A good technical project, but honestly useless in like 90% of scenarios.
You want to use an NVidia GPU for LLM ? just buy a basic PC on second hand (the GPU is the primary cost anyway), you want to use Mac for good amount of VRAM ? Buy a Mac.
With this proposed solution you have an half-backed system, the GPU is limited by the Thunderbolt port and you don’t have access to all of NVidia tool and library, and on other hand you have a system who doesn’t have the integration of native solution like MLX and a risk of breakage in future macOS update.
Nvidia GPUs were usable on Intel Macs, but compatibility got worse over time, and Apple stopped making a Mac Pro with regular PCIe slots in 2013. People then got hopeful about eGPUs, but they have their own caveats on top of macOS only fully working with AMD cards. So I've gotten numb to any news about Mac + GPU. The answer was always to just get a non-Apple PC with PCIe slots instead of giving yourself hoops to jump through.
Until there is official support for Mac coming from nvidia, I don't think anything will happen.
> the hardware wasn't usable on macOS
This eGPU thing is from a third-party if I understand correctly. I don't see why nvidia would get excited about that. If they cared about the platform, they would have released something already.
The point is that if nvidia cared about Mac platform they would have done something to make eGPU usable on Mac a long time ago.
Even on Intel Macs using eGPU with nvidia cards was near impossible. nvidia just doesn't care about it after the breakdown of the two companies' relationship.
Whether a third party has created a signed driver or not doesn't matter much until there is more interest from the GPU maker. This barely moves the needle.
If a model can run on a 512GB M3 Ultra via MLX or CUDA, but simultaneously benefit from the memory bandwidth of something like an RTX 6000 Pro; that would save my company hundreds of thousands of dollars. That's $20,000 for roughly 600GB of VRAM, and enough token generation speed to fulfill the needs of any enterprise that's not a hyperscaler or neocloud.
I'll let someone else do the math for you on what it costs to put together a 10U server to get that kind of performance without the $10K M3 Ultra Studio.
What we're paying for five old 80GB A100s is criminal, but it's nothing compared to what these GB200 Blackwell setups are going to cost in 2030. Market economics aside, the fact that they require sophisticated liquid cooling infrastructure and draw 3x the power of the A100s, will make these cards unattainable for small to medium organizations.
So yeah, if there's some outside chance that we can pair NVIDIA's speed with a an arm-powered machine that offers 512GB Unified Memory while drawing 50W -- you better believe it's a big deal. We'll see. Sounds too good to be true.
"Nvidia." Not NVidia or nVidia, or the other ways. I feel that I can frequently figure out if someone is going to express a negative view about this company based only on whether they picked a weird way to write their name.
There's more to peripheral limits than the protocol used. Thunderbolt connections offer higher latency and limits on bandwidth. Both, either, or neither of those things may be much of an actual problem (depending on the use case) but they are some examples of limits vs native PCIe.
Evidence that NVIDIA has even been trying? My understanding is that Apple didn’t allow 3rd parties to write graphics drivers past 10.13, but they could’ve done a non-graphics driver like this.
i emailed jen sen huang at the very tail end of the maxwell era and p much begged for maxwell support on macos. i didnt expect a reply, especially since i guessed his email based on some "how to find ceo emails" google search result.
he actually did reply weeks later and said "i didnt realize people wanted this, my team has added them. go check now". pretty sure that was the last time nvidia drivers came to macos.
there's a lot of assumptions made with this topic, particularly the assumption that apple is blocking them. at least in my experience the opposite was true, nvidia just flat out wasn't making them. however i don't doubt the truth lies somewhere in between: nvidia and apple have a pretty much nonexistant relationship now. i dont know whats required here but i also don't doubt apple makes this experience suck butt for any interested parties.
The government doesn’t care? They’re a minority of the market? The vast majority of their computers didn’t have slots to put Nvidia GPUs in, and now none of them do?
An internal PCIe slot can be had in up to 16x 5.0, whereas Thunderbolt 5 maxes out at 4x of 4.0.
Plus you have another Thunderbolt controller in between the CPU and the hardware, and it takes more energy to push that many bits 1m over a cable vs a few dozen cm over traces.
Also Thunderbolt is trivially disconnected, which in many critical workflows is not a positive, but an opportunity for ill-timed interruptions. Plus I don't have to buy a fucking dongle/dock for a real goddamn slot, make room for external power supplies, etc.
It depends how you define the market. In the 2001 microsoft case [0], the courts ruled Microsoft had a monopoly over the "Intel-based personal computer market".
Apple has a monopoly over the "M-chip" personal computer market. They have a monopoly over the iOS market with the app store. They have a monopoly over the driver market on macOS.
Like, Microsoft was found guilty of exploiting its monopoly for installing IE by default while still allowing other browser engines. On iOS, apple bundles safari by default and doesn't allow other browser engines.
If we apply the same standard that found MS a monopoly in the past, then Apple is obviously a monopoly, so at the very least I think it's fair to say that reasonable people can disagree about whether Apple is a monopoly or not.
I wouldn’t say it is obvious. Apple does not have the monopoly of ARM based PCs. Labeling it as a monopoly of M chips is not fair or accurate when comparing to MS on Intel. It’s also probably relevant that MS was not selling PCs or their own hardware. They had a monopoly on a market where you effectively had to use their software to use the hardware you bought from a different company. Because Apple is selling their own hardware and software as a single product, the consumer is not forced into restricting the hardware they bought by a second company’s policies.
> Labeling it as a monopoly of M chips is not fair or accurate when comparing to MS on Intel.
The relevant thing here isn't the chips, it's tying things to the chips, because those would otherwise be separate markets. If you could feasibly buy an iPhone and install Android or Lineage OS on it or use Google Play or F-Droid on iOS then no one would be saying that Apple has a monopoly on operating systems or app stores for iOS since there would actually be alternatives to theirs.
The fake alternative is that you could use a different store by buying a different phone, but this is like saying that if Toyota is the only one who can change the brake pads on a Toyota and Ford is the only one who can change the brake pads on a Ford then there is competition for "brake pads" because when your Toyota needs new brake pads you can just buy a Ford vehicle. It's obvious why this is different than anyone being able to buy third party brake pads for your Toyota from Autozone, right?
> It’s also probably relevant that MS was not selling PCs or their own hardware.
This is the thing that unambiguously should never be relevant. It can't be a real thing that you can avoid being a monopoly by owning more of the supply chain. It's like saying that Microsoft could have avoided being a monopoly by buying Intel and AMD, or buying one of them and then exterminating the other by refusing to put Windows on it. That's a preposterous perverse incentive.
> It can't be a real thing that you can avoid being a monopoly by owning more of the supply chain.
Move the most important aspects of your software to hardware. Hard for MacOS but for a Chromebook style thing you could write the browser into its own pice of wafer.
Well “had to use” is a strong phrase here. Linux was already around and you could have used it too with your hardware. I think you can always bend an argument to fit your point.
I don't think any of what you're describing are legal "monopolies". I don't have a single Apple product in my life but I'm fairly sure there's nothing I'm prevented from doing because of that.
And back in the "Microsoft has a monopoly on IE6" ruling's days, I did not use Windows or Internet Explorer, and I was not prevented from doing anything because of that. Netscape Navigator on Linux worked fine. Sure, I occasionally hit sites that were broken and only worked in IE, but I also right now frequently hit apps that are "macOS only" (like when Claude Cowork released, or a ton of other YC company's apps).
Microsoft was found guilty, so clearly the bar is not what you're trying to claim.
Microsoft was found guilty of using their market power to do product bundling, which is illegal. The fact that they had dominance in the market is not what they got popped for, nor is it illegal.
Yes, but that was coupled with other factors like them strongarming vendors, already being hugely dominant on desktops and abusing that position et al. I don't see this as being the same. Maybe my bar here is wrong, but it doesn't change whether they are a monopoly or not.
The issue was never "Microsoft has a monopoly on IE6". That's obviously nonsense.
The monopoly that Microsoft held was the home computer operating system market, first through DOS, then later through Windows. Holding a monopoly like that isn't illegal unto itself. What they were actually found guilty of was unfairly leveraging their monopoly on the OS market to gain the upper hand in a different market (the browser market). The subsequent range of issues we had with IE6 (compatibility, security, etc) was a result of Microsoft succeeding in achieving a monopoly on the browser market through illicit means.
Likewise, "Apple has a monopoly on the App Store" is just the same amount of nonsense. What you could argue is that Apple has a monopoly on the home computer market, or the mobile phone market, and that the way they integrate the App Store should be considered illegal leveraging of that monopoly, but that argument simply doesn't hold water — Microsoft's monopoly on the OS market at the time was pretty much incontrovertible, you simply couldn't walk into a shop and buy a computer running something else (except maybe a Mac at a more specialised place). Today, just about any shop you walk into that sells computers will probably have devices for sale running three different OSes (macOS, Windows, ChromeOS). Any phone place will have iPhones and Android devices, and probably a few more niche options. Actual market share percentage is nowhere near the high 90s that Microsoft saw in its heyday. At most, Apple is the biggest individual competitor in the market, but I don't think it hold an outright majority in any specific product class.
Mind you, I think that there is a good argument to be made that the Apple/Google duopoly on mobile devices does deserve scrutiny, but that's a very different kettle of fish.
You were not prevented from doing anything, but that doesn’t mean others weren’t. For example, OEMs were not allowed to offer any other preinstalled OS as a default option. That effectively killed Be and I’m sure hindered RedHat.
That’s not how monopoly definitions work. That makes about as much sense as saying Nintendo has a monopoly on Nintendo consoles or Ford has a monopoly on Mustangs
> Apple has a monopoly over the "M-chip" personal computer market. They have a monopoly over the iOS market with the app store
When a company is deemed an illegal monopoly, the DoJ basically becomes part of management. Antitrust settlements focus on germane elements, e.g. spin offs. But they also frequently include random terms of political convenience.
I don’t think we want a precedent where companies having a product means they have an automatic monopoly on said product.
More to the point: having a monopoly isn't de facto illegal (just look up natural monopolies), it's using the monopoly power in an anti-competitive way that's illegal. Microsoft wasn't charged with having a monopoly, they were charged because they used that monopoly to exclude Netscape Navigator and force bundling of IE.
There’s no such thing as “monopoly on Apple-produced processors” because that’s absurd. The monopoly for MacBook would be “consumer laptops” most likely. Apple does not have a monopoly in consumer laptops to the best of my knowledge.
> Apple has a monopoly over the "M-chip" personal computer market
lmao what ? the "M-chip" is literally their chip that they designed, built relationships with TSMC over and bankrolled into production to put in their products. literally hardware by apple for apple. this was a decade plus long thing in the making, this is the risk/gamble apple took and invested heavily into. that is apples innovation. any other manuf is free to go do this themselves for their own devices, they just didn't and for the most part still don't. that just like isn't a monopoly at all, i'm amused you even got to that point in the first place. seems to carry some broad misunderstandings of what the M-series chips are or carries an assumption that cpus are supposed to be shared to any interested parties just because that was intels business model. intel was historically slacking & their one-size-fits-most approach wasn't meeting the engineering requirements apple was after generation after generation, so apple took the cpu destiny into their own hands and made their own. if you feel like non-apple laptop chips aren't living up to that kind of perf/ppu.... well yeah you'd be right. but that's not really apples fault. that's not a monopoly thing, like at all. either laptop manufs need to go make their own chip (unlikely) or intel/qualcomm/etc need to catch up.
It isn't just about monopoly or unfair competition. This can also be covered under consumer rights - the Right to Repair. No OS provider should be allowed to dictate what software you can or not run on your own device and / or OS you have paid for.
> It isn't just about monopoly or unfair competition. This can also be covered under consumer rights - the Right to Repair.
If we have a right to repair (we broadly do not, AFAICT), then that doesn't necessarily mean that we have a right to modify and/or add new functionality.
When I repair a widget that has become broken, I merely return it to its previous non-broken state. I might also decide to upgrade it in some capacity as part of this repair process, but the act of repairing doesn't imply upgrades. At all.
> No OS provider should be allowed to dictate what software you can or not run on your own device and / or OS you have paid for.
I agree completely, but here we are anyway. We've been here for quite some time.
Courts have already ruled it does in the iOS app store market. You can disagree of course but then you'd be disagreeing with legal experts who know more about anti-trust law than you do.
You can, but that doesn't mean your opinion is as valid as those who study the subject. Otherwise we might as well follow the sovereign citizen believers.
Internet Explorer Mobile is a YouTube client. You're describing a client-server disagreement when the user is talking about an entirely client-based conflict.
> That's normal behavior when your server is being reverse-engineered or abused. Video bandwidth is not free.
Microsoft rewrote their Windows Phone native client to pass through Google's ads. Google still blocked it.
Was it normal behavior when Google blocked Amazon Fire devices from connecting to YouTube with a web browser during the Google/Amazon corporate spat?
To be fair, Google did back down almost immediately when the tech press picked up on it.
Not allowing a native client for your monopoly market share video service on Amazon devices while also blocking Amazon's web browser on those devices is making things a bit too obvious.
Yeah I'm pretty sure Nvidia just doesn't care to make Mac drivers. For years there was no SIP, Apple sold the Mac Pro which could take Nvidia GPUs, but you basically couldn't use Nvidia because of how bad and outdated the drivers were. I had a GTX 650 in my Mac Pro for a while, it was borderline unusable.
Such a shame both companies are big on vanity to make great things happen. Imagine where you could run Mac hardware with nvidia on linux. It's all there, and closed walls are what's not allowing it to happen. That's what we as customers lose when we forego control of what we purchase to those that sold us the goods.
Unfortunately, Apple still won't release iMessage for Android or Linux (unlike every other messenger platform, like whatsapp, telegram, wechat, microsoft teams, etc, which are all cross-platform).
Because of that, you need an apple device around to be able to deal with iMessage users.
Then it would be more correct to say that we "lose when we forego control" when our friends push the iMessage on us.
In my bubble literally noone uses iMessage. More tech savvy use Signal/GroupMe, less tech savvy use SMS/Email. Family use Signal to chat with me, as I can steer my own family a little.
Also I sometimes open web-interface of Facebook, but any attempts to offer WhatsApp I answer "sorry no Facebook apps on my phone, no Instagram/Messenger either". Never had any issues with that. Although I heard some countries are very dependent on Facebook, so might be hard there.
By the way, I noticed it's not hard to use multiple messengers actually, sometimes it's even faster to find a contact as you always remember what app to look at in recents.
UPDATE: My point is that you can also influence your life and how people communicate with you. Up to a point of course, but it's not like you can do nothing with it.
My social circle is the complete opposite. We're all on iMessage (except for one group of extended family on Messenger), and we like it that way. I was the last holdout for years while I went from Android -> Windows Phone -> Android -> iPhone.
But you don't need an Apple device to contact iMessage users. Every iMessage ID is a phone number (SMS/RCS) or email.
You've listed a whole bunch of alternatives available to you, but for some reason you demand that Apple change its unique offering into just another one of those for you. Why? Is that not a completely enforced monoculture?
Apple has always been off to the side, doing their own thing, and for some reason that fact utterly enrages people. They demand that Apple become just like everyone else. But we already have everyone else! And in every single field Apple is in, there is more of everyone else than there is of Apple.
Have you considered people like Apple products precisely because they're not like everything else? That making Apple indistinguishable from Facebook or Google is no victory, but a significant loss for customer choice?
I have been an Android user for the past 15 years, and somehow iMessage has never been a problem. Most of the time I don't even know if someone uses iMessage or not.
Good thing that iMessage is only popular in the US. I have never seen anybody using it, I don't even know how it looks, and if someone told me to use it I would laugh at them.
That is no longer true. https://bluebubbles.app/
Well… it’s not exactly no longer true, you do need an Apple VM but it doesn’t have to be the end device.
I don't understand the logic for downvotes.
We vote with our wallets.
When I could not update the Ram on my personal Dell machine I asked for a Frame.work in my new job. As my Intel based FW at work had thermal throttling problems, for my next personal purchase I got an AMD one. As Ubuntu had shady practices, I installed Fedora, as Gnome forced UX choices I did not want, I used KDE. As I wanted my machine to be even more stable I use an immutable spin.
The machine I'm using now represents my choices and matches what matters to me, and works closer to perfectly than all my machines in the past
And yes, I have worked with macs, and no, the UX and the entire tyranny in the Apple ecosystem was not something I could live with
And yes, this machine is fast, predictable, a joy to work with and is a tool I control, not a tool to control me. If something happens to it, I can order the part with the same price that goes into a new machine, and keep using my laptop
"We vote with our wallet, so don't complain" is a bad take in my opinion.
Like, for phones, I want a phone which runs Linux, has NFC support, and also has iMessage so my friend who only communicates with blue-bubbles and will never message a green-bubble will still talk to me. I also want it to have regulatory approval in the country I live in so I can legally use it to make calls.
Because apple has closed the iMessage ecosystem such that a linux phone can't use it, such a device is impossible. I cannot vote for it.
As such, I will complain about every phone I own for the foreseeable future.
> Like, for phones, I want a phone which runs Linux, has NFC support, and also has iMessage so my friend who only communicates with blue-bubbles and will never message a green-bubble will still talk to me. I also want it to have regulatory approval in the country I live in so I can legally use it to make calls.
I actually agree with you, but I also suggest getting better friends.
What is the blue and green bubble thing? I've never used an iPhone so don't understand the term. Does it classify messages as iMessage and non-iMessage?
iOS has two built-in messaging apps. Like all phones, they have SMS built in, and hardly anyone uses it for anything except SMS 2FA codes.
And then they have iMessage, aka blue bubbles, which are kinda like Signal or Whatsapp or Telegram. Everyone in Europe uses whatsapp, and a lot of people in the US use iMessage. If you don't use whatsapp in europe, you'll have a rough time communicating with some social groups, and the same thing for iMessage in the US.
However, unlike every other messenger app I can think of, iMessage isn't cross platform.
Also unlike every other messenger I can think of, it comes installed by default and for some reason uses the same app as the SMS app, and also claims encryption but randomly switches to SMS and breaks encryption making it obviously the least secure of all the apps (and also backs up your keys to iCloud in a way apple can access them by default, neither here nor there).
Blue bubbles are when iMessage is acting as the iMessage app, and has encryption and can use features like sending high resolution photos, location, invites, and a bunch of other apple-specific features.
Green bubbles are when the iMessage app has converted itself into the SMS and RCS app, and has a reduced feature set, like being unable to remove people from group chats.
It's frankly a quite confusing decision to have two quite different apps built into the same app and indicate which feature-set is active based on the color of a UI element. I think everyone would prefer if apple split it into the 'Messages' app (SMS + RCS) and an optional 'iMessage' app which doesn't come installed by default, but you can download on the app store from Apple. I'm frankly surprised the EU hasn't forced apple to show a prompt for "default messenger app" on startup with the options being "Whatsapp", "iMessage", etc etc, like they do for default browser.
Woah, this is exciting. I'm traveling but I have a 5090 lying around at home. I'm eager to give it a go. Docs are here: https://docs.tinygrad.org/tinygpu/
I hope it'll work on an M4 Mac Mini. Does anyone know what hardware to get? You'll need a full ATX PSU to supply power, right? And then tinygrad can do LLM inference on it?
I own one of these, the cage is just a piece of plastic. Anyway, I don't think 80$ is that big of a difference here. I can't really afford a 4k Nvidia GPU. Intel is my only hope.
Almost twice the price and simply more accurate info regarding price and features.
Brand is TH3P4G3. Egpu.io has decent eGPU comparisons.
I wouldn't want all that dust in my GPU fans, prefer that near my case fans. I also don't like it given I got cats and want to store/box hw. I do use the eGPU in the fuse box. If I had a larger house, I'd use a server rack.
I was recently in the market for an eGPU but for a different niche (not eGPU/eNPU/eTPU but getting a HBA via TB to connect a LTO-6 drive via SAS). I went for a Sonnet instead, very low profile and small. I also bought an Asus one. Slightly bigger, came with more fans but TB4 instead of TB3 on the Sonnet. The cages are aluminium. Those eGPU were second hand (also without warranty but quicker S&H than Chinese New Year) but came with PSU. As you also gotta buy a PSU for it which came with the eGPUs I mentioned. For me no biggie, as I got a decent PSU lying around.
Maybe I’m lacking imagination. But how will a GPU with small-ish but fast VRAM and great compute, augment a Mac with large but slow VRAM and weak compute? The interconnect isn’t powerful enough to change layers on the GPU rapidly, I guess?
> But how will a GPU with small-ish but fast VRAM and great compute, augment a Mac with large but slow VRAM and weak compute?
It would work just like a discrete GPU when doing CPU+GPU inference: you'd run a few shared layers on the discrete GPU and place the rest in unified memory. You'd want to minimize CPU/GPU transfers even more than usual, since a Thunderbolt connection only gives you equivalent throughput to PCIe 4.0 x4.
My Mini is actually the smallest model so it actually has "small but slow VRAM" (haha!) so the reason I want the GPU for are the smaller Gemmas or Qwens. Realistically, I'll probably run on an RTX 6000 Pro but this might be fun for home.
based on your card, it should be using a decent bit of the 600w or so passed through the new 16-pin connector. goes without saying, a proper PSU (doesn't have to be ATX, but at least 750W to be on the safer side) is a must.
“Lying around”. I’ve got an unopened 5090 in a box that I know will suffer the same fate, so I’m sending it back. So privileged to have the money to impulse buy a 5090 and yet no time to actually do anything with it.
I followed the instructions link and read the scripts...although the TinyGPU app is not in source form on GitHub, this looks to me like the GPU is passed into the Linux VM underneath to use the real driver and then somehow passed back out to the Mac (which might be what the TinyGrad team actually got approved).
Or I could have totally misunderstood the role of Docker in this.
My read of everything is that they are using Docker for NVIDIA GPUs for the sake of "how do you compile code to target the GPU"; for AMD they're just compiling their own LLVM with the appropriate target on macOS.
I think that metal isn’t double precision; so that limits some serious physics simming; but if you’re doing that I guess you just rent a gpu somewhere.
I would definitely be into this if adding an egpu was first class supported.
Well, to be fair, the whole shebang is from a completely different company, that have their own ML library and such, so that isn't that surprising. Although I agree that some CUDA shim or similar would be a lot more interesting, still getting to the place of running inference and training with your very own library is pretty dope already.
What do you mean when you say "run"? Low graphics 45 FPS at 720p? Or ultra graphics 120 FPS at 4k? My assumption is that a fairly large part of that space is inaccessible with the integrated GPU.
> This all feels like NVidia trying to be relevant versus ARM.
Except it's done by a third group, tinygrad, so it's more non-nvidia people wanting to use nvidia hardware one Apple hardware, than "nvidia trying to be relevant".
I'm writing scientific software that has components (molecular dynamics) that are much faster on GPU. I'm using CUDA only, as it's the eaisiest to code for. I'd assumed this meant no-go on ARM Macs. Does this news make that false?
No, MLX is nothing like a Cuda translation layer at all. It’d be more accurate to describe MLX as a NumPy translation layer; it lets you write high level code dealing with NumPy style arrays and under the hood will use a Metal GPU or CUDA GPU for execution. It doesn’t translate existing CUDA code to run on non-CUDA devices.
Well, for starters, PCIe 5.0 x16 would do something like about 60 GB/s each way, while Thunderbolt 4 does 4 GB/s each way, TB 5 does 8 GB/s each way. If you don't actually hit the bandwidth limits, it obviously matters less. Whether you'd notice a large difference would depends heavily on the type of workload.
I hooked up a Radeon RX 9060 XT to my Feodra KDE laptop (Yoga Pro 7 14ASP9) using a Razer Core X Chroma (40Gbps), and the performance when using the eGPU was very similar to using the Radeon 880M built into the laptop's Ryzen 9 365 APU.
So at least with my setup, performance is not great at all.
On paper, TB4 is capable of pushing 5GB/s, which is somewhere between 4x and 8x of PCIe 3.0, while a 16x PCIe 4.0 link can do ~31.5GB/s.
For gaming, lots of things can affect Thunderbolt eGPU performance.
First, you need to connect the display directly to the eGPU rather than to the laptop.
Second, you need to make sure you have enough VRAM to minimize texture streaming during gameplay.
Third, you'll typically see better performance in terms of higher settings/resolutions vs higher framerates at lower settings/resolutions.
Fourth, depending on your system, you may be bottlenecked by other peripherals sharing PCH lanes with the Thunderbolt connection.
Finally, depending on the Thunderbolt version, PCIe bandwidth can be significantly lower than the advertised bandwidth of the Thunderbolt link. For example, while Thunderbolt 3 advertises 40 Gbps, and typically connects via x4 PCIe 3.0 (~32 Gbps), for whatever reason it imposes a 22 Gbps cap on PCIe data over the Thunderbolt link.
Even taking all this into account, you'll still see a significant performance drop on a current-gen GPU when running over Thunderbolt, though I'd still expect a useful performance improvement over integrated graphics in most cases (though not necessarily worth the cost of the eGPU enclosure vs just buying a cheap used minitower PC on eBay and gaming on that instead of a laptop).
Thunderbolt is its own protocol, electrically incompatible with PCIe. Its purpose is to encapsulate data traffic from multiple other protocols (PCIe, DisplayPort, USB) and multiplex that over the same wires. It cannot function exactly like an external PCIe port because it's solving a bigger, more complicated problem.
My main thought is would this allow me to speed up prompt process for large MoE models? That is the real bottleneck for m3ultra. The tokens per second is pretty good.
tinygrad does have pretty neat support for sharding things across various devices relatively easy, that'd help. I'm guessing you'd hit the bandwidth ceiling transferring stuff back and forth though instead.
Doesn't Apple support the major standard device categories: NVMe, XHCI, AHCI, and such, like most operating systems do? The challenges are all for hardware that needs a vendor-specific driver instead of conforming to a standard driver interface (which doesn't always exist). Lots of those can be supported with userspace drivers, which can be supplied by third parties instead of needing to be written by Apple.
Not for the past decade; it's been no connectors for most products, but standard PCIe connectors for the Mac Pro, and NVMe over Thunderbolt works fine.
>> XHCI
> Not on Lightning.
Again, not relevant to any recent products. And I'm pretty sure you're misunderstanding what XHCI is if you think anything with a Lightning connector is relevant here (XHCI is not USB 3.0). You can connect a Thunderbolt dock that includes an XHCI USB host controller and it works out of the box with no further driver or software support. I assume you can do the same with a USB controller card in a Mac Pro.
>> AHCI
> How exactly would Apple not support AHCI?
This might be another case of you not understanding what you're talking about and are lost in an entirely different layer of the protocol stack. Not supporting AHCI would be easy, since they're no longer selling any products that use SATA, and PCIe SSDs that use AHCI instead of NVMe died out a decade ago. But as far as I know, a SATA controller card at the far end of a Thunderbolt link or in a Mac Pro PCIe slot should still work, if the SATA controller uses AHCI instead of something proprietary as is typical for SAS controllers.
> Why does Apple need to make the drivers in a walled garden?
Isn't that the whole point of the walled garden, that they approve things? How could they aim and realize a walled garden without making things like that have to pass through them?
I think the OP is asking why Apple is enclosing macs in a walled garden when that concept is generally associated with iPhones, not general-purpose computers.
> Why does Apple need to make the drivers in a walled garden?
Because third party drivers usually are utter dogshit. That's how Apple managed to get double the battery life time even in the Intel era over comparable Windows based offerings.
Macs and PCs are fundamentally different. Their architectures have always been distinct though the Intel Mac era has somewhat blurred the line.
Modern Mac is Macintosh descendants and by contrast PC is IBM PC descendants (their real name is technically PC-clone but because IBM PC don’t exist anymore the clone part have been scrapped).
And with Apple silicon Mac the two is again very different, for example Mac don’t use NVMe, they use just nand (their controller part is integrated in the SoC) and they don’t use UEFI or BIOS, but a combination of Boot ROM, LLB and iBoot
Not using NVMe is not a difference. It's not a different architecture. It's just simple circuit/space optimization, and has been done in other platforms as well. The controller, instead of being on the ssd module; it's present onboard the device.
Probably because you can't actually read anything more than the initial post without getting a login-wall: "Join X now to read replies on this post." (Not to mention "X" is a trash site now)
The opportunity cost of Apple refusing to sign Nvidia's OEM AArch64 drivers is probably reaching the trillion-dollar mark, now that Nvidia and ARM have their own server hardware.
Apple got out of the server game long before they adopted aarch64, so that's a trillion worth of server hardware they never would have sold anyway. And probably not actually a trillion.
Almost everyone including myself had MacBook Pros at my last place of work.
If Apple was in the high-end server market, I see no reason why the company I was working for would not be running macOS on Apple hardware as servers, instead of the fleet of Linux based servers they had.
Why wait? You can go run macOS as a server right now. It will take you a few hours to get Docker working, and disable mdworker_shared() and turn off SIP, and then install a package manager/XCode utilities, and finally configure macOS to run as a headless UNIX box, but it's attainable.
Despite how easy Apple makes it, nobody is really using Macs as a server in production. Apple[0] is not using them as a server in production. They would need a radically different strategy to replace Linux, because their efforts on macOS still haven't replaced Windows.
You want to use an NVidia GPU for LLM ? just buy a basic PC on second hand (the GPU is the primary cost anyway), you want to use Mac for good amount of VRAM ? Buy a Mac.
With this proposed solution you have an half-backed system, the GPU is limited by the Thunderbolt port and you don’t have access to all of NVidia tool and library, and on other hand you have a system who doesn’t have the integration of native solution like MLX and a risk of breakage in future macOS update.