Is it important if XMOS is a MCU or something else?

Off topic discussions that do not fit into any of the above can go here. Please keep it clean and respectful.
ale500
Respected Member
Posts: 259
Joined: Thu Sep 16, 2010 9:15 am

Post by ale500 »

Jonathan: they sort of banned Leon of saying XMOS there ;-)... and I started using YMOS instead so they would not complain...

What is great about the propeller is its friendliness to the hobbyist to get it up and running. XMOS is a much more powerful option with tighter requirements... in tiny packages ;-)


User avatar
octal
XCore Addict
Posts: 228
Joined: Thu Jan 27, 2011 3:30 pm
Location: Argenteuil - France
Contact:

Post by octal »

I agree with Folknology about the comparison (BOM and hard) MCU vs XMOS, and I agree with Jonathan also.
But what we should take into account is that there are no MCU that does 400MHz are sold less than $5 (unless it's sold in qtt of million).
XMOS is not done for low priced solutions, it's not suitable to make general purpose low cost solutions. When it comes to make a little board that read a sensor and send data to PC, XMOS is not the ideal solution for such as you can do that with almost any extremely low cost MCU.
I think that conditions when we need Extreme computational power (like signal proc, filtering, ...) XMOS can find its place easily as this kind of solutions are never extremely low priced.
When some companies use XMOS for audio processing, the final product price will not be weighted too much by the XMOS chip.
I think that most of the time we will find XMOS chips in products where final product price will make the XMOS $5 almost insignificant!

XMOS is not there to replace low cost MCUs as it have too much power to be underused. And when a product need too much power and so much capabilities, its price is generally so high that the XMOS price will not be siginificant!

regards
User avatar
Interactive_Matter
XCore Addict
Posts: 216
Joined: Wed Feb 10, 2010 10:26 am
Contact:

Post by Interactive_Matter »

Thanks folknology for you answers. I can personally understand and live with all of them. Only that is is a bit unnice:
Folknology wrote:Interactive_Matter there are also a number of limiting factors in the XS1 vs MCU position to be considered during the design phase

1) The majority of shipping MCUs contain on board rewritable ROMs for code either EEprom or Flash. The XS1 currently has only a limited 8K OTP (write once). This often means adding an external flash/eeprom part to the BOM.

2) Because of above there is a 2nd side effect, code and static storage in flash/eeprom is not memory mapped, thus it must be copied to the SRAM at runtime reducing the memory available for actual processing. A good example of this is when running large code libraries such as xTCP/IP which leave very little left for your actual application.

5) All of the above increase cost in parts, layout and testing putting XS1 at a competitive disadvantage from the out.

regards
Al
The high BOM and the somewhat lack of a nice starter pack (XS1-G4 with only the neccessary) is missing - but there are a number of projects in the making solving this problem.

I was not aware of the RAM problem - and this is a major one. So I want an external RAM interface which is considered in the compiler (wouldn't it be great to use a default SRAM interface or similar and tell the compiler how much it is and you directly got more RAM?

I think the msot important views on that topic came from jonathan:
jonathan wrote:All customers have an interest in a company's positioning. It's normal. If you don't care you may find yourself in a lot of trouble. Imagine XMOS went out of business for whatever reason tomorrow and you could no longer buy the chips. One of the main reasons young companies fail is poor marketing. If you want to just be an engineer and ignore this, that's fine, but make sure you have someone who understands this working with you.

Rubberneck: your reference to the Parallax forums/community is dubious. The community there does care about "marketing" - certainly market position. A great example of this is the bickering that went on when XMOS came on the scene and Leon started mentioning it in the Parallax forums.

Another example is here:

https://www.xcore.com/forum/viewtopic.php?f=18&t=1105

There are people from very big companies also very interested in "positioning" and "roadmap" - public statements of intent and direction.

Positioning is a mechanism of signalling to both current and future customers who you are and what you want to be. By positioning itself as a microcontroller, XMOS positions itself as competing in one of the toughest electronics markets in existence - one where there is already true competition and where the incumbents understand their market incredibly well.
Beside the point that I share his oint of views in the parts I left out. I think my question came from the unclear messaging XMOS seems to be using. I asked myself: What kind of beast is the XMOS and got no clear answer from the web site.

So I think it could help a lot of the marketing messaging would be more clear.

My personal point of view is that XMOS is a (massively) parallel processor:
  • It has more constrained (no external components higher BOM) and higher cost than a complete microcontroller - therefore it is just a processor.
  • It is highly parallel requiring a slightly different software archiecture for you solution. So you have to redo a lot of things. As a benefit you get very high performance and very easy system design.
  • It can compete with FPGA in some areas (done properly FPGAs are presumablystill faster) without the pain of an FPGA. As stated above it is quite hard to implement something in XMOS that is not working at all and the platform and tools make it quite easy to improve your solution over time. So it is good for a first prototype and can be refined until your software is production ready
So bottom line: If you want to invest in some external components XMOS gives you a real high performance platform which is realy joy to work with. You got more output than you can dream of and addign external components does not slow you down in any way.

Ok there an always be more 1bit ports - but let's hope for further silicone releases.

I think we still need a broader soultion base (read more prefabricated and tested software modules for default application) - but we are working hard on this on github.

But it take some time for me to realize what XMOS realy is - and I think it is up to the marketing to make the messaging so clear that I do not need the time ;)
User avatar
tautic
Member
Posts: 11
Joined: Sat Jun 05, 2010 4:10 pm
Contact:

Post by tautic »

Folknology wrote:Interactive_Matter there are also a number of limiting factors in the XS1 vs MCU position to be considered during the design phase

Compared to and competing with existing MCUs, XS1 selection can be at a disadvantage because :

1) The majority of shipping MCUs contain on board rewritable ROMs for code either EEprom or Flash. The XS1 currently has only a limited 8K OTP (write once). This often means adding an external flash/eeprom part to the BOM.

2) Because of above there is a 2nd side effect, code and static storage in flash/eeprom is not memory mapped, thus it must be copied to the SRAM at runtime reducing the memory available for actual processing. A good example of this is when running large code libraries such as xTCP/IP which leave very little left for your actual application.

3) Although many MCUs come standard with ADC and or DAC not such option is available with XS1, this again leads to an external chip addition and increased BOM as well as adding to layout complexity.

4) XS1 unlike most MCUs requires dual supply regulation 3.3v for IO and 1v for the core, this adds complexity and BOM count as well as layout.

5) All of the above increase cost in parts, layout and testing putting XS1 at a competitive disadvantage from the out.

regards
Al
You make a lot of valid points here. Going back to the original post question, "is it important if XMOS is a MCU or something else?" I think it is absolutely critical, especially from a marketing/positioning standpoint as many have already pointed out. If it's not an MCU (which I don't think it technically falls under due to the lack of on board memory), it won't be as successful being marketed as one IMO. Is it as good as an MCU? Sure, perhaps better than a lot of others out there - depending on what it will be used for.

One of the things that originally drew my interest towards these "things" (until we know what to call them this will have to do), was their ability to be customizable to your application - i.e. they're great at data processing, and parallel processing. Sure, it doesn't have A/D, CAN or a lot of other peripherals, but if I needed a 12 bit A/D converter and there was only a 10 bit on board it wouldn't meet my needs anyway. I'd rather design my circuit to contain the core components to complete its designed purpose as a whole.

I googled the definition of "Microprocessor", and the first result is pretty definitive, and sounds like it has potential for use here: "http://wordnetweb.princeton.edu/perl/we ... oprocessor" (from http://wordnetweb.princeton.edu/perl/we ... oprocessor). In my opinion, XMOS has the capability to be positioned quite nicely right between an MCU and an FPGA, it a real sweet spot.

If I had to sum it up, XMOS silicon is/are a straight to the point, no frills, high performance (parallel) processing unit. No overhead like CAN, USB, A/D, D/A, Ethernet, UART, etc that you may not need.

Regards,

Jayson
User avatar
shawn
XCore Addict
Posts: 238
Joined: Thu Dec 17, 2009 5:15 am

Post by shawn »

Hello All,
Long time no post...
The Answer is of course..........
Far more than a MCU. Xmos is a displacing technology
And displaces MCU's, CPU's, Fpga's, DSP's, plus a load of other typeO electronics.
As far as positioning go's @approx 5$/core you have 1/2gigaMIP slicable to eight event and or interupt driven threads.
And these are not ordinary threads either. They are channeled and wounderfully, stacked and very reentrent.
It's the Instruction set with its tricked out XC macros that are the real bonus.
Its a streaming proccessor and there are not many as general or as universal as an Xmos Core.
what MCU can MIMD very well?
what CPUs stream very well?
what Fpga's have instruction sets or macros? does an fpga with an embedded proccessor or MCU change
the fact that the're really just great blanksheets of paper, fancy ram looking for IP.
What DSP are universal," not many"...
On chip ram has been so keenley designed into the archetecture with switching channels for IO it would not be
a big deal to design onboard ram that could be paged or setup for say a VLIW or content addressed.
It's not a UniC, im mean Unix, and yet it can thread C or C++ better than a SPARC.
Hmmmm...
MCU's CPU's FPGA's DSP's TYPE0's will serve the XMOS core well. None of could replace an XMOS core well.
ITS a very raw and VAXie archetecture for computer science that has exscaped from the lab.
XMOS and their developers are the coolest, a class act...

Shawn 8-)
User avatar
matrix
Active Member
Posts: 62
Joined: Sat Sep 17, 2011 12:05 pm

Post by matrix »

Hi all,

obviously it DOES matter a lot. Otherwise how to explain more
than 1500 post views ?? :lol:
Post Reply