SEA C18 Forth
deSilva
Posts: 2,967
A little bit OT but we do like to comment on other people's parallel architectures don't we?
This is a 24 core FORTH machine with unclear price tag ($20 p.1000?), based on a 18,000 MIPS 18bit SEA C18 chip.
www.intellasys.net/templates/trial/content/SEAforth-24A_Data_Sheet.pdf
It was announced nearly 2 years ago and could have looked to the general audience like a Propeller double
some reference
The company had to change its name BTW, as Intelasys sounded as "Intel Assist", so its IntellaSys now
www.intellasys.net
Post Edited (deSilva) : 3/7/2008 9:39:29 AM GMT
This is a 24 core FORTH machine with unclear price tag ($20 p.1000?), based on a 18,000 MIPS 18bit SEA C18 chip.
www.intellasys.net/templates/trial/content/SEAforth-24A_Data_Sheet.pdf
It was announced nearly 2 years ago and could have looked to the general audience like a Propeller double
some reference
The company had to change its name BTW, as Intelasys sounded as "Intel Assist", so its IntellaSys now
www.intellasys.net
Post Edited (deSilva) : 3/7/2008 9:39:29 AM GMT
Comments
1. The prop has general-purpose i/o accessible by any cog. The Seaforth dedicates i/o to certain cpus on it's periphery. Want that pin?, ask the cpu that owns it!
2. Seaforth is a 1.8V (cpu and i/o) power device in a 100-pin QFP pack so it's bigger and harder to interface. Prop's I/O can drive real stuff real hard.
3. Seaforths C18 cpu core only has 64 words of RAM and 64 words of ROM on the 24A version available. Though it does pack 3 or 4 5-bit instructions/word.
4. It's Chuck's tinkering toy and as such they are rather hard to get hold of because they don't sell them through distributors etc like Parallax do. Sign the NDA etc etc etc.
The 24B version looks more promising but once again these are esoteric and generally unobtainable devices. Prop chips are on the shelf, the information and support are readily available and Parallax did the right thing in releasing a tested and stable device. After all, people might actually want to use them in production rather than just tinker with them.
*Peter*
I think it is interesting that the "parallel paradigm" now rises its head at all places.
Whatever an instruction may be worth , having 18 billions per second of them for $20 is tremendious...
Post Edited (deSilva) : 3/7/2008 10:29:29 AM GMT
This discussion is starting to look like this
http://forums.parallax.com/forums/default.aspx?f=25&m=204280
That was last year July ...
Funny enough some of the people in that thread still are actively supporting this wonderful forum.
Cheers Ron Mel OZ
So a lot is going on in looking for some useful "compromises" between local computing power and inter-communication bandwidth. The Propeller is ONE try (over-emphasizing the first aspect and under-serving the second)
Edit:
The enclosed main architecture diagram shows how it is in some way "complementary" to the Propeller: Its "COGS" are dedicated to specific tasks.. Peter explained that already above..
Post Edited (deSilva) : 3/7/2008 10:57:00 AM GMT
Do these chips look like they are targeted at hobbyists or the little guy? Is there an equivalent programming environment akin to the Propeller IDE? Is there a place on the website that I can click for samples or to order some chips? There is a market for these chips, not in these shallows but some faraway ocean.
The website always ends up directing you to email for everything. They have an evaluation board called the FORTHdrive but they are only available to "qualified" parties. Once again if you would like to be "considered" you can ... email. Sorry, it's a bit too exclusive for me though I am curious about the device as I can envisage applications that do require some real speed and flexibility.
Steven, you mentioned the nine bit counter but you have to realize that the 24A is the smaller brother of the 24B which does/will have 512 words of memory per cpu. As a clockless cpu it only uses a clock to synch to the outside world so the silicon will run faster or slower depending upon the temperature and the phase of the moon and it does run at around 700mHz/cpu. Clocks are just used to simplify the design of digital systems to avoid metastable and race conditions etc. Part of the reason that the SEAforth runs so fast too is the silicon process and operating voltage. If the prop were to be fabricated in this same process it would run a lot faster than it does but this creates other problems as well. The trouble with 1.8V I/O logic levels is that things like leds would need to be buffered and driven from higher voltages as would many other components and devices.
You are right DeSilva when you mention about useful compromises and when it comes to inter-communication the prop is sadly lacking. But bear in mind that this is Chip's first chip and it is a fantastic chip whereas Chuck has been tinkering with silicon for quite some time now but unfortunately us poor mortals can never get our hands on any of it.
I'm very happy with the Propeller, it's put the fun back into processors, and I am really looking forward to Prop II as well.
*Peter*
www.intellasys.net/phpBB/viewtopic.php?f=7&t=63
To me the Propeller was straight-forward to understand and adopt once I got my head round the fact it was simply just eight CPU's, some shared RAM/ROM, and each micro booted an interpreter at reset. It's "evolution" not "revolution" and no different to use than a set of single micros are -- That's not to be dismissive of the superb way it's put together as a whole (*)
SEAforth seems to be a whole new ball game, not least that it uses Forth which is alien to me. The closest I can visualise it as is a collection of Transputers on a single chip. To me, SEAforth requires a completely different design paradigm. I have to admit that I cannot see what market it is aiming for, or what purpose this massive parallelism is designed to serve, given that each 'CPU' seems quite constrained.
(*) I'm quite saddened that the Propeller hasn't achieved the market penetration I think it should have, especially outside the US. It was definitely at the top of my list for "best innovation of 2006/7" and something I think Parallax should seriously set about addressing.
I read some hours through the data sheet and I think I understand fully what it is meant for and how useful it can be for a lot of algorithms... I should not say: "Give me one, and I program it", but it would not take more than two or three days... Maybe I know a lot of "alternative" architectures to sort such things out easier... Of course it will not run "out of the box", it is "building block", whereas the propeller comes closer to an SOC...
But as Peter said: Its market is not us!
I just read that Renesas just deemed it essential to announce another 20 different versions of there M16C controllers: That's the market!
The Propeller is not underestimated! It just has not the attractiveness to the masses, and for good reasons.
The same thing applies to Tilera's TILE64: which is a most excellent design!
Post Edited (deSilva) : 3/7/2008 3:25:59 PM GMT
It's really more that I cannot visualise the application usefuleness; and you perhaps have the Forth advantage over me
I don't personally see the lack of attractiveness. To me it's more missed marketing and promotional opportunities and lack of cost-effective sales channels. Perhaps people need to say what the perceived lack of attractiveness is ?
Perhaps I'm too 'old school', used to picking up a new processor, having few tools to work with and having to make things work rather than only taking on new processors when they meet criteria I impose on their adoption. Taking what I'm given, rather than rejecting until what I'd like is provided. That also explains why I don't see the fuss or need over Propeller C compilers.
Hippy! Your analysis is to the point and your description is disclosing! You are the best example of a person who will not use the propeller if he wouldn't already!
You also gave the best pointer possibiliy to IntellaSys: What they say sounds familiar:
In Germany we have an influential TV program "ComputerClub2" and an influential elctronics magazine ("ELEKTOR") Both are not against the Propeller, on the contrary. There had been a small chance that they could make a common series "How to use the Propeller..."
Of course they didn't. Though I shouldn't I am deeply disapointed.
But such things will repeat. I have a vivid imagination and a good (though shrewd) knowledge of people... Will the BBC make a series about the Propeller?
P. "John, let's make something with the Propeller. That will interest people."
J. "Peter, I think we'll get more viewers when we take the Atmel chip."
P. "Right, I still have some material from the 90ies that we can use then."
J. "Also, I know a person who will sponsor that, and he makes his money with Atmel."
--- end of OT ----
Post Edited (deSilva) : 3/7/2008 5:15:10 PM GMT
http://www.falvotech.com/blog/index.php?/archives/200-Forth-Day-Report.html