Hacker News Reader

> So what will happen? The same thing that happened with all of Apple's bad products- they'll continue to offer it for sale at incredibly high margins for eternity and never attempt to make it more competitive.

There's another thing Apple do with bad products, though. The original 13" Aluminum Macbook launched in November 2008. It was discontinued in June 2009, in favour of the 13" Macbook Pro, which offered a permanently built-in battery (the previous model had a removable battery), and dropped Firewire. That stayed on sale, subject to regular spec bumps, until October 2012, when they bumped the screen to a retina display ...

I'm not an insider, just a long-term Apple watcher (been using Macs since 1991): what seems to happen is they used to get an internal goal to fill a niche in the product range by a certain date, and if the performance target isn't quite achievable they ship whatever they've got -- then iterate on it, possibly with a model branding change.

The Mac Pro looks like a classic case of this: Tim Cook committed Apple to being 100% Apple Silicon across the range by the end of 2022, they missed, but they had to ship something to keep the Mac Pro brand alive. So they shipped a turkey (a Mac Studio with PCI slots and a higher price tag).

What happens next is: either they ship the real Apple Silicon Mac Pro in a year or two (with who-knows-what changes, but something to differentiate it from the top of the Mac Studio range other than price), or they admit defeat and kill the Pro range altogether, or maybe downgrade it to be the name of the top end Mac Studio.

But this model won't be available for more than two years, max.

What was bad about the 13" Aluminum Macbook? It was weirdly named, but I don't know what was "Mac Pro" levels of bad about it?

My 2009 Macbook Pro 13 (5,5) had firewire. Could you be referring to a different model?

Just checked on MacTracker -- yes, the 5,5 Pro had firewire; it's immediate predecessor was the Macbook (no pro) 5,1, which didn't have FW but had the removable battery -- the 5,5 added Firewire.

Oh yeah, I'd forgotten about that aluminum Macbook (sans pro). I was working in a multimedia lab at the time and we were pretty dependent on firewire for student projects and putting the Macs into targeted disk mode.

> What happens next is: either they ship the real Apple Silicon Mac Pro in a year or two (with who-knows-what changes, but something to differentiate it from the top of the Mac Studio range other than price), or they admit defeat and kill the Pro range altogether, or maybe downgrade it to be the name of the top end Mac Studio.

There were a lot of rumors around a quad-chip (a ultra is just 2 M* Max's) that was 2 ultras fused together but the yields were terrible and they scrapped it. I've heard speculation that they might be trying again with M2/M3's but yeah, as it stands now the Mac Pro is very lack luster.

Honestly even if they manage a quad-chip design it's still missing things professionals want (more ram, more GPU power). I'm not at all in the market for this device but I really hope Apple pulls it off. Apple is better when it has devices spanning the whole spectrum even if the edges (low-end phone, high-end desktop) don't have blockbuster sales. You sweet spot of sales is empowered by those 2 extremes and benefits from it.

I think they can solve for more RAM pretty easily with the M3. Apple silicon is pretty dedicated to having the GPU cores integrated with shared memory though so I don’t see them offering graphics card options.

An M3 extreme which is 2 or even 4 M3 ultras sounds like it might be doable. M3 is already going to be a huge performence step because it’s a new process though.

I honestly don’t see what demographic the Mac Pro serves that the Mac Studio doesn’t.

I think it's really more simple than that - they have made architectural choices for Apple Silicon that simply don't fit the Mac Pro paradigm. Unified Memory Architecture primarily.

The Mac Pro lives on as a Mac Studio for people who need PCIe cards for their audio/video hardware to connect to. That's it.

I think that is enough. Sometimes I want everything in one box. To be able to spec a computer with the expansion peripherals on the inside is very convenient.

Especially for media studios with lots of data to digest. They eat all of these PCI slots up probably.

They probably gonna kit whole studios out with these things.

The author of this Verge article had the same assumption, but couldn't find anyone in the VFX or video production business who uses Mac Pros like this.

The reasons given by the studio people interviewed make sense. They get a lot more graphics bang for the buck with Nvidia-equipped PCs, and they can upgrade piecemeal.

Additionally for video pros, they were burned by Apple's consumer-oriented FCP X product overhaul over a decade ago and the mistrust still lingers:

'And while Apple’s decision to overhaul Final Cut Pro was over a decade ago, studios still haven’t forgotten. Almost all of the video professionals I spoke to brought it up. “Every filmmaker in the world was using this,” Lebensfeld complained. “And they lost that whole market. They just don’t take it seriously.”

'“They really did screw us over on that,” Ford, the documentarian, agreed. “I was really upset.”

'I searched high and low and ended up connecting with over 20 professionals for this story in order to locate someone who enthusiastically wanted to buy the Mac Pro. I found exactly one: Drofa. He loves the cheese-grater design.

'“The killer feature is when somebody comes and says, ‘Okay, you have a Mac Pro,’ and I say, ‘Yeah, I can make a cheesesteak,’” he explained.'

They didn't talk to me.

I may have a poor understanding of the mechanisms involved but it doesn’t seem immediately obvious why unified memory should hold back the implementation of high bandwidth PCI-E. Except of course if they just haven’t consciously designed their processors with enough PCI-E capability

The only issue I see is their lack of support for video cards integrating their output into the thunderbolt ports. Which is a problem they pretty much solved in the last Mac Pro with their MPX extension to PCI-E, which conveniently provided additional power and allowed the injection of displayport signal into the thunderbolt outputs.

All of the problems with the Mac Pro are theoretically solvable. They could even add expandable memory if they really wanted to. But since the Mac Studio is so ridiculously fast, why bother? The number of people who are going to be held back by the Mac Studio is a minuscule market not worth serving, they would never make a profit on it, and ultimately Apple is a business not a charity.

The Mac Pro is Apple saying: “if you need more than 192GB of RAM, we can’t help. But if you need PCI-E, we have a lot of old Mac Pro enclosures left over from the past 3 years when we didn’t sell any Intel Mac Pros, here’s a Mac Studio in a bigger box, enjoy”

I completely agree - a computer like any system is a series of design choices, compromises and decisions. Apple's vision of the Mac appears to be of a small, svelte and powerful system with limited extensibility. These are the systems of which they sell the most.

Regarding socketable memory - I believe that their current memory setup is significantly faster than any system with socketed memory. Given that they market the memory bandwidth heavily in their specs I imagine this would lead to a perceived performance drop for what is essentially their most expensive system.

My one gripe with the new pro is that they have reintroduced a problem they themselves solved, because they chose to not support discrete GPUs. MPX looks like a really neat solution to both internal system cable management, and the routing of thunderbolt and display data through the system. It's a shame it only existed in one generation of one product

Interestingly enough from looking at pictures of the new system board - they have less power cable connectors than the 2019. Is the new system using a smaller capacity PSU than the old?

Apple and others could just as well create an accelerator card with a bunch of M2s on it.

They could also create some sort of blade server.. they print their own cups in literally everything. Phone, laptops, headset, apple cinema displays. The production cost of an M1 is about $30.

Remember that the 5-year $1.5B contact with AWS will expire next year. They already have XCode cloud, and have enabled virtualization natively for a while

Either large Mac Pro based rack-servers or accelerators for rendering, transcoding, and AI/ML.

They could market those 10x$30 chips on a single board as something that is "10x" the power of the Mac Pro, at “only” $20000 a pop.

You can be very certain that NVDA did not get unnoticed, as Apple has the same capabilities, and sees a future in AI/ML.

Apple is not going to enter the commodity server business, that is just a race to the bottom and probably the reason they got out of it in the first place.

The only reason XCode Cloud (or XCode at all) exists is because people need it to develop for ios, if developers could choose anything else they would.

At this point they are a luxury consumer product company, that is where the margins are that have made them successful for the last 10 years.

I'm not oracle so I can't say if you are wrong but wouldn't that be a significant investment to go compete in a market where margins are not that great while risking to dilute their brand?

First step would be to run their own hardware. Why? $1.5B is a substantial amount of money. They can claim to be more green/energy efficient.

As for a public cloud, maybe not yet.. the problem with it is that it can be difficult to change/unwind.

However, they might just want to sell those sweet high margin AI/ML cycles running on both the backend and the client devices

The thing is, the trash can Mac Pro kinda fits the market segment the Mac Studio is now in. If Apple had created the "Studio" range back then for it, rather than suggest it was the successor of the Pro, it would have been better received.

Apple don't tend to make the same mistake twice, this time they created the new marketing category for that product. 99% of people who would have been in the "Pro" category are happy getting a "Studio", plus a bunch upgrading from an iMac.

The Studio is the new Pro to them, it's just for marketing they needed to change the name and have a "continuation line" for the old product. As you said it's here for the long haul but mostly has a legacy range.

This is exactly me. I have a 2017 27' iMac that's really starting to feel sluggish. I've been holding out for a new iMac but I have a new Studio on the way in a couple of weeks.

The "trashcan" Mac Pro (late 2013) was my favourite Mac of all time - it actually fit my professional needs perfectly, and performed incredibly well in a small package for a reasonable price (I couldn't spec out a comparable Xeon system at the time that was cheaper). I don't think it could be considered a failure by any means. It was a sweet spot when it came to price/performance for many of us in tech/science industries, and perhaps that was the reason it wasn't replaced for a long time.

The Mac Pro is actually very competitive against similar $7,000 - $10,000 workstations; especially if you need a lot of high bandwidth memory.

However, it's not competitive with workstations that are above that because you'll quickly approach workstations that have Epyc and professional Nvidia cards.

We'll need to wait for an Mx "Extreme" SoC, if Apple ever makes one, to compete above the $7,000 price point.

I feel like I'd have a very hard time designing a $7000 system which doesn't completely blow the Mac Pro out of the water in terms of CPU and GPU performance. Do you have some examples of $7000 systems which are beaten by the $7000 Mac Pro's cut-down M2 Ultra in terms of CPU or GPU power?

EDIT: I don't get why this is getting downvoted. The M2 Ultra really doesn't have very impressive GPU power compared to even consumer offerings from nvidia and AMD. The competitive $7000 workstations would have to have something weaker than a 4080.

RTX 6000 Ada alone is $7000. Don’t see how you’re “completely blowing the Mac Pro out of the water” at that price point.

The 4090, which seems like it's blowing the M2 Ultra out of the water, is $1600 though.

Great for Cyberpunk 2077, but not great if you need more than 24 GiB.

Considering the RTX 6000 ADA is like $8k, you could add three 4090s for half the cost.

You know what, that's actually fair. If your workload needs tons of GPU memory, nothing probably beats the Mac Pro's up to 192GB.

Except, I guess, for the Mac Studio.

Is there a workload that actually needs more than 24GB of VRAM that can't be offloaded to regular RAM.

I honestly don't have any examples out of my head. LLM inference is compute bound AFAIK.

LLM inference is GPU bound and VRAM bound. Given quantization however, 2x3090 or 4090 (48GB) is enough VRAM to load 65B quantized llama derivatives (30-40GB). With exllama you can get maybe 20-30 tok/s with dual 4090s on ~1200W. With an M2 Ultra maybe 10 tok/s on ~178W.

(I don't have these devices but have been researching LLM inference performance in the interest of buying a machine for them.)

Advantage goes to M2 Ultra if you ever might need more than 48GB VRAM. I think it's unlikely nvidia is going to release a significantly higher VRAM consumer card anytime soon, since that's what their A100s are for.

If only the big ML stuff didn't all use CUDA. Guessing that world won't transition to Metal compute any time soon.

Are there any such non AI use cases? For AI anyway its all Nvidia.

You'll still be limited by memory. If Apple were to create a Jade 4c chipset, (two Ultra SoC), you'd only have 384GB of RAM.

Depends on what you do. Let's say Apple makes an M3 "Extreme" and sets max memory to 512GB of RAM.

That's 512GB of high bandwidth RAM that you'd have to piece together 7 Nvidia H100 80GB cards to match. Each H100 80GB card costs $30,000+. Seven of them would cost $210k alone.

Now of course 7 H100s would have more compute power, but if what you need is a lot of high bandwidth RAM such as a large LLM or highly detailed scene rendering, then maybe an M3 "Extreme" would have the best value out there, by far.

> So what will happen? The same thing that happened with all of Apple's bad products- they'll continue to offer it for sale at incredibly high margins for eternity and never attempt to make it more competitive.

That’s one of the good things about Apple- you know the product will stick around, rightly or wrongly. It’s not as bad now, but other big companies would try to moonshot with a new product then kill it if it didn’t get traction.

Things like the MS Surface and Google Pixel are the recent-ish exceptions to this where you could actually have confidence buying them as they are established product lines and they won’t get cancelled next month ( well you shouldn’t buy a pixel due to the chronic emergency call crash problems, but that’s a separate issue ).

> [Apple] doesn't have a chip with a tonne of transceivers on it that it can use to hook up a massive GPU

What do you mean? It has an x16 PCIe 4 slot, that's more than enough for any video card.

You can’t run video cards in the new Mac Pro. https://www.macrumors.com/2023/06/11/apple-exec-discusses-ma...

I know. I'm saying the problem isn't a lack of PCI bandwidth or anything like that as my parent comment suggested.

Is that a hardware limitation, or just the lack of willingness to provide software support? I can't really imagine the former.

Apple has said that it isn't a direction they've wanted to pursue.