SiFive Announces The Performance P550 As The Fastest RISC-V Processor Yet

[phoronix]

Phoronix: SiFive Announces The Performance P550 As The Fastest RISC-V Processor Yet

SiFive today is announcing the Performance P550 as the fastest RISC-V processor yet while also announcing the Performance P270 too...

https://www.phoronix.com/scan.php?page=news_item&px=SiFive-Performance-P550-P270

[Dawn]

The A75 is a pretty old core, and mainstream ARM licensables have progressed massively since then (an A710 should be doing close to double the same-clock perf of an A75); I wonder how area efficiency compares. Still, good to see progress.

[Jabberwocky]

RISC-V is the right tool to fight IP. This is pretty awesome!

Regarding the IP: I don't see the relavance to your statement. This is about the P550 core by SiFive.

There's no fight against IP or not anymore. SiFive now has IP, patents, and massive investors just like any other big company. They have customers like Intel, Samsung, Renesas...

Regarding the fight. The SiFive engineers are obviously working extremely hard, yet they still have a very long way to go to be competitive in general processing. This comment by Dr. Ian Cutress:

https://www.anandtech.com/show/16226/apple-silicon...
https://www.anandtech.com/show/16214/amd-zen-3-ryz...

A13 scores 19.85/GHz
A14 scores 21.1/GHz
R9 5950X scores 13.57/GHz
i9-10900K scores 11.08/GHz

P550 is probably peak score/GHz, whereas all of these points are at the most extreme/least efficient point for the high-end products. The overall performance also matters when you factor in frequency, as cores won't scale linearly with core frequency as other parts of the SoC will be fixed.

I wonder what's going to happen with Intel that's starting to use SiFive (there were talks about Intel buying them). I favor RISC-V ISA over x86, but not a fan of SiFive's "new" appoach to RISC-V. I'm hoping that open implementations will become popular: https://github.com/riscv/riscv-cores-list AFAIK Raptor Computing Systems still has the fastest open hardware, it's just expensive for a hobby.

Looking forward to seeing something with the performance of a Raspberry Pi 3 or faster and completely open source. Chips Alliance for the win!

[sykobee]

Getting better. Maybe Intel could create a small cheap Pi-level SoC to demonstrate their foundry services based around a quad P550 and some other Intel IP (GPU, etc)...

[ehansin]

I just wanted to see if anyone had any opinions on the quality of the ISA for RISC-V. I read something once were someone complained about it, but of course you know how the internet and opinions go! I do think it is pretty cool to see the progress of the RISC-V ecosystem. And I am guessing the extensibility of the ISA gives it some growing room, just hate to see it become a hairball mess. But I really don't have enough understanding in that area. Anyway, cool to see it gain traction.

[Almindor]

I just wanted to see if anyone had any opinions on the quality of the ISA for RISC-V. I read something once were someone complained about it, but of course you know how the internet and opinions go! I do think it is pretty cool to see the progress of the RISC-V ecosystem. And I am guessing the extensibility of the ISA gives it some growing room, just hate to see it become a hairball mess. But I really don't have enough understanding in that area. Anyway, cool to see it gain traction.

I've been working on Rust support for their Hifive1 board early when getting the thing booting was a challenge. This got me pretty deep into risc-v. I'd say there's definitely some validity to some of the criticism but a lot is just hot air "NIH" or fanboy-ism over something else. It's not perfect but I think the ISA has a really good chance and some of the problems can be fixed by additional extensions.

Ironically the extension system is also it's biggest weakness. It's like openGL all over again, keeping track is a PITA.

As a side note people need to understand that an open ISA is great it has 0 impact on the open-ness (and non-spy-backdoorness) of the actual hardware.

[Dawn]

I just wanted to see if anyone had any opinions on the quality of the ISA for RISC-V. I read something once were someone complained about it, but of course you know how the internet and opinions go! I do think it is pretty cool to see the progress of the RISC-V ecosystem. And I am guessing the extensibility of the ISA gives it some growing room, just hate to see it become a hairball mess. But I really don't have enough understanding in that area. Anyway, cool to see it gain traction.

It's... fine. It lacks the brain-damages of the first-gen RISCs like MIPS or SPARC, and that's a good thing. Relatively little of it is egregiously mis-designed, though I have complaints about some little things. OTOH, there's a lot of useful things missing - advanced addressing modes (especially pre/post-increment, which are very useful for loops), bit ops, especially rotate (Bitmanip has been coming Real Soon Now for a while), embedded-oriented SIMD (also Real Soon Now with the P extension), and some other small stuff.

I don't particularly love it, compared to ARM64 or PowerPC, and I think it's telling how many RV vendors have ended up heavily extending the ISA (the Andestar V5 extensions, for instance) But it is what it is.

[brucehoult]

The A75 was announced 29 May 2017, almost exactly four years ago. I differ with a previous commentator who says it’s “a pretty old core”. Not the latest from ARM, certainly, but not that long ago.

When SiFive announced the U74 at the end of October 2018 that was six years after ARM announced the A53.

When SiFive announced the U84 at the end of October 2019 that was 4.75 years after ARM announced the A72.

Now SiFive are four years behind the A75.

SiFive are steadily catching up to ARM.

[wizard69]

RISC-V is the right tool to fight IP. This is pretty awesome!

this doesn’t make sense, IP is a good thing. It is how you protect your IP that sometimes is a problem. SFive owns plenty of IP which frankly enables them to have a chance of success.

[lyamc]

I'm definitely looking forward to seeing RISC-V suddenly hit mainstream thanks to the support of Intel