8-bit uC @ 100MHz?

[miguelvp]

Divide is an expensive operation. Multiplication not so much.

I understand.   But on CM3, CM4, x86, etc, they both complete in the same time...

True, but the CM0 doesn't have a hardware divide. It probably takes a lot of fabric on their asic to have a divide that only takes two cycles. I'm not sure if the standard CM0 core has even a hardware multiplication, I know that in for example PSoC 4 series Cypress added a 32x32 hardware multiplier.

Edit: just looked at the C0 TRM and MUL takes 1 cycle to 32 cycles depending of the hardware implementation. DIV is no where to be seeing.

[edavid]

Divide is an expensive operation. Multiplication not so much.

I understand.   But on CM3, CM4, x86, etc, they both complete in the same time...

This is not true, divide is slower, especially on most X86 implementations.

[miguelvp]

It's also slower on CM3 and CM4 but just by one cycle, 1 for MUL 2 for DIV but it's so insanely fast that didn't want to point it out.
on x86 at least on the FPU, the divide is 6 time slower than the multiply for integer operations, maybe they are more in par for floating point or maybe he was referring to the non FPU math which I didn't look t the cycles.

Of course it got way faster after the Pentium, before that things were a lot slower even with the math coprocessor.

[free_electron]

8051 is license-free.

M0 is royalty free now.

Royalty free yes, after paying 40k to get i to the arm club...

[f4eru]

Royalty free yes, after paying 40k to get i to the arm club...

Good price

[coppice]

8051 is license-free.

M0 is royalty free now.

Royalty free yes, after paying 40k to get i to the arm club...

It is actually royalty free? The publicity about the free for pre-commercial work and $40k to get up and running deal doesn't seem to clearly state that there are no ongoing royalties.

[coppice]

I have encountered single cycle implementations of the 8051 running at 480MHz. I have encountered Z80 clones running at 1GHz. Semiconductors is a big industry, and there is a lot of weird and wonderful stuff out there, mostly existing in some niche most of us never notice.

[tszaboo]

8051 is license-free.

M0 is royalty free now.

Royalty free yes, after paying 40k to get i to the arm club...

It is actually royalty free? The publicity about the free for pre-commercial work and $40k to get up and running deal doesn't seem to clearly state that there are no ongoing royalties.

Well this is the official statement:
https://www.arm.com/about/newsroom/arm-offers-free-access-to-cortex-m0-processor-ip-to-streamline-embedded-soc-design.php
40K is nothing I guess, unless there is a hidden catch, like it is a price per seat or product. IDK, I'm not in the semiconductor business.

[JoeN]

8051 is license-free.

Oddly, the mentioned SILabs part is like $20 in small quantities.  How's that license-free working out for the user?   

[abyrvalg]

We have a 250MHz 8051 extracted from a cheap USB flash drive - Phison PS2303. The original firmware even runs >64K code by demand-paging it from NAND.

[dannyf]

Many of the highly specialized 8051 parts tend to be far more powerful than a generic 8051 and at a far lower price. You just have to buy them in the millions.

[jesuscf]

How comes my EFM8UB2 runs only at 48MHz?  What SFR do I have to change to make it go at 100MHz?

[matseng]

How comes my EFM8UB2 runs only at 48MHz?  What SFR do I have to change to make it go at 100MHz?

Why would you think that all mcus from Silabs can run at 100MHz?

The datasheet clearly states that the EMF8UB2 have a:
Pipelined 8-bit 8051 MCU Core with 48 MHz maximum operating frequency

[jesuscf]

You are right, I got confused by this sentence from the article: "but we have an 8051 based device that runs at 100MHz and a 72MHz part in our EFM8 range".   The 72MHz part is the EFM8 Laser Bee Family.

Why would you think that all mcus from Silabs can run at 100MHz?

The datasheet clearly states that the EMF8UB2 have a:
Pipelined 8-bit 8051 MCU Core with 48 MHz maximum operating frequency