lkcl
Posts: 18
Joined: Tue Dec 27, 2011 7:42 am
Contact: Website

Re: if its good copy it?

Tue Dec 27, 2011 7:50 am

Jongoleur said:


@Wooloomooloo:

As you say, its a module for a different purpose to the Pi,


no it's not.  the first module is stand-alone, could run the exact same software as the rbpi and do exactly the same job.  it *also* happens to be possible to plug it into any EOMA-PCMCIA-compliant motherboard, for example mass-produced laptops, tablets, STBs, IPTVs or low-power rack-mounted servers.  in this way, the chances of reaching the mass-volume pricing levels are increased, and everybody else benefits by being able to ride on the back of that mass-volume pricing "wave".


however, the Rhombus-Tech website ( http://rhombus-tech.net/ )specifically mentions and knocks the Raspberry Pi.


yes that was the slashdot moderators "hyping" the post.  i did not put the words "steal" or "thunder" into the original slashdot submission.  little toe-rags.




lkcl
Posts: 18
Joined: Tue Dec 27, 2011 7:42 am
Contact: Website

Re: if its good copy it?

Tue Dec 27, 2011 8:11 am

Jongoleur said:


Quote from tntexplosivesltd on December 18, 2011, 08:11
I don\"t think it has a PCMCIA connecor, it\"s jst PCMCIA-sized.

Have a look at the FAQ page (linked to from the middle of the Rhombus main page), they\"re clearly thinking about using PCMCIA.

http://rhombus-tech.net/faq/

see also

http://elinux.org/Embedded_Ope.....ion_1.0.29

Its clear the intention is for a \"future proof\" intelligence module for consumer electronics.


absolutely.  the pace of change right now is just staggering.  the rpbi - and every single device which has a fixed CPU from a fixed supplier - is in danger of becoming irrelevant within under 4-6 months.  ok, that's assuming that the motivation is profit [which it isn't in the rbpi case, it's education].

six months ago the leading CPU was the AML-8726-M Cortex A9.  $15.  one year ago it was briefly the motorola IMX53 Cortex A8.  $15.  eighteen months ago it was the TCC8901 ARM11.  $11.   two months ago the Allwinner A10 caused a temporary major recession in Shenzen as _everyone_ cancelled orders en-masse and went for the A10.  Cortex A8, 1ghz.  price: $7 (seven) in mass-volume.

that's _going_ to happen again [inside information witheld].


Referencing Raspberry Pi is just an attention grabbing, \"look, we\"re faster and have more RAM\" ploy. They\"re just talking up a storm and piggybacking on some Chinese proposals for a modular PC board.


yeah - did it work?





Good luck to \"em, but I\"m not holding my breath. Remember the problems with the Globescale Guruplug systems? Sold on pre-order (it took 6-7 months to deliver from first orders taken). Hurried into production. Under developed. Prone to overheating and PSU failure when used in the real world. I think they\"ve been withdrawn after about a year on sale (the UK distributor isn\"t listing them any more).



yes, we learned from that one.

a) don't use a marvell CPU which has gigabit interfaces that cause the CPU to overheat when you actually start using them in anger.

b) don't put in a stupid low-cost switch-mode PSU that's only 25% efficient.

one of my associates used to work for LSI Logic, and then Samsung.  he knows his stuff, here.  he says that if heat dissipation becomes a problem, you flood the PCMCIA case with thermal gel, to get a good contact, and if that's not enough, a beryllium spring can be added which draws heat down to the main PCB.  anyway this is a max of 3.5 watts we're talking, in a metal-shielded case, not a lump of plastic like the guruplug.


Unless Rhombus\"s Chinese partners are going to do a lot of the design heavy lifting, I don\"t see any progress with this particular proposal for at least a year (being charitable).


ah, right, ok, what you may not be aware of is that the SoC vendors - all of them - provide "Reference Board" schematics.  full schematics.  unlike in the USA (with the exception of boards like the beagleboard, IMX53QSB, Origen etc.) there is no "we are not going to give you schematics in case you get it wrong and sue us" attitude.

in other words, there is a *working* set of schematics *already* in the hands of the little factory we're using.  all they have to do is cut/paste the relevant parts off of that - remove all the Evaluation Board's peripherals - and the job's 90% completed.

we would be absolutely stupid to have picked a partner that wanted to do this entirely from scratch.  if we had gone down the "let's do this entirely from scratch" route, you would be absolutely correct: the project would be at least one year out, even before it had begun.

also we have on the arm-netbooks list at least one hardware engineer who has offered to do a schematics review, and there's one more whom i think could probably be roped into doing a review as well.



Prometheus
Posts: 308
Joined: Tue Dec 13, 2011 11:09 pm

Re: if its good copy it?

Tue Dec 27, 2011 8:14 am

Hey lkcl! Nice to see you over here!

lkcl said:


Jongoleur said:


@Wooloomooloo:

As you say, its a module for a different purpose to the Pi,


no it's not.  the first module is stand-alone, could run the exact same software as the rbpi and do exactly the same job.


I'm guessing that Jongoleur might have been referring to the fact that the RPi is from an education-focussed charity, whose primary goal for the device is getting it into educational settings, rather than disputing that it could do the same job.

Well, that's how I read it, anyway...

lkcl said:

yes that was the slashdot moderators "hyping" the post.  i did not put the words "steal" or "thunder" into the original slashdot submission.  little toe-rags.
Wow, that is disgraceful... Makes me glad that I stopped reading Slashdot quite some years ago, really. I don't like that style of editing at all...

lkcl
Posts: 18
Joined: Tue Dec 27, 2011 7:42 am
Contact: Website

Re: if its good copy it?

Tue Dec 27, 2011 8:29 am

JamesH said:


I\"m quite surprised that someone thought \"we could do that, better and cheaper\" which appears to be the aim.


yup.

actually, the aim is to split out motherboards + CPUs so that each time a new CPU comes along that disrupts the market (see other post just earlier), our client's won't get shafted by having to abandon a half-designed product or worse a fully-designed one through broken contracts where the components have already been ordered.

Software (Libre) Developers get to ride on the back of that.  actually, they're kinda crucial to the whole thing.  ok, you can read the front page of the rhombus tech web site as well as anyone



The Raspi is cut down to the bone, pricewise, in the market, at the moment.


what's the CPU price, then?  i estimate it to be around.... $12 in mass-volume.  even if it's $8 (which it almost certainly won't be), that's still $1 MORE than the 1ghz Allwinner A10, which, because of its ultra-high volume in China, can be sold at $7 in mass-volume.

then, there is the RAM.  what RAM ICs can the rbpi's CPU take?  can it take the lower-cost DDR3 RAM ICs which are lower-cost because of x86 PCs moving those ICs in a mass-volume "wave"?

paradoxically, the price of smaller (older) RAM becomes *more* than the faster, larger RAM ICs thanks to mass-volume ordering.


Not sure how people thought they could do better against a charity with no profit motive. Yes, there will be better chips out for the same price in a years time. Not now.


that's factually incorrect.  actually there have been better chips out for *less* money, for at least... *thinks*... ok, the Ingenic jz4770 came out 4-6 months ago ($7, 1ghz), the Ingenic jz4760 came out 18 months ago ($7.50, 700mhz), the Allwinner A10 came out 2-3 months ago ($7, 1ghz, Cortex A8), the AML-8726-M came out 4-6 months ago ($15, 800mhz, Cortex A9)

so, apologies, but you are under-informed.


Remember that the BRCM2825 is well over a year old now, so its not leading edge tech, and the price reflects that, and also that the cost of the SoC is not the majority cost of the board.....





definitely correct!  the other major component is the RAM IC.  i'm not completely up on RAM IC prices but you can get 1gbyte DDR3 RAM SO-DIMMs for £12 if you look hard enough.  back-calculate the mass-volume pricing on that and you start to appreciate that yes, $15 is an achievable BOM.

ironically, LCD screens, being a massive silicon jobbie with over a million transistors are the major cost component.  ok, with the exception of these damn capacitive touchpanels



lkcl
Posts: 18
Joined: Tue Dec 27, 2011 7:42 am
Contact: Website

Re: if its good copy it?

Tue Dec 27, 2011 8:37 am

Prometheus said:


Hey lkcl! Nice to see you over here!


:)  yeah only just discovered this discussion.


I'm guessing that Jongoleur might have been referring to the fact that the RPi is from an education-focussed charity, whose primary goal for the device is getting it into educational settings, rather than disputing that it could do the same job.


yehh, acknowledged.  my engineer-driven brain went "wark, wark, missing info equals misleading conclusion"


Well, that's how I read it, anyway...

lkcl said:


yes that was the slashdot moderators "hyping" the post.  i did not put the words "steal" or "thunder" into the original slashdot submission.  little toe-rags.


Wow, that is disgraceful... Makes me glad that I stopped reading Slashdot quite some years ago, really. I don't like that style of editing at all...


achh, pffh - it's the first time it's happened.  cmdrtaco's retired, and the new people are trying, well... new things.  grr.  never mind.  i'll try something different next time.

lkcl
Posts: 18
Joined: Tue Dec 27, 2011 7:42 am
Contact: Website

Re: if its good copy it?

Tue Dec 27, 2011 8:50 am

Burngate said:


The Rhombus-Tech board is aimed at a slightly different market. However they also see it in our (geek and education) sectors. Also their business model is different. But it\"s interesting that they\"re following the Pi philosophy of trying to be as open as possible about their progress!


*deep breath* and the openpandora, and the openmoko, and the beagleboard, IMX53QSB, guruplug, hardkernel.com odroid, flow g1.5, leaflabs maple, and... i think my brain has imploded with the verbal diarrhea of spitting out all those project names

the lesson from the openpandora: *don't* get your own casework done, and *don't* get involved with R.F. (openpandora paid $50k to an ex T.I. engineer to *fail* to get the WL1251 part of the PCB design correct.  that's $50k of NREs that has to be amortised onto everybody's units, for nothing)

openmoko: get it done, y'all!  release some damn software already, pick somebody else's if you have to, but get at least something that you can do early testing of the PCBs on.  the openmoko failed primarily because they didn't listen to sound R.F. engineering advice, but also they spent so long developing an OS that they couldn't test the boards and find out that the R.F. engineers they weren't listening to were right!  then a couple of the major components went end-of-life...

leaflabs maple: low-speed CPUs and open hardware schematics ROCK!

flow g1.5: modular computing and open hardware schematics ROCK!

beagleboard/IMX53QSB etc: open hardware schematics ROCK!   we have someone who has offered to morph the beaglebone (TI AM3357/8) into an EOMA-PCMCIA form-factor.  that simply wouldn't be possible for a single hardware engineer to even _consider_ offering that, if the full ORCAD schematics weren't available.

so, yes, there are a ton of projects to learn from, here.  and yes, open-ness is crucial.  people are _fascinated_ by being able to get involved, see the nitty-gritty and so on.  as things progress, i will change the focus of the rhombus-tech web site from being one which is designed to attract Software (Libre) Developers - wiki style web site, mailing list - into one which is more end-user-orientated (forums.  gah.  PHP.  wordpress.  i wonder if i can bring myself to do it

lkcl
Posts: 18
Joined: Tue Dec 27, 2011 7:42 am
Contact: Website

Re: if its good copy it?

Tue Dec 27, 2011 8:57 am

sylvan said:


Quote from emercer on December 18, 2011, 05:56
Quote from sylvan on December 18, 2011, 04:38
I do wonder how much it will cost for the breakout board (with the socket for the card and connectors for power, USB, LAN, ...).


I\"m not waiting on one, I\"ll just rig my own breadboarded connectors >D

I\"ve never found a good way to connect to a PCMCIA card. The connectors are too fine for me to solder wire to, and sticking wires into the card has never worked well. I suppose if there was a PCMCIA connector breakout I\"d be good to go...

*nods*.  you want this: http://elinux.org/Embedded_Ope.....eringBoard

the plan is: first PCB is just a bunch of connectors, 2-layer board.  the next PCB is that, extended to include the components of the http://leaflabs.com/devices/maple/

they'll likely be open hardware projects, because, duh, the maple already is, but just putting together a bunch of connectors i mean even i could do that, using gEDA or KiCAD

lkcl
Posts: 18
Joined: Tue Dec 27, 2011 7:42 am
Contact: Website

Re: if its good copy it?

Tue Dec 27, 2011 9:04 am

sylvan said:


Quote from jamesh on December 18, 2011, 19:27
Very true, but I wouldn\"t want to be the one coughing all that dosh in the hope they all sell......!!


Which brings us full-circle right back to why they are taking pre-manufacture commitments...


that was - is - to give the factory confidence that yes, there really are people out there who want to buy and work with them.  china-based factories have absolutely no software expertise.  at all.  it was - is - very very important to the success of the project to get Software (Libre) Developers involved.  we've managed to do that.  there are now approximately 5 (and increasing) people willing to work on the linux kernel source code, ready for when hardware comes out, so that the next level of people (betas) have something to work with, can start tinkering, and it goes from there.


The next message will be about fraud.


*drily* thanks for that.


(which is why I\"m not committing until I see product)


you have a deal.  i'll hold you to that committment.  if you _really_ do not wish to trust, you will always be able to buy units in the supermarkets (at retail prices).  *sigh* that's the plan / roadmap, anyway

lkcl
Posts: 18
Joined: Tue Dec 27, 2011 7:42 am
Contact: Website

Re: if its good copy it?

Tue Dec 27, 2011 9:11 am

JamesH said:


I wonder if the people designing the board will bring out the VGA? or decide there are other more important interfaces to add. Remember, this is a small board and space limited for connectors. The Raspi has the same problem, there are a limited number of connectors so not all interfaces are pushed out.

I\"d add VGA if I were them.



rrrmmmm..... *wavering*.... naah.  VGA is analog signallling at 75 ohms, and thus is automatically power-hungry.  things have moved on: HDMI does the same job (ok, better - higher resolution) and because it's balanced-lines and digital signals, is less power.

the TV-out however we'll likely be putting onto the expansion header.  the reason is: TV-out is less pins than VGA.

it's a real... interesting challenge... picking the right interfaces!  in the RHT case, we have the additional challenge of having to cover a wide range of mass-volume products, in order to achieve the goal of creating that synergy (what a joyous word, that...) between china-based factories and software (libre) developers that we're looking for.

loovely

lkcl
Posts: 18
Joined: Tue Dec 27, 2011 7:42 am
Contact: Website

Re: if its good copy it?

Tue Dec 27, 2011 9:23 am

JamesH said:


I think they would have trouble making a board with two layers and getting all the interfaces out. We needed to use a 6 layer board on the Pi to route everything to the edges.


yes.  i've advised the factory that, because of the tight space requirements, they may have to go to 8-layer, we'll see.  the reference board that they'll be cut/pasting from is 6-layer, but you can see from photos of it (http://hands.com/~lkcl/a10_evb.png) that they have the CPU quite widely spaced apart, and all the DRAM and NAND flash tracks are clearly visible on the 1st layer: we won't be able to do that (higher part density). i did a mock-up for them, here: http://rhombus-tech.net/allwin.....module.png

*but*…. butbutbut – when it gets to actual products, all this is completely irrelevant as far as factories that *use* this PCB are concerned.  any factory – or in fact any Open Hardware Project – that uses this EOMA-PCMCIA-compliant card would be able to create an incredibly simple 2 to 4 layer board, from scratch probably in about one or two weeks.

some of these china-based factories don't have any real expertise – they just do trial-and-error: the labour costs are so low that they can do that   and, because the hard part has been done (the high-speed signals and the critical parallel tracks – DDR RAM timings are a bitch) they are only dealing with lower-speed signals and balanced lines (SATA, Ethernet).  as long as they get that right, they're good to go.

[edit] - so, yes: the main CPU card is 6 maybe 8 layer.  but the rest of the product(s) are, if they actually need one rather than using flying-leads off the expansion header(s), much lower-cost 2 to 4 layer boards.

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 27721
Joined: Sat Jul 30, 2011 7:41 pm

Re: if its good copy it?

Tue Dec 27, 2011 9:47 am

Couple of points, why are you using the Raspi website to advertise your own product?

And secondly, your figures for SoC prices on the previous page are WAY OUT. I'll leave you to figure out where.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed.
I've been saying "Mucho" to my Spanish friend a lot more lately. It means a lot to him.

baybal
Posts: 1
Joined: Tue Dec 27, 2011 12:05 pm

Re: if its good copy it?

Tue Dec 27, 2011 1:08 pm

Hello,


The R-Pi project has been going for 4 or 5 years now. When you aren't getting paid to work on something like that full-time then it takes a while to make it happen. I wish them the best of luck in making it happen, but we all have plenty of reason to be skeptical of their claims. It is a lot of features for a small amount of coin, and like the old man used to say "Something too good to be true, usually is."


The EOMA initiative haven't got to this point without a substantial amount of homework done either. For over two years, the group behind the project have been evaluating all possible solutions and shaping the strategic vision. Our original cause started as an effort to make a design for proper and affordable ARM or MIPS based smartbook. Since then, LKCL approached countless companies of the OEM world. At certain points of time, group had prototype designs shipped to it, torn apart, and rejected after finding critical issues with designs of all previous contenders. After all of this so far, I'm sure that this time, if the sky wouldn't fall apart, EOMA will make it to the market and will be commercially viable.


manufacturing in the UK is more expensive than China. BUT… then you need to come up with 10x or 100x the initial cash outlay.


It"s not always a rule.

First, PCBs would be made at real established factory, that already manufactures widgets for consumer market, with clear commercial goal.

Second, RT is for profit organization, thus are more viable and there are much more confidence from supplier side than if it was styled like mozilla foundation or something like that.

Third, there are pre-orders for prototypes and interest from the industry after all.


that was – is – to give the factory confidence that yes, there really are people out there who want to buy and work with them.  china-based factories have absolutely no software expertise.


I can certify that. Often companies simply pass on hardware down the OEM chain, never seeing the final product, to the company that flashes it and markets it.

I have been watching over all of development going on in Shenzhen area for many years since I first tried my fortune at trading small consumer microelectronics in Russia. In last two or three years growth of microelectronics industry there were especially intensive.

Many of local market pioneers started with modular designs seven or six years ago, because they were seen as cost savers and were more flexible than getting designs at ODM. This trend somehow subdued to to inflow of Western companies that can easily order tens of millions of widgets a year that made designs more mass production friendly and "dumber".

Nowadays, firms are realizing benefits of modular designs again. Much shorter and cheaper design to market transition, independence from ODMs, higher commercial flexibility for legacy products. These are major factors to consider for any technology producing firm, and are potential sources for capitalization.

I know that there are 4 or 5 groups working on COMs in Shenzhen area, none of them advanced even close to a stage where Rhombus Tech is now.

User avatar
johnbeetem
Posts: 945
Joined: Mon Oct 17, 2011 11:18 pm
Location: The Mountains
Contact: Website

Re: if its good copy it?

Tue Dec 27, 2011 7:41 pm

lkcl said:


... as things progress, i will change the focus of the rhombus-tech web site from being one which is designed to attract Software (Libre) Developers - wiki style web site, mailing list - into one which is more end-user-orientated (forums.  gah.  PHP.  wordpress.  i wonder if i can bring myself to do it


You can always follow BeagleBoard's example and use Google Groups for discussion and elinux.org for wiki, so you can concentrate on other things. One thing I find weak with BeagleBoard is that IFAIK there is no central repository for "recommended software", which makes it hard for newbies (and others) to choose which version of which software to install. It will be interesting to see how RasPi deals with this. Since RasPi is eventually targeted to a non-hacker audience, it is a problem they must solve to succeed IMO.

Wooloomooloo
Posts: 92
Joined: Fri Nov 25, 2011 10:52 am

Re: if its good copy it?

Tue Dec 27, 2011 8:12 pm

lkcl said:


JamesH said:

I\"d add VGA if I were them.
rrrmmmm..... *wavering*.... naah.  VGA is analog signallling at 75 ohms, and thus is automatically power-hungry.  things have moved on: HDMI does the same job (ok, better - higher resolution) and because it's balanced-lines and digital signals, is less power.

the TV-out however we'll likely be putting onto the expansion header.  the reason is: TV-out is less pins than VGA.

it's a real... interesting challenge... picking the right interfaces!  in the RHT case, we have the additional challenge of having to cover a wide range of mass-volume products, in order to achieve the goal of creating that synergy (what a joyous word, that...) between china-based factories and software (libre) developers that we're looking for.

loovely


I can see why you'd shy away from bringing out VGA - the RasPi followed the same path. But I have to warn you, one of the most common complaints - if not the most common one - related to RasPi is that missing VGA. Quite a few people (myself included) seem to be stuck without a HDMI capable monitor and in no mood to buy one just for a low-cost dev-board.

Also, as I mentioned elsewhere - I've just bought a new LCD monitor as a present for someone, and rather interestingly none of the entry-level models in our stores had anything other than a VGA input...

Joules
Posts: 87
Joined: Sun Dec 25, 2011 8:50 pm

Re: if its good copy it?

Tue Dec 27, 2011 8:28 pm

Hopefully someone will come up with a virtual desktop that allows you to run the Raspi screen on what ever platform you may choose to program it from. That solves the VGA problem for a lot of people. The old TV should work for almost everyone else.

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 27721
Joined: Sat Jul 30, 2011 7:41 pm

Re: if its good copy it?

Tue Dec 27, 2011 8:44 pm

Joules said:


Hopefully someone will come up with a virtual desktop that allows you to run the Raspi screen on what ever platform you may choose to program it from. That solves the VGA problem for a lot of people. The old TV should work for almost everyone else.


Well, X windows works fine and does exactly that.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed.
I've been saying "Mucho" to my Spanish friend a lot more lately. It means a lot to him.

Prometheus
Posts: 308
Joined: Tue Dec 13, 2011 11:09 pm

Re: if its good copy it?

Wed Dec 28, 2011 12:57 am

lkcl said:


the lesson from the openpandora: *don't* get your own casework done, and *don't* get involved with R.F. (openpandora paid $50k to an ex T.I. engineer to *fail* to get the WL1251 part of the PCB design correct.  that's $50k of NREs that has to be amortised onto everybody's units, for nothing)


Whilst there are certainly lessons that can be learned, I do have a correction to make here, if I may - the issue with the engineer was nothing to do with the PCB (which was designed by Michael Weston, not any ex-T.I. folks) - it was to do with WiFi drivers.

Benedict White
Posts: 224
Joined: Sat Dec 24, 2011 12:24 am

Re: if its good copy it?

Wed Dec 28, 2011 1:12 am

Ah Vapour ware! So much so that there appears to be no actual website to link to!

I have a whole host of imaginary friends who are all using vapour ware! They claim it's great!

Before this project was announced there was at least some proof of concept boards, then some alpha boards.

I know Cortex chips can be cheap, but seriously fast? Not seen one. Perhaps someone could point me to one.

User avatar
cnxsoft
Posts: 191
Joined: Sat Oct 15, 2011 2:33 pm
Location: Chiang Mai, Thailand
Contact: Website

Re: if its good copy it?

Wed Dec 28, 2011 2:33 am

Benedict White said:

I know Cortex chips can be cheap, but seriously fast? Not seen one. Perhaps someone could point me to one.
Cortex are the faster ARM cores right now. There are faster than the ARM11 used in R-Pi.

I suppose the fastest you can find is the quad core Cortex A9 Tegra 3, but it's not really cheap anymore.

User avatar
johnbeetem
Posts: 945
Joined: Mon Oct 17, 2011 11:18 pm
Location: The Mountains
Contact: Website

Re: if its good copy it?

Wed Dec 28, 2011 3:52 am

cnxsoft said:


Benedict White said:


I know Cortex chips can be cheap, but seriously fast? Not seen one. Perhaps someone could point me to one.


Cortex are the faster ARM cores right now. There are faster than the ARM11 used in R-Pi.

I suppose the fastest you can find is the quad core Cortex A9 Tegra 3, but it's not really cheap anymore.


Wikipedia is a good place to start looking: http://en.wikipedia.org/wiki/ARM.

ARM Cortex-A are the fastest, but are complex chips with long pipelines.  Cortex-A9 is the fastest shipping now I believe, but there's a Cortex-A15 in the queue that's even faster by using many classic pipelined architecture tricks.  BeagleBoard has Cortex-A8 processor.

ARM Cortex-M are cheap, but not as powerful.  They typically don't have cache, but use on-chip Flash and SRAM for low latency.  They're designed for low-end embedded applications.  The ARM Cortex-M0 is designed to be cheap enough to replace 8-bit processors.  The M4 is the latest, with DSP instructions for real-time signal processing applications.

There's also a Cortex-R series for high-end embedded applications, but I've never seen it sold by anyone as a general-purpose chip.

lkcl
Posts: 18
Joined: Tue Dec 27, 2011 7:42 am
Contact: Website

Re: if its good copy it?

Wed Dec 28, 2011 8:40 am

obarthelemy said:


Yep the ICs fit.
And the traces are such a small detail, right guys ? ^^


  if you have software that takes care of calculating the lengths correctly: yyep, it's a small detail.  if not, woo-hoo are you in for some fun and games.  DDR2 is very very fussy about timings.  bottom line: get some decent PCB CAD/CAM software, or be prepared to take a risk, or, the other possibility: ramp the clock speed down so low that it doesn't matter.

in china, if they don't have the CAD/CAM software, they just take a risk, and go through an unbelievable number of iterations on the PCB.  why should they care: labour is cheap, right?

lkcl
Posts: 18
Joined: Tue Dec 27, 2011 7:42 am
Contact: Website

Re: if its good copy it?

Wed Dec 28, 2011 8:48 am

cnxsoft said:


Benedict White said:


I know Cortex chips can be cheap, but seriously fast? Not seen one. Perhaps someone could point me to one.


Cortex are the faster ARM cores right now. There are faster than the ARM11 used in R-Pi.


i hear it's about 2.5x faster for the equivalent clockspeed.  the reason is the pipelining (harvard architecture) and the out-of-order instruction execution.  the cortex a8, that's quite a short look-ahead but the cortex a9 it's a really large one, something like that.

linaro got paid by a number of SoC vendors to add in optimisations into gcc to cater for the lack of look-ahead in the cortex a8, and managed to get a 30% performance improvement just by putting the assembly instructions in the right order for the execution pipeline to be happiest.

ARM11 on the other hand is an absolute dog.  there are state-delays all over the place, which even if you tried to increase the CPU speed you simply couldn't.  Intel did a total redesign called PXA (then sold it to marvell).  they were _supposed_ to hand back any modifications made, back to ARM (that was the deal) but the "modifications" were such a total redesign that Intel refused.

bottom line is: ARM11 is a hopelessly shit CPU design, to be frank; by comparison, the Marvell series and the Cortex series CPUs are absolutely superb.

lkcl
Posts: 18
Joined: Tue Dec 27, 2011 7:42 am
Contact: Website

Re: if its good copy it?

Wed Dec 28, 2011 8:50 am

Prometheus said:


lkcl said:


the lesson from the openpandora: *don't* get your own casework done, and *don't* get involved with R.F. (openpandora paid $50k to an ex T.I. engineer to *fail* to get the WL1251 part of the PCB design correct.  that's $50k of NREs that has to be amortised onto everybody's units, for nothing)


Whilst there are certainly lessons that can be learned, I do have a correction to make here, if I may - the issue with the engineer was nothing to do with the PCB (which was designed by Michael Weston, not any ex-T.I. folks) - it was to do with WiFi drivers.



ahh, thank you for the correction.  i shall go beat my friend about the head (the one who reads the openpandora forums)

lkcl
Posts: 18
Joined: Tue Dec 27, 2011 7:42 am
Contact: Website

Re: if its good copy it?

Wed Dec 28, 2011 8:58 am

JamesH said:


Joules said:


Hopefully someone will come up with a virtual desktop that allows you to run the Raspi screen on what ever platform you may choose to program it from. That solves the VGA problem for a lot of people. The old TV should work for almost everyone else.


Well, X windows works fine and does exactly that.


i discovered last year something called xrdp.  it's a server-side implementation of the MS "Terminal Server" protocol, matching up with rdesktop.

the problem with running X-Windows over a network is that it's designed for running over very _fast_ networks.  someone created NX (software-libre version called FreeNX, predictably) to cater for this but it is seriously, seriously complex to set up.

xrdp on the other hand, when i tried it, it just... worked.  apt-get install xdrp, job done.  beautiful.

so, you could apt-get install xrdp onto the rbpi, then use an terminal server client from anywhere in the world.  PDAs, smartphones, android, GNU/Linux desktop PCs or even shock horror a w32 system.

that'd be a hoot.  using a desktop pc with an OS that's so bloated it forces a minimum requirement of 4gb of RAM and at least a 2ghz Dual-Core CPU just to function merely as a "thin client" in order to fire up a desktop where the applications are all running on a 700mhz ARM11 with only 256mb of RAM.  love it

User avatar
riffraff
Posts: 307
Joined: Tue Sep 27, 2011 3:17 am
Location: Newnan, Georgia, US

Re: if its good copy it?

Wed Dec 28, 2011 9:10 am

Please stop thinking IBM-PC. VGA will soon go away. EGA/VGA/SVGA were adequate solutions for a time due to the lack of a proper display device.

Portable systems have LCD displays now, and the home computer is back where God intended it to be - connected to the television.

All of those monitors will only be remembered as a very tragic chapter of our history - when we wasted our money and cluttered our landfills.

Return to “General discussion”