Qualcomm Acquires RISC-V Chip Designer Ventana Micro Systems

Qualcomm has acquired RISC-V startup Ventana to strengthen its CPU ambitions beyond mobile, "reinforcing its commitment and leadership in the development of the RISC-V standard and ecosystem," the company said in a press release. CRN Magazine reports: The San Diego-based company said Ventana's expertise in RISC-V, a free and open alternative to the Arm and x86 instruction set architectures, will enhance its CPU engineering capabilities and complement "existing efforts to develop custom Oryon CPU technology." Financial terms of the deal were not disclosed.

Qualcomm, which has already been using RISC-V for some products outside the PC and server markets, said Ventana's contributions will boost its "technology leadership in the AI era across all businesses," indicating the broad impact expected by this acquisition. "We believe the RISC-V instruction set architecture has the potential to advance the frontier on CPU technology, enabling innovation across products," Durga Malladi, executive vice president and general manager of technology planning, edge solutions and data center for Qualcomm, said in a statement. "The acquisition of Ventana Micro Systems marks a pivotal step in our journey to deliver industry-leading RISC-V-based CPU technology across products."

Further reading: Qualcomm Is Buying Arduino, Releases New Raspberry Pi-Esque Arduino Board

Will this be for RISC-V, or ARM?

[unixisc]

I recall when Apple acquired PA-Semi, which at the time was a designer of PowerPC based CPUs. Obviously that changed once they were in Apple, and started turning out ARM CPUs

I wonder if the same thing will happen w/ Qualcomm, which has its ARM chip Snapdragon. Will they repurpose Ventana to design ARM CPUs instead, or is this their way of entering the RISC-V market? If it's the latter, it would be good for RISC-V. I'd like to see some of the heavy hitters in the semiconductor industry - Qualcomm, NVID

Re:Will this be for RISC-V, or ARM?

[rickb928]

Here's the trick...

RISC-V is ostensibly an open source ISA. So as designers build new implementations, they may be advancing the capabilities of the ISA and contributing to the RISC-V universe.

But history teaches us that despise licensing and such, open source advances often get locked behind commercial license forks, and it is a fight to get these outfits to obey the true license. ARM suffered from this occasionally, but not like I expect RISC-V to. This chip ISA has the potential to upend the whole business.

Unless the big stuff gets locked away.

Combine Qualcomm's IP and expertise with the RISC-V platform, a nearly blank slate, and we could see cool stuff. Giving back to the RISC-V community? Not Qualcomm's strength from experience.

But RISC-V could win, if the innovators aren't locked out or patent-trolled into oblivion.

Re:Will this be for RISC-V, or ARM?

[tlhIngan]

Except they are. I think SiFive accomplished a lot of their speed by implementing their own extensions to get around issues with the RISC-V ISA w.r.t. addressing mode. They are sufficiently popular that compilers do support those extensions as the code runs faster.

Qualcomm has also made RISC-V moves, as has nVidia and Western Digital.

But given Qualcomm bought Arduino, it woiuld not surprise me if they were going to release a bunch of RISC-V variants that require a compiler that can handle the Qualcomm extensions.

Re:

[unixisc]

RISC-V is under an MIT license, not GPL. Nobody has to make designs "free" after improving and optimizing the base designs

Re:

[arglebargle_xiv]

I'd say they're just hedging their bets, they make umpty bazillion each year off their Arm offerings but want to make sure that if Risc-V ever takes off they'll have a stake in it. All without having to divert any of their own resources to bothering with it.

How is RISC-V better than ARM?

[evil_aaronm]

I like you guys; you're almost like family. Ya know, your "weird cousing" types, but hey. I highly value your insight. So while I could probably look the answer up elsewhere, I'm looking to "family" to tell me truly, what does RISC-V have that would lead me away from ARM? I've been in a long, stable relationship with ARM; she's a charming gal. She's got all that uncle John needs. Why would I consider trashing all that for RISC-V hotness?

Are we talking speed? Speed per watt? Cost? It's kinda hard to beat Pico at ~$5. Development tools? Industry support? More cores per chip? What?

Re:

[alvinrod]

Really there's nothing outside of the ISA being open and freely extensible for anyone who wants to do so. The ISA itself is similar to ARM, MIPS, or other RISC ISAs. If you wanted to build a custom microcontroller for some purpose it would be cheaper to use RISC-V since it doesn't have licensing costs. Most companies aren't doing this though and just by OTS components that work with their codebase. If you wanted to create some dedicated hardware paths for computationally expensive operations, the RISC-V ISA

Re:

[kriston]

The current owner of MIPS is creating a RISC-V variant using the MIPS technology underneath. It's sufficiently similar, evidently.

Re:

[unixisc]

In fact, MIPS has pretty much gone wholly w/ the RISC-V architecture for all future designs and implementations. I looked, and only 2 instructions are different b/w MIPS and RISC-V

Re:

[kriston]

I kinda wondered why RISC-V didn't just adopt the MIPS ISA. All of the patents have expired and it has a 64-bit variant. Plus, it can do both big-endian and little-endian, something RISC-V supports but nobody has implemented.

Re:

[_merlin]

It's really only better if you want to make a really low-end CPU with minimal gates/silicon. If you're trying to do a multi-issue out-of-order CPU, ARMv8 is better.

Re:

[unixisc]

Currently, there are 32, 64 and 128 bit versions. Would they need a 16 bit version for really low end, embedded applications?

Re: How is RISC-V better than ARM?

[_merlin]

It doesn't need a 16-bit variant. We're at the point where the very low end is mostly 32-bit anyway. It's competing with various MIPS derivatives (e.g. PIC32) and 32-bit ARM in that space. Just being royalty-free and open isn't really all that exciting. SPARC and PowerPC are both free and open. All the patents on MIPS and Super-H (up to SH-5 at least) have expired, although the current owner of the MIPS trademark will threaten to sue you if you call it MIPS. RISC-V is probably going to displace MIPS derivat

Re:

[unixisc]

I had read that MIPS has pretty much abandoned its ISA in favor of RISC-V. Not a major deal, since only 2 instructions were different. Also, in addition to OpenSparc and OpenPower, ARM itself is "open" in the sense that it can be licensed, after a company pays an upfront membership fee (According to Steve Furber, it was to keep ARM from going into the red).. I daresay one could even do what Cyrix and Centaur, in addition to AMD, used to do in the 90s w/o running afoul of either Intel or AMD ISA patents

Re:

[_merlin]

The MIPS instruction set is used in a lot of places because it's fairly simple and has decent compiler support. A lot of things like home routers, electronic toys, etc. use MIPS-like CPUs, particularly in China. These aren't actual MIPS CPUs, just embedded CPUs with MIPS-like instruction sets. These applications are slowly moving towards RISC-V. There are other MIPS-like CPUs, e.g. the SunwayMPP supercomputer CPU which is like MIPS with a bunch of vector processors tacked on (SunwayMPP is to MIPS as Cel