Exophase
Nothing good will ever come of Exophase.
Hey lkcl, you forgot i.MX51 again. You should check it out sometime.
Help Needed Replicating Success Of Openpandora For A Netbook/laptop
- Thread starter lkcl
- Start date
darkborn said:
yep, that's us
the price in 20k volumes is $102.that $65 laptop is actually fantastic, but there's a _reason_ why it's $65. the first is that it's a truly dreadful ARM9 which will eat your mind. the second is that the 800x480 LCD will be connected directly to the TTL interface on the ARM9 CPU. there will be _no_ LVDS driver IC, which is pretty much standard fare for anything above 800x600 such as 1024x600, 1024x768, 1200x768 etc. you _just_ don't get "direct" 18-pin or 24-pin TTL panels at those resolutions (not in mass-volume quantities, anyway).
so, yeah - it's a completely different market, that one. ironically, the 2100mAh battery is identical to the CT-PC89E's one, and that gets 3 hours out of the same sized battery! a 667mhz ARM11 @ 65nm + a 2.5 watt 1024x600 LCD almost beats a 533 mhz ARM9 @ 130nm + a 2 watt 800x480 LCD, how weird is that? _and_ the S3C6410 does 3D (OpenGL ES 2.0)
Last edited by a moderator:
Polossatik
I reject your reality and substitute my own
nikudorei said:
I'm a trackball fanatic myself, I thought they where a dying race.... incredible to see there are other survivors.
not sure if it works IRL , but just a stupid idea (sorry for the lefties):
A nice trackball off center to the right next to the cursor keys and a additional pair of mouse keys on the not totally left end of the keyboard ("under" the "keyline" ) please.
grab the net/notebook, use right thumb to manipulate trackball, left hand/thumb to click.
If you incorporate one of those little nipple mouse thingy's in the middle of the keyboard, make sure your keyboard doesn't flex. they work great on Thinkpads but i had to disable on quite a few (much cheaper) laptops that stuff because it simply reacted on keyboard flexing.
Ha and +1 for a scrolling thingy (if it can be made to work properly) - a kind of scroll wheel or touch based strip next to the page up down keys?
Last edited by a moderator:
Exophase said:
bleeegh, noo, i hadn't forgotten. i have erased freescale from my memory banks, due to them being desperate oil-slicked barrel-scrabblers. translation: freescale don't put enough money into the design and implementation of their CPUs; they chase after customers who will pay lots of money; they rush products out the door before they're ready but they're _still_ late and/or behind the times...
so i erased them from my memory for damn good reasons: they're simply too risky to let loose on poor unsuspecting end-users. the OMAP3530 Cortex A8 came out _over two years_ ago, so far back that even the open-pandora could pick it and it's _still_ a damn good CPU, very mature and reliable; the S5PC100 Cortex A8 came out what... over one year ago? and freescale announce the iMX51 Cortex A8 like... three weeks ago? you _got_ to be kidding
i fink i advise people to wait for a year and see if that iMX51 is any good in 1 year, neh?
Last edited by a moderator:
Prometheus
Well-Known Member
- Joined
- Mar 8, 2008
- Messages
- 9,472
TrackPoint fanatic, here - my favourite pointing device of all time, bar none.
I would gladly pay extra for such a thing - not enough machines have them, and on the occasions where I've had to settle for machines with trackpads, I've always had to connect an external mouse and make some extra room on my desk to use it.
Attachments
Monk said:
no, these are all good questions. the summary is: this is _not_ a "we dictate what you will get" thing: the intention is for it to be a _multi_ phase project, where people first get in and find out what everyone else wants, too, and under what conditions; _then_ the 2nd phase is "let's go do it". thus, far from being a project with lots of surprises, the idea is for everyone to go in with eyes open and if people _like_ surprises and risks great, they get to _choose_ to take that route.
about the only thing i am going to lay down the law on is that there will be NO proprietary firmware blobs, NO proprietary software, NO ndas are to be signed that stop people from writing GPL drivers for any of the hardware components, and NO proprietary bioses. also: i will _advise_ people on which CPUs or components that i am aware are somewhat reckless decisions (such as the freescale ones) or are non-free and thus out of the question (such as anything from NVidia) and i will change the questionnaire/survey accordingly as i learn more and as people give me feedback and other options they've heard about but i will NOT be "dictating" the final technical specs or even how the machine is ultimately constructed.
so:
1) i've seen what happened with the 2-year delay, and went "wtf??"
so one of the key questions on the questionnaire - right on the first page - will be do you want to i) know about any off-the-shelf units which can be bought ii) ready-made reference designs which can simply be bought en-masse iii) get something "knocked up" by e.g. using an existing x86 netbook/laptop case and replacing the Motherboard (thus saving $100k in design costs) iv) go the whole hog and get it completely designed from scratch2) there will be options for people to pre-order, donate or invest. going into the 2nd phase, after analysing the questionnaire answers, i will tryyy to make heads of it and advise which "critical thresholds" are passed, first, depending on whether people _actually_ are prepared to put money-where-mouth-is. it really _will_ be up to people to do that, not me, thus i _cannot_ dictate what they want - only advise.
the "invest" option will be an interesting one. if someone wants to invest, and wants a 10% return on their money, one of the things that will need to be up for discussion is: do we _need_ that investor's money, and are people prepared to accept the resultant 10% price-hike? are there other people willing to offer better terms, with a lower return? do we need the money _at all_, because of some donations or the pre-orders? are there enough people willing to pay significantly over-the-odds, paying sample-level pricing And Then Some in order to be early adopters, thus effectively paying for the development, such that the mass-production runs are achievable without any investment at all?
this we will all find out, pretty soon, as the first phase progresses - _if_ it progresses. we just can't say for sure - but i'm pretty certain that there are absolutely tons of people out there who really do want a decent competitively-priced non-intel [net|smart|lap|cookery][book|top] and i'm also confident that the ARM SoCs are at the right price and design-simplicity levels to make this a really viable and exciting project.
l.
Last edited by a moderator:
Polossatik said:
polossatik, these are great comments. lots of things to look up. i particularly wanted to come back to the SATA thing, because as i mentioned in another post this evening, finding SATA PHY ICs seems to be a total BITCH. atmel's offerings have been bought by another company, who have then stuffed the IP into some half-arsed product of their own in order to re-sell it. marvell are sods about NDAs. there is _one_ TI OMAP ARM, 300mhz, which has built-in SATA - i was thinking of recommending it and connecting the two CPUs back-to-back, using this OMAP _just_ for its SATA controller... but it's $USD 17! forget memory or support infrastructure - just the 300mhz older CPU alone is more money than a samsung 833mhz Cortex A8 by about 25%!
michelle, one of the hardware engineers i know, said that it's easier to do General-Purpose Memory Bus (MAC) to PCI-e, then PCI-e to SATA, than it is direct GPMB to SATA. so, i looked up TI for "PCI-e PHY ICs" and came up with this:
http://focus.ti.com/docs/prod/folders/print/xio1100.html
it's $USD 7 in 1k volumes.
now - the thing is: if you're going to bother to put a PCI-e on (full spec PCI-e not just the USB2 special-interest-group extension) then, well.... sod the SATA, you can put in a perfectly good 50mm x 30mm PCI-e SSD and be done with it.
anyway: all these kinds of things will be encoded in the (constantly evolving) questionnaire. "do you want SATA, yes, then relatively speaking it will cost you an arm and a leg: we recommend PCI-e instead, how about it?"
Last edited by a moderator:
Exophase
Nothing good will ever come of Exophase.
lkcl said:bleeegh, noo, i hadn't forgotten. i have erased freescale from my memory banks, due to them being desperate oil-slicked barrel-scrabblers. translation: freescale don't put enough money into the design and implementation of their CPUs; they chase after customers who will pay lots of money; they rush products out the door before they're ready but they're _still_ late and/or behind the times...
so i erased them from my memory for damn good reasons: they're simply too risky to let loose on poor unsuspecting end-users. the OMAP3530 Cortex A8 came out _over two years_ ago, so far back that even the open-pandora could pick it and it's _still_ a damn good CPU, very mature and reliable; the S5PC100 Cortex A8 came out what... over one year ago? and freescale announce the iMX51 Cortex A8 like... three weeks ago? you _got_ to be kidding
Okay, first of all, I find your brand prejudice a little off-putting. Perhaps you should give a product a chance regardless of whatever conception you have of the corporation behind it?
Second, your timelines are off by astronomical proportions. i.MX51 was announced 16 months ago, dev boards have been available for a year, and you've been able to purchase them from several resellers for at least several months. In fact, they're much easier to buy than Samsung's SoCs. And unlike Samsung they also make datasheets available plain as day on their website.
So there's your year, huh?
And OMAP3530 sure as hell wasn't available to the public 2 years ago. It was announced a ridiculously long time before anyone could buy it, and it suffered quite tremendous errata.
BTW, your "no proprietary whatever" rule means nothing from IMGTech. Enjoy your inferior graphics IP I guess? Well, by your comments you probably think the IP in the S5PC100 is all the same anyway, I mean, it's OpenGL ES 2 right?
Last edited by a moderator:
Exophase said:
hi exophase, i'm very pleased to stand corrected. i knew there was an announcement of _some_ freescale cpu three weeks ago (quietly) i just couldn't remember which one it was. ok, so.... let's look this up.
http://www.h-online.com/newsticker/news/item/Freescale-hints-at-Apple-netbook-Updated-739529.html
ah ha! references to the iMX51 being used in the pegatron (asus subsidiary) netbook reference design; pricing $199 retail and there's even a reference design desktop PC for $150 ooooo
http://www.pogledaj.name/150-desktop-pc-made-by-pegatron-using-the-freescale-arm-i-mx51-processor-platform/video/ymYGIAtvA2o
http://www.notebooks.com/2009/01/07/new-generation-of-netbooks-199-and-299-eight-hour-battery-sexy-design/
so.... yes, ok, that's a good one. also, yes: the iMX51 does OpenGL ES 2.0 and OpenVG so it does 3D and 2D accelerated graphics.
ok, so yes - thank you for prodding me to investigate, because these reference designs have been available for well over a year, which means that pegatron might accept an in-bulk order and simply roll them off the production line.
regarding IMGtech: i don't know who they are. a reference to what or whom you're referring, or a clearer explanation, would help me - and thus help the people whom i am in turn trying to help, by providing accurate information which i myself might not know about. in this instance, it's hard to tell, but you seem to be indicating that S5PC100 OpenGL is proprietary, is that correct? if so, i would prioritise/recommend the iMX51 over the S5PC100, despite my concerns about freescale's financial situation.
bottom line: keep me informed so i can inform others, ok?
thanks exophase.
l.
Last edited by a moderator:
This ARM netbook project interests me. It was why the Pandora originally caught my eye (since if you ignore the excellent gaming controls and retro-enthusiast community, the Pandora is, or will be a small ARM netbook-style device).
One thing which I think is key that you touched on is that documenting this process is actually a product in itself, it gives people something to read, something to cheer on, something to think about. An alternative to gossip mags or sports competitions. Documenting what you discover for others that follow is a useful service too, even if it's not something you can hold in your hand in exchange for cash. I, for example, was intrigued to hear about your comments saying that as pixel density increases so does wastage and therefore costs and at the other end of the scale as screens get larger the backlight power begins to cancel out any battery-saving effect of good software and hardware choices. So screens are pushed from both sides into a sweetspot of price vs power draw. I'd be interested to hear how that effects Pixel Qi screens which I believe have a monochrome hi-res mode so you could get higher resolution for text, with full colour for video.
I would say a key lesson of OpenPandora is that the manufacturers in the east are not going to magically produce what you want. They have the technical capability but lack the end-user knowledge of what is important for a gaming handheld or a fully free software device. Eric von Hippel talks about this in his excellent (and free) book Democratizing Innovation (http://web.mit.edu/evhippel/www/democ1.htm) giving the example of specialist tool makers that can propose stronger, lighter alloys and so provide one kind of innovation compared with surgeons or sculptors who are actual end-users and know that they need a tool shaped just so. The ideal is of course to enable both types of innovation.
Has any thought been given to trying to buy a batch of OLPC innards and piggy back on their efforts? They kind of lost their way for a bit but have been talking about a switch to ARM and there seems to still be some momentum there. The opposite would be to try and find organisations that can put in orders for thousands of machines to run the OLPC sugar software on the same kind of hardware you want to build, effectively poaching OLPC customers (or those just outside their target demographics) to expand your customer base.
I think the software thing is key. Obviously you want people to be able to do anything with it, and put any linux distro of their choice on it. But by combining software and hardware you get something special. Again this is a lesson from the Pandora, twin analogue sticks plus emulators is a USP that no-one is likely to ever challenge even as chips get smaller and faster.
The litl project had some interesting ideas in this regard: http://litl.com with their flip over stand set up allowing you to use it as a picture frame or family notice board. They also went HDMI only with the idea it could show photos on the big screen. I thought it could be a good alternative to Apple TV or Boxee Box where, if you're not using it as those you can use it as it's own portable screen.
One thing which I think is key that you touched on is that documenting this process is actually a product in itself, it gives people something to read, something to cheer on, something to think about. An alternative to gossip mags or sports competitions. Documenting what you discover for others that follow is a useful service too, even if it's not something you can hold in your hand in exchange for cash. I, for example, was intrigued to hear about your comments saying that as pixel density increases so does wastage and therefore costs and at the other end of the scale as screens get larger the backlight power begins to cancel out any battery-saving effect of good software and hardware choices. So screens are pushed from both sides into a sweetspot of price vs power draw. I'd be interested to hear how that effects Pixel Qi screens which I believe have a monochrome hi-res mode so you could get higher resolution for text, with full colour for video.
I would say a key lesson of OpenPandora is that the manufacturers in the east are not going to magically produce what you want. They have the technical capability but lack the end-user knowledge of what is important for a gaming handheld or a fully free software device. Eric von Hippel talks about this in his excellent (and free) book Democratizing Innovation (http://web.mit.edu/evhippel/www/democ1.htm) giving the example of specialist tool makers that can propose stronger, lighter alloys and so provide one kind of innovation compared with surgeons or sculptors who are actual end-users and know that they need a tool shaped just so. The ideal is of course to enable both types of innovation.
Has any thought been given to trying to buy a batch of OLPC innards and piggy back on their efforts? They kind of lost their way for a bit but have been talking about a switch to ARM and there seems to still be some momentum there. The opposite would be to try and find organisations that can put in orders for thousands of machines to run the OLPC sugar software on the same kind of hardware you want to build, effectively poaching OLPC customers (or those just outside their target demographics) to expand your customer base.
I think the software thing is key. Obviously you want people to be able to do anything with it, and put any linux distro of their choice on it. But by combining software and hardware you get something special. Again this is a lesson from the Pandora, twin analogue sticks plus emulators is a USP that no-one is likely to ever challenge even as chips get smaller and faster.
The litl project had some interesting ideas in this regard: http://litl.com with their flip over stand set up allowing you to use it as a picture frame or family notice board. They also went HDMI only with the idea it could show photos on the big screen. I thought it could be a good alternative to Apple TV or Boxee Box where, if you're not using it as those you can use it as it's own portable screen.
Exophase
Nothing good will ever come of Exophase.
lkcl said:
To be more specific it uses an AMD Z430, the same accelerator as in the current Snapdragon SoCs.
lkcl said:
IMGTech make the most popular 3D accelerators in the mobile market, ie the MBX and SGX lines of TBDR GPUs. SGX accelerators are present in OMAP3530, iPhone, iPad, Z series Atom chipsets, etc. They're currently generally regarded as the best choice vs AMD Z430, Samsung's OGL2 ES, and probably Mali (which I'm not sure is in any mainstream SoC). It's probably less certain how, say, SGX 535 stacks against Tegra 2, but that's in a different ballpark altogether.
lkcl said:
No, S5PC100 uses Samsung's in-house video which has information available in their user guide (which itself isn't really publicly available but doesn't appear to be confidential). I think Samsung is killing this off, though, as iPhone and their next generation A8 SoCs (like S5PC110) use SGX.
Last edited by a moderator:
I have also been looking for a cheap SATA solution for an open hardware project (portable music player). Not really for disk bandwidth, just to get access to higher capacities. I was considering a PATA-SATA bridge such as JM20330, which is used in cheap adapters found on ebay. I've no idea if this chip can be easily sourced, I was thinking of buying the adapter and hacking it.lkcl said:5) SATA. if you do a search for SATA controllers (google "SATA PHY IC") you find a ton of references to forums where people have been looking for _years_ for SATA controllers esp. ones to connect to FPGAs. there are two ICs which come up: one is from ATMEL and is 1.5gbps and the other is by Marvell and is 3.0gbps. the trouble with these ICs is that if you order a million, they fire up a special run, just for you, no problem. everyone else can simply get stuffed, as far as they're concerned. so, some embedded systems engineers simply write their own SATA controllers, as an FPGA or using a low-speed CPU! it has to be said though: even 1.5gbps is 187mbytes per second; 3gbps is 375mbytes/sec... running an ARM CPU's General-Purpose Memory Bus at up to 375mbytes/sec when the CPU is 833mhz or 1ghz? mmm... there had _better_ be a DMA engine handling the data transfers, that's all i can say about that. but - i quite like the idea floated by michelle (an embedded designer) of using a general-purpose 75mhz 8051 CPU with lots of GPIO and an on-board 32k PROM, and implementing SATA or even PATA in assembler! that would be just absolutely wicked (and probably a lot cheaper than asking IC manufacturers to fire up the boilers just for us). it will take quite a bit of research, but it's such a cool idea that it just has to be investigated
An embedded software solution is an interesting idea - if software USB can be done on a PIC (slowly), maybe same idea can be used for SATA? I wouldn't use 8051 though, but there are quite a lot of 32 bit controllers in the low end now, like Cortex M3.
Last edited by a moderator:
Polossatik
I reject your reality and substitute my own
lkcl said:
Personally I don't care about SSD to much (yet), I have modded CF stuff in laptops and that was fast enough for me ( little form factor, rather low power and price and relative low capacity/speed) to usable for the application intended.
I was more on the line of "to have choise in one box"
I have no doubt some people would like have or
* cheap, power hungry but plenty space for the $ -> HDD (pata is so on it's way out it's silly to even consider it)
* fast, rather expensive solid state -> SSD (aldo I see a "SATA OCZ miniPCI-Express Solid State Drive (SSD) " of 16gb for only 90USD - no idea if this stuff is any good)
* slow, lower power and not so big then hdd, but cheaper then SSD and solid state -> SD
If sata is costing an arm and a leg (or the "open source sideprocessor" implementations are not very performant) then that is as such not a big issue for the have it canned for a PCI-e solution.
But for example for photographers "disk size does matter" to consider a laptop as "image tank" (it needs to be workable also - the transfer times from SD to HDD/SSD) , if you go for a better/higher res LCD then that stuff is right in the alley of photographers I know who are put of by the crappy LCD (resolution/image quality) of most netbooks.
If a solution like the JM20336 (family of the JM20330 mentioned but USB to SATA) more or less works at acceptable rates (no idea what the actual speed is of that stuff) and cost little (which I doubt) then I think it might be an idea to forsee it.
No idea what a PCI-e slot costs, but it looks like a quite groovy thing to have it on board...
Beeing cheap is one thing (it really helps of course) but to be unique is IMHO a better thing then let's say 50 usd cheaper at then end.
It all depends also on the form factor of course, an 2.5 HDD takes up relative a lot of space (and heats), is it usefull/possible to forsee space for this?
Would that space be better used for a bigger battery?
questions, questions...
Last edited by a moderator:
Exophase
Nothing good will ever come of Exophase.
I don't think the idea to implement an SATA controller in a cheap microcontroller is workable. From end to end it just doesn't seem to work at several points along the way.
SATA's input/output lines are differential and at very low voltage (250mV). It wouldn't be possible to accept this over GPIOs, nor is it electrically compatible with typical USART interfaces on MCUs.
There's no way an MCU can pull 1.5gbit/s of data off of anything, regardless of hardware interface. USARTs only tend to go up to around 4.5mbit/s. Trying to do it over bit-banging would give you something even slower, and would tie up 100% of your CPU time (leaving you nothing for processing the data and sending it out).
You then have to decode up to 150MB/s of data from 10-bit words to 8-bit words. Even with LUTs I doubt any conventional cheap MCU (up to 100MHz) can do this.
Then the data has to be streamed off the unit. You'd need a high speed externally bus mastered parallel interface that connects internally to DMA. I don't know of any low cost MCU that implements this (PIC24 has external bus mastered parallel interfaces but they only go to FIFOs).
All of this is just for streaming data, and from a single channel - no actual control work is done. I assume it's okay to just support a single channel though.
Of course I could be totally wrong, so I guess I'd like to know what Michelle is thinking exactly.
SATA's input/output lines are differential and at very low voltage (250mV). It wouldn't be possible to accept this over GPIOs, nor is it electrically compatible with typical USART interfaces on MCUs.
There's no way an MCU can pull 1.5gbit/s of data off of anything, regardless of hardware interface. USARTs only tend to go up to around 4.5mbit/s. Trying to do it over bit-banging would give you something even slower, and would tie up 100% of your CPU time (leaving you nothing for processing the data and sending it out).
You then have to decode up to 150MB/s of data from 10-bit words to 8-bit words. Even with LUTs I doubt any conventional cheap MCU (up to 100MHz) can do this.
Then the data has to be streamed off the unit. You'd need a high speed externally bus mastered parallel interface that connects internally to DMA. I don't know of any low cost MCU that implements this (PIC24 has external bus mastered parallel interfaces but they only go to FIFOs).
All of this is just for streaming data, and from a single channel - no actual control work is done. I assume it's okay to just support a single channel though.
Of course I could be totally wrong, so I guess I'd like to know what Michelle is thinking exactly.
Exophase said:
ouaah, thank you exophase - very enlightening, and quite important information. i've added the iMX51 to the survey: i'll make sure that this kind of info goes into the "report" that people get at the end.
pandorra mentioned that yes, the documenting of the process is itself a product: i'll actually be encoding the information as a database-driven python application (it's complicated
python code-fragments are taken from the database and executed on-demand, and the results determine what goes into the report, or what goes onto subsequent pages of the survey). i'm not entiirely sure i want absolutely anyone editing that, but it will certainly be published as a free software application - excluding peoples' email addresses etc.l.
Last edited by a moderator:
Exophase said:
yehh, simply looking at the numbers i kinda worked that out - not at the level of detail you're aware of - more in the "fuzzy" kind of way which you've probably already noticed
so, the alternative is PCIe PHY (TI XIO1100) followed by PCI-e to SATA or PCI-e to PCI and PCI to SATA. crazed and crazy. the TI XIO1100 is $USD 7. The XIO2000A (PCI-e-to-PCI) is $9. found an SATA to PCI-e and also SATA to PCI: SiL3132 and SiL3114 http://www.sequoia.co.uk/components/product.php?d=1&c=1&f=7&p=950&fmt= http://www.sequoia.co.uk/components/product.php?d=1&c=1&f=7&p=951&fmt=
so... all i can say about adding SATA is: it'll cost about... $20 to $25 in components. and that's fine, as long as there are at least 1,000 other people who want SATA.
this kind of thing - encoding and being able to express that kind of logic - is the purpose of the survey. being able to pledge "i will pay for this feature if 1,000 other people also say that they will pay for it" is entirely the point.
Last edited by a moderator:
Exophase
Nothing good will ever come of Exophase.
Gah I swear I posted here already, I wish gp32x wouldn't log me out every other minute -_-
Anyhow. i.MX51 and S5PC100 both support ATA/ATAPI-6. It seems to me that most small harddrives - the kind people would be most interested in for a netbook - are still PATA so they don't need additional support chips when used with these SoCs. If you need SATA you can get the JMicron bridge chip mentioned by bobc, which is probably not that much. You do have to make sure that UltraATA/133 (ATA/ATAPI-7) is not required.
Anyhow. i.MX51 and S5PC100 both support ATA/ATAPI-6. It seems to me that most small harddrives - the kind people would be most interested in for a netbook - are still PATA so they don't need additional support chips when used with these SoCs. If you need SATA you can get the JMicron bridge chip mentioned by bobc, which is probably not that much. You do have to make sure that UltraATA/133 (ATA/ATAPI-7) is not required.
Exophase
Nothing good will ever come of Exophase.
Do you have any price and availability information on S5PC110? It'd be very attractive compared to any of the other Cortex-A8 offerings out.
