Could Apple step in the enterprise chip market with Apple Silicon?

[senttoschool]

I think Apple will sell cloud versions of their chips as complete platforms. Think about how you will one day be able to sit in a coffee shop with a Macbook Air and then with one swipe, switch to a remote M3 Extreme with 384GB of RAM, 40-core CPU, and 160-core GPU to finish a task.

Only Apple could do this kind of integration to make renting a much more powerful cloud instance seemless.

I've written extensively how an 'Extreme' version of Apple Silicon makes little financial sense unless Apple plans to open the market up by offering cloud versions of Apple Silicon.

 

[falainber]

• If Apple becomes a cloud service provider, it will compete with AWS/Azure/GCP. What innovations could Apple bring to the table to compete against companies that have been developing an infrastructure, software stack and customer base for 15 years?

If Apple sells SoCs for data centers, it will compete with AWS (Graviton) or Ampere (Altra) in CPUs, with AWS (Trainium/Inferentia), GCP (TPUs), Habana Labs (Gaudi/Greco) in ML accelerators, and Nvidia in GPUs. Could Apple offer a more cost-effective and efficient solution than any of those companies?

If Apple sells macOS, it will compete with several Linux-based operating systems. What would be the advantage of using macOS over RHEL or Ubuntu? Would programs run faster or be more cost-effective on macOS? Would Apple provide better support than Red Hat or Canonical?

In recent years, Apple has launched new products and quickly captured significant market share because those products created new markets. But data centers and HPC is a mature market and Apple will find it much more difficult to capture significant market share.

P.S: I seem to have misunderstood the topic of the thread. What does Windows on ARM64 have to do with data centers and HPC?

In addition, why would any enterprise decide to use a solution from a company known for:
* changing processor architecture every decade
* dropping OS version support after just 5 years or so
* being allergic to providing any information about their road map
* being allergic to cooperation with other companies
* being allergic to industry standards
Just the cost of rewriting the software every time Apple requires you to do so would have bankrupted all but a few companies.

 

[falainber]

I think Apple will sell cloud versions of their chips as complete platforms. Think about how you will one day be able to sit in a coffee shop with a Macbook Air and then with one swipe, switch to a remote M3 Extreme with 384GB of RAM, 40-core CPU, and 160-core GPU to finish a task.

Only Apple could do this kind of integration to make renting a much more powerful cloud instance seemless.

I've written extensively how an 'Extreme' version of Apple Silicon makes little financial sense unless Apple plans to open the market up by offering cloud versions of Apple Silicon.

Then, of course, one can already do just that (use cloud computing resources from the coffee shop) using a laptop from any other vendor but Apple. Macs strong feature is integration with iPhone not cloud or enterprise applications.

 

[bobcomer]

There is no server version of macOS for any architecture, period.

That's definitely true, but all I meant by saying that, is there isn't any Windows Servers with ARM Processors. (at least sold to the public! There's no technical reason there couldn't be though, and I think there is Azure ARM servers. (VM servers)

 

[Xiao_Xi]

I think Apple will sell cloud versions of their chips as complete platforms. Think about how you will one day be able to sit in a coffee shop with a Macbook Air and then with one swipe, switch to a remote M3 Extreme with 384GB of RAM, 40-core CPU, and 160-core GPU to finish a task.

Only Apple could do this kind of integration to make renting a much more powerful cloud instance seemless.

Currently you can connect to the cloud directly through apps (for example, you can send your render to a render farm directly from the app). How could Apple improve this?

there isn't any Windows Servers with ARM Processors. (at least sold to the public! There's no technical reason there couldn't be though, and I think there is Azure ARM servers. (VM servers)

It appears that Windows-only software, such as SQL Server, is being ported to Linux. What Windows-only software would you run on Windows Server?

 

[senttoschool]

Currently you can connect to the cloud directly through apps (for example, you can send your render to a render farm directly from the app). How could Apple improve this?

So why does the Mac Pro even exist?

 

[senttoschool]

Then, of course, one can already do just that (use cloud computing resources from the coffee shop) using a laptop from any other vendor but Apple. Macs strong feature is integration with iPhone not cloud or enterprise applications.

So you're basically saying the Mac Pro is useless?

 

[Gudi]

The lack of any decent ARM chips to run Windows is the problem there.

Not even close to the problem.

1. Microsoft's exploitative and extortive business conducts are known. Nobody wants to work with them and rely on them unless they must. That's why they are forever stuck on the desktop PC paradigm, where they still enjoy the last days of the old Wintel monopoly.

2. Windows is just a knockoff of Macintosh OS. Microsoft never learned how to innovate. Zune, Bing, Office. All just clones of preexisting products from other companies. Unless you give them a proven concept to steal from, they've never learned to come up with their own idea and make it work.

3. Twelve years ago the A4 in the iPhone 4 was Apple's first in-house designed silicon in a long time. But they still had the spirit of a garage-founded company. Bill Gates on the other hand is the son of a lawyer. He first had a signed treaty with IBM to supply an OS for the IBM-PC and then went and bought 86-DOS. The leopard can't change its spots and suddenly become an apple tree.

 

[Philip Turner]

Apple certainly has the resources and the technology to make a big splash in the AI and HPC markets.

HPC? Nevermind that, just being able to run HPC applications on the M1's GPU.

 

[leman]

Could Apple offer a more cost-effective and efficient solution than any of those companies?

This is the key point, I think. I just don’t see any way how Apple can be more cost-efficient here. Their tech is very very expensive.

P.S: I seem to have misunderstood the topic of the thread. What does Windows on ARM64 have to do with data centers and HPC?

Nothing. Just few people derailing a thread as usual.

There is no server version of macOS for any architecture, period.

What would a „server version“ of macOS be anyway? Headless? Windows sells a separate product called „Windows Server“, but there is no such individuation among Unix and Unix-alikes.

 

[kitKAC]

Yes, just look at the sales numbers. They sell *far* less of them than x86 PCs. I know I wouldn't buy one (not even used and cheap) and I'm more friendly towards ARM than most.

Looking at Dell.com, they have exactly no ARM-based laptops for sale, HP has 4 out of many, many x86 laptops. If non-Apple Silicon ARM processors were more performant, manufacturers would create more devices with them and sales would increase.

No, it's still the same problem. Compatibility is important in the x86 market, and the most compatible, is, you guessed it, x86. Apple doesn't care about backwards compatibility, it's a whole different theory than what drives the x86 market.

Apple doesn't care? Was Rosetta 2 an accident? Of course they care, they just aren't slavish to it like Microsoft is. ARM Windows also has its own compatibility layer:

https://learn.microsoft.com/en-us/windows/arm/apps-on-arm-x86-emulation

For most average people, that'll be good enough once ARM machines are performant enough to be worth buying (and manufacturers start producing them).

 

[Xiao_Xi]

HPC? Nevermind that, just being able to run HPC applications on the M1's GPU.

Which Mac should the CERN Data Center buy to process and store the petabyte of data it creates every day?

[T]he CERN Data Centre used to process on average one petabyte (one million gigabytes) of data per day during LHC Run 2.

Storage

home.cern

How long would it take you to train Stable Diffusion on an M1 GPU if it took 200,000 A100 hours to Stability?

From AWS re:Invent 2022 - How Stable Diffusion was built: Tips & tricks to train large AI models

Being able to run toy examples on your computer doesn't mean it can run the demanding workloads that the cloud and HPC can run.

 

[Richu]

Not even close to the problem.

1. Microsoft's exploitative and extortive business conducts are known. Nobody wants to work with them and rely on them unless they must. That's why they are forever stuck on the desktop PC paradigm, where they still enjoy the last days of the old Wintel monopoly.

2. Windows is just a knockoff of Macintosh OS. Microsoft never learned how to innovate. Zune, Bing, Office. All just clones of preexisting products from other companies. Unless you give them a proven concept to steal from, they've never learned to come up with their own idea and make it work.

3. Twelve years ago the A4 in the iPhone 4 was Apple's first in-house designed silicon in a long time. But they still had the spirit of a garage-founded company. Bill Gates on the other hand is the son of a lawyer. He first had a signed treaty with IBM to supply an OS for the IBM-PC and then went and bought 86-DOS. The leopard can't change its spots and suddenly become an apple tree.

This is outdated from the Balmer era. The perception of Microsoft amongst developers (e.g. me who’s still an Apple fan boy) and businesses (e.g. me as a business owner) has changed significantly since the new CEO and their partnering strategy. Microsoft is great these days. VSCode is great, github is great, azure is great.

Apple is not the poster child for a reliable business partner. Plus their whole niche is consumer centric, which puts a lot of emphasis (and cost) on other priorities than businesses care about.

 

[unrigestered]

Enterprises want longevity, stability and security first and foremost and not a new gimmick OS each year, every time with new bugs, incompatibilities and glitches introduced.

also, depending on the type of enterprise (industry for example), some fancy-schmancy designer objects might not be the first choice, even when not considering the price tags

 

[Oculus Mentis]

So why does the Mac Pro even exist?

With the advent of AS, the Mac Pro product line is in a limbo. It can either go two ways in order to fulfil those market requirements of increasing high memory and high data bandwidth:
it could eventually be scrapped or expanded into the enterprise beyond Apple own product lines.

With high watering development costs, Apple won’t be able to keep the Mac Pro alive on a drip as it has done under the wide spread Intel architecture.
In the workstation market, with AS, it’s either make it, by increasing market adoption to root the ecosystem where data is created/processed or break it (as it happened to the fragmented RISC/Unix high end workstations market during the ‘90s.

In my opinion, without a serious market presence of AS in big data, the new Mac Pro will be toasted especially if it won’t be able to support workflows requiring the same amount of memory, precision and data bandwidth of its predecessors.

 

[maflynn]

So why does the Mac Pro even exist?

I think that's a very good question. Right now, what would a Mac Pro offer that is markedly different then the studio?

One of the advantages the M1 has is that nearly everything on the silicon, which in turn means nothing to upgrade. The Mac Pros are Apple's pro grade desktop that is modular and upgradeable.

This generation Mac Pro was designed for that use case, as they made the trash can Mac Pro not modular or as upgradeable and gave a mea culpa over their missteps. How will a M2 Mac Pro be modular and upgradable when everything is on the silicon? I'm not saying apple cannot re-architect the design but 2022 has come and gone and we've yet to see a Mac Pro.

 

[Xiao_Xi]

In my opinion, without a serious market presence of AS in big data, the new Mac Pro will be toasted especially if it won’t be able to support workflows requiring the same amount of memory, precision and data bandwidth of its predecessors.

Big data relies on commodity hardware and distributed fault-tolerant software. How could the Mac Pro fit into that market?

Apple seems to be particularly fond of Apache Cassandra.

United States - Jobs - Careers at Apple

Explore all United States jobs at Apple. Create a profile and apply today.

jobs.apple.com

Apple's secret NoSQL sauce includes a hefty dose of Cassandra | TechRepublic

If you want to scale like Apple, you might need to consider Cassandra. Matt Asay explains.

www.techrepublic.com

 

[bobcomer]

It appears that Windows-only software, such as SQL Server, is being ported to Linux. What Windows-only software would you run on Windows Server?

We run SQL Server and internal applications. (Our own)

 

[bobcomer]

Looking at Dell.com, they have exactly no ARM-based laptops for sale, HP has 4 out of many, many x86 laptops. If non-Apple Silicon ARM processors were more performant, manufacturers would create more devices with them and sales would increase.

I doubt it, people just don't buy them. No reason to.

 

[bobcomer]

Apple doesn't care? Was Rosetta 2 an accident?

You betcha they don't care. They will discontinue Rosetta 2 in not to long a time and a lot of software, especially custom software wont work anymore. That's just what they did with the x86 transition.

For most average people, that'll be good enough once ARM machines are performant enough to be worth buying (and manufacturers start producing them).

But why bother? There's certainly no reason I'd buy one over an x64 PC.

 

[Pezimak]

Now that Apple has proven its capability in designing and producing Silicon for their own products and with Intel struggling on the innovation front, wouldn’t it make sense for Apple to enter the enterprise data center chip market for AI and HPC?

Just like Nvidia, Apple could design CPU and GPUs to be produced by TSMC then to be deployed in cloud HPC data centres of their own or to be sold to AMZN, meta, googl etc.

Warren Buffet announced a few weeks ago that he’s investing in TSMC, since he already is a large Apple shareholder, I do wonder if he knows something is moving in that direction…

Apple certainly has the resources and the technology to make a big splash in the AI and HPC markets.

No, Apples entire ethos and business model is the complete opposite almost as to what the data centre and enterprise server markets want and need.

 

[Pezimak]

I think that's a very good question. Right now, what would a Mac Pro offer that is markedly different then the studio?

One of the advantages the M1 has is that nearly everything on the silicon, which in turn means nothing to upgrade. The Mac Pros are Apple's pro grade desktop that is modular and upgradeable.

This generation Mac Pro was designed for that use case, as they made the trash can Mac Pro not modular or as upgradeable and gave a mea culpa over their missteps. How will a M2 Mac Pro be modular and upgradable when everything is on the silicon? I'm not saying apple cannot re-architect the design but 2022 has come and gone and we've yet to see a Mac Pro.

Exactly, the latest rumours being they scrapped the top end Mac Pro chip due to its performance being too low to justify its high retail price. To me that reads as a none upgradable system, with everything soldered down or built into the chips. The complete opposite to what the current Mac Pro is.
We shall see but I won't be surprised to see any new Mac Pro literally just being a more powerful more expensive Mac Studio at this point.

Not what the data centre or enterprise markets wants.

 

[HobeSoundDarryl]

Right now, what would a Mac Pro offer that is markedly different then the studio?

Flexility in expansion.

I have Mac Studio and I can passionately argue how great it is for my needs. However, it's basically a completely, locked-down Mac mini... a fat Mini if you will.

Mac mini has filled the same basic role of self-contained, headless Mac at the low (price) end for a long time. At the high end, Mac Pro has been the other version of a self-contained, headless Mac. Mac Studio plops in somewhere in the middle though it is more like the Mini than the Pro. Why? What has Mac Pro offered vs. Mac Mini all along?

Slots. "We" will marginalize slots like crazy now ("99% don't need...") to help push Studio to those worried a new Mac Pro won't have slots, but slots open up many niche uses for a Mac that are not fulfilled by what Apple decided to jam into a single, locked down SOC. Anyone who needs those things need a slot(s) into which to put that extension hardware. No Mac with slots- no way to use what they need to do whatever they do with Mac Pros. Hello PCs with slots if Apple doesn't deliver a Mac with them... not because people want to defect to Windows/UNIX but because Apple has no Mac to give them what they need.

Big RAM. Some need RAM beyond the upper end of what Apple chooses to supply on the SOC. Mac Pro has tended to offer substantial RAM expandability and slots offer even more of such opportunity. Those who need big RAM can't magically hop on Studio Ultra and everything that demands the huge RAM just happily adapts to the MAX config. RAM is key to fastest processing and more RAM typically means faster processing. For many computer buyers, SPEED to completion of computer tasks trumps what "we" spin to argue against anyone needing more RAM than what Apple chooses as max config options now.

Evolution. Traditionally, the best graphics card at any given time is supplanted by better ones within a year or two at most. Mac Pro can "keep up" vs. "throwing baby out with the bathwater" because the graphics capabilities get too far behind the times. Similarly, some people want a computer that can evolve with their unanticipated future needs vs. trying to guess exactly what they will ever need and buy it up front. Mac Pro is the ONLY Mac that can evolve inside the case to keep up with advancing, supporting technologies.

Horsepower. Mac Pro has traditionally been the most powerful Mac... not the most power efficient Mac. Power Per Watt is not generally a consideration with this Mac. It is supposed to be the most powerful Mac. Those who pay the most for a Mac are generally wanting a Mac that can compute FASTER than any other Mac and, ideally, any other PC. "Our" push for PPW so that Apple can win a head-to-head contest vs. PC comparisons misses the point here. Computing speed translates into getting tasks done FASTER. If that takes a little more electricity to get things done faster, this kind of buyer is good with that.

My own Mac Studio Ultra is great and all but there are no slots for specialized hardware, RAM is locked at the choice I made at purchase no matter how much I might ever need, the only possible Evolution will be in software (it will NEVER have a better graphics "card," nor adapt to unanticipated future needs unless whatever that might be can be external via thunderbolt/USB (which is NOT a lock at all)). MAX horsepower is already inside; it has no path to become more (hardware) powerful at any point in the future. Objective reviews show that higher end, newer Intel hardware is already more powerful (yes, at the expense of using slightly more electricity)... but, bottom line, more computing horsepower gets more done FASTER and sometimes getting work done faster is much more important than shaving a dollar or three each month off of the electric bill.

My first Mac was "pro"-like: the PowerMac G4 "tower." During it's useful life, I did eventually encounter use situations where the card slots came in very handy. I didn't anticipate those needs up front but it was nice to get that added value of it when such needs arose. Computer owner situations change/evolve- sometimes in unanticipated ways- and it's nice to have a computer that can evolve with whatever happens. As is, ALL Silicon Macs are completely locked down... basically peaked in their capabilities for their buyers on delivery, unless some want can be covered in software updates only. Even my Mac Studio Ultra is all it will ever be on the hardware front. Should I ever need a computing something that can't be addressed in software only, I have to buy a whole new computer.

That last line is why there has long been a group want for the so-called x-Mac... a Mac between Mini and Pro with a few slots... the equivalent to a PC mini-tower if you will. Instead, Mac Studio is slotted in there now and it is locked down. Want/Need to update the graphics? Nope. Want/Need to add some additional RAM? Nope. Have some niche hardware need on an industry-standard card? No card slots for that inside or outside of Mac Studio. Etc.

I hope this helps.

 

[maflynn]

Flexility in expansion.

My point is the current architecture does not give you flexibility in expansion - everything is on the silicon. I know apple could re-architect an M series processor and logic board to allow for upgradeable components but right now they don't have such an architecture to .allow expansion of components.

 

[maflynn]

Exactly, the latest rumours being they scrapped the top end Mac Pro chip due to its performance being too low to justify its high retail price.

I agree, and from what rumors I've seen, I'm guessing M2 Extreme yields were so low that it didn't make financial sense. If history is an indicator, the M2 Extreme was going to be even larger in size then the M1 ultra - if that is the case, then each wafer produces less M2s driving the unit costs up