Imagination Announces "Catapult" RISC-V CPU Family

[phoronix]

Phoronix: Imagination Announces "Catapult" RISC-V CPU Family

With Imagination Technologies having sold off what was MIPS Technologies several years ago and that CPU architecture having been abandoned now, Imagination today announced "Catapult" as their new family of RISC-V processor IP...

https://www.phoronix.com/scan.php?page=news_item&px=Imagination-RISC-V-Catapult

[microcode]

Hey, if they can land a quality Vulkan driver in Mesa for their GPUs, the application processors may actually be worth something.

[willmore]

Hey, if they can land a quality Vulkan driver in Mesa for their GPUs, the application processors may actually be worth something.

A stable driver from Imagination? +1 Funny

[Awesomeness]

Oh great, if Imagination Technologies gets on board, RISC-V is dead....

[StarterX4]

Catapult SDK will support Windows

Windows for RISC-V? Would be interesting.

[uid313]

While it is exciting with a company making RISC-V CPU and SoC, it is hard to get excited when that company is Imagination Technologies.

[ms178]

A Chinese GPU brand uses their newest GPU architecture with multiple chips and if they decide to compete in the consumer segment again, there is currently a lot of unmet demand to get some sales. More competition is certainly a good thing, but my last GPU from them was a Kyro II and got a bit burned with mediocre driver support and I have my doubts that they can afford enough talent to compete successfully with the big GPU companies (plus Intel soon). As they have focused on the low-power chips, RISC-V seems to be the way to go for that market as CPU ISA. But I wouldn't expect them to lead the way in terms of high-performance designs.

[tuxd3v]

Until recently, a lot of RISCv extensions were only drafted and not on paper, very recently a lot of them were approved..
RISCV born from the desire to create a simple cpu to teach kids in the university.. yes it come a long way more than that.. but it still has some weak points..

Microcontrollers, is were riscv will be more competitive at the moment..
Microcontrollers are already being shipped in Western Digital, Seagate, Andes, SiFive, Gigadevice, Imagination, Mikron( already being produced but will only enter market in 2022 ), STMicroelectronics is already assembling a team, just only for that, and probably a lot more companies do the same..
And some of this microcontrollers already include or AES or GOST or Chinese accelerators, for cryptography..

We can take already the Case of Gigadevice, that delivers the double of performance, while consuming only half the power of a ARM cortex m3..

Now, because RISCV was made initially for academic purposes, they left a lot of parts unattended, like the interrupt controller that is very basic in RISCV, and usually each vendor needs to implement their own superior solution. Well this will lead to fragmentation since each vendor will provide each own.. I hope RISCV foundation take more care of details, the current interrupt controller is garbage, ARM own interrupt controller, is light years ahead of Riscv one..

High performance application cores, will be more difficult to get, and they will have to compete with cpus on the market that have 10-20 years or even more of development on them..but code density wise, they are in the same ballpark, for what I have seen already, even tough in some cases they lose a bit, to ARM, but they gain marginally in other places..

The problem will be in Standardization, in my opinion, since some components does not deliver in terms of specs, the interrupt controller is one of those sever cases..
Also their are some lack of instructions on some extensions that would maybe make sense, but were not included because you can use others to complement the ISA.
They assured that are their still a lot of space, that can be used to encode new instructions, so on that side its more or less good.

I also need to check their new approved extensions to see if something really god comes out of them..

But for me, the biggest place were RISCV can shine is the embedded/Low Power one.. here the price of microcontrollers are very low, and RISCV has the capability to create microcontrollers that are very small in size, and so they have a lot lower cost to produce..only this aspect will be a great factor, if you add to it the compressed extension, you are in a good position to win the market.
ARM will lose the microcontrollers market, right now, its only a question when that will happen..

[tildearrow]

Imagination + RISC-V = Big No

[lyamc]

Not sure why so many people are complaining, this looks great!