News [Bloomberg] Chipmaker SiFive Is Said to Draw Intel Takeover Interest

DisEnchantment

Golden Member
Mar 3, 2017
1,590
5,722
136

SiFive Inc., a startup that designs semiconductors, has received takeover interest from investor Intel Corp., according to people familiar with the matter.
Intel offered to acquire SiFive for more than $2 billion, one of the people said, asking not to be identified because the matter is private.

Big if true. It could mean Intel could find its way into all sorts of low power MCUs and myriad of other things which is unnoticed by most people but form the back bone of most smart devices we have today.
I think it is a not bad move, let SiFive run independently and fab their stuff on cheap nodes.
Pocket change for Intel. Hopefully not XScale 2.0.
And with RISC-V being open could be a benefit to the entire industry designing low complexity custom devices.

One thing immediately comes to mind is if Intel is going to bring to the Auto Industry.
 
Last edited:
  • Like
Reactions: Tlh97 and moinmoin

DisEnchantment

Golden Member
Mar 3, 2017
1,590
5,722
136
Really? SiFive? That'll drive the RISC-V crowd nuts.
I think it is a good thing. If SiFive/Intel were to contribute back to the RISC-V consortium it is a big win for all RISC-V adopters. Lots of organizations around the world developing with RISC-V.
RISC-V is a quick way to royalty free, open and customizable design for low complexity MCUs for a myriad of use cases.
I hope this deal materialize, and Intel not being NV, I dont see this being so bad.
At most it will be a mixed bag, on the upside we would have Intel/SiFive contribution to RISC-V and the downside of lopsided competition.
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
I think it is a good thing. If SiFive/Intel were to contribute back to the RISC-V consortium it is a big win for all RISC-V adopters. Lots of organizations around the world developing with RISC-V.
RISC-V is a quick way to royalty free, open and customizable design for low complexity MCUs for a myriad of use cases.
I hope this deal materialize, and Intel not being NV, I dont see this being so bad.
At most it will be a mixed bag, on the upside we would have Intel/SiFive contribution to RISC-V and the downside of lopsided competition.

Well

From a strictly technical angle, I'm not sure what will be Intel's contribution to SiFive. Will it improve RISC-V CPU development? If they offer the existing design teams more resources, then maybe. If they meddle obsessively with the design initiatives then maybe not. Admittedly, SiFive's current offerings that can be bought on the open market aren't all that impressive (yet), so it's not likely that Intel will make them worse, but it's unclear if the acquisition will make them any better.

The main problem up front is that anyone that was hoping SiFive would remain independent from the Five Eyes is gonna have to look elsewhere for RISC-V implementations. RISC-V enthusiasm is at times intensely political in the abstact sense as well as the practical.
 

Gideon

Golden Member
Nov 27, 2007
1,608
3,573
136
This is an interesting take. While it's a random person on reddit, it does make some sense. With Pat being in the lead I can even see this ending up as something good, rather than the usual buy-to-shelve move.

TEXzLIB said:
I know several people very close to me at SiFive and many of them ex-Intel.
The game plan for the past few years was to get out of Intel, make a product, and get acquired by Intel. They already work very closely with Intel. Many of these guys wanted to work on RISC/WISC architecture and obviously, Intel was not the place to do so.
Intel is making some serious moves internally to open up pathways outside of X86. This actually ties in with their new foundry model very well. I don't know people super high up at Intel so it remains to be seen whether they are getting SiFive as a hedge or a genuine interest, in reality it's probably a bit of both.

Intel definitely needs a new ISA in the long- medium term as x86 is getting really long in the tooth with it's legacy and design decisions far from ideal by now.

It's very hard to get a new ISA accepted, and RISC-V is already gaining considerable interest and traction (particularily by the chinese). I don't believe Intel could really beat ARM with any kind of a new licensable ISA, let alone proprietary. The next best thing is to be the leading force behind the new open-source one.
 

DisEnchantment

Golden Member
Mar 3, 2017
1,590
5,722
136
Intel definitely needs a new ISA in the long- medium term as x86 is getting really long in the tooth with it's legacy and design decisions far from ideal by now.

It's very hard to get a new ISA accepted, and RISC-V is already gaining considerable interest and traction (particularily by the chinese). I don't believe Intel could really beat ARM with any kind of a new licensable ISA, let alone proprietary. The next best thing is to be the leading force behind the new open-source one.
This is another take.

This is my personal take on the issue.

What I would like to see is Intel/SiFive deal go through and to see them put some more effort in their Auto offerings and contribute back to the RISC V consortium.
Intel is the one of the top contributor to open source, if not the top (commit statistics across the OSS world indicate Intel is top). If they keep that same approach also to RISC-V ISA design it would really benefit a lot of adopters.

While Intel marketing is not the best, if I have to ask myself the question, with whom I would rather like to work with, B2B, engineer to engineer.
Would you work with a company who shares their code, send their Field Engineers and willing to debug their issues, listen to your concerns etc...?
Or would you rather work with a company who share absolutely nothing, whose docs need to be vetted and diluted before you can see? Who puts condition like 10x the FAE charges/hour if reported issue is your problem and normal charges if their problem? who thinks they are in charge of the solution not you, even though you are the customer?
I know what I would chose, and my experience with Intel was very good, both on the backend side and on the SoC side.
While we have moved away from Intel for some time due to different issues, I wish they could come back with something much more interesting because our cooperation was quite smooth. We worked with the Silvermont, Goldmont and other non public stuffs for a bunch of our products

I have been waiting for something like RISC-V a long while. By the time such a thing arrived, I was a bit far away from developing with such devices though.
I had a long enough history in Auto industry starting with small MCUs projects to full fat distributed compute platorms for highly Assisted Driving backed by V2X cloud backbone. And in my early days, If we ever have to take a look at which MPC5XXX series to choose and which one is best for our use case we would go nuts. There were just too many part numbers, more than 200+ just for MPC5XXX series, no joke, or probably even more. After selecting a specific part number we still need to do some custom CPLD/FPGA for some use cases like early collision detection which we cannot implement in software because the system needs to operate instantly from power on, which is not possible without waiting for the MCU to boot up.
Now, with RISCV you can choose a kit from Synopsys to build a custom SoC, the so called Domain Specific Processor.

With SiFive, Intel could offer me some IPs and differentiators to go along and also fab the SoC. Better than letting the competition have everything.

You could guess in which industry I work in :blush:
 
Last edited:

NostaSeronx

Diamond Member
Sep 18, 2011
3,683
1,218
136
Just should be mentioned:
ARM controls the ARM ISA.
SiFive doesn't control the RISC-V ISA.

If Intel was buying RISC-V International I would be worried. However, SiFive being taken over isn't much of a dent to RISC-V.

Overall, Intel has been eyeballing SiFive since 2018. They also do hiring for SiFive on Intel Capital: https://careers.intelcapital.com/jobdetail.php?jobid=1713097

Soft ownership to hard ownership isn't a huge deal in my opinion.
 
Last edited:

Nothingness

Platinum Member
Jul 3, 2013
2,371
713
136
Intel is the one of the top contributor to open source, if not the top (commit statistics across the OSS world indicate Intel is top). If they keep that same approach also to RISC-V ISA design it would really benefit a lot of adopters.
It doesn't seem Intel are at the top: https://opensourceindex.io. But they surely are doing rather well.

Now the track record of Intel buyouts is very poor. They fail at properly integrating companies by trying to make them fit their own way of doing things; all the people I know who worked at Intel after a buyout just hated it. But it seems Mobileye is an exception, so perhaps did they change the way they integrate companies?
 

gdansk

Golden Member
Feb 8, 2011
1,975
2,354
136
Embrace, extend, extinguish
Though in this case it hardly applies as Intel+SiFive's shared goal is to extinguish ARM.
Why pay licensing fees for an embedded system control chip inside of a GPU/CPU/chipset? It's a logical route to go since RISC-V has reasonably good compiler and RTOS support already. If Intel ships enough units it is cheaper than paying ARM fees per core.
 
Last edited:

DisEnchantment

Golden Member
Mar 3, 2017
1,590
5,722
136
It doesn't seem Intel are at the top: https://opensourceindex.io. But they surely are doing rather well.

Now the track record of Intel buyouts is very poor. They fail at properly integrating companies by trying to make them fit their own way of doing things; all the people I know who worked at Intel after a buyout just hated it. But it seems Mobileye is an exception, so perhaps did they change the way they integrate companies?
Sorry for a bit of nitpicking, but I would like to just add a note that this data source is GitHub only and Microsoft open source activity is massive on dotnet/xamarin/asp.net/ml.net and such MS products. Also Github is not the repo used by freedesktop, GNU etc. Not denying the above data, but it is one of many datapoints to consider.
Intel is active in kernel, drm, llvm, gcc, gstreamer and countless repos which are the backbone of a modern linux system. On top of that, they are active in openembedded, GENIVI and others as well which are usually community projects not owned by Intel.

Now the track record of Intel buyouts is very poor. They fail at properly integrating companies by trying to make them fit their own way of doing things; all the people I know who worked at Intel after a buyout just hated it. But it seems Mobileye is an exception, so perhaps did they change the way they integrate companies?
That is a problem I know, and I have close friends at Intel who were caught in all of this, but our company is no better, but part of this is also down to the nature of being big companies, (Our company is way bigger than Intel btw)
We buy companies instead of developing something organically. It is a big letdown for the guys with vision, our ideas don't get approved because someone feels we can just buy everything. Because of this there is resistance to integrate and cooperate because people feel let down, sidelined and marginalized.
It is seen as the fastest way to address competition or when the guy at the top has a new vision. Or we just buy companies to get patents and trademarks.
 
Last edited:

SarahKerrigan

Senior member
Oct 12, 2014
339
468
136
Though in this case it hardly applies as Intel+SiFive's shared goal is to extinguish ARM.
Why pay licensing fees for an embedded system control chip inside of a GPU/CPU/chipset? It's a logical route to go since RISC-V has reasonably good compiler and RTOS support already. If Intel ships enough units it is cheaper than paying ARM fees per core.

Intel generally doesn't use ARM for embedded controllers anyway, AFAIK. Movidius uses SPARC (LEON, which is free) auxiliary cores. So do at least portions of the Atom family, though I don't know if it's still all of them. Mobileye uses MIPS, which is presumably a candidate for replacement. Auxiliary cores on non-Atom x86 product families (ME/IE) have moved from ARC to inhouse (the same 486 variant that high-end Quark used.)

There's probably some ARM as embedded controllers in Intel parts, somewhere, but not US$2bn worth - likely not even if you stretch it over ridiculously long periods of time. There's ARM for embedded apps processors in FPGAs that I guess could be potentially replaced with SiFive cores, though.
 
  • Like
Reactions: Tlh97 and ElFenix

Doug S

Platinum Member
Feb 8, 2020
2,201
3,405
136
Intel definitely needs a new ISA in the long- medium term as x86 is getting really long in the tooth with it's legacy and design decisions far from ideal by now.

It's very hard to get a new ISA accepted, and RISC-V is already gaining considerable interest and traction (particularily by the chinese). I don't believe Intel could really beat ARM with any kind of a new licensable ISA, let alone proprietary. The next best thing is to be the leading force behind the new open-source one.

China's government and big tech companies are looking to push Loongson instead of RISC-V, so I wouldn't count on China to make RISC-V happen. The only purpose it has for them right now is avoiding ARM due to western control and having to pay for it. Intel buying the biggest backer of RISC-V would only act to speed up China's push towards Loongson.

From what I understand from others who are quite familiar with RISC-V, it has more issues than it should for a new ISA (which may be due to its roots as an academic project not intended to be commecially viable) and is pretty fragmented due to how much is "optional" in the spec.

If Intel is just looking for a small core they can stick into larger designs where they currently shoehorn x86 (like how they had x86 cores in their cellular modems) and to build a range of embeddable cores they can add to their library for the foundry clients they are hoping to snag, RISC-V is fine.

However, if they are looking to leave x86 behind eventually as you speculate, I think they would be way better off designing a new clean ISA and making it freely licensable like RISC-V. Yes, Intel has a history of utterly failing with new ISAs, but they were always trying to do something well outside the mainstream like i432, i860 and IA64 and also kept them proprietary. So just because they failed in the past doesn't mean they'd fail if they designed a simple clean and mostly important non-proprietary RISC ISA.

If you're going to make a "bet your business" decision to go with a new ISA, don't choose one that was designed by grad students who lack the decades of experience that can inform architectural designs that will avoid pitfalls that limit future implementation flexibility. I'm skeptical there is even the slightest inclination anywhere in Intel to consider a non-x86 future. I think they would have to see x86 drop to less than a 30% share of the PC/server market before they'd do anything so drastic. There would be massive internal resistance to even discussing the possibility.
 

gdansk

Golden Member
Feb 8, 2011
1,975
2,354
136
Intel generally doesn't use ARM for embedded controllers anyway, AFAIK. Movidius uses SPARC (LEON, which is free) auxiliary cores. So do at least portions of the Atom family, though I don't know if it's still all of them. Mobileye uses MIPS, which is presumably a candidate for replacement. Auxiliary cores on non-Atom x86 product families (ME/IE) have moved from ARC to inhouse (the same 486 variant that high-end Quark used.)

There's probably some ARM as embedded controllers in Intel parts, somewhere, but not US$2bn worth - likely not even if you stretch it over ridiculously long periods of time. There's ARM for embedded apps processors in FPGAs that I guess could be potentially replaced with SiFive cores, though.
Gelsinger talked about not forcing everything to be x86 and if rumors are true it seems he is serious. This is about going forward. Consider all the disparate micro-cores they have. If they want to switch to more efficient cores, RISC-V with different feature configurations available as needed for each product is a compelling option. Plus they wouldn't have to hack up compilers for their own internal architectures anymore. SiFive would be a way to skip designing their own and given them other options.

But also it is not a bad plan to offer it as free IP or low cost IP for customers using their new contract fabrication.
 
Last edited:

DisEnchantment

Golden Member
Mar 3, 2017
1,590
5,722
136
Intel generally doesn't use ARM for embedded controllers anyway, AFAIK. Movidius uses SPARC (LEON, which is free) auxiliary cores. So do at least portions of the Atom family, though I don't know if it's still all of them. Mobileye uses MIPS, which is presumably a candidate for replacement. Auxiliary cores on non-Atom x86 product families (ME/IE) have moved from ARC to inhouse (the same 486 variant that high-end Quark used.)

There's probably some ARM, somewhere, but not US$2bn worth - likely not even if you stretch it over ridiculously long periods of time.
It is not about Intel replacing ARM in their products, but more like Intel through SiFive offering its customers some DIY kits like Domain Specific Processors and then fabbing them.
From personal experience, lots of OEMs in Auto Industry do highly complex CPLD/FPGAs because there is no possibility to have custom silicon for their applications.
RISC-V is a paradigm change and it enables you to go to somebody like Synopsys and assemble some lego bricks. A TSMC CyberShuttle TO is pocket change for most Auto OEMs and it is not like they lack the expertise in ASIC design. Even a new comer like Tesla is already far beyond this having already developed inhouse designs for some critical components of their ADAS systems.

An analogue to something like what many OEMs are already doing with FPGAs, it is like shopping from the Xilinx store for ready made blocks to quickly assemble potent object collision detection systems.
Applies to similar industries like Manufacturing and Aerospace .
 
Last edited:

clemsyn

Senior member
Aug 21, 2005
531
197
116
If BK was the CEO, I would be pessimistic of the move but since it's Pat Gelsinger, I am optimistic. I think this would be a good long term move for Intel with him running the show.
 
  • Like
Reactions: CHADBOGA

dullard

Elite Member
May 21, 2001
24,998
3,326
126
It doesn't seem Intel are at the top: https://opensourceindex.io. But they surely are doing rather well.

Now the track record of Intel buyouts is very poor. They fail at properly integrating companies by trying to make them fit their own way of doing things; all the people I know who worked at Intel after a buyout just hated it. But it seems Mobileye is an exception, so perhaps did they change the way they integrate companies?

That is a problem I know, and I have close friends at Intel who were caught in all of this, but our company is no better, but part of this is also down to the nature of being big companies, (Our company is way bigger than Intel btw)
We buy companies instead of developing something organically. It is a big letdown for the guys with vision, our ideas don't get approved because someone feels we can just buy everything. Because of this there is resistance to integrate and cooperate because people feel let down, sidelined and marginalized.
It is seen as the fastest way to address competition or when the guy at the top has a new vision. Or we just buy companies to get patents and trademarks.
It is not just an Intel thing. It is a known issue with just about any buyouts in any industry. Small, nimble, innovative companies can turn on a dime and implement a new great idea the instant that the idea is generated. Then they get bought by large, slow behemoths that have many layers between new ideas and implementation. Meeting, committees, layers of management, interoffice politics, pet projects, old habits, etc. Any one person who doesn't take the time to fully understand the idea can kill it.

I'm in the biotech industry and the company that I worked at was bought out by a much larger entity. Sure it is great to work for a company that is financially stable. And sure the quality systems, standards, etc are very helpful. But I keep seeing great idea after great idea die in committee only to be later "invented" by another company that successfully takes it to market. That does really drain motivation and creativity. Yes, I do feel marginalized when a new idea or a change to an existing process is proven to be far better but I have no power to implement them.
 

Thala

Golden Member
Nov 12, 2014
1,355
653
136
The problem with SiFive is, that they are the main drivers behind the RISC-V specification - so a buyout might turn out to be problematic. If Intel just wants a core, they are free to develop one using the open RISC-V license - as many in the industry are doing. Intel does not need SiFive for the know-how either.
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
The problem with SiFive is, that they are the main drivers behind the RISC-V specification - so a buyout might turn out to be problematic. If Intel just wants a core, they are free to develop one using the open RISC-V license - as many in the industry are doing. Intel does not need SiFive for the know-how either.

If the above quotes from reddit are to be believed, SiFive has a lot of ex-Intel people working there. Possibly people that would not willingly go back to work for Intel proper. This move may be Gelsinger's way of getting back better people than would be currently available to Intel otherwise. Hiring people to work on a RISC-V core as a side project might be more than Intel can handle right now, depending on whom you believe.

For all I know, a certain ex-Intel forum member around here might even work for SiFive.
 
  • Like
Reactions: CHADBOGA