A man with ...Ideas (troublemaker ?? :-) )

XCore Project reviews, ideas, videos and proposals.
Post Reply
vanjast
Member++
Posts: 30
Joined: Sat Jan 16, 2010 9:57 pm

A man with ...Ideas (troublemaker ?? :-) )

Post by vanjast »

Going thorugh all this over the past few months, I started thinking (miracles do happen :D ).

The main problem I have with the CPUs is the BGA pin thing. It has it's advantages, but for 'joe-soap' it becomes a mission, possibly plagued with unseen dry joints only visible by X-Ray inspection.
This I think will restrict interest to mainly the L1 and L2 series...

So enter the 'troublemaker'
How about a nice reduced pin QFP packaged quad core chip. If you think about it, this type of cpu doesn't really require more than 40 pins, plus Pwr & Cfg. The 40 pins can be used for I/O, but their main purpose is for fast links. Also make the chip real estate larger, it's not that much of a big deal and heat dissipation is important at higher speeds (Sorry image error... LCC = QFP )
Image

Another thing lacking is extended Memory.. always need more. How about another reduced pin QFP with a dedicated memory I/F
Image

.. and lastly, generic I/O Extender for I/O control... reduced pin QFP of course :shock:
The L1 or L2 can fill this gap
Image

These ideas are not new and would have come out with the T9000... many years ago.
Reviving these old ideas again would definitely take Xmos beyond where the Transputer died.
Maybe to Infinity and Beyond ;)


User avatar
Folknology
XCore Legend
Posts: 1274
Joined: Thu Dec 10, 2009 10:20 pm
Contact:

Post by Folknology »

Some good ideas, the last one is something I thought about as a protocol I/O adapter and could be packaged in 24/32/48 pin packages. In my version the boot was internally locked to boot over link only, this also enabled pricing to be kept low without dragging down the prices of other L1 packages as protocol adapters could only operate in conjunction with other L1s.

They all have to be L1 core based though otherwise they can't easily talk together.

regards
Al
User avatar
Bianco
XCore Expert
Posts: 754
Joined: Thu Dec 10, 2009 6:56 pm
Contact:

Post by Bianco »

If you need four cores you problably already have a complex board with 4 or more layers and BGA's won't be that much of an issue (that in mass production). It is nice to have 4 cores in an easy to solder package for hobbyists but will it sell enough to get a return on investment? Is there enough interest from larger manufacturers?
vanjast
Member++
Posts: 30
Joined: Sat Jan 16, 2010 9:57 pm

Post by vanjast »

From what I can gather, the work load input for the current BGA device is far greater than what it would be for, say a 80pin QFP with wider pin spacing. Wider pin spacing = Larger chip but the clock speed is going to increase in the future, so having a larger chip will help with heat dissipation. We have for arguments sake, a 20mm squared chip instead of a 10mm one. I'll take the 20mm any day knowing it'll survive better in rugged environments.

Not only will the 80pin QFP make layouts easier, but more appealing to the hobbyist/user, resulting in a lot more interest and projects. This has a cumulative effect and manufacturers will follow the demand.
BGA demands a lot more layers and with most high performance QFPs you can get away with 4 layers with careful planning... reducing costs significantly. BGA is great for big companies who have the tools, $$ and means to work with it, but how many big companies are there and using Xmos. I think Xmos must look at base levels first to get their products established. The L1/L2 is a start and from what I see, generating more interest, although I don't profess to know the whole story ;)

Just thinking multifaceted...
:)
User avatar
franksanderdo
Experienced Member
Posts: 69
Joined: Sat Apr 30, 2011 6:19 pm

Post by franksanderdo »

Hi & Hello,

so we are min. 2 troublemakers :lol:

as I am totally new to the Xmos chips, I am more reading than saying some thing. :?
I was struggling over the BGA chips as well (not only the Xmos ones).
To over come it I have started a chat with a PCB manufacturer and we are thinking about some BGA adaptor boards. The Idea behind it is exactly the same you are discussing here:
Make silicon with BGA Packages available for "hobby".

We are only in the very beginning of collecting Ideas how we can do it best (cheapest) and most flexible for different kind of silicon. One of the issues we are facing is that in most cases a minimum circuit needs to be done.
Even though, we do not really look into producing "stamps".

As soon as I understood a little more about the difficulties and differences of L and G Xcores I will open a related Project here.

See you around
Frank
Post Reply