such a thing. But I wonder if TB6 will.

I&#x27;m sure someone will figure out how to make thunderbolt switch&#x2F;router

Very cool. It requires a fully-connected mesh so the scaling limit here would seem to be 6 Mac Studio M3 Ultra, up to 3TB of unified memory to work with.

Now we need some hardware that is rackmount friendly, an OS that is not fidly as hell to manage in a data center or headless server and we are off to the races! And no, custom racks are not &#x27;rackmount friendly&#x27;.

Networking is part of the OS&#x27;s job.

IS this... good? Why is this something that the underlying OS itself should be involved in at all?

I didn&#x27;t know they skipped 10 version numbers.

If it was still around, it would probably still be stuck on M2, just like the Mac Pro.

Imagine if the Xserve was never killed off. Discontinued 14 years ago, now!

Yeah, even decent 40 Gbps QSFP+ DAC cables are usually $30+, and those don&#x27;t have active electronics in them like Thunderbolt does.The ability to also deliver 240W (IIRC?) over the same cable is also a bit different here, it&#x27;s more like FireWire than a standard networking cable.

In this case Thunderbolt is much much cheaper than 100G Ethernet.(The cord is $50 because it contains two active chips BTW.)

Remember when they enabled egpu over thunderbolt and no one cared because the thunderbolt housing cost almost as much as your macbook outright? Yeah. Thunderbolt is a racket. It’s a god damned cord. Why is it $50.

Just for reference:Thunderbolt5&#x27;s stated &quot;80Gbps&quot; bandwidth comes with some caveats. That&#x27;s the figure for either Display Port bandwidth itself or in practice more often realized by combining the data channel (PCIe4x4 ~=64Gbps) with the display channels (=&lt;80Gbps if used in concert with data channels), and potentially it can also do unidirectional 120Gbps of data for some display output scenarios.If Apple&#x27;s silicon follows spec, then that means you&#x27;re most likely limited to PCIe4x4 ~=64Gbps bandwidth per TB port, with a slight latency hit due to the controller. That Latency hit is ItDepends(TM), but if not using any other IO on that controller&#x2F;cable (such as display port), it&#x27;s likely to be less than 15% overhead vs Native on average, but depending on drivers, firmware, configuration, usecase, cable length, and how apple implemented TB5, etc, exact figures very. And just like how 60FPS Average doesn&#x27;t mean every frame is exactly 1&#x2F;60th of a second long, it&#x27;s entirely possible that individual packets or niche scenarios could see significantly more latency&#x2F;overhead.As a point of reference Nvidia RTX Pro (formerly known as quadro) workstation cards of Ada generation and older along with most modern consumer grahics cards are PCIe4 (or less, depending on how old we&#x27;re talking), and the new RTX Pro Blackwell cards are PCIe5. Though comparing a Mac Studio M4 Max for example to an Nvidia GPU is akin to comparing Apples to Green OrangesHowever, I mention the GPU&#x27;s not just to recognize the 800lb AI compute gorilla in the room, but also that while it&#x27;s possible to pool a pair of 24GB VRAM GPU&#x27;s to achieve a 48GB VRAM pool between them (be it through a shared PCIe bus or over NVlink), the performance does not scale linearly due to PCIe&#x2F;NVLinks limitations, to say nothing of the software, and configuration and optimization side of things also being a challenge to realizing max throughput in practice.This is also just as true as a pair of TB5 equipped macs with 128GB of memory each using TB5 to achieve a 256GB Pool will take a substantial performance hit compared to on otherwise equivalent mac with 256GB. (capacities chosen are arbitrary to illustrate the point). The exact penalty really depends on usecase and how sensitive it is to the latency overhead of using TB5 as well as the bandwidth limitation.It&#x27;s also worth noting that it&#x27;s not just entirely possible with RDMA solutions (no matter the specifics) to see worse performance than using a singular machine if you haven&#x27;t properly optimized and configured things. This is not hating on the technology, but a warning from experience for people who may have never dabbled to not expect things to just &quot;2x&quot; or even just better than 1x performance just by simply stringing a cable between two devices.All that said, glad to see this from Apple. Long overdue in my opinion as I doubt we&#x27;ll see them implement an optical network port with anywhere near that bandwidth or RoCEv2 support, much less a expose a native (not via TB) PCIe port on anything that&#x27;s a non-pro model.EDIT: Note, many mac skus have multiple TB5 ports, but it&#x27;s unclear to me what the underlying architecture&#x2F;topology is there and thus can&#x27;t speculate on what kind of overhead or total capacity any given device supports by attempting to use multiple TB links for more bandwidth&#x2F;parallelism. If anyone&#x27;s got an SoC diagram or similar refernce data that actually tells us how the TB controller(s) are uplinked to the rest of the SoC, I could go in more depth there. I&#x27;m not an Apple silicon&#x2F;MacOS expert. I do however have lots of experience with RDMA&#x2F;RoCE&#x2F;IB clusters, NVMeoF deployments, SXM&#x2F;NVlink&#x27;d devices and generally engineering low latency&#x2F;high performance network fabrics for distributed compute and storage (primarily on the infrastructure&#x2F;hardware&#x2F;ops side than on the software side) so this is my general wheelhouse, but Apple has been a relatively blindspot for me due to their ecosystem generally lacking features&#x2F;support for things like this.

They were only a thing to do ci&#x2F;compilation related to apples os because their walled garden locked using other platforms out. You&#x27;re building an iPhone or mac app? Well your ci needs to be on a cluster of apple machines.

It&#x27;s in a similar vein to the PS2 linux cluster or someone trying to use vape CPU&#x27;s as web servers...It might be cost effective, but the supplier is still saying &quot;you get no support, and in fact we might even put roadblocks in your way because you aren&#x27;t the target customer&quot;.

You might ignore this but, for a while, Mac Mini clusters were a thing and they were capex and opex effective. That same setup is kind of making a comeback.

Nobodies gonna take them seriously till they make something rack mounted and that isn&#x27;t made of titanium with pentalobe screws...

does this means an egpu might finally work with macbook-pro or studio?

That&#x27;s nice butLiquid (gl)ass still sucks.

MacOS does wash out SDR content in HDR mode specifically on non-Apple monitors. An HDR video playing in windowed mode will look fine but all the UI around it has black and white levels very close to grey.Edit: to be clear, macOS itself (Cocoa elements) is all SDR content and thus washed out.

Works well on Linux, just toggle a checkmark in the settings.

Really? I thought it&#x27;s always been that HDR was notorious on Windows, hopeless on Linux, and only really worked in a plug-and-play manner on Mac, unless your display has an incorrect profile or something&#x2F;<a href="https:&#x2F;&#x2F;www.youtube.com&#x2F;shorts&#x2F;sx9TUNv80RE" rel="nofollow">https:&#x2F;&#x2F;www.youtube.com&#x2F;shorts&#x2F;sx9TUNv80RE</a>

AI is arguably more important than whatever gaming gimmick you&#x27;re talking about.

Can we get proper HDR support first in macOS? If I enable HDR on my LG OLED monitor it looks completely washed out and blacks are grey. Windows 11 HDR works fine.

inference accelerators ... It’s why Apple is so far ahead of their peers in local device AI compute, and will be for some time.Not really. llama.cpp was just using the GPU when it took off. Apple&#x27;s advantage is more VRAM capacity.this introduction isn’t just for large clustersIt doesn&#x27;t work for large clusters at all; it&#x27;s limited to 6-7 Macs and most people will probably use just 2 Macs.

You&#x27;re thinking, &quot;You can&#x27;t put modern models on that sort of distributed compute network&quot;, which is technically correct.I was thinking, &quot;How could we package or run these kinds of large models or workloads across a consumer&#x27;s distributed compute?&quot; The Engineer in me got as far as &quot;Enumerate devices on network via mDNS or Bonjour, compare keys against iCloud device keys or otherwise perform authentication, share utilization telemetry and permit workload scheduling&#x2F;balance&quot; before I realized that&#x27;s probably what they&#x27;re testing here to a degree, even if they&#x27;re using RDMA.

The bandwidth of rdma over thunderbolt is so much faster (and lower latency) than Apple&#x27;s system of mostly-wireless devices, I can&#x27;t see how any learnings here would transfer.

Consumers generally have far more compute than they think; it&#x27;s just all distributed across devices and hard to utilize effectively over unreliable interfaces (e.g. Wi-Fi). If Apple (or anyone, really) could figure out a way to utilize that at modern scales, I wager privacy-conscious consumers would gladly trade some latency in responses in favor of superior overall model performance - heck, branding it as &quot;deep thinking&quot; might even pull more customers in via marketing alone (&quot;thinks longer, for better results&quot; or some vaguely-not-suable marketing slogan). It could even be made into an API for things like batch image or video rendering, but without the hassle of setting up an app-specific render farm.There&#x27;s definitely something there, but Apple&#x27;s really the only player setup to capitalize on it via their halo effect with devices and operating systems. Everyone else is too fragmented to make it happen.

I think you are spot on, and this fits perfectly within my mental model of HomeKit; tasks are distributed to various devices within the network based on capabilities and authentication, and given a very fast bus Apple can scale the heck out of this.

This doesn’t remotely surprise me, and I can guess Apple’s AI endgame:* They already cleared the first hurdle to adoption by shoving inference accelerators into their chip designs by default. It’s why Apple is so far ahead of their peers in local device AI compute, and will be for some time.* I suspect this introduction isn’t just for large clusters, but also a testing ground of sorts to see where the bottlenecks lie for distributed inference in practice.* Depending on the telemetry they get back from OSes using this feature, my suspicion is they’ll deploy some form of distributed local AI inference system that leverages their devices tied to a given iCloud account or on the LAN to perform inference against larger models, but without bogging down any individual device (or at least the primary device in use)For the endgame, I’m picturing a dynamically sharded model across local devices that shifts how much of the model is loaded on any given device depending on utilization, essentially creating local-only inferencing for privacy and security of their end users. Throw the same engines into, say, HomePods or AppleTVs, or even a local AI box, and voila, you’re golden.EDIT: If you&#x27;re thinking, &quot;but big models need the higher latency of Thunderbolt&quot; or &quot;you can&#x27;t do that over Wi-Fi for such huge models&quot;, you&#x27;re thinking too narrowly. Think about the devices Apple consumers own, their interconnectedness, and the underutilized but standardized hardware within them with predictable OSes. Suddenly you&#x27;re not jamming existing models onto substandard hardware or networks, but rethinking how to run models effectively over consumer distributed compute. Different set of problems.

macOS 26.2 enables fast AI clusters with RDMA over Thunderbolt

CapROS: The Capability-Based Reliable Operating System

2002: Last.fm and Audioscrobbler Herald the Social Web

Anthropic Outage for Opus 4.5 and Sonnet 4/4.5 across all services

JSDoc is TypeScript

Hashcards: A plain-text spaced repetition system

Rio de Janeiro's talipot palm trees bloom for the first and only time

What are you working on? Any new ideas that you&#x27;re thinking about?