News More Than 16 Billion RISC-V Chips Forecasted by 2030

[Admin]

RISC-V is used for everything from microcontrollers to high-end application processors, and the RISC-V International organization predicts up to 16 billion chips by 2030.

More Than 16 Billion RISC-V Chips Forecasted by 2030 : Read more

 

[ezst036]

A RISC-V future would be more beneficial for us as users and consumers than an ARM future would be. These companies and the things they hide in their proprietary cloaks cannot be trusted.

None of them.

 

[HaninTH]

A RISC-V future would be more beneficial for us as users and consumers than an ARM future would be. These companies and the things they hide in their proprietary cloaks cannot be trusted.

None of them.

Unfortunately, there's little we can do to ensure certain govt agencies don't enforce either a watered down security system or backdoors (in some fashion) in these systems as they can not allow us to do anything that doesn't serve them nor have us doing anything they don't know about.

Alas, I fear paranoia will never elude us, for good reason.

 

[bit_user]

"RISC-V is the most profound technical revolution of our time."

It might be an industry revolution, but it's not a technical revolution.

On a technical level, it's all fairly derivative. It pretty much has to be, given that it's royalty-free. There almost can't be any cutting-edge tech in the standard parts of the ISA, by definition. Anything novel, is all covered by patents.

 

[brucehoult]

On a technical level, it's all fairly derivative.

Sure. So is Arm64. They both stand on the very solid base of things such as MIPS in 1985, or at least 64 bit MIPS III and DEC Alpha in 1991 -1992, with mistakes dropped and just a few innovations.

Alpha was in fact a very good design which would have been viable today, but it died because of company mismanagement (as did MIPS), not for any technical reasons. Being proprietary, neither can survive their owners -- but MIPS has now switched to the RISC-V ISA and may see a revival.

That is part of the great industry revolution of RISC-V. It doesn't have to be the best, it just has to be Good Enough (and it's better than x86!), and the fact that no one company's misfortune can take it down means that you'll never have to move to yet another new ISA.

RISC as a marketing name has existed in commercial machines for almost 40 years (just over 40 in academia), but the principles are virtually unchanged from the CDC6600 in 1964 and Cray-1 in 1975 (and the new RISC-V Vector extension builds on what Cray did). Sixty years of experience shows that the RISC ideas stand the test of time and are never a bad idea, in any implementation technology.

 

[bit_user]

Alpha was in fact a very good design which would have been viable today, but it died because of company mismanagement (as did MIPS), not for any technical reasons.

Jim Keller has said in interviews that he believes Alpha's relaxed memory model (which ARM & others have copied, though I think nobody else has gone quite so far) was partly to blame for its poor adoption. He recounted how they'd often get calls from Windows developers at Microsoft, claiming the CPU was broken. DEC would reply back that they had similar code in their own operating systems they internally developed & sold, and it works just fine. The result was that when Microsoft eventually did get Windows support working, it was probably littered with way more performance-robbing memory barriers than it really needed.

If Microsoft encountered such difficulties, I'm sure independent software vendors were similarly plagued. That would've hurt 3rd party support for the platform and limited adoption.

That is part of the great industry revolution of RISC-V. It doesn't have to be the best, it just has to be Good Enough

Only because it's up against fairly weak competition (x86) and there are geopolitical concerns and economic encumbrances around ARM. If not for those, I'd be less rosy about RISC-V's prospects.

you'll never have to move to yet another new ISA.

I hope we'll see some real innovation in the ISA realm, in coming decades. I think current approaches are too inefficient.

Sixty years of experience shows that the RISC ideas stand the test of time and are never a bad idea, in any implementation technology.

OpenGL was good enough, until it wasn't. There comes a point where the level of complexity underneath becomes so great that you pay a real penalty for maintaining such a simplistic interface. It forces a lot more work to be done on on each side, to try and recover concurrently that's lost in the translation, and detect invariants which are known on the other side of the ISA boundary.