News SiFive Lays Off Hundreds of RISC-V Developers

[Admin]

In a major blow for RISC-V, SiFive is laying off hundreds of chip designers.

SiFive Lays Off Hundreds of RISC-V Developers : Read more

 

[ezst036]

Should have designed at least one ATX motherboard. Even if the performance is lower than x86, there would be takers.

Too bad.

 

[Findecanor]

Should have designed at least one ATX motherboard.

Not ATX but the mITX Milk-V Oasis was announced a few days ago.
The SG2380 SoC has 16×SiFive P670 cores. (12 at 2.5 GHz, 4 at 1.6GHz).
Plus a "NPU" with eight SiFive X280 cores with 512-bit vector units. GPU, NVMe, SATA, LPDDR5 slots, PCIe slot ...

Each P670 core is out-of-order. RVA22 + V 1.0 + V cryptography. and has 4-wide decode. 2×ALU, 2×branch, 2×LDST, 2×FP, 2×V units,

SiFive announced the larger P870 core this year, which should be quite competitive. I hope it comes out.

 

[pug_s]

Don't see this as a big issue. There's plenty Chinese companies developing Risc-V hardware and software and not just American Ones.

 

[TCA_ChinChin]

Don't see this as a big issue. There's plenty Chinese companies developing Risc-V hardware and software and not just American Ones.

Overall its not a big issue for RISCV, but I think its still a small obstacle to more adoption for RISCV overall, since SiFive is one of the more notable or well known RISCV hardware companies. At least from my limited perspective.

 

[ThomasKinsley]

Wasn't SiFive's CEO recently complaining about US sanctions affecting their industry? Perhaps this is connected to that.

 

[Steve Nord_]

What are these champions going to do next? Lead Tabata workouts? Build check and wisdom check NPUs? Power efficiency workshop a new GPU? Hack ATSC 3 frameworks?

 

[bit_user]

Should have designed at least one ATX motherboard. Even if the performance is lower than x86, there would be takers.

Well, the HiFive Unmatched was mini-ITX and the upcoming HiFive P550 is micro-ATX.

SiFive Boards - The best way to Develop RISC-V Software.

One-stop download for documentation, software development kits, toolchains, utilities, and software ecosystem solutions for each SiFive RISC-V development board.

www.sifive.com

 

[bit_user]

Wasn't SiFive's CEO recently complaining about US sanctions affecting their industry? Perhaps this is connected to that.

You raise a good point, but this doesn't necessarily mean SiFive is in trouble.

In the startup game, being first-to-market can be everything. So, that naturally leads you to staff up quickly and build/refine your products to the point where they're viable. Once that happens, you might no longer need so many staff to keep the business on its natural growth trajectory until a liquidity event (i.e. IPO or acquisition). Furthermore, to make yourself more attractive to investors or would-be acquirers, you want to reduce costs to improve your financials. That means cutting headcount.

It's brutal, but it's too often a part of the game. Not to sound cold - and I do extend my sympathies to all affected - but people who like job security shouldn't go to startups.

I've been there myself, once having gotten let go from a startup that was not doing well (many years ago). I also have a friend who got brutally cut from a viable startup, pretty much immediately after they wrapped up the first version of their product. Kinda like how some game or graphics studios will have layoffs right after they finish a title. At least with a startup, they usually give you the opportunity to exercise your stock options when you're laid off.

 

[Kamen Rider Blade]

Hopefully AMD picks up some of the SiFive engineering staff that got laid off.

 

[Jagwired]

Shouldn't Intel and AMD be all in on RISC-V? Otherwise, they're going to be paying ARM for the next 50 years.

 

[bit_user]

Shouldn't Intel and AMD be all in on RISC-V? Otherwise, they're going to be paying ARM for the next 50 years.

It's not that simple, of course.

If they jump in too soon, they risk alienating their existing x86 customers by signalling their priorities lie elsewhere. That will cause some customers to drop x86, at which point there's no particular reason why they wouldn't also switch vendors. At the very least, they might switch to whichever x86 vendor that moves last.

The other risk of jumping too soon is that they produce products before there's a real market for them. I think that's arguably what happened with AMD's Opteron A1100 ARM-based server CPU, which came and went without a whole lot of fanfare. The ARM software ecosystem wasn't quite there, yet. The same is true of the RISC-V, today.

While we're talking markets, you should consider that ARM is in the #2 spot for servers and cloud. For chip makers, you have to produce the chip people want to buy, and that has a lot to do with what software they want to run. That, of course, ties back into the question of the RISC-V ecosystem which Intel has actually done a lot to support.

So, it's basically a question of building the right CPU for the right markets, at the right time. For Intel and AMD, the benefits of getting into ARM or RISC-V relatively early are outweighed by the need to convince customers they are committed to & focused on x86.

 

[wbfox]

Shouldn't Intel and AMD be all in on RISC-V? Otherwise, they're going to be paying ARM for the next 50 years.

Yes, the companies that own ARM are going to end up paying ARM....

Why waste time propping up a political hot potato when you can just buy the thing you actually want and have all the experience, overhead and employees to deal with?

 

[wbfox]

Hopefully AMD picks up some of the SiFive engineering staff that got laid off.

Yep, new owners of ARM are going to hire people that drive the risc-v hype train. why would any of the companies that bought ARM be like, hey, people that know how to do stuff with this group that is going to be squarely on the US black list, come work for us?

 

[Kamen Rider Blade]

Yep, new owners of ARM are going to hire people that drive the risc-v hype train. why would any of the companies that bought ARM be like, hey, people that know how to do stuff with this group that is going to be squarely on the US black list, come work for us?

AMD has had the ARM license for a VERY long time, long before the cancelled K12.

They aren't attached to the hip at it.

They're more than willing to make RISC-V chips as well.

AMD ready to diversify its portfolio with ARM-based chips

Some AMD customers are asking for ARM-based processors, and AMD is willing to fulfill this demand. Moreover, AMD seems to be open to explore other architectures like RISC-V, as well. It is not yet clear what markets would benefit from the increased ARM demand, but, seeing that AMD is already planning to use Xilinx tech for its datacenter products, we could infer that ARM may be destined for Team Red’s desktop and laptop product stacks.

 

[bit_user]

Why waste time propping up a political hot potato when you can just buy the thing you actually want and have all the experience, overhead and employees to deal with?

Successful companies function in the world as it exists, not the world as they wish it were.

ARM has dominated the phone market and has a strong #2 position in the cloud. It's present in the Chromebook market, the hobbyist market (Raspberry Pi & similar) and Microsoft is pushing it into the mainstream laptop market. If AMD wants a piece of that action, they need to implement ARM.

That doesn't mean they can't also engage with RISC-V, but the market for it just isn't there yet. Most of SiFive's business is in the embedded market, which AMD has no real presence other than through Xilinx. Yes, I know there are embedded Ryzens, but I'm pretty sure their sales volume is quite low by comparison with ARM or perhaps even RISC-V based SoCs.

Yep, new owners of ARM are going to hire people that drive the risc-v hype train.

Huh? ARM IPO'd. It's now an independent company, though I think Softbank still owns a chunk of it.

 

[bit_user]

AMD has had the ARM license for a VERY long time, long before the cancelled K12.

Source?

AMD ready to diversify its portfolio with ARM-based chips

I just posted that same link in the thread about AMD & Nvidia designing their own ARM cores.

 

[Kamen Rider Blade]

Source?

AMD Finally ARMs Itself for New War on Intel
Jun 13, 2012 12:13 PM

On Wednesday, AMD announced that it has indeed signed a deal to license ARM's chip architecture. But the pact isn't quite the revolution many were expecting. That revolution may still come, but for the moment, AMD says it will merely use ARM technology to build new security tools into the chips it's designing for tablets and other devices. It will not build entire processors based on the ARM architecture -- at least not yet.

On September 18, 2015 Keller left AMD
AMD Cancelled K12 after Jim Keller left AMD for Tesla

I just posted that same link in the thread about AMD & Nvidia designing their own ARM cores.

I'm borrowing your link to reinforce a point about RISC-V

 

[bit_user]

AMD Finally ARMs Itself for New War on Intel
Jun 13, 2012 12:13 PM

"But the pact isn't quite the revolution many were expecting. That revolution may still come, but for the moment, AMD says it will merely use ARM technology to build new security tools into the chips it's designing for tablets and other devices. It will not build entire processors based on the ARM architecture -- at least not yet.

AMD has licensed ARM Cortex-A5 processor architecture -- an architecture much simpler than what you find in, say, an iPhone -- and this will be used to add new security tools into AMD's APUs, or accelerated processing units, chips that include both a CPU and a graphics processor. Including mechanisms for encrypting data, these security tools are designed to prevent outsiders from accessing sensitive information handled by a wide range of computing devices."
​

Okay, so it seems like that's not an architecture license, but rather AMD licensing ARM's TrustZone cores & software technology to embed in AMD processors. That's the equivalent of Intel's Management Engine.

The reason you buy an "Architecture License" is to design your own ARM ISA cores. You wouldn't need it just to embed a Cortex-A5 in your CPUs.

 

[Kamen Rider Blade]

​

"But the pact isn't quite the revolution many were expecting. That revolution may still come, but for the moment, AMD says it will merely use ARM technology to build new security tools into the chips it's designing for tablets and other devices. It will not build entire processors based on the ARM architecture -- at least not yet.​

​

AMD has licensed ARM Cortex-A5 processor architecture -- an architecture much simpler than what you find in, say, an iPhone -- and this will be used to add new security tools into AMD's APUs, or accelerated processing units, chips that include both a CPU and a graphics processor. Including mechanisms for encrypting data, these security tools are designed to prevent outsiders from accessing sensitive information handled by a wide range of computing devices."​

​

Okay, so it seems like that's not an architecture license, but rather AMD licensing ARM's TrustZone cores & software technology to embed in AMD processors. That's the equivalent of Intel's Management Engine.

The reason you buy an "Architecture License" is to design your own ARM ISA cores. You wouldn't need it just to embed a Cortex-A5 in your CPUs.

It seems like AMD was working on a full ARM CPU, but they cancelled the K12 as Jim Keller said.

They focused their "$$$ & Time" on Zen when AMD was at it's low point.

Whether or not they will come back to a full ARM CPU is TBD (To Be Determined), but it's not out of the question.

 

[bit_user]

It seems like AMD was working on a full ARM CPU, but they cancelled the K12 as Jim Keller said.

Yes. That's when I think they first got an architecture license.

If you recall, it was preceded by the Opteron A1100, which I think was them trying to feel out the maturity of the ARM server market.

 

[Kamen Rider Blade]

Yes. That's when I think they first got an architecture license.

If you recall, it was preceded by the Opteron A1100, which I think was them trying to feel out the maturity of the ARM server market.

They were WAY ahead of their time in terms of ARM being ready for the Server market

Spoiler: ARM's recent market share in Enterprise

Amazon has more than half of all Arm server CPUs in the world
Over 50% of that small ARM Server MarketShare belongs to Amazon alone for their AWS deployments.
Amazon owns (> 50% of 8.1%), that means Amazon owns (≥ 4.05% of the Enterprise Server CPU's MarketShare).

Spoiler: ARM's MarketShare in Consumer CPU's

ARM really only mattered in DeskTop Consumer CPU's & Server CPU's starting in 2020.
While Apple owns 90% of ARM-Based PC's today, Competitors are gearing up to adopt Qualcomm's ARM processors for PCs in 2024
Apple owns (90% of 14.8% = 13.32%) vs "Team x86"

 

[wbfox]

Successful companies function in the world as it exists, not the world as they wish it were.

ARM has dominated the phone market and has a strong #2 position in the cloud. It's present in the Chromebook market, the hobbyist market (Raspberry Pi & similar) and Microsoft is pushing it into the mainstream laptop market. If AMD wants a piece of that action, they need to implement ARM.

That doesn't mean they can't also engage with RISC-V, but the market for it just isn't there yet. Most of SiFive's business is in the embedded market, which AMD has no real presence other than through Xilinx. Yes, I know there are embedded Ryzens, but I'm pretty sure their sales volume is quite low by comparison with ARM or perhaps even RISC-V based SoCs.


Huh? ARM IPO'd. It's now an independent company, though I think Softbank still owns a chunk of it.

Yes, it IPO'd and who bought the most shares....I'm leading you here.....these giant blue chip tech stock companies such as.....you can do it....
As an example. If I own a large chunk of lets just say, Coca-Cola, and I buy a coke at a store, is it really buying it when I get it all back since I own a chunk of the company or is it just an accounting ledger version of paying myself?

 

[TJ Hooker]

Yes, it IPO'd and who bought the most shares....I'm leading you here.....these giant blue chip tech stock companies such as.....you can do it....
As an example. If I own a large chunk of lets just say, Coca-Cola, and I buy a coke at a store, is it really buying it when I get it all back since I own a chunk of the company or is it just an accounting ledger version of paying myself?

What do you consider a "most shares"/"a large chunk"? From what I can tell, those companies bought less than a 1% stake each, and only up to 10% combined.

Exclusive: Arm signs up big tech firms for IPO at $50 billion-$55 billion valuation

Customers of Arm Holdings Ltd including Apple Inc , Nvidia Corp , Alphabet Inc and Advanced Micro Devices Inc have agreed to invest in the chip designer's initial public offering, according to people familiar with the matter.

www.reuters.com

"Apple, Nvidia and the other strategic investors have agreed to invest between $25 million and $100 million each in the blockbuster IPO, the sources said. Arm and SoftBank have set aside 10% of the shares to be sold in the IPO for its clients, Reuters has previously reported."

If you own a 0.1% stake of Coca Cola and you buy a coke at a store, you're hardly going to "get it all back".

 

[bit_user]

As an example. If I own a large chunk of lets just say, Coca-Cola, and I buy a coke at a store, is it really buying it when I get it all back since I own a chunk of the company or is it just an accounting ledger version of paying myself?

Just to add to what @TJ Hooker said:

Let's say you owned 100% of a business which makes widgets and all of the profits ultimately end up in your pocket. Then, you go to a store and buy one of those widgets, the only thing you're paying yourself is the margin your company makes on each widget. That's it. You're still losing the materials cost, production costs, distribution costs, and the store's markup. So, maybe you'd do well to get back 10% of what you're spending. That'd be high for some certain commodities, perhaps a bit low for premium brands. Now, let's say you owned just a 10% stake. Okay, so you get roughly 1% back.

However, the main thing that's rotten about your analogy is that shareholders don't necessarily get anything back, until they sell their shares. At that point, the shares are priced on the basis of future earnings expectations, not based on what the company earned up until that point. ARM could pay a dividend or do share buybacks, but as long as they're focused on growth, those aren't terribly likely.