Discussion

bigyabai: 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.

chuckadams: 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.

wmf: Pretty misleading. This driver is only for compute not graphics.

polotics: As a sizable share of the market is going to want to use this for local LLMs, I do not think this is that misleading.

arjie: 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?

eoskx: Interesting, but cannot run CUDA or more to the point `nvidia-smi`.

manmal: 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?

arjie: 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.

frankc: 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.

bangonkeyboard: I don't know how Apple has avoided regulatory scrutiny for their refusal to sign Nvidia's eGPU drivers since 2018.

lowbloodsugar: “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.

embedding-shape: 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.

GeekyBear: We've seen many recent projects to stream models direct from SSD to a discrete GPU's limited VRAM on PCs.How big a bottleneck is Thunderbolt 5 compared to an SSD? Is the 120 Gbps mode only available when linked to a monitor?

brcmthrowaway: What are the limitations of USB4/Thunderbolt compared with a regular PCIe slot?

embedding-shape: 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.

givinguflac: I think you missed a zero, TB5 does 80GB/s.

mch17: No, it does 80 Gb/s. With encoding loss it’s closer to 8GB/s

Tepix: No. It does 80Gbps.https://www.convertunits.com/from/Gbps/to/GB/s

zozbot234: > 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.

the__alchemist: 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?

manmal: But isn’t the Mac Mini the weak link in that scenario?

zozbot234: It has way more unified memory than your typical dGPU.

GeekyBear: The same way Google evaded regulatory scrutiny for refusing to allow a YouTube client for Windows Phone?

bigyabai: Internet Explorer Mobile is a YouTube client. You're describing a client-server disagreement when the user is talking about client-client disagreement.

Keyframe: 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.

MBCook: 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?

justincormack: It carries pcie, but only at x4. Thunderbolt 4 is pcie gen 3 and Thunderbolt 5 is pcie gen 4.

vondur: If you could get Nvidia driver support on Mac’s I bet Apple would have sold more MacPro’s.

wmf: This driver doesn't support CUDA.

hgoel: They said eGPU

999900000999: You can buy a cheap GPU enclosure for about 100$ off ali express.Takes a standard PSU. However, Mac Minis don't have occulink. So you might be a bit limited by whatever USB C can do.Now if Intel can get there Arc drivers in order we'll see some real budget fun.https://www.newegg.com/intel-arc-pro-b70-32gb-graphics-card/...32 GB of VRAM for 1000$. Plus a 500$ Mac Mini.

Fnoord: Those $100 ones don't come with a cage. If you do want a cage, you'll end up with $180 in total, with zero warranty.Article mentions: "Apple finally approved our driver for both AMD and NVIDIA"Does not mention Intel (GPUs). Select AMD GPUs work on macOS, but...Macs (both Intel and ARM) support TB, but eGPU only work on Intel Macs, and basically only with AMD.Good news is for medium end gaming choices are solid, and CUDA works on AMD these days.

999900000999: Fortune favors the bold my friend.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.

mlfreeman: 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.

mulderc: Apple doesn’t have a monopoly in any market they are in.

TheDong: 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.[0]: https://en.wikipedia.org/wiki/United_States_v._Microsoft_Cor....

Fnoord: [delayed]

raw_anon_1111: 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

dd_xplore: Why does Apple need to make the drivers in a walled garden? Atleast they should support major device categories with official drivers.

embedding-shape: > 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?

deepsun: Don't purchase? I don't own any Apple devices, everything works fine.

TheDong: 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.

sunnybeetroot: That is no longer true. https://bluebubbles.app/

GeekyBear: > Why does Apple need to make the drivers in a walled garden?For the same reason that Microsoft requires Windows driver signing?Drivers run with root permissions.

Underphil: 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.

TheDong: 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.

tensor-fusion: As more people carry ARM laptops and keep the GPU somewhere else, I think the interesting UX question becomes whether the GPU can "follow" the local workflow instead of forcing the whole workflow to move to the GPU host. That's the problem we've been looking at with GPUGo / TensorFusion: local-first dev flow, remote GPU access when needed. Curious whether people here mostly want true attached-eGPU semantics, or just the lowest-friction way to access remote compute from a Mac without turning everything into a remote desktop / VM workflow.

mort96: I mean when it comes time to output the image from the GPU, I don't want to add a hundred milliseconds of network latency...

whalesalad: This is re gpu for compute not graphics.

mort96: Oh. Weird use for a graphics unit.

manmal: Graphics was not what came to mind when I saw the headline.

mort96: Graphics is typically what comes to my mind when people talk about graphics processing units

selectodude: 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.

kllrnohj: Why? Just make iMessage users put up with green bubbles if they want to talk to you?Thanks to Apple co-opting phone numbers, there's literally no need to ever have iMessage for anyone

thisislife2: 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.

mattnewton: Still undesirable latency for a lot of compute use cases, like image or video editing; it’s really only negligible for LLMs.Since that’s definitely a big enough use case all on its own, I wonder if such a product should really just double down on LLMs.

jonhohle: 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.

MrArthegor: 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.

gsnedders: https://docs.tinygrad.org/tinygpu/ are their docs, and https://github.com/tinygrad/tinygrad/tree/4d36366717aa9f1735... is the actual (user space) driver.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.

satvikpendem: 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.

hilsdev: Credentialism to prevent discussion of political and government entities is incredibly dangerous

MrArthegor: 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