Looks great! It's a good thing to heat up the competition. ARM makes good cores but too many engineers (or their managers) put too much emphasis on "wanting an ARM core". Truth is that with different peripheral sets between ARM chips and compilers taking care of differences between cores, there really isn't much difference between porting code from one ARM chip to another and porting code from an ARM chip to another core, at least for RTOS and bare metal systems that these smaller cores tend to be used for.
I learned of a new core today with this article and was surprised! I thought I knew most of the players but I have to confess to not even knowing anything about Andes until reading about it here. It is nice to be surprised (once in awhile anyway). Thanks for bringing me up to speed!
To have a look at their website it appears that they still thinking larger lithography by using clock gating. I do not see power gating mentioned which seems to say that they have not gotten to the current technology in their IP.
As anyone who evaluates CPU core technology or chip/SoCs knows it will be difficult for anyone to build the ecosystem needed to challenge ARM (well unless they are Intel). So, I look forward to seeing Andes replace the 8bit systems which can reduce much of strain on those low end chips I cannot be certain that it will replace ARM in the nearer term (unless ARM chooses to ignore the challenge or EOL some current products).
In the very low end MCUs where this will get used, it is more than just the CPU - it is the subsytem around it including interrupts, low power features, interface to embedded flash and the software tool chain and ecosystem. ARM is hard to beat in this area.
The only advantage as Linley points out to Andes it the cost. This will keep ARM honest and that is always good for the market.
I'm heavy ARM user, but I think the success key is the easiness to understand then to use, and the price. We already have many option about performance, power and extendability to choose core, not saying only ARM. Old engineers tend to use their favorite core. Almost young engineers like new cores, normally student. So Easiness. And they should survive until these young engineers grow up. So Price. I think ARM Cortex series is already too complex, even for M series.
The key part for me was “Given what ARM is charging, I wouldn’t be surprised if you could cut your costs in half or more”
As usual, it's about the money. The Andes core can be had for less in licensing costs than an equivalent ARM core, and there are relentless cost pressures in the semi-conductor market.
Is the Andes core as good as ARM? It doesn't matter. The proper question is "Is it good *enough* for the job to which it will be applied?". It appears an assortment of customers think it is.
I don't see ARM exactly quaking in their boots. They have a commanding position in various segments of the market, so their question will likely be "Do we cut licensing costs to preserve and possibly extend market share?" That will depend upon what they believe the net effect on revenues and profits will be from doing it.
“Given what ARM is charging, I wouldn’t be surprised if you could cut your costs in half or more” using the Andes cores, said Linley Gwennap, principal of market watcher The Linley Group (Mountain View, Calif.).
hmm.. how much is ARM charging per chip.. IINW, the royalties are less than 1~2% which I feel is pretty low already.
My first thought was that if an ARM Cortex M0 can be purchased for under a dollar in quantity, there's not a lot of room for undercutting. But then I thought more about the "Internet of Things."
When connected embedded systems are being added to products that sell retail for, say $10.00, the difference between $0.80 and $0.40 starts to matter. When building an embedded system into something half that retail price, that same difference will likely be the determinant factor as to possible or impossible. I'm real interested in seeing where Andes prices their MCUs.
To Duane: you can get Cortex-M0/M0+ based MCU now for much less than 0.50$ from NXP (LPC800), Freescale (KL series), ST (STM32F0). Even Infineon announced on Embedded World 0.25EUR for their XMC1000. There is not so much gap to go lower ;-)!
Cost, cost, cost.... then code-density will be even more key than licensing cost.
I suspect a couple of kB of extra on-chip memory will quickly erase the 1-2% of royalties metionned above by Eewiz. Andes seems to be the only provider of tools for his processor, can the code size be really be as optimizied as the one of the big names like IAR, Green-Hills and all the others that compete for the Cortex family?
BTW, I find it weak to resume the story about "better DMIPS and lower cost". It's no secret it is easy to tweak DMIPS (which version?, inlining on?) and cost will be surely be all about negociation.
Now please Mr Journalists and Analysts, what about real investigation on what is really important for deep embedded systems?
* certified "difficult to tweak" CoreMark?
* what about interrupt latency or nesting?
* extra debug goodies like trace?
* what about family concept and compatibility with the other Andes processors?
Please don't treat embedded topics as you would cover apps processors. Catchy title is always good, content is even better ;-)!
Considering that there are free cores available, Andes has to offer a compelling value, presumably in system integration, peripherals, software tools and other ecosystem facilities. It'd be interesting to see how well they'll succeed in that---Far Eastern companies do not yet have a big track record in creating open communities and ecosystems.