SoftBank considering selling ARM

[MK14]

No problem at all, because 125MB/s (gigabit ethernet speed) is less than the max speed of SATA(3), so the disk(s), even if connected through USB3, won't be the bottleneck.

Software RAID is fully functional in Linux and works very very well, it's called mdadm (apt install mdadm). You can make any raid you want with that, look: https://linux.die.net/man/8/mdadm

That's useful to know, thanks.

I prefer software raid (if using raid), over hardware raid. Because if the hardware breaks, with softraid, you are not forced to buy/find, the exact original hardware. Which can be difficult.

[nctnico]

[ ... ]
As far as I know, an ancient 8088 program, would still run on the latest X86 cpus.
[ ... ]

You seem to think in terms of .exe(s), like a windows user... but Linux doesn't work like that. You apt install things, and if it isn't in the apt repository, you download the sources and ./configure and make it. That's how it works in Linux. And it works very very well, 99.999% of the times.

Edit:
A .exe program has no value in Linux... because the source code is available. You don't need to keep it saved like if it were a treasure, a .exe, because you can always recompile the thing for the latest and greatest CPU you just bought.

No. Not at all. Linux programs depend heavily on library versions and compiler versions. Compiling a Linux program 5 years later can be a major challenge. The best way by far is to pack the binary together with all the libraries it uses and store that. You'll have a much bigger chance of being able to run the program later on.

[SiliconWizard]

NVidia is another story. They don't have the same potential issues with competitors, and they are a (albeit fabless) semiconductor company. They sell ICs. They have a few products based on them, but mostly for show-off purposes IMO. Designing and selling ICs is their core business. Apple is NOT a semiconductor company, they probably wouldn't know how to run one and I don't think they ever want to be one.

Apple IS a fabless semiconductor company since quite a while ago. Their CPUs use ARM instruction set but have custom cores since A6 SoC.

No they aren't.
They have been designing their own CPUs (at least for their mobile products) based on ARM cores for a while, but AFAIK, that's strictly for their own use. I don't think they sell a single Apple processor to any third-party. Or do they?
So that doesn't make them a semiconductor company. They don't sell ICs. They sell products. So, yeah, they have the internal resources and know-how to design CPUs around ARM cores, but that's not their core business and they still wouldn't want nor probably be able to become a semiconductor company, marketing and selling processors to third parties, handling the tech support for that, even architecturing processors for other uses than their own, and so on. That's a whole other business.

You have some strange logic. How by not selling ICs to 3rd party you stop being a semiconductor company? Their ICs go into every single phone/tablet/laptop they sell, and that's a huge amount. Way more than many "true semiconductor companies" produce. Not a core business? If production is halted they will have nothing to sell at all. Also those are not "ARM" cores in a meaning that they are cores designed by ARM. They design their own cores. If they wanted to start selling SoC to 3rd parties, they could start doing so in a blink of an eye.

That's not strange logic. Making your own chips doesn't make you a "semiconductor company" when it's absolutely not your core business. Just like when they design specific cables at Apple, and have them manufactured elsewhere, they are not a cable company. They just design specific components for their products that they can't find off-the-shelf. They design a myriad of parts actually. They have always designed they own cases. Does that make them a case/enclosure company? Nope.

Additionally, even for CPUs, they still do use off-the-shelf CPUs for some products, they have in the past, and IMHO they are likely to in the future as well. Sure they design their own ARM-based CPUs these days for some of their products, but if tomorrow better off-the-shelf CPUs come up, at a better price, I'm pretty sure they could abandon their own CPU line and move on.

And no, they wouldn't get into the business of selling their CPUs that easily. It's a completely different thing. Once you do that, you need a different mindset. You need to know how to promote and sell semiconductors. You need to design them with compromises in mind - something I'm sure Apple would never want to do - you need to take customers' requests into account, you need to provide support, a lot of things that would completely derail Apple from what they do and are good at.

And with that in mind, you seem to have focused on that term instead of seeing the difference between Apple and NVidia. Apple's business is again not to sell semiconductors, it is for NVidia. It's a completely different business, and it's easy to see why Apple wouldn't want to do that (and probably why it would be likely to be a very bad move for Apple IMHO.)

[GeorgeOfTheJungle]

No. Not at all. Linux programs depend heavily on library versions and compiler versions. Compiling a Linux program 5 years later can be a major challenge. The best way by far is to pack the binary together with all the libraries it uses and store that. You'll have a much bigger chance of being able to run the program later on.

You can always touch the source code, use the (new) library headers, look at the (compiler) errors and do what it takes to match the two things to compile, can't you? Unless the new libraries' API is now completely different (very rare!), that's easy and the right way to do it, IMHO.

[Bassman59]

SoftBank fund has made clown mistakes costing billions, i.e. WeWork, Wirecard etc. so they likely need to raise some money.

This is the whole of the truth.

The rest, like the following three pages, is, as the rabbi said, commentary.

[novicefedora]

NVIDIA uses ARM in Nintendo Switch(they are selling their Tegra which they couldn't sell as a tablet), they also use it in their Jetson devices. What would NVIDIA buying ARM mean for consumers? Good or bad?

I'm surprised Qualcomm and Broadcomm are not interested in buying ARM.

[MK14]

NVIDIA uses ARM in Nintendo Switch(they are selling their Tegra which they couldn't sell as a tablet), they also use it in their Jetson devices. What would NVIDIA buying ARM mean for consumers? Good or bad?

I'm surprised Qualcomm and Broadcomm are not interested in buying ARM.

I don't think it would suit, "Qualcomm and Broadcomm"'s style.

Arm seems to need to go into lots of different companies (round the world), and support them. Such, that they buy arm licence(s), along with all the technical support they need (hopefully), to get the arm cores successfully implemented, into the various silicon dies, that the customer ends up using/choosing.
I think, customer support, is a big part of the business model.

Whereas "Qualcomm and Broadcomm", seem to have a business model (I have been led to understand), which has very few customers, but they are very big ones, who buy huge number's of their products. So, "Qualcomm and Broadcomm"'s, only has to support a relatively small number of very big customers.

Which is why (I have also been led to believe), they ("Qualcomm and Broadcomm"), don't sell (or allow to be sold), their chips (as in bare ones, e.g. as used on Raspberry PI's) to one-off consumers, such as hobbyists, or small, one man band businesses.
Unlike, e.g. Microchip with their PIC (and other) chips, who do (probably actively) support such things.

[GeorgeOfTheJungle]

I'm surprised Qualcomm and Broadcomm are not interested in buying ARM.

A curiosity: Sophie Wilson is currently a director at the technology conglomerate Broadcom Inc.

[SiliconWizard]

NVIDIA uses ARM in Nintendo Switch(they are selling their Tegra which they couldn't sell as a tablet), they also use it in their Jetson devices. What would NVIDIA buying ARM mean for consumers? Good or bad?

This is oddly put. *Nintendo* uses NVidia Tegra SOCs. NVidia doesn't "use" anything in Nintendo products, they "sell" SoCs to Nintendo.

Jetson boards are NVidia products - which IMO are in the category I mentioned earlier, show-off products rather than products meant to sell on a very large scale, make lots of cash and really be used widely in any end-product. I'm pretty sure they are just meant to give more exposure to their Tegra SOCs which they'd probably like to sell more than they actually do.

ASUS has also used Tegra SOCs though, both in phones and tablets. I don't know if they still do in their latest models?

As to what the merge would mean in the end, nobody knows. Whereas that could be opportunity to release some interesting processors/SOCs in the short to mid-term, I'm not sure it would be that good in the long term. ARM model, which has been to only sell IPs, not actual chips, has had the big benefit (IMO) of not having to deal with end-products. That means (IMO) they could better focus on architectures. If ARM merges with an actual semiconductor company, that may mean some internal "conflicts" of interest, and decreased innovation in the long run. So whereas NVidia looks like a better candidate than Qualcomm, and of course even a lot more so than Apple, I'm not too sure this would be good news, and least in the long term. But if the merge ever happens, I'm not sure it will last forever either anyway.

Your mention of Tegra SOCs is interesting as Tegra may be an indication that NVidia may not manage to do much wonders owning ARM - after all, Tegra doesn't seem to have been all that successful so far, for instance. And if, over time, NVidia realizes they can't really manage to sell enough ARM-based SOCs even when owning ARM (which I can't predict, but is not unlikely), they would just end up having to "maintain" ARM just as it is now, selling IPs to third parties, and at some point may then decide to sell it.

Point is, I'm wondering whether companies like ARM are not better off being completely independent.

[GeorgeOfTheJungle]

What I thought when the japanese (softbank) bought ARM: that's the end of it.

[nctnico]

NVIDIA uses ARM in Nintendo Switch(they are selling their Tegra which they couldn't sell as a tablet), they also use it in their Jetson devices. What would NVIDIA buying ARM mean for consumers? Good or bad?

This is oddly put. *Nintendo* uses NVidia Tegra SOCs. NVidia doesn't "use" anything in Nintendo products, they "sell" SoCs to Nintendo.

Jetson boards are NVidia products - which IMO are in the category I mentioned earlier, show-off products rather than products meant to sell on a very large scale, make lots of cash and really be used widely in any end-product. I'm pretty sure they are just meant to give more exposure to their Tegra SOCs which they'd probably like to sell more than they actually do.

Well, the Jetson modules seem to be fairly popular. NVidia has a reasonably sized eco-system around it (with a lot of active users), there is a broad range of models with different price/performance levels and there is long term availability. I designed a Jetson module into a board of a customer recently. The modules seem easier to get in small quantities compared to Tegra chips. Still the total solution isn't cheap but for lower volume it is viable.

[nctnico]

No. Not at all. Linux programs depend heavily on library versions and compiler versions. Compiling a Linux program 5 years later can be a major challenge. The best way by far is to pack the binary together with all the libraries it uses and store that. You'll have a much bigger chance of being able to run the program later on.

You can always touch the source code, use the (new) library headers, look at the (compiler) errors and do what it takes to match the two things to compile, can't you? Unless the new libraries' API is now completely different (very rare!), that's easy and the right way to do it, IMHO.

That is a major PITA to do and often impossible to do in a reasonable time frame. The compiler can be incompatible too. Recently I had to downgrade GCC on a Linux machine (VM) to compile a huge project.

[MK14]

I'm surprised Qualcomm and Broadcomm are not interested in buying ARM.

A curiosity: Sophie Wilson is currently a director at the technology conglomerate Broadcom Inc.

As regards, Sophie Wilson, it gets complicated.
They did (arguably with others) invent/start/create the arm processor chip. In her case, especially the original instruction set.
But, they were working for the 'original' arm company, that had produced the BBC microcomputer in the UK.
As later arm cores began to be created, they (arm) invented/created the 'Thumb' instruction set(s). But, ironically, Sophie Wilson, (reportedly), actually disagreed with the 'Thumb' instruction set concept, when it was first proposed.

So, historically, they carried on and decided to go with the 'Thumb' instruction set incarnations. Also, at some point, Sophie Wilson, left the company.
I think that was BEFORE, the arm company existed, as we know it.
So, the arm core licencing company, didn't actual have Sophie Wilson, working for it, ever, as such.
If I remember my arm history, correctly. (I could look it up, but hope I don't need to).
tl;dr
Sophie Wilson, never actually worked for 'arm', as we think of the modern 'arm' companies. They worked for the original one (Acorn computers, or whatever it was called, a long time ago, originally).
Best to seek out and read the history, if you want to know more.
They did work on the original, ancient, first arm chips (and maybe a bit later than that).
But not the later, licenced arm cores, we think of these days, when someone mentions 'arm'.
I.e. They were different companies, one followed on from the other, but not all employees, stayed with the new company.
So, Sophie Wilson, was one of the people that left, and didn't actually join the new company (a quick lookup, they left in 1990, when the new company was split off (Arm Ltd), but did do some consultancy work, with them.  Source:
 https://en.wikipedia.org/wiki/Sophie_Wilson  ). A very long time ago, before most people had even heard of 'arm' processor chips, unless they were involved with the early UK computer scene, especially related to educational computers (Archimedes).
https://en.wikipedia.org/wiki/Acorn_Archimedes

[novicefedora]

I don't think it would suit, "Qualcomm and Broadcomm"'s style.

Arm seems to need to go into lots of different companies (round the world), and support them. Such, that they buy arm licence(s), along with all the technical support they need (hopefully), to get the arm cores successfully implemented, into the various silicon dies, that the customer ends up using/choosing.
I think, customer support, is a big part of the business model.

Whereas "Qualcomm and Broadcomm", seem to have a business model (I have been led to understand), which has very few customers, but they are very big ones, who buy huge number's of their products. So, "Qualcomm and Broadcomm"'s, only has to support a relatively small number of very big customers.

Which is why (I have also been led to believe), they ("Qualcomm and Broadcomm"), don't sell (or allow to be sold), their chips (as in bare ones, e.g. as used on Raspberry PI's) to one-off consumers, such as hobbyists, or small, one man band businesses.
Unlike, e.g. Microchip with their PIC (and other) chips, who do (probably actively) support such things.

But it would align perfectly with what ARM is doing, they sell their IPs to others and let those companies deal with the customers, similar to how Qualcomm and Broadcomm to operate.

[novicefedora]

This is oddly put. *Nintendo* uses NVidia Tegra SOCs. NVidia doesn't "use" anything in Nintendo products, they "sell" SoCs to Nintendo.

Jetson boards are NVidia products - which IMO are in the category I mentioned earlier, show-off products rather than products meant to sell on a very large scale, make lots of cash and really be used widely in any end-product. I'm pretty sure they are just meant to give more exposure to their Tegra SOCs which they'd probably like to sell more than they actually do.

ASUS has also used Tegra SOCs though, both in phones and tablets. I don't know if they still do in their latest models?

As to what the merge would mean in the end, nobody knows. Whereas that could be opportunity to release some interesting processors/SOCs in the short to mid-term, I'm not sure it would be that good in the long term. ARM model, which has been to only sell IPs, not actual chips, has had the big benefit (IMO) of not having to deal with end-products. That means (IMO) they could better focus on architectures. If ARM merges with an actual semiconductor company, that may mean some internal "conflicts" of interest, and decreased innovation in the long run. So whereas NVidia looks like a better candidate than Qualcomm, and of course even a lot more so than Apple, I'm not too sure this would be good news, and least in the long term. But if the merge ever happens, I'm not sure it will last forever either anyway.

Your mention of Tegra SOCs is interesting as Tegra may be an indication that NVidia may not manage to do much wonders owning ARM - after all, Tegra doesn't seem to have been all that successful so far, for instance. And if, over time, NVidia realizes they can't really manage to sell enough ARM-based SOCs even when owning ARM (which I can't predict, but is not unlikely), they would just end up having to "maintain" ARM just as it is now, selling IPs to third parties, and at some point may then decide to sell it.

Point is, I'm wondering whether companies like ARM are not better off being completely independent.

The reason Tegra didn't do well is because of the price and the segment they were targeting, they positioned it as a premium product with a premium price. No company was interested in creating a tablet around it because it's cost would have been similar to an iPad. Tegra wasn't really about showing off their ARM design, it was meant to show NVIDIA's graphics capabilities, back then there weren't any tablets which had the same graphical capabilities as Tegra.

[MK14]

But it would align perfectly with what ARM is doing, they sell their IPs to others and let those companies deal with the customers, similar to how Qualcomm and Broadcomm to operate.

The companies that arm licence their cores to. Do, handle/service their own customers.
But, what I meant was those companies (the licensee's), need/get paid (I suspect), support from arm. To put those cores, into the silicon devices, that those companies are using.
Such help, can get extremely complicated, because it is dealing with complicated integrated circuits, which can cause all sorts of unforeseen problems and issues.
I watched some kind of TV (or youtube) programme about it, a long time back. It explained all about it. Possibly on UK TV, I can't remember precisely.

[novicefedora]

The companies that arm licence their cores to. Do, handle/service their own customers.
But, what I meant was those companies (the licensee's), need/get paid (I suspect), support from arm. To put those cores, into the silicon devices, that those companies are using.
Such help, can get extremely complicated, because it is dealing with complicated integrated circuits, which can cause all sorts of unforeseen problems and issues.
I watched some kind of TV (or youtube) programme about it, a long time back. It explained all about it. Possibly on UK TV, I can't remember precisely.

If you find the link to the programme, let me know.

What I thought was ARM create it's processor architecture and ISA and sells licenses to use it, companies buy it and fabricate the processor and put in their devices or they may buy prebuilt processors. And software is built around by those same companies or collaboration of companies and used in smartphones, etc. I don't think ARM has to deal with providing support to make ARM processor work with other integrated circuit, etc. Maybe in the past they had to do it but now they are literally the world leaders in smartphones, and they don't have to provide such support. What other alternative do smartphone and SBC makers have to ARM.

RISC-V might have to provide such support if they want their processor to be used. Because I don't think it will work well with modem, etc.

[MK14]

If you find the link to the programme, let me know.

I tried, and couldn't find it. EDIT: But the 4th (last video I added, is rather similar (or even the same, probably not) video). But it has similar content, to what I was looking for, and is around 9 minutes or so, long.

As a consolation prize (I'd suggest, quickly skimming through bits of it, if it seems good, watch that video), I did find these 2 videos (+ 3rd bonus one, edited in). The first seems interesting, I've only glanced at the second one, which is a much longer one. So, I'm not sure about it.

Seems reasonably worth watching, talk by CEO of Arm, 2013:



Not sure, brief glances seems promising, but rather long:



Unwatched, bonus video. To me, it seems rather interesting (from tiny glances):



Closest find so far, here:

[donotdespisethesnake]

There is a good history of ARM here



The real genius at ARM seems to have been Dave Jaggar. Sophie Wilson helped create the original inefficient design, which Dave fixed in an ingenious way, maintaining some back compatibility at the same time providing a future path. Sophie Wilson then tried to sabotage his work (some time after she had left the company), which I find quite unethical behaviour (and with hindsight technically inept). We only get Dave's side of the story, but Sophie Wilson went seriously down in my estimation after watching that video.

[unitedatoms]

The CEO Of Chinese branch went out of control

https://www.techpowerup.com/270474/arm-china-goes-rogue-ex-ceo-blocking-the-business

Will it reduce the value? Its hard to think in whose interest is it happening. The sell of ARM to japanese was a mistake.

[Marco]

I think NVIDIA want their GPU as the standard GPU and they want to tie directly into the processor (cache wise etc). Which they would never be able to license out as a SoC to third parties at the moment (because the customer would need a flexible access license, a step up from even an architecture license). They need to be able to define what is an ARM architecture to do it, so they need to own ARM.

[kaz911]

SoftBank fund has made clown mistakes costing billions, i.e. WeWork, Wirecard etc. so they likely need to raise some money.
2016 they paid $32B for ARM, a 43% premium, seems overpriced for what, a design firm?
ARM's earnings fell 21% in 2019. It's not the cash cow people think.

hmmm..... I can't talk much about this subject due to where my better half works ..

But SB has not lost any cash on Wirecard. SB does not need to raise money but is busy buying back stock (read their IR documents - it is all in there) - WeWork - hmm yes not the "at the time" smartest decision - but I actually think they will have a giga revival post covid.. SB still own 25% of Alibaba with $705b current market value.

And the CEO - Masa-san has a financial degree but studied computer science as well while at Berkeley.

He is actually an amazing person - almost lost everything in the Dotcom crash and worked his way back to the top. For the old timers - he used to own Comdex - the at the time largest IT exhibition running in Las Vegas.

So the last thing one should do is underestimate the drive of Masa-san.

[floobydust]

Careful with the kool-aid SoftBank Vision Fund Posts $17.7 Billion Loss on WeWork, Uber  "The losses are the worst ever in the company’s 39-year history." "SoftBank under-reported FY2019 income by $380 million"  OOPS.

I'm not a fan of mega-funds because the money they toss around is simply too big, too fast. Great for scandals, scams- but ultimately they wreck companies because either a stupid acquisition gets carelessly unloaded (ARM?) or they throw billions into a stupid business model (WeWork) that is all hype and fiction. The greed needs to slow down and look at the business case, the fundamentals instead of how to spew cash to instantly make cash.

I think ARM is massively overvalued- because the core is in trillions of devices does not mean it's a cash cow.

ARM Co-Founder: Sale to Nvidia Would Be a Disaster  "It’s one of the fundamental assumptions of the ARM business model that it can sell to everybody” BBC Dr. Hermann Hauser

[coppice]

ARM Co-Founder: Sale to Nvidia Would Be a Disaster  "It’s one of the fundamental assumptions of the ARM business model that it can sell to everybody” BBC Dr. Hermann Hauser

Every silicon vendor realises this, which means nVidia would have to have lost touch with reality to buy ARM. I doubt that even a consortium of silicon vendors buying ARM, and setting it up like SRC, could work.

[floobydust]

It's too late, ARM china has gone rogue and the fired CEO has refused to step down. I'm sure the IP is safe. Thanks Softbank /s.