Will Intel and the others catch up to match M1

[ian87w]

Shutting down overnight may work for you but engineers and scientists require multi-day processes to run uninterrupted.

The Windows forums are filled with threads from lawyers, accountants, writers, engineers-- all of them are losing work from Windows rebooting out from under them. Some researchers in those threads have lost days worth of computation because of it. It's happening during Active Hours too. One person got up to get a glass of water and came back to find all of his open files shut down and Windows installing updates.

The only solution is setting up a domain but that's not always an option. There's a lot of talk over there about leaving Windows for good and switching to Linux or macos.

Being a long time Windows user, I agree on this. I don’t know what Microsoft is doing, but the way they do updates is just ultra annoying these days. Worse, their updates length take a ridiculously long time. Windows Pro allows updates to be differed, but that will only make the downtime even longer when those updates have piled up and are forced to be installed.

Heck, I have had a Microsoft employee with his laptop suddenly rebooted and went into updating in the middle of his own presentation. We had a good laugh, but it was embarrassing.

The problem with Apple is product selection. Their cheapest laptop is already $1k plus (add 30 to 40% mark up in my country). That’s unattainable by most people. With Linux, the problem is apps as majority of people are more familiar with the standard apps on Windows, and it can cost more to retrain a slew of workers just for the basic workflow.

There’s a reason why Intel and Windows can be so mediocre for so long. It’s because they can. Most people don’t really have the choice. The only real threat is from the mobile space, as more and more people are attempting to be productive just using Android. But that brings a new set of issues.

 

[pshufd]

x87 co-processors.

Do you remember when they added MMX and overlayed the x87 register file with the MMX registers?

 

[KPOM]

Intel has likely been at the mercy of their customers on some of this stuff. I can't imagine they wanted to retain x87 compatibility this many years, considering that they told people to stop using it around 2000.

I’m old enough to remember when Intel released a “486sx” that was just the 486 with the floating point coprocessor disabled. Intel actually sold a “487” co-processor that was really just a fully enabled 486 that disabled the “main” processor if installed.

 

[KPOM]

Being a long time Windows user, I agree on this. I don’t know what Microsoft is doing, but the way they do updates is just ultra annoying these days. Worse, their updates length take a ridiculously long time. Windows Pro allows updates to be differed, but that will only make the downtime even longer when those updates have piled up and are forced to be installed.

Heck, I have had a Microsoft employee with his laptop suddenly rebooted and went into updating in the middle of his own presentation. We had a good laugh, but it was embarrassing.

The problem with Apple is product selection. Their cheapest laptop is already $1k plus (add 30 to 40% mark up in my country). That’s unattainable by most people. With Linux, the problem is apps as majority of people are more familiar with the standard apps on Windows, and it can cost more to retrain a slew of workers just for the basic workflow.

There’s a reason why Intel and Windows can be so mediocre for so long. It’s because they can. Most people don’t really have the choice. The only real threat is from the mobile space, as more and more people are attempting to be productive just using Android. But that brings a new set of issues.

But in the US most people can afford to spend $1000 on a laptop. Minimum wage in half the country is $30,000/year, which matches the per capita GDP of over half the world. For that $1K the base MacBook Air is an incredible machine.

 

[pshufd]

But in the US most people can afford to spend $1000 on a laptop. Minimum wage in half the country is $30,000/year, which matches the per capita GDP of over half the world. For that $1K the base MacBook Air is an incredible machine.

Most people can't afford to spend $1,000 on a laptop. At Best Buy, their cheapest laptop is $299 and it's the second entry on their page.

At Walmart's laptop page, their Black Friday laptop deals run from $169 to $849 (five models) on the first page.

The people in Mac Forums are generally affluent or figure out how to get them cheaper (used is a good way to go).

There are people in my state that don't have broadband at home. They have to go somewhere to get broadband or a bus with broadband may go to a neighborhood to provide it.

 

[thekev]

I forgot about those instructions. I think that modern compilers use SIMD instructions for FP operations these days.

They use either VEX encoded (AVX,AVX2) or SSE instructions, both of which have scalar versions of most things. The former uses a standard 3 operand form, whereas with the SSE types, the destination of an arithmetic instruction would be one of the inputs. the three operand form has also seen an increase in the number of operations that allow folding loads, so you have something of the form [REG REG PTR].

I actually prefer ARM's Neon conventions here. They offer fewer choices, which is good, considering that register allocation is considered NP-complete and compilers don't necessarily maintain SSA for that pass.

But in the US most people can afford to spend $1000 on a laptop. Minimum wage in half the country is $30,000/year, which matches the per capita GDP of over half the world. For that $1K the base MacBook Air is an incredible machine.

$30k/yr is before tax, as is the $1k base. If you make those adjustments and account for housing inflation in the past decade or two, it seems a bit tight to spend $1k on a laptop.

 

[ian87w]

But in the US most people can afford to spend $1000 on a laptop. Minimum wage in half the country is $30,000/year, which matches the per capita GDP of over half the world. For that $1K the base MacBook Air is an incredible machine.

I’m not saying the base MacBook Air is not a good value, but US market is not a growing market. There’s a reason why mobile is the next platform since smartphones are THE affordable “computers” in the emerging (growing) markets. Thus Apple themselves is investing in the mobile space, with the M1 being the extra fruit of that labor.

When we translate that base MacBook Air cost to an enterprise level, any management in my country will balk (again, our Apple prices have extra 30 - 40% markup), especially when SSD equipped Windows laptop can be had for much less, and with most businesses IT infrastructures are built around Windows servers. For a mature US company, they might be able to run long enough to realize the TCO benefit of Apple computers. In my country, majority of the enterprises don’t have that long term realization to gain the TCO savings benefit. Add on again the extra markup just to buy Apple stuff, it’s a no go from the start.

Right now, the base Intel MacBook Air (the i3 model) is on “sale” for USD $1300 in my country. That’s the problem with Apple here. Tariffs is one thing, but Apple’s reluctant to enter the country as an entity only allows the current official distributors to mark up Apple products to their hearts contents.

 

[KPOM]

Most people can't afford to spend $1,000 on a laptop. At Best Buy, their cheapest laptop is $299 and it's the second entry on their page.

At Walmart's laptop page, their Black Friday laptop deals run from $169 to $849 (five models) on the first page.

The people in Mac Forums are generally affluent or figure out how to get them cheaper (used is a good way to go).

There are people in my state that don't have broadband at home. They have to go somewhere to get broadband or a bus with broadband may go to a neighborhood to provide it.

I meant it in a relative sense. Minimum wage is $15/hr in most of the country now (Florida just approved it). Most people making $15/hr pay no federal income tax, but they do pay payroll taxes. $1K is a stretch but on a monthly basis is affordable. Credit is freely available here. $42/month for a MacBook Air is <3 hours of work per month.

 

[shady825]

The people in Mac Forums are generally affluent or figure out how to get them cheaper (used is a good way to go).

The monthly payments with 0% interest on the Apple Card is a game changer to me. Sure I could pay outright for a Mac but if I have the option of zero interest and small monthly payments, it almost feels like you aren't even paying a lot for it.

 

[warp9]

The problem with Apple is product selection. Their cheapest laptop is already $1k plus (add 30 to 40% mark up in my country). That’s unattainable by most people. With Linux, the problem is apps as majority of people are more familiar with the standard apps on Windows, and it can cost more to retrain a slew of workers just for the basic workflow.

I'm from the old world when tech used to have a minimum $10,000 entry cost. Having cheap computers today means we have people using them that normally couldn't afford them. That's great on the face of it, but it comes with a huge set of problems, not the least of which is a ceiling on price.

Look at Microsoft with their "free" Windows 10. They know that price keeps people from committing to Windows upgrades so they made it basically free. In exchange, it's loaded with tracking and advertising and constant popups pushing their paid products. Security Center wants me to know I'm in danger unless I buy a OneDrive subscription. That's a tactic of criminals from the XP days and now Microsoft has fully baked it in because they know it works.

Looking for the cheapest of anything is always a bad idea and doubly so with computers. Apple is right in sticking to "high" prices. They know that cheap computers provide cheap experiences and will reflect badly on the company.

All this to say, expensive computers like the mac have allowed Apple to create a computing experience that doesn't rely on exploitation of the user and has given them the capital needed to R&D their way out of x86 and forge a path that doesn't rely on catering to the lowest common denominator.

 

[ian87w]

I'm from the old world when tech used to have a minimum $10,000 entry cost. Having cheap computers today means we have people using them that normally couldn't afford them. That's great on the face of it, but it comes with a huge set of problems, not the least of which is a ceiling on price.

Look at Microsoft with their "free" Windows 10. They know that price keeps people from committing to Windows upgrades so they made it basically free. In exchange, it's loaded with tracking and advertising and constant popups pushing their paid products. Security Center wants me to know I'm in danger unless I buy a OneDrive subscription? That's a tactic of criminals from the XP days and now Microsoft has fully baked it in because they know it works.

Looking for the cheapest of anything is always a bad idea and doubly so with computers. Apple is right in sticking to "high" prices. They know that cheap computers provide cheap experiences and will reflect badly on the company.

All this to say, expensive computers like the mac have allowed Apple to create a computing experience that doesn't rely on exploitation of the user and has given them the capital needed to R&D their way out of x86 and forge a path that doesn't rely on catering to the lowest common denominator.

I’m not asking Apple to race to the bottom. All I’m asking is reasonable pricing, according to Apple’s own product positioning. Apple can do this by entering some markets officially as an entity. That will at least put a hold on those greedy distributors. And have like the education only cheaper models to be made available for consumers in emerging markets. That way Apple will have a lesser SKU for these markets without changing their lineup in developed markets. There are ways to do this without resorting to race to the bottom.

 

[pshufd]

I meant it in a relative sense. Minimum wage is $15/hr in most of the country now (Florida just approved it). Most people making $15/hr pay no federal income tax, but they do pay payroll taxes. $1K is a stretch but on a monthly basis is affordable. Credit is freely available here. $42/month for a MacBook Air is <3 hours of work per month.

That didn't sound right so I ran it through a tax calculator. I generally recommend avoiding debt if possible unless the revenue generated is greater than the cost of credit assuming equivalent risk. I do see people in libraries in cities using public access computers. I assume that they do that to save money as the network performance at those places is typically awful.

 

[MEJHarrison]

I once had a computer that would fail, and fail regularly when compiling a Fortran program with a particular line of code in it. It wasn't the fault of the OS, or the software, or the Fortran code. It turned out to be a faulty RAM chip.

I ran into something like that in college. It was my final project of the year. I was crushing the class. Got 100% on the first two big projects. For a challenge, I decided to do it in Modula-2. Eventually hit a wall. I discovered a bug in the compiler. I was able to replicate the bug and demonstrate the bug. Out of time, I was unable to complete the project and it didn't work. Up to that point, I was on track for another 100%. The professor gave me a D on the project and just said that's how things are in the real world. Things don't always fail at convenient times. He probably taught me more in that moment that the silly project did. Obviously since I can't recall the project.

 

[KPOM]

That didn't sound right so I ran it through a tax calculator. I generally recommend avoiding debt if possible unless the revenue generated is greater than the cost of credit assuming equivalent risk. I do see people in libraries in cities using public access computers. I assume that they do that to save money as the network performance at those places is typically awful.

View attachment 1675400

You proved my point. Massachusetts taxes are more than enough to pay for a base model Air even after taxes.

 

[pshufd]

The monthly payments with 0% interest on the Apple Card is a game changer to me. Sure I could pay outright for a Mac but if I have the option of zero interest and small monthly payments, it almost feels like you aren't even paying a lot for it.

My philosophy is to create long-term income streams that require little to no labor and then to take that income and create additional income streams that require little to no effort. This a popular way to grow wealth.

Buying on credit is the opposite of that. Zero-percent interest loans are made up somewhere. In general, I avoid recurring payments if I can.

 

[pshufd]

You proved my point. Massachusetts taxes are more than enough to pay for a base model Air even after taxes.

This is why the stores in NH are out of most models. People in MA want to save the $50 in sales taxes too.

 

[shady825]

My philosophy is to create long-term income streams that require little to no labor and then to take that income and create additional income streams that require little to no effort. This a popular way to grow wealth.

View attachment 1675415

Buying on credit is the opposite of that. Zero-percent interest loans are made up somewhere. In general, I avoid recurring payments if I can.

I understand what you're saying. I don't know if I explained it right or not. I have the money to go out and buy a new Mac but 12 payments of $100 with no interest is just easier for me to justify. Especially when I didn't "need" a new computer. I was just excited about the M1 and figured what the hell I'll grab one.

 

[pshufd]

I understand what you're saying. I don't know if I explained it right or not. I have the money to go out and buy a new Mac but 12 payments of $100 with no interest is just easier for me to justify. Especially when I didn't "need" a new computer. I was just excited about the M1 and figured what the hell I'll grab one.

I know the feeling. Credit makes it easier to buy things. It's not a problem if you are disciplined but buying with payments for one thing can lead to additional things to where you can't just pay it all off with cash.

https://www.cnbc.com/2018/04/18/warren-buffett-buys-breakfast-from-mcdonalds-for-under-3-point-17.html

 

[vigilant]

Now that we know the M1 chip is giving Apple a big advantage to the Intel/windows combo. When will Intel/windows start going the Arm route?

This entry level M1 is beating some of Intel/AMD’s desktop chip. When the next iteration of the M1 chip comes out for the 16mbp, i mac and mac pro, I think its safe to say it will run circles round Intel and AMD’s most expensive chips.

So when will Intel or AMD or Samsung fab make an ARM based CPU for the windows platform? 3 years? 5 years?

Samsung fab is/will be making 5nm ARM based chips for their phones now. Can samsung work together with windows to make an ARM chip for the windows PC platform?

Trying to think about this pragmatically... I don’t think Intel will do anything to try to compete on ARM whether it’s making an x86 processor with similar performance at same power levels, or by making an ARM specific chip to compete. I am not 100% certain, but I’m pretty close to it.

Intel is in a very interesting bind. Intel can’t just wake up one day and say “Sure, we’re making ARM chip nows!” And expect their entire ecosystem to follow suit. They’ve tried something like this before with Itanium and VLIW, and it all just sucked.

Intel isn’t relevant because they are the best at all times. In fact, frequently they haven’t been. They’ve been relevant because the x86 ecosystem has carried on for decades, and the cost of moving off of it is really tough to do without heavily incentivized platform holders helping them. Microsoft will probably put a greater focus on ARM as a result of this. Qualcomm will probably double down on their work with Microsoft.

Intel has prided itself on being the near exclusive provider of processors that has a very rich ecosystem. It isn’t much of an exaggeration to say every attempt to try to do something ARM like (performance at low power) or anything not x86 has literally failed. Just a frame of reference for the ARM like, while still being x86, look at the Atom. They are dirt cheap, and perform badly. Badly is the most polite word I can use for them.

Other companies will eat Intels lunch here. Microsoft has a version of Windows that runs on ARM. If the software is compiled for ARM, I’ve heard good things about the performance. Not stellar things, like we’re hearing on the M1 but good things. Microsoft has already started investing in helping getting code made for x86 to work on ARM. You should expect this to expand aggressively over the course of the next few years. Microsoft needs it more than Intel.

The problem the PC industry is going to face is going to be very telling in the next few years. Apple just released a MacBook Pro at a reasonable price that best I can tell from what I’ve seen, just spanks every Intel chip for it’s form factor, at arguably any price. The M1 is not perfect, I could tell you a few of my own personal thoughts. But it is better than anything else you can get at any price for a 13 inch laptop. Arguably maybe higher.

The larger truth of the matter is this. Intel cares less about the PC market then it did in the past. It just does. It can give Dell, HP, whoever, whatever they want and because the people buying those devices are tied to the processor to get whatever work they are doing done, Dell and HP will keep buying them, and putting them into whatever feeds their business. Intels much bigger focus is clearly Cloud. Amazon wants to buy hundreds of thousands to put into their data centers to charge by the hour? Sure! Intels got plenty of chips for them, and they love it. Amazon and Microsoft aren’t paying sticker for those. Even at the sticker massively inflated prices selling enough to fill data centers across the globe is very profitable for them.

I don’t mean any of this to be disparaging of Intel. I‘ve got a lot of x86 hardware in my house, including my Xbox Series X. It’s not going anywhere any time soon.

Intel is a smart company, but they are purely reactive. If you need an example, last year Microsoft did a preview of the Surface Neo, and I admit it’s a very interesting device. It has 4 Atom cores, and a single Core core, and it’s marketed as “hybrid”. How do you think this will perform compared to the M1? I’m sorry, but it’s dead on arrival. I don’t think theres any point of even starting to fab them.

Intel will do what they did when the Pentium 4 was released, and that was just a train wreck. They’ll listen for 2 years, and they’ll regroup and have something to market in a couple years. It wouldn’t surprise me if they do a lot of architecture backtracking like they did back then. What got Intel out of that slump was the Core architecture. Most people don’t realize it, but the Core architecture started off by going back to the Pentium 3m design that was very energy efficient, and started retooling that after it was abandoned. The history of it is truly fascinating, I’d type out what I remember but I don’t think anyone here would read it.

 

[dmccloud]

Intel's biggest problem is that they can not get out of their own way in terms of getting their lineup to the 10nm process, let alone 7nm like AMD is using and 5nm that Apple is using. When Intel had to announce earlier this year than 10nm processors would be pushed back to 2022, it really exposed the company's weaknesses with regard to fabrication and manufacturing of their processors. Meanwhile, both AMD and Apple use TSMC to fabricate their processors, and while Apple is the first to 5nm, TSMC is already working on both 3nm and 2nm processes. Since Apple and AMD just design the processors, TSMC can devote its resources to perfecting new (and smaller) processes to build these chips, which leads to power efficiencies such as what Apple is promoting with the M1 Macs.

 

[pshufd]

Trying to think about this pragmatically... I don’t think Intel will do anything to try to compete on ARM whether it’s making an x86 processor with similar performance at same power levels, or by making an ARM specific chip to compete. I am not 100% certain, but I’m pretty close to it.

Intel is in a very interesting bind. Intel can’t just wake up one day and say “Sure, we’re making ARM chip nows!” And expect their entire ecosystem to follow suit. They’ve tried something like this before with Itanium and VLIW, and it all just sucked.

Intel isn’t relevant because they are the best at all times. In fact, frequently they haven’t been. They’ve been relevant because the x86 ecosystem has carried on for decades, and the cost of moving off of it is really tough to do without heavily incentivized platform holders helping them. Microsoft will probably put a greater focus on ARM as a result of this. Qualcomm will probably double down on their work with Microsoft.

Intel has prided itself on being the near exclusive provider of processors that has a very rich ecosystem. It isn’t much of an exaggeration to say every attempt to try to do something ARM like (performance at low power) or anything not x86 has literally failed. Just a frame of reference for the ARM like, while still being x86, look at the Atom. They are dirt cheap, and perform badly. Badly is the most polite word I can use for them.

Other companies will eat Intels lunch here. Microsoft has a version of Windows that runs on ARM. If the software is compiled for ARM, I’ve heard good things about the performance. Not stellar things, like we’re hearing on the M1 but good things. Microsoft has already started investing in helping getting code made for x86 to work on ARM. You should expect this to expand aggressively over the course of the next few years. Microsoft needs it more than Intel.

The problem the PC industry is going to face is going to be very telling in the next few years. Apple just released a MacBook Pro at a reasonable price that best I can tell from what I’ve seen, just spanks every Intel chip for it’s form factor, at arguably any price. The M1 is not perfect, I could tell you a few of my own personal thoughts. But it is better than anything else you can get at any price for a 13 inch laptop. Arguably maybe higher.

The larger truth of the matter is this. Intel cares less about the PC market then it did in the past. It just does. It can give Dell, HP, whoever, whatever they want and because the people buying those devices are tied to the processor to get whatever work they are doing done, Dell and HP will keep buying them, and putting them into whatever feeds their business. Intels much bigger focus is clearly Cloud. Amazon wants to buy hundreds of thousands to put into their data centers to charge by the hour? Sure! Intels got plenty of chips for them, and they love it. Amazon and Microsoft aren’t paying sticker for those. Even at the sticker massively inflated prices selling enough to fill data centers across the globe is very profitable for them.

I don’t mean any of this to be disparaging of Intel. I‘ve got a lot of x86 hardware in my house, including my Xbox Series X. It’s not going anywhere any time soon.

Intel is a smart company, but they are purely reactive. If you need an example, last year Microsoft did a preview of the Surface Neo, and I admit it’s a very interesting device. It has 4 Atom cores, and a single Core core, and it’s marketed as “hybrid”. How do you think this will perform compared to the M1? I’m sorry, but it’s dead on arrival. I don’t think theres any point of even starting to fab them.

Intel will do what they did when the Pentium 4 was released, and that was just a train wreck. They’ll listen for 2 years, and they’ll regroup and have something to market in a couple years. It wouldn’t surprise me if they do a lot of architecture backtracking like they did back then. What got Intel out of that slump was the Core architecture. Most people don’t realize it, but the Core architecture started off by going back to the Pentium 3m design that was very energy efficient, and started retooling that after it was abandoned. The history of it is truly fascinating, I’d type out what I remember but I don’t think anyone here would read it.

One of their Israeli teams developed it. I have a Pentium 3 laptop sitting on my old washing machine right now. Maybe two of them. Merom, Conroe and (don't recall server name) were the response to AMD64. I remember AMD stock rocketing from single digits to the high 40s or 50s and then falling right back down when Core benchmarks were released. I recall a lot of us not believing the leaked benchmarks. And then they came out next years and AMD couldn't execute. Hector Ruiz (I think that's his name) was full of Hubris and things went badly for a while.

Intel has a pretty good lock on Cloud servers. You would think that it would be easy to go to ARM because it's just a service, not an architecture or even an operating system. But it isn't.

Most of the world doesn't know how good M1 is right now. They'll find out when Apple reports earnings next year. And M1 is just the start.

 

[vigilant]

One of their Israeli teams developed it. I have a Pentium 3 laptop sitting on my old washing machine right now. Maybe two of them. Merom, Conroe and (don't recall server name) were the response to AMD64. I remember AMD stock rocketing from single digits to the high 40s or 50s and then falling right back down when Core benchmarks were released. I recall a lot of us not believing the leaked benchmarks. And then they came out next years and AMD couldn't execute. Hector Ruiz (I think that's his name) was full of Hubris and things went badly for a while.

Intel has a pretty good lock on Cloud servers. You would think that it would be easy to go to ARM because it's just a service, not an architecture or even an operating system. But it isn't.

Most of the world doesn't know how good M1 is right now. They'll find out when Apple reports earnings next year. And M1 is just the start.

Reading your post, all of that sounds just about right. It’s been a long time for me so I’m going off of pure spur of the moment memory, but yes. I definitely remember Hector.

ARM is coming in the Cloud space too. It’s probably not going to hurt Intel for 5 years. I do think 5 years is probably being generous. It would shock me if total 100% net new projects for most companies don’t start being focused on ARM in 2 years. AWS is being aggressive. For non .Net projects it could be a thing soon.

I’ll tell you that I have very specific reason to believe that Amazon is investing heavily in ARM long term. I don’t know how real it is, but right now they seem to be talking the talk.

It wouldn’t surprise me if in 5 years you see projects going live that run .Net on IIS on ARM that are supported by some form of x86 backed SQL server. Put the front end servers on something cheap and resilient enough, keep the data on something that is predictable, and don’t worry if anything changes. I think thats very very real.

 

[Bearbrickster]

What chunk Apple may or may not take in the premium market is a tough call

The very people who buy in this price bracket have already been alienated by lack of dual monitor, external GPU and 90% of the worlds software, now no bootcamp

It's not like Tiger lake latest mobile chip releases are too shabby, maybe 10-20% down on single/muilti core benchmarks, comparable read/writes and better openCL on a mobile chip with just 4 cores that also have instant on and 12 hours plus endurance

For many W10 users the gains vs the sacrifices and alienation and limited options even on screens 120hz/OLED

I think many over estimate the appeal of Apple M1 for dedicated Apple users to the preferences and choice for W10 users

Samsung little Book S with arm that's been out for a year needs to step up for sure but then it is at the lower end of pricing and if MS get the 64bit support sorted will help

Interesting times for sure but a bit over hyped with some of these silly comparisons we see IMO

Windows will no doubt still be the market leader for the foreseeable future.
Dual monitor support, higher ram, mini led, even faster chip should be available in the new 14, 16 mbp, imac and mac pro. Professionals that buy expensive high spec pc or mac pro, might just buy the next imac which cost considerably less and thus able to reach more consumers. Which in turn might lead developers to making native app for the mac. And vice versa. Interesting times indeed...

 

[pshufd]

Reading your post, all of that sounds just about right. It’s been a long time for me so I’m going off of pure spur of the moment memory, but yes. I definitely remember Hector.

ARM is coming in the Cloud space too. It’s probably not going to hurt Intel for 5 years. I do think 5 years is probably being generous. It would shock me if total 100% net new projects for most companies don’t start being focused on ARM in 2 years. AWS is being aggressive. For non .Net projects it could be a thing soon.

I’ll tell you that I have very specific reason to believe that Amazon is investing heavily in ARM long term. I don’t know how real it is, but right now they seem to be talking the talk.

It wouldn’t surprise me if in 5 years you see projects going live that run .Net on IIS on ARM that are supported by some form of x86 backed SQL server. Put the front end servers on something cheap and resilient enough, keep the data on something that is predictable, and don’t worry if anything changes. I think thats very very real.

I think the Cloud guys will put in hardware and platforms if there is a demand for it. It may be that AWS wants to provide the supply to drive demand there with lower costs. I worked on Oracle Autonomous Database Cloud and it was Exadata servers (engineered x86 database servers) and I don't really see a change from that in the near future. I saw the major ports of the Oracle RDBMS and the are still Linux, Windows and Solaris. Porting and testing would be a considerable amount of work and there's already a lot of money invested in x86. Oracle bought Sun Microsystems to try to compete with x86 and lost. If you can't beat them ...

I'll be in the gallery watching.

 

[vigilant]

I think the Cloud guys will put in hardware and platforms if there is a demand for it. It may be that AWS wants to provide the supply to drive demand there with lower costs. I worked on Oracle Autonomous Database Cloud and it was Exadata servers (engineered x86 database servers) and I don't really see a change from that in the near future. I saw the major ports of the Oracle RDBMS and the are still Linux, Windows and Solaris. Porting and testing would be a considerable amount of work and there's already a lot of money invested in x86. Oracle bought Sun Microsystems to try to compete with x86 and lost. If you can't beat them ...

I'll be in the gallery watching.

Ah! I almost feel like you are family now that you’ve said that. I won’t name names, but I was fortunate enough to know a lot of Sun people during the initial build of Sun Cloud, and some went on to work at Oracle, and are still there, and others went on to other things.

These are the conversations I’ve had back when I was essentially a child. I truly love it, it does my heart good.

So I will say this, theres a lot of companies that I work with today that will never change their ways. It doesn’t matter whats put in front of them.

And as I at least tried to allude to, I can see very specific workloads never moving off of x86. Something highly transactional, that has thousands of companies with each with millions of dollars invested in one very specific solutions. I don’t think we’ll see that go away.

I did try to very intentionally say that companies with something that is 100% greenfield, net new, they maybe relying on something that isn’t x86 for things around that that could be ARM. I was talking to someone from that same circle thats over at AWS and I don’t know how true it is, but I think Netflix is running 100% on ARM.

ARM isn’t the overnight thing that everyone will be buzzing about like we’ve seen in many areas. But I do think it will be something that finds it’s way into projects where technical debt is low, or non-existent. I don’t think by any means that Intel is going to go the way of BlackBerry. I was told the other day that BlackBerry is now a security company.

Intel will react in time, and I expect them to do something interesting. But I think they need to focus on core IP to do something that shakes things up.

If I were to wager, and I’d be willing to do it, I don’t think the M1 is great because Intel sucks. I’d wager that because Apple controls where framework calls go, they are routing them to pieces of IP, be it the “Neural Engine” which I don’t fully think is 100% original, but they’ve done great work with it, or their GPU, which I have reason to believe has a lot of things from AMD implemented through licensed patents, to do things better by not focusing solely on the CPU, and treating the GPU as a distant neighbor that can do it’s own thing. I think thats why the unified memory is important, because you don’t have to wait to copy. They can control where things go, and they can build those parts to accommodate with shared memory in mind. I’m happy to be wrong, but I think thats the true magic.

And my friend, by the way, I don’t know how that works at scale in multitenet environment. I don’t think it can be done as quickly. If you have a however many ARM chip cores, and 6 or 20 are actively working for one customer, how do you minimize latency going to another shared utilization thing such as a GPU? I’m not saying it can’t happen. There could be something in AWS Nitro that could accommodate it? If it exists, it’s beyond me, but I’m eager to read it.