Apple Sues SoC Startup Rivos for Poaching Employees and Stealing Trade Secrets

[kc9hzn]

No company in it's right mind poaches and then asks the new employee to steal on their way out. Not that it can't happen but that is an extreme exception as someone that has been in the industry for three decades

Apple's cores are ARM. Taking verilog from an ARM core and applying that code to a RISC-V is a non starter. More applicable would be the idea of a.specific feature such as a cache structure and designing a similar structure for RISC-V let's say.

The lawsuit looks specious because tech companies routinely try to enforce "non-compete" agreements. They aren't worth the paper they are printed on.

Also if it was truly a theft case it would be criminal not civil.

It does look like this firm was targeting a specific set of Apple employees, with high access privilege to Apple’s CPU designs. Besides, it’s a stealth mode startup, they don’t have an established product or name recognition, all they have is VC funding. Corporate espionage isn’t something I’d put past a new firm looking to rapidly increase VC funding (especially if they’re intending to get a lot of VC money and run), particularly in an industry as competitive as chip design. I find Apple’s story plausible because of the amount and sort of data stolen, and, reading between the lines, it seems that it’s more than just the two named employees, it’s just that the two named employees are either the most open and shut case or the only ones Apple caught red handed. Apple indicates that they accessed this info after accepting an offer from Rivos. The linked lawsuit contains more of Apple’s argument, including that the sort of information taken could be used to kickstart development of RISC processors of other architectures.

 

[name99]

A simulation environment for ARM would be useless for RISC-V.

Also if it's a performance simulator it's going to be very ARM specific.

If It's simulating the actual architecture then the simulator is going to be off the shelf Synopsys, Cadence, etc. The environment and testbench code would not be portable and neither would the tests or instrumentation.

RISC-V is so different than ARM both at an architectural level and system level implementation.

If they did steal all the information claimed, most of it would be useless. Architectural ideas would be more valuable. More of have you thought about this implementation vs that?

No it wouldn't. Jesus Christ!
Most of what goes into designing (and simulating) a fast CPU is generic, especially if the two use similar to identical memory and IO models.

Look, for example, at most of the PA Semi Patents -- they all discuss what they are doing as relevant to any ISA, and list Power, ARM, even x86 as possible use cases. Same with the PA Semi simulation codes.

 

[RedWeasel]

Probably impossible for someone holding the title of CPU implementation engineer.

You have to interface with FPGAs at some point.

You could restrict that to dedicated desktop machines instead of personal laptops (for
example).

 

[JPack]

You could restrict that to dedicated desktop machines instead of personal laptops (for
example).

And then what? Watch each of them like a hawk for 12 hours/day? You would need engineers to watch engineers. For example, Intel has several thousand people involved in CPU design and maybe 10,000 or 20,000 people doing verification, optimization, debugging, etc.

 

[RedWeasel]

And then what? Watch each of them like a hawk for 12 hours/day? You would need engineers to watch engineers. For example, Intel has several thousand people involved in CPU design and maybe 10,000 or 20,000 people doing verification, optimization, debugging, etc.

Then you can implement measures like physical checks to ensure no one brings in disks or USB sticks into the offices where those desktops are.

Also usually you can control what kind of device can be connect to USB or restrict to specific device IDs.

 

[kc9hzn]

For people who didn’t read the linked lawsuit (and/or can’t deal with Scribd), it looks like the reason Apple is suing Rivos is because this pattern of behavior was more widespread. Apple says it lost 40 engineers to Rivos, and it seems like it was more than the two named here in the lawsuit who were up to shady business. Apple alleges that several, if not most, of the Rivos poached engineers downloaded Signal to discuss Rivos related business, including at least one message sent via iMessage from one to another not to discuss Rivos over iMessage. Apple alleges that many of them suspiciously cleared web browser history (one after having searched for online legal advice). It seems like the pattern of behavior of the two named defendants was common to most, if not all, of the engineers Rivos poached, which is why Apple alleges Rivos was actively engaged in the theft of trade secrets. The two named defendants seem to be the most egregious cases, the ones that Apple feels are a slam dunk. (They’re also the two who were the worst at covering their tracks. Skilled at processor design, perhaps, but kind of lacking in common sense about computers from a usage perspective.) If Apple’s allegations are true*, it definitely paints a picture of Rivos engaging in conspiracy to exfiltrate a wealth of design documents that could be used to get their own system online. (Apple notes that much of the data would be very useful for that purpose, especially for other reduced instruction set architectures.)

* And really, I have no reason to doubt it. The idea that a stealth mode startup would be engaged in such a conspiracy isn’t particularly weird, especially if the firm feels like it needs a breakthrough to get the next round of VC funding. And having a RISC-V chip with Apple Silicon performance would be a breakthrough that would certainly net them considerable amounts of funding. And given the current bubble state we’re in, malicious behavior (like pump and dumps or ginned up prospectuses) is distressingly common at bubble peak, the goal becomes snatching up as much money as one can (instead of delivering a sustainable business model).

Now, I have no evidence of Rivos specifically engaging in investment fraud, but the whole idea of trying to get information that will lead to them having an unusually fast time to market (or at least a dramatic prototype to show off) just makes me think of the get rich quick mentality that’s driving a lot of crypto and fintech startups. (For what it’s worth, it seems like, every other month, I see an ad on the subway for yet another new crypto startup, yet another fintech providing retail banking, and yet another fintech providing retail investing. Never mind that, with all the existing fintechs engaged in retail banking and investing, there’s serious competition in those fields, and these guys all seem to be delivering largely undifferentiated products [and never mind that the big financial firms are already players in app based retail banking and app based/online retail investing].)

 

[JPack]

Then you can implement measures like physical checks to ensure no one brings in disks or USB sticks into the offices where those desktops are.

Also usually you can control what kind of device can be connect to USB or restrict to specific device IDs.

I think you underestimate the type of work CPU engineers need to do. If it were possible, these restrictions would have been implemented a long time ago.

 

[macfacts]

"good artists copy, great artists steal."

 

[DeepIn2U]

did anyone else think this was odd:

Not odd.

Because they cannot take Rivo’s to Civil/Criminal court - it would MUCH harder to prove they’re in the wrong. BUT since they have this data, make them pay for it - perpetually. And before making similar patents to avoid such litigation.