Major manufacturers -- like Intel, AMD or ARM -- have spent years upons years developing methodologies, technologies, micro-architectural optimizations, and testing+development+evaluation infrastructure. This is the key for understanding how to navigate and ensure good performance across the many different trade-off and optimization domains in performant-processor design.

This is not easy to acquire or build. Some is wisdom from decades of design and iteration. Some is hundreds of thousands of engineering hours. Some is big money.

RISC V is a new open source project. Who knows how close they will ever get.

I suppose it's true for currently available chips that implement the RISC-V ISA.

Nothing says the ISA itself is a barrier to performance on par with popular existing processors though. The RISC-V BOOM implementation is supposed to be close to an ARM Cortex A9 in performance.

> Can anyone say thy the performance of RISCV is so lacking?

"Because no one has manufactured a high performance RISC-V implementation yet" isn't answer enough for you? All that exists for purchase at the moment are microcontrollers aimed at the ARM Cortex-M market niche.

There's nothing about the ISA that says you couldn't make a deeply pipelined, six-way issue implementation with three levels of cache running at 4 GHz. But that fact doesn't make such a machine appear from nothing either.

Many good answers already, but I want to mention RISC-V Boom (superscalar, out-of-order), which seems to, in theory, beat ARM Cortex A9 quite nicely:

https://www.youtube.com/watch?v=HVPSlS2v1F0

The whole toolchain they are working with looks quite awesome tbh.

Commercially, the Freedom U500 platform seems to be really interesting:

https://dev.sifive.com/documentation/freedom-u500-platform-b...

Nobody has invested the money and work to make it happen.

You can take RISC-V ISA specification or ready Verilog/VHDL and produce working ASIC chip relatively easy, but it's not fast.

If you want fast general purpose chip in that competes toe-to-toe with AMD and INTEL, it will take huge amount of chip and physical design, simulation and verification. AMD has spent tens of millions to get new x86 ISA compatible architecture out. Doing the same for RISC-V without enough demand would be economic suicide.

Likely because there's no major consumer devices shipping them (or at least high-end versions of them) that could help them hit a scale that brings the cost down, which makes them less viable for a general public consumer standpoint, which means there's less time spent optimizing it.

I remember a wihle back when Google was shopping around for Intel replacements (likely a negotiation tactic), people were saying they should buy the POWER division from IBM (IIRC). That would have been really interesting...

>"I remember a wihle back when Google was shopping around for Intel replacements (likely a negotiation tactic), people were saying they should buy the POWER division from IBM (IIRC). "

Funny, I was speaking to some IBM engineers a few months ago and brought up the POWER chips and they kind of laughed and said something to effect that biggest use case for POWER was Google as a means of keeping Intel pricing in check.