02:58.55 | *** join/#arm-netbook mikey_w (~mike@pool-74-110-218-2.rcmdva.fios.verizon.net) |
02:59.58 | *** join/#arm-netbook eduprey (~chatzilla@ip70-173-29-15.lv.lv.cox.net) |
03:25.57 | *** join/#arm-netbook mikey_w (~mike@pool-74-110-218-2.rcmdva.fios.verizon.net) |
05:23.29 | *** join/#arm-netbook eebrah (~Ibrahim_K@41.215.224.88) |
05:48.39 | *** join/#arm-netbook sectionme (~shift@cpc1-benw1-0-0-cust440.gate.cable.virginmedia.com) |
05:59.43 | *** join/#arm-netbook sectionme (~shift@cpc1-benw1-0-0-cust440.gate.cable.virginmedia.com) |
06:17.07 | *** join/#arm-netbook A2Sheds (~ly@unaffiliated/l84supper) |
09:18.34 | *** join/#arm-netbook gordan (~gordan@78-86-195-86.zone2.bethere.co.uk) |
09:23.10 | eduprey | A2Sheds: just looked at the mock-up for EOMA-CF.. wow. |
10:21.46 | A2Sheds | eduprey: too small? :) |
10:22.51 | eduprey | A2Sheds: no, just impressive. thinking a cluster board for CF's would be better than EOMA68's |
10:23.34 | eduprey | A2Sheds: how much further behind is CF in terms of making boards, do you know? |
10:23.45 | A2Sheds | it would be nice to fit a NAND to DDR, with a few GB of swap space as well |
10:24.13 | A2Sheds | it's more about interest |
10:24.48 | eduprey | A2Sheds: also mouser has cheap CF connectors |
10:24.59 | A2Sheds | a10 first, then maybe x86 or CF or dual core or quad core a9 or ?? |
10:25.30 | A2Sheds | even cheaper in China :) |
10:25.32 | eduprey | well a10 CF is most exciting for me |
10:26.05 | eduprey | even more so than EOMA68 IMO but EOMA68 is admittedly a good starting point and can have more stuff for getting started developing |
10:26.20 | eduprey | makes sense as a first board even though the size of the CF one is awesome |
10:26.32 | A2Sheds | a10 CF maybe quad a9 or dual a15 CF card next |
10:27.10 | A2Sheds | it doesn't take long to make cards, if we have the parts available |
10:27.16 | A2Sheds | A10 no problem |
10:27.29 | A2Sheds | other ARM soc's ??? |
10:27.36 | A2Sheds | x86, no problem |
10:27.47 | eduprey | think allwinner will be doing bigger SoC's in future? they seem to be awesome on price point :) |
10:28.08 | eduprey | and good features and basically unbrickable it seems |
10:28.23 | A2Sheds | don't know, we might even design our own open SOC's for this purpose |
10:28.53 | A2Sheds | that way we aren't locked into vendors with personality disorders or worse |
10:29.41 | eduprey | wow IC fab, the final frontier :) |
10:30.09 | A2Sheds | some ARM soc vendors aren't willing to trade $$ for parts unless you jump through other strange hoops |
10:30.31 | A2Sheds | "what will you do with out parts?" |
10:31.37 | A2Sheds | who cares, we trade you $$ for parts, we might decide to turn them into a sculpture or whatever we decide |
10:31.47 | A2Sheds | out/our |
10:32.30 | eduprey | yeah seriously |
10:32.49 | A2Sheds | but they do care, they seem to want to control the parts after they are purchased, it's like some sort of M$ disease |
10:33.10 | A2Sheds | control control control |
10:33.42 | A2Sheds | you wonder how they were raised and by what family of wolves |
10:35.57 | A2Sheds | so don't be surprised if an EOMA SOC shows up later this year |
11:59.30 | *** join/#arm-netbook A2Sheds (~ly@unaffiliated/l84supper) |
14:07.48 | *** join/#arm-netbook A2Sheds (~ly@unaffiliated/l84supper) |
15:18.10 | lkcl | eduprey: i've done a pinout design for a SoC, it's 304 pins. that means it would fit into a Quad Flat Package, about 30sqmm on a side. |
15:19.20 | lkcl | in 28nm, each 12in wafer costs about $10,000 (you have to get 16 done at a time) and contains about 8,000 chips. |
15:19.56 | lkcl | assuming you get a 95% yield, that's approx... $1.32 per wafer. |
15:20.30 | gordan | <PROTECTED> |
15:20.45 | lkcl | then you add on testing of the chip (appx $1), packaging (QFP is about $0.20 or so, whereas BGA is about $1.50 and you lose some to yield problems) |
15:20.55 | lkcl | gordan: yep :) |
15:21.08 | lkcl | so that's $1.32 + 1 + $0.20 = about $2.50 |
15:21.31 | lkcl | meaning that it could be sold for about $3 to $3.50 |
15:21.44 | A2Sheds | now if we could only get the cost to produce masks at 28nm down to $1000 :) |
15:22.34 | lkcl | weeeellll... TSMC normally charge $1m to $2m to gwailo foreigners just to get them to go away, not waste their time. |
15:23.02 | lkcl | if however it's a taiwanese company that's asking, they charge "at cost" for the masks, which are about $100k to $200k. |
15:23.32 | lkcl | ergo, it's very very important to find a taiwanese partner for the project. |
15:24.12 | lkcl | assuming it takes 3 attempts to get it right, that would be $600k *not* $6m |
15:24.12 | gordan | We need more manufacturing in Europe. :( |
15:24.18 | A2Sheds | http://semiaccurate.com/2011/09/08/exclusive-tsmc-raises-prices-on-amd-and-nvidia/ |
15:26.16 | lkcl | yeah, what the author of that article doesn't realise is that, irrespective of whatever has been signed, there is one "Law" for gwailo foreigners, and another Law for taiwanese companies. |
15:26.54 | lkcl | taiwan, as a country that _still_ isn't recognised as a Sovereign Nation even by the U.N., basically has to "look after its own" |
15:27.14 | gordan | Re: Article, I heard Apple were being dropped by a lot of their suppliers because they place large speculative orders, then cancel a large chunk of most of them. |
15:27.32 | lkcl | but hmmm it's good to know that the yields aren't too good on 28nm yet. |
15:28.23 | A2Sheds | gordan: thats how traders game the markets |
15:28.43 | lkcl | assuming a 40% yield, that $10k to do 8,000 chips and only get 40% puts the price up to $3.125 + 1 + $0.20 |
15:29.04 | lkcl | $4.33 *lol* that's funny, right there :) |
15:29.23 | lkcl | has been watching "Cars".... |
15:31.02 | gordan | Yeah, but if you screw your supplier over by ordering 100K chips and then only by 10K and leave 90K of chips on their hands to shift however they can, they are fully justified in telling you to go forth and multiply next time you want to do business with them. |
15:31.21 | lkcl | yeah that's just... dumb. |
15:31.22 | gordan | And all things considered, Apple aren't that big a part of the market for most companies. |
15:31.42 | A2Sheds | http://www.globalfoundries.com/ |
15:31.46 | lkcl | yeah, not when it's locked in to only a few large suppliers. |
15:32.09 | lkcl | nobody else really benefits |
15:32.56 | gordan | They may suffer from delusions of self-importance, but all things considered, when you look at things like, e.g. Intel CPUs, most supliers can afford to lose their business without really noticing too much. |
15:33.07 | lkcl | i'm going to speak to tensilica on monday. initial reaction is very favourable. |
15:33.19 | gordan | Who's that? |
15:33.23 | lkcl | The Plan is to put down 8 XTensa CPUs... |
15:33.36 | lkcl | tensilica gahh, i'm stunned that i never encountered them before. |
15:33.45 | gordan | Sorry, I obviously missed something important on the list or here... |
15:33.55 | lkcl | you should see the list of their clients. they're a RISC CPU design company |
15:33.55 | gordan | Is a diet summary / link possible? |
15:34.01 | lkcl | i haven't mentioned this on the list, gordan. |
15:34.11 | gordan | That'd be why I missed it, then. :) |
15:34.14 | lkcl | :) |
15:34.37 | gordan | BTW, apoligies the AC100 isn't with you, yet - work has been _nuts_ the past 10 days and I haven't gotten around to wiping one clean. |
15:35.02 | gordan | Hoping to pull one from the farm this afternoon. |
15:35.22 | lkcl | not a problem - tell you what: after nvidia's rather lacklustre response when i approached them, don't worry about it. |
15:35.38 | gordan | Oh, OK. |
15:35.47 | lkcl | _why_ is it that every american IC design company is clearly clinically insane?? |
15:35.52 | gordan | Let's face it, they never were going to play the OSS game. |
15:36.10 | A2Sheds | it's part of the culture here |
15:36.20 | lkcl | they're going to design modules based around the Q7 standard. |
15:36.22 | A2Sheds | not just IC co's |
15:36.22 | gordan | What is part of the culture where? |
15:37.11 | lkcl | gordan: _every_ american company i approach about the EOMA initiative has gone completely ape-shit. the only exception is bari, bless 'im. |
15:37.35 | A2Sheds | they just took out disorders from the DSM since the majority of the population has them Narcissistic Personality Disorder for one |
15:37.46 | A2Sheds | some disorders |
15:37.47 | gordan | BTW, guys, if you are interested in varying degrees if attrociousness of performnce of various flash modules (SD/CF/USB), you may want to take a look here if you haven't seen it already: http://www.altechnative.net/?p=408 |
15:38.01 | lkcl | lol |
15:38.23 | gordan | The problem is that most companies over a certain size just don't produce anything. |
15:38.32 | A2Sheds | http://well.blogs.nytimes.com/2010/11/29/narcissism-no-longer-a-psychiatric-disorder/ |
15:38.59 | gordan | And what little they do produce is overpriced, underdesigned by a committee and generally unusable. |
15:39.02 | A2Sheds | don't treat the disorder, just change the diagnosis |
15:39.12 | gordan | LOL |
15:39.43 | gordan | It's basically why in the IT industry there's a long standing saying: "If you want something done, hire a contractor." |
15:40.27 | lkcl | gordan: there's a book about that, called "The Other side of Innovation". |
15:40.40 | gordan | Regarding the benchmark article link I mentioned above, I'm updating the benchmark results list all the time (daily most of the time). |
15:40.59 | gordan | I have a sizeable number of modules still to test and post results. |
15:41.12 | gordan | But it should show the general gist of what you can expect. |
15:41.14 | A2Sheds | "Mask costs for a typical system-on-chip (SoC) design in the first year of process availability have gone from $800,000 at 65 nm to $2.8 million at 28 nm. Mask costs are predicted to be more than $10 million at 20 nm." |
15:41.43 | lkcl | it's brilliant: well worth reading. basically it goes to some lengths to praise the "efficiency engine" behind most large companies... but points out that trying to shoe-horn innovation into that is very very hard |
15:41.51 | gordan | The only problem is that the least shit SD card I have found so far (Ingegral Endurance SLC) is discontinued as of the beginning of this year. |
15:42.02 | lkcl | gordan: doh! |
15:42.12 | lkcl | probably because they simply can't compete |
15:42.44 | gordan | And yes, my test is deliberately harsh (/dev/urandom overwrite priming passes, plus testing done on the back half of the card to eliminate distortion of some cheats like putting SLC at the front where the FAT tables live). |
15:42.45 | lkcl | A2Sheds: yep. but that's prices to keep people away, because most companies will fold if they don't get it right first time. |
15:43.16 | gordan | Well, they can't compete on price, but the performance was at least 3-4x better on writes, and in theory with SLC endurance should be 10x better, too. |
15:43.28 | A2Sheds | it's good to have friends to share with in Taiwan |
15:44.00 | gordan | If anybody wants to contribute benchmark results, I will happily accept them for the list. |
15:44.15 | lkcl | A2Sheds: i have someone who's willing to make an introduction to an appropriate company... it's just that we have to wait until next week |
15:44.20 | lkcl | jumps up and down |
15:46.51 | lkcl | eduprey: anyway, yes - the plan is to start the ball rolling, to get a SoC made that will be out some time mid-2013: 8-core SMP using Xtensa RISC cores running at over 1.2ghz, in 28nm. it'll be entirely software-programmable. unlike AMD or NVidia CPUs, it will be absolutely tiny, so there will be 1000s per wafer. |
15:49.28 | lkcl | for interfaces, i've managed to cram USB-OTG, USB2 and USB3, SATA, RGMII (gigabit ethernet), 2x LVDS, RGB/TTL and 2x PCIe (x2) |
15:49.49 | A2Sheds | business has decided to make money by gambling vs manufacturing, so there is little incentive to actually promote science and engineering |
15:50.28 | lkcl | HDMI, 3 UARTS, NAND Flash 8-bit, 76-pin DDR3, I2S, AC97 and a few others. |
15:51.12 | lkcl | it's massively multiplexed (4 functions per pin for 50% of the pinouts) so it's crammed down to 304 pins which is ridiculously low. |
15:51.50 | lkcl | A2Sheds: i don't know whether to laugh or cry at that one. |
15:51.53 | A2Sheds | and if you look into the internal dynamics of the large semiconductor OEM's you'll wonder how they even stay in business |
15:52.30 | lkcl | A2Sheds: normally i would say "please don't tell me", but i _do_ want to know, so i don't make the same mistakes. |
15:55.03 | A2Sheds | I'll put up the red flags and signal flares when you start asking people what they are planning on doing with the EOMA68 cards before they can buy any |
15:56.31 | A2Sheds | customer: "Hello company X, we'd like to trade you $$$ for chips" |
15:56.35 | lkcl | A2Sheds: whoops, i already did that :) i asked people to put something on the preorders, but in my defense it was more so that other people could see what other people are doing with them |
15:57.31 | A2Sheds | company X : "What do you need them for? What will you make with them? What market? Who are your customers?" |
15:57.58 | lkcl | ah yes... |
15:58.00 | A2Sheds | customer " We'd like to trade you $$$ for chips" |
15:58.11 | lkcl | oh wait... hang on... i think i get it. |
15:58.20 | A2Sheds | company X " Sorry we don't do business like that" |
15:58.25 | lkcl | something just occurred to me. |
15:58.30 | lkcl | the fuckers are a cartel. |
15:58.46 | A2Sheds | yeah, in many ways |
15:58.47 | lkcl | they're a fucking cartel. |
15:59.19 | lkcl | they've already divvied up the market, and if you don't fit what they've decided between them you can fuck right off. |
15:59.19 | A2Sheds | price fixing is another thing they do |
15:59.27 | lkcl | no wonder we're not getting anywhere. |
15:59.30 | A2Sheds | yup |
15:59.38 | lkcl | *sigh* |
16:00.30 | lkcl | doesn't matter that it's illegal. |
16:00.55 | lkcl | the EOMA initiative doesn't fit _any_ of that, because it's a general-purpose standard. |
16:01.05 | lkcl | any processor, in any system. |
16:01.34 | lkcl | *sigh* well, awareness of the problem is half the solution. |
16:01.52 | gordan | Well, bottom line - screw them. If the EOMA project rattles their cages - GOOD. |
16:02.01 | lkcl | :) |
16:02.05 | gordan | There is little they can provide that Allwinner can't. |
16:02.11 | gordan | So seriously - why bother. |
16:02.31 | gordan | Get a product out with The A10 and the JZ4760 and that should be most things covered. :) |
16:02.33 | lkcl | well... i don't want it to be said that i didn't offer them a chance, that's all, gordan. |
16:02.44 | A2Sheds | the same is true in the printer market |
16:02.54 | gordan | Fair enough. |
16:03.02 | A2Sheds | cost of ink, getting cartridges and printheads |
16:03.22 | gordan | But seriously, us lot in the OSS community need to stop bending over backwards to help save these companies from themselves. |
16:03.38 | gordan | On one hand, yes, I'd love to see an OSS Tegra Xorg driver. |
16:03.43 | lkcl | achh don't get me started, A2Sheds :) DMCA takedown notices for companies doing compatible printheads??? |
16:03.45 | A2Sheds | it's all controlled by NDA's and by partnerships to lock in the supply chains |
16:04.11 | gordan | OTOH, that is only until one of them opens up their spec so we can have a proper OSS driver based on that. |
16:04.24 | gordan | After that, the rest of them can go to hell. |
16:04.27 | lkcl | gordan: when there's no choices... |
16:04.47 | lkcl | yes, that's why that MALI 400MP reverse-engineering is so damn important. |
16:04.48 | gordan | Indeed, at the moment we're pretty much stuck with 4 equally bad choices. |
16:05.26 | gordan | In fairness, the Tegra2 stuff seems to map the hardware functionality and GLES functionality pretty much 1:1, so duing a driver based on that would be a no brainer. |
16:05.34 | lkcl | hence the reason why i am starting an initiative to get a decent SoC manufactured. it'll be entirely software-programmable and will NOT come with any DRM features. |
16:06.16 | gordan | And what GPU will be in it? |
16:06.31 | A2Sheds | at what point the governments step in and requires a backdoor is yet to be seen |
16:07.10 | lkcl | gordan: no GPU - just a fuck-off massive SIMD floating point unit... per CPU |
16:07.36 | gordan | You do realize that won't actually be competitive performance-wise with a proper GPU, right? |
16:07.37 | A2Sheds | stream processor |
16:07.39 | lkcl | A2Sheds: well, they can fuck off too.... or they can go fiddle with the masks while i'm not looking, i don't care if they do that, to be honest |
16:07.55 | lkcl | gordan: http://www.mips.com/products/architectures/mips-3d-ase/ |
16:09.28 | A2Sheds | I'm sure the gpu driver devs will chime in on what would be nice to have as well |
16:09.37 | lkcl | they basically divide the problem down into 2... |
16:09.42 | lkcl | A2Sheds: ah good point. |
16:10.02 | lkcl | ohh, that reminds me: i _did_ actually ask, on various lists, last year... didn't get any traction. |
16:10.19 | lkcl | but that MIPS-3D-ASE document goes into quite a lot of detail. |
16:10.23 | A2Sheds | they didn't take you seriously |
16:10.25 | gordan | Seriously, if you're baking a custom chip wouldn't it be a good idea to put something like an OGP GPU on it? |
16:10.31 | gordan | http://wiki.opengraphics.org/tiki-index.php |
16:10.42 | lkcl | 1) floating-point - the critical operation is 1/(x*x). |
16:11.23 | lkcl | 2) shader "edge detection" needs 4-word "AND" to detect zeros as fast as possible and 4-word "NOR" to detect all 1s as fast as possible |
16:11.35 | lkcl | gordan: yeahh, that project's not finished. |
16:12.03 | A2Sheds | the OGP project designed around the constraints of working with FPGA's |
16:12.10 | lkcl | plus, the cost of the silicon to do it, you need an _actual_ CPU to back it up. why not just put the functions directly into the CPU itself? |
16:12.14 | gordan | Better than nothing. |
16:12.22 | gordan | At the end of the day - it doesn't _really_ matter. |
16:12.42 | gordan | If it'll do video playback at HD res without stuttering everything else is secondary. |
16:12.50 | gordan | 3D functionality is overrated. |
16:13.06 | gordan | And it is only relevant to a tiny, tiny number of ARM users anyway. |
16:13.30 | lkcl | gordan: you should look at tensilica's video DSP extensions. they're just... stunning. |
16:13.37 | A2Sheds | hardware video decoders are no problem |
16:13.41 | gordan | If I'm playing a game, I do so with all knobs turned up to 11 on a machine that burns through the thick end of a killowatt. |
16:13.58 | A2Sheds | ATI uses tensilica |
16:14.01 | lkcl | yeah precisely, gordan! |
16:14.07 | gordan | But that happens for 1-2 hours every 1-2 weeks. |
16:14.56 | lkcl | A2Sheds: i don't think there isn't a significant company that _doesn't_ license Xtensa. i laughed yesterday when i saw a block diagram of an Atheros WIFI IC with an Xtensa RISC core on it |
16:15.23 | lkcl | http://www.tensilica.com/products/video.htm |
16:15.52 | A2Sheds | once my dilithium crystal battery is done, we can have the ultimate petaflop handheld game console (with built in cooling system) |
16:17.00 | A2Sheds | lets examine the world 6-9 months from now |
16:17.47 | A2Sheds | if 1 million eoma68 modules are in use, they will hardly notice |
16:18.07 | lkcl | A2Sheds: *lol* |
16:18.20 | lkcl | has to make some food for lilyana, brb |
16:19.00 | A2Sheds | they will still be selling their $800 quad core tablets to people that need to feel better about themselves by buying one |
16:22.28 | A2Sheds | the technology that Apple has developed is now the tech inside the iwhatever, it is the mind control technology that gets people to settle for their stuff |
16:22.36 | A2Sheds | not/now |
16:27.03 | gordan | Yes, but since such people only use Apple hardware why should we care? It's not like they are normal people. :) |
16:27.21 | gordan | Oh wait - I forgot that the definition of "normal" has changed... |
16:27.24 | gordan | As per the link above. |
16:29.53 | A2Sheds | heh |
16:31.26 | A2Sheds | it really is evil genius |
16:35.02 | A2Sheds | like this mindset http://i.usatoday.net/news/opinion/cartoons/2009/December/e091207_pett.jpg |
16:35.47 | lkcl | i love it! i think i will make that my .sig :) |
16:36.28 | lkcl | rhombus-tech: what if it's a big hoax and we create better technology, for nothing? |
17:16.51 | A2Sheds | last week I had a vendor tell me that they were only focusing on Android support so Linux might tie up their resources if there are problems |
17:17.06 | A2Sheds | I had a hard time not laughing out loud on the phone |
17:53.28 | eduprey | lkcl: wow, I'm impressed with the SoC plans, I would have considered it out of reach of a project like this but I suppose it all depends on volume and the skill of the people contributing |
17:54.43 | eduprey | surprised about QFP too, that would make it a lot more home solderable, though I suppose you'd still need multi layer board for routing signals |
18:02.03 | lkcl | eduprey: it's a matter of building blocks. the actual design work is so complex that you simply can't do things any other way. |
18:02.24 | lkcl | _actual_ design is about 10 gates per day per person |
18:02.40 | lkcl | so you pick off-the-shelf parts and literally sling them together. |
18:03.12 | lkcl | throw them through the (massively expensive) simulation and design tools, and out pops a cpu at the end. |
18:03.57 | lkcl | if you've seen the stuff on http://opensparc.net you can even, gosh, wow, specify some macros which say "how many cores, how large the cache, is there an FPU" etc. |
18:04.26 | lkcl | but Xtensa take that to the absolute Nth degree extreme, hence why i was so excited to find their stuff |
18:05.29 | lkcl | yes it pretty much goes without saying that you'd need a multi-layer board :) you _might_ however be able to get away with a 4-layer because, unlike with a BGA, the pins aren't so ridiculously close. |
18:06.03 | eduprey | yeah, that's actually hobbyist-approachable |
18:07.11 | eduprey | I'm not sure how massively multiplexing the pins works though, do you have to have some kind of demuxing step afterward, or only use one or another feature |
18:08.28 | eduprey | <PROTECTED> |
18:10.18 | lkcl | eduprey: it's becoming increasingly common practice. samsung's SoCs have to use it, TI's as well. there simply isn't enough space otherwise. |
18:10.59 | lkcl | ah no basically you pick what functions you want the pins to be, connect them to the hardware and then in software you say "right, these pins are a UART" |
18:11.14 | eduprey | ah ok, cool |
18:11.25 | lkcl | because, connected to those pins in the hardware is a device which talks RS232. |
18:11.41 | lkcl | i mean you _could_ multiplex externally, but what would be the point? |
18:12.13 | eduprey | lkcl: so I assume the simulation software has to be really really good, because actually fabricating bad dies would be an incredibly expensive blunder? |
18:12.21 | lkcl | you'd need an external multi-channel routing IC (eurgh) which could cope... |
18:12.24 | gordan | This may sound crazy, but what about standardizing on a CPU socket? |
18:12.28 | lkcl | eduprey: yyyep, you got it. |
18:12.37 | lkcl | gordan: oh ah thank you for reminding me! |
18:12.46 | lkcl | must add that to the elinux.org page |
18:13.01 | gordan | What I have in mind is something like a modern x86 CPU socket. |
18:13.05 | lkcl | we had a crazy idea about re-use of something like AMD Socket 9 or whatever |
18:13.06 | lkcl | yes :) |
18:13.17 | eduprey | nice |
18:13.17 | lkcl | but, obviously, not pin-compatible |
18:13.30 | gordan | You have your actualy CPU that is tiny, soldered onto a PCB that fits into a big fat socket with sane pin spacing that breaks it all out. |
18:13.53 | gordan | You can then route out of that large pin pitch micro-mobo (for want of better word) onto a full size mobo. |
18:13.56 | lkcl | or, did you have in mind actually being pin-compatible with x86? (hope not!) |
18:14.00 | gordan | LOL! |
18:14.26 | gordan | I'm not sure there would be that much mileage in it, but the idea may not be as utterly daft. |
18:14.41 | eduprey | gordan: well why not just have the actual CPU packaged in a case with pins like x86 stuff is? |
18:14.53 | gordan | I hadn't thought of it, but now that you mention it, it would make sense if the SoC has PCIe and 64-bit RAM channel so it could take normal DIMMs. |
18:15.26 | gordan | eduprey - that's sort of what Luke and I are talking about here. |
18:15.56 | gordan | A modern x86 CPU is a small chip on a large PCB with a shiny metal heat plate on it. |
18:16.18 | eduprey | heh, fair enough, didn't consider that |
18:16.26 | lkcl | yeahh, that's the 2nd IC - 64-bit DDR3 but that needs 152 pins for signals, 30 pins for power and 30 for GND |
18:16.48 | gordan | The main problem is that since x86 doesn't have on-board USB and everything else that SoCs do, you couldn't make it pin compatible. |
18:16.55 | lkcl | it's a f*** of a lot of pins for DDR RAM. |
18:17.18 | lkcl | gordan: yeah i know. which increases cost, because now you need an external hub controller chip. |
18:17.20 | gordan | Yes, but if you have an x86 style socket with lots of pins, then it doesn't really matter. |
18:17.37 | gordan | Having an external hub controller for a SoC is just silly. :) |
18:17.42 | lkcl | oh, and because you now *have* to have 64-bit RAM bus-width, you've now just massively increased the power consumption. |
18:18.10 | gordan | "We have a chip that does everything, so we won't use any of it's extra features and instead have those features in an external chip." |
18:18.14 | lkcl | ok, not massively. well... yes, massively: it's about 400mA or something |
18:18.17 | lkcl | hurrah! :) |
18:18.27 | lkcl | drops an own-goal.... |
18:18.36 | lkcl | hmmm... |
18:18.55 | gordan | So no, I don't think x86 compatibility would make sense. :) |
18:19.00 | lkcl | *lol* |
18:19.16 | gordan | But that doesn't mean you couldn't use the same socket. |
18:19.18 | lkcl | however... re-using the low-cost x86 sockets etc. does. |
18:19.20 | lkcl | yehhh |
18:19.26 | gordan | :) |
18:20.11 | gordan | As long as the socket allows for all the features of any plausible currently available or maybe in the future available SoC, then it doesn't matter. |
18:20.30 | lkcl | oh god. 800-pin compatibility. |
18:20.32 | lkcl | quails |
18:21.00 | gordan | e.g. enough USB lines, enough GPIO lines, GPU, PCIe, etc, etc, etc. |
18:21.28 | gordan | BTW, re: RAM and 64-bitness - how wide are the RAM channels on the A10 (and other similar SoCs for that matter)? |
18:21.58 | gordan | Could you not do a bodge to invserse-multiplex it so you can address the 64-bit RAM bank in two 32-bit chunks? |
18:22.07 | lkcl | 32-bit |
18:22.12 | lkcl | no! :) |
18:22.16 | eduprey | this project is totally mind blowing, I was just getting excited about EOMA68 and building a small cluster with it when I saw the EOMA-CF stuff and was like, oh wait, even better / smaller, and you guys are even talking about SoC's for next year. I don't think it's overstating things to say this is going to change the world at least re OSS and open hardware |
18:22.33 | gordan | lkcl: Why not? |
18:22.36 | lkcl | eduprey: well, it's just a logical step. |
18:23.14 | lkcl | gordan: the signals are running at 400mhz, they're balanced line pairs that are required to be accurate to under 0.5mm |
18:23.21 | eduprey | lkcl: yeah, well looking at it this way, I'm just amazed that nobody's done it yet now |
18:23.24 | lkcl | you _really_ don't want to f*** with that. |
18:23.45 | gordan | What about using PC133 SDR RAM? |
18:23.48 | lkcl | eduprey: it's either confidence or stupidity :) |
18:24.07 | lkcl | gordan: ooo, ouch. 10x slower RAM. ouch |
18:24.12 | lkcl | brb, daughter fell over.... |
18:24.23 | gordan | Oh dear... |
18:25.04 | gordan | Well, yes and no re: 10x slower. |
18:25.11 | gordan | DDR is actually rated at half the MHz speed. |
18:25.18 | gordan | i.e. DDR2 666 == 333 MHz. |
18:25.36 | gordan | And my ARM machines generally manage 300-400MB/s in terms of RAM I/O. |
18:26.02 | gordan | This is very similar to my old 1GHz P3 machine that had PC133 RAM in it. |
18:26.29 | lkcl | well... yes and no, gordan :) |
18:26.36 | gordan | Anyway, if you're looking at, say, DDR2-533, that is only 2x the clock speed of PC133 SDR DIMMs. |
18:27.05 | lkcl | no, that would be a bit silly to have in 2013 |
18:27.18 | gordan | Probably would be more expensive, too. |
18:27.23 | gordan | Just an idea, though. |
18:27.30 | lkcl | i think this IC needs to aim for a minimum of DDR3-1333 (667mhz x2) |
18:27.32 | gordan | Being able to use normal DIMMs would be a huge boon. |
18:28.06 | lkcl | depends on the market. if the market being aimed for is $3.50 per CPU then having DIMM sockets is a bit... silly |
18:28.07 | gordan | Anyway, we are getting way ahead here. |
18:28.29 | lkcl | but yes there will be 2 versions. |
18:28.57 | gordan | Since none of the SoCs can handle more than 4GB of RAM anyway, at current DRAM prices you might as well just max it out for the extra few $ and be done with it, no need to have it upgradable because it's not going to work anyway. |
18:29.14 | lkcl | one BGA probably about 400 pins |
18:29.30 | lkcl | yyep. |
18:29.45 | lkcl | *sigh* yes the Xtensa RISC cores are all 32-bit. |
18:29.52 | lkcl | can't have everything. |
18:30.15 | gordan | 64-bitness is overrated anyway,. |
18:30.32 | gordan | In general, if your application doesn't fit into 3GB of RAM, you are most likely doing it wrong. |
19:00.53 | A2Sheds | cpu soc die in a POP with DDR3 on top, the only issues with that is cooling the cpu sandwich |
19:04.30 | gordan | It's not really that much of an issue on ARM. |
19:04.48 | gordan | Most POP setups with ARMs don't even have a heatsink. |
19:05.23 | gordan | Going from no heatsink to a heatsink makes a _huge_ difference, as does going from completely passive cooling to even a very slow, quiet fan. |
19:06.12 | lkcl | well the Xtensa CPU core i worked out that each one would be about 50mW in 28nm at 1ghz |
19:06.19 | lkcl | x8 that's 400mW for 8 cores. |
19:06.43 | gordan | You don't think 8 cores is a little excessive? |
19:06.45 | lkcl | so the whole thing would be pushing about 1 to 1.5 watts |
19:07.04 | gordan | 4 cores is still more than 90% of use-cases would benefit from. |
19:07.04 | lkcl | for a CPU in 2013? naah. |
19:07.15 | gordan | Remember the 4GB of RAM limitation. |
19:07.25 | gordan | There's only so many things you'll be running before you run out of RAM. |
19:07.52 | lkcl | yeah, but this is for a CPU that could be doing the job of a GPU _and_ 1080p video decode _and_ running the OS as well |
19:08.09 | gordan | Not at the same time, though. |
19:08.30 | gordan | It's not like you'll be playing 3D game, while playing a 1080p video AND compiling the kernel at the same time. |
19:08.50 | A2Sheds | 8 50mW cores + DDR3 would work, all the heat will be from the DDR3 |
19:09.06 | lkcl | A2Sheds: pretty much... yeah. |
19:09.13 | A2Sheds | you have to design for worst case |
19:09.27 | gordan | As long as you have decent thermal compound between the sandwich layers, a heatsink will solve any probmlems you may or may not have. |
19:09.30 | A2Sheds | if you want it to last more than a few months |
19:09.37 | gordan | Nah. |
19:09.48 | gordan | Seriously, heat isn't a problem. |
19:09.54 | A2Sheds | ever notice why Sat receivers die so quickly that use broadcom? |
19:09.56 | lkcl | and the amount of area per core, it's like a tiny blob in the top left hand corner, under 1% of the actual die area |
19:10.09 | A2Sheds | no heat sinks on the chips |
19:11.28 | A2Sheds | electromigration kills them soon after the end of the warranty |
19:13.36 | lkcl | ouch |
19:13.59 | lkcl | TV etc. products are notorious for being massively power-hungry. |
19:14.05 | lkcl | relatively speaking |
19:14.09 | A2Sheds | add a fan or heatsinks and they last a few years |
19:14.58 | A2Sheds | so we have to use real crappy cases for eoma68 or people won't buy eoma68+ next year :) |
19:18.36 | ManoftheSea | wow, talking |
19:18.52 | lkcl | ManoftheSea: hehe |
19:19.07 | lkcl | A2Sheds: tsk tsk |
19:21.45 | A2Sheds | so whats the toolchain for tensilica cores? |
19:22.36 | A2Sheds | who gets to write the compiler? |
19:22.50 | mnemoc | anyone interested in making "industrial" eoma products? or it's 100% CE? |
19:23.12 | *** mode/#arm-netbook [+ooo A2Sheds lkcl gordan] by mnemoc |
19:23.21 | A2Sheds | mnemoc: let me know |
19:23.47 | mnemoc | A2Sheds: meaning? |
19:23.59 | A2Sheds | the package is only qualified for CE, but it doesn't really mean much unless you need it for GM or the military qual |
19:24.29 | A2Sheds | the A10 can be made to operate well into industrial temp range |
19:25.17 | A2Sheds | GM = general motors etc, automotive temp range |
19:25.59 | mnemoc | i mean din-rail serial controllers, dumb and ugly but heavy duty touch panels, etc |
19:26.43 | mnemoc | for factories, warehouses, ... |
19:26.59 | A2Sheds | sure |
19:27.22 | A2Sheds | I've been waiting for a suitable soc for EMC for machine controllers |
19:27.52 | mnemoc | does the a10 qualify? |
19:28.03 | A2Sheds | http://code.google.com/p/miniemc2/ |
19:28.12 | A2Sheds | mnemoc: that is the plan! |
19:28.16 | mnemoc | :D |
19:28.25 | mnemoc | CNCs are fun |
19:28.30 | eduprey | A2Sheds: yep, looking forward to working on that too |
19:28.38 | A2Sheds | the AMD APU version as well |
19:29.09 | mnemoc | die x86 die |
19:29.23 | eduprey | A2Sheds: I think EOMA-CF would be even better for the electronics baord form factor we want for 3d printer hardware but EOMA68 will be fine to start with |
19:29.42 | A2Sheds | i heard there was a high volume LCD panel already considering the A10 for CE use |
19:30.13 | A2Sheds | eduprey: yeah, for the price. Other cards will follow |
19:30.29 | A2Sheds | LCD panel vendor that is |
19:30.30 | mnemoc | the upgradability of EOMA is a very interesting point in this field were the LCD panel itself is by far the most expensive part |
19:30.31 | lkcl | A2Sheds: yes. monitor company. have to wait for them to get back from chinese new year before confirming, but yes they were definitely excited |
19:30.50 | lkcl | mnemoc: ironic, isn't it. |
19:30.53 | eduprey | EOMA68 + LCD panel = full computer |
19:30.59 | A2Sheds | yes, just swap the eoma68 card and you upgrade the power, features etc |
19:31.20 | lkcl | eduprey: now you're getting it |
19:31.24 | A2Sheds | but industrial versions are planned |
19:32.00 | mnemoc | lkcl: but sad that the focus seems to be in fancy HD, SPIF, ... |
19:32.04 | A2Sheds | industrial touch panels, POS systems (point of sale) etc |
19:32.19 | lkcl | eduprey: what's hilarious is that you can create a "dumb monitor card"... which is actually in EOMA-CF or EOMA-68 form-factor. |
19:32.21 | A2Sheds | ever price a POS touch panel system? |
19:32.41 | mnemoc | the current prices are absurd |
19:32.42 | A2Sheds | they generally are a P.O.S. and cost >$500 |
19:32.52 | mnemoc | and include very junky x86 machines |
19:33.03 | mnemoc | it makes me nuts |
19:33.03 | lkcl | that you can swap out with a computer.... but that "dumb" card, with its HDMI in etc. could actually be plugged into a laptop! |
19:33.51 | lkcl | and you could use its screen on another computer :) hell, if you put a USB connector on the "dumb" card as well you could even use the laptop's keyboard and mouse.... on another machine as well. |
19:34.16 | lkcl | ok, well, i think that's hilarious, anyway. |
19:34.20 | eduprey | heh |
19:35.07 | A2Sheds | I have to check the google wayback machine, there used to be modular laptop vendors that were driven out of business |
19:35.17 | eduprey | yeah I wonder about mainstream laptop and tablet manufacturers, they stand to sell fewer upgrades if people can upgrade the computer part of their tablet with a card swap |
19:35.28 | gordan | It's almost like applying "reductio ad absurdum" to the concept of a computer. |
19:35.34 | A2Sheds | yeah, thats the mindset $$$ |
19:36.35 | A2Sheds | http://techreport.com/discussions.x/19013 not the best design, but you get the idea |
19:36.36 | eduprey | but obviously swapping it is way more "sustainable", environmentally friendly etc, why continually create obsolete stuff when screens keyboards and batteries are fundamentally the same as they've been for many years |
19:37.12 | lkcl | eduprey, but they are being absolutely killed on profit margins right now. the bottom's dropped completely out the tablet market (thanks to the A10) |
19:37.17 | A2Sheds | remember VIA's BS openbook initiative? |
19:37.29 | lkcl | dell's abandoned netbooks, and so have a couple of other companies. |
19:37.55 | A2Sheds | http://www.viaopenbook.com/ |
19:38.00 | A2Sheds | not very open |
19:38.18 | A2Sheds | also not modular |
19:38.20 | lkcl | at least if they sell modular parts, they stand a chance of re-entering a market that is moving so f****g fast it's a blur |
19:38.53 | A2Sheds | Intel pretty much dictates what gets built in notebooks |
19:39.40 | A2Sheds | look at the ultrabook drama |
19:40.20 | A2Sheds | http://en.wikipedia.org/wiki/Ultrabook |
19:40.53 | eduprey | hmm wonder how hard it would be to open up an older laptop, take out system board, wire into LVDS screen (guess I would need a serializer IC) and keyboard etc |
19:41.04 | mnemoc | i totally fail to understand why G made the chromebooks use atom |
19:41.42 | gordan | They used to work on ARM. |
19:41.58 | lkcl | eduprey: quickembed will do a motherboard design doing exactly that, for $USD 2,000 |
19:42.02 | gordan | A lot of patches that made AC100 work so well with full fat Linux come from the ChromeOS kernel. |
19:42.06 | A2Sheds | eduprey: it's possible but we have already talked to netbook vendors about retooling |
19:42.32 | eduprey | nice |
19:42.49 | gordan | Also I was looking at what it would take to take something like a Pandaboard and re-wire a laptop chasis of something big like a Clevo M860TU to fit it in. |
19:42.54 | gordan | It wouldn't be all that difficult. |
19:42.59 | A2Sheds | so eoma68 may be easliy dropped into existing designs |
19:43.15 | gordan | But the problem is that you're still limited to 1GB RAM because that's what's on the Pandaboard. |
19:43.24 | gordan | And you're still limited on battery life by the TFT screen. |
19:43.35 | gordan | So the actual net gain for an awful lot of work isn't that much. |
19:44.08 | A2Sheds | the panda was designed not to be used in products |
19:44.19 | ManoftheSea | this is all facinating |
19:44.23 | ManoftheSea | just caught up. |
19:44.23 | gordan | You're better off modifying an AC100 with a higher res screen, a bit of OC-ing and other upgrades I've come up with. It's more cost effective. |
19:44.31 | A2Sheds | it was a way to get free development for their platform and charge devs for the boards at the same time |
19:44.42 | ManoftheSea | not only are screens not changing, they're going backwards. |
19:44.55 | ManoftheSea | The older generation of Dell screens went 1920 x 1280. |
19:44.56 | gordan | It's actually not bad as far as using it in products goes. |
19:45.02 | ManoftheSea | Now, it's like 1366 x 1080 |
19:45.13 | gordan | Yeah, everybody has been cutting that corner of late. |
19:45.15 | A2Sheds | that is my only resistance to a beaglebone respin, why help TI? |
19:45.34 | gordan | This is why I very rarely change my laptops. |
19:45.46 | ManoftheSea | Oh, there was a mention of no DRM... |
19:45.54 | gordan | The only recent one that I deemed worthy was a Clevo M860TU, that has a 1920x1200 15.4in screen. |
19:45.57 | ManoftheSea | I, for one, want a TPM on my computer. |
19:46.11 | ManoftheSea | A TPM that I can own, to be sure... |
19:46.14 | eduprey | treacherous platform module |
19:46.20 | lkcl | ManoftheSea: cost-cutting galore. the low margins are eating them alive. |
19:46.21 | ManoftheSea | But encryption is the only protection available to data-at-rest. |
19:46.31 | eduprey | heh but yeah using it to ensure your own stuff is signed by you can be useful |
19:46.55 | ManoftheSea | The TPM tries to handle the Evil Maid attack. |
19:46.57 | gordan | And it went to my girlfriend. I'm still using my 5 year old ThinkPad T60. It's decent enough spec (2.33GHz dual Core2, 3GB of RAM) and I have fitted a 2048x1536 panel into it. |
19:47.05 | ManoftheSea | not that it's solved yet... |
19:47.36 | A2Sheds | I guess an A10 eoma68 version with an FPGA could handle the machine control and NAND-DDR3 interface |
19:48.06 | A2Sheds | 1GB ddr3 main memory with 4GB DDR3 ram drive |
19:48.24 | gordan | The only thing that'll come anywhere near displacing my ThinkPad is the new Transformer Prime HD _if_ that ever gets releases (1920x1200 in 10", but I'll have to see if I can live with the pixel count shrinkage for serious work. |
19:48.27 | A2Sheds | + IO for servos, steppers etc |
19:48.41 | ManoftheSea | eduprey: as you said: I'd want to be sure the boot environment is "good" for when I open up my "personal life computer" |
19:49.02 | ManoftheSea | Where I store my world domination plans, of course. |
19:49.18 | A2Sheds | I wish vendors of LCD would make smaller pitch screens, it has stalled at 1920 |
19:49.33 | A2Sheds | and for what 20" and > |
19:50.01 | A2Sheds | 1920x1200 in 10" would be fantastic! |
19:50.02 | gordan | Not true. |
19:50.13 | gordan | 1920x1200 is available on 15.4in. |
19:50.21 | gordan | And 1920x1080 in 15.6in |
19:50.33 | gordan | I have a 2048x1536 on my 15" ThinkPad. |
19:50.39 | A2Sheds | nice |
19:50.50 | gordan | Sony Vaio Z series has 1920x1080 on a 13" panel. |
19:50.58 | A2Sheds | gordan: who makes those panels? |
19:51.02 | gordan | Which ones? |
19:51.14 | A2Sheds | the 3 mentioned above |
19:51.32 | A2Sheds | sony probably makes the sony :) |
19:51.39 | gordan | I _think_ the 15.6" 1920x1200 is a Samsung. |
19:52.40 | A2Sheds | "Chimei Innolux (CMI) CEO Hsing-chien Tuan will lead a negotiation team to hold debt settlement talks with its creditor banks on January 20" :( |
19:53.11 | ManoftheSea | I also like smart cards. |
19:54.00 | A2Sheds | once we get the cards out we should attract more top developers |
19:54.07 | A2Sheds | everybody wants these |
19:54.35 | gordan | The ThinkPad panel is IDTech. |
19:54.45 | gordan | http://www.thinkwiki.org/wiki/Installing_a_QXGA_display_in_a_R/T60_or_61 |
19:55.18 | A2Sheds | http://www.linuxfordevices.com/c/a/News/VDC-COM-and-mezzanine-report/ |
19:55.30 | A2Sheds | good timing as well |
19:55.46 | A2Sheds | gordan ; thanks |
19:56.56 | A2Sheds | and that report was written without any knowledge of EOMA68 |
20:00.32 | lkcl | iiinteresting. |
20:01.49 | A2Sheds | is the official launch still planned for April 1st :) |
20:02.03 | A2Sheds | talk about a bad date for a product launch |
20:03.19 | A2Sheds | webos isn't going to be released into the wild until the fall at the earliest, anybody taken a look at it yet? |
20:04.16 | A2Sheds | mnemoc: http://www.linuxfordevices.com/c/a/News/Ibase-IPPC1500/ only with a EOMA68 slot |
20:05.36 | mnemoc | :) |
20:14.33 | eduprey | A2Sheds: I'm a long time webos hacker |
20:14.48 | eduprey | A2Sheds: getting webos going on EOMA68 stuff is actually one of the things I want to do |
20:17.20 | A2Sheds | eduprey: are many apps available? |
20:17.56 | eduprey | A2Sheds: not compared to android but yeah a pretty decent number now |
20:18.10 | A2Sheds | http://www.webosnation.com/app-gallery this a good overview |
20:18.24 | A2Sheds | or is there a better app "market"? |
20:19.58 | eduprey | A2Sheds: yeah unfortunately I don't know of a good way offhand to show you HP's app catalog without a webos device |
20:20.06 | A2Sheds | eduprey: does it run on top of Linux or how is it bastardized? |
20:20.35 | eduprey | I'm part of a project called webos internals -- see http://www.webos-internals.org/wiki/Main_Page and freenode #webos-internals |
20:20.50 | A2Sheds | eduprey: Thanks |
20:21.38 | eduprey | A2Sheds: it's Linux based on openembedded but the UI (Luna) is a currently closed source binary called LunaSysMgr |
20:21.51 | A2Sheds | eduprey: if you have working linux kernel source it's pretty much guaranteed to work with webos? |
20:21.56 | mnemoc | i like tizen's idea that it's better to provide an standarized API for javascript for accessing the hardware and make all apps be html5 instead of "native" |
20:22.42 | eduprey | A2Sheds: yeah when we get the webos source code. without webos source code we still managed to get it working on shooter (evo3d) so some progress can be made with only binaries |
20:22.47 | mnemoc | it makes no sense to keep adding new app platforms |
20:23.01 | eduprey | mnemoc: that's webos's idea too |
20:23.11 | eduprey | and has been for a lot longer. See Enyo |
20:23.25 | eduprey | Enyo is currently open source |
20:23.28 | A2Sheds | neat! |
20:25.01 | A2Sheds | I've been wondering how low we could keep the BOM cost for a smartphone using eoma68 |
20:25.10 | eduprey | A2Sheds: so people are running ubuntu chroots and X apps via opengl-es |
20:25.15 | eduprey | in webos |
20:26.46 | eduprey | A2Sheds: the palm pre uses a comm board that can be swapped between devices, pre, pre+ and pre2 all used the same comm board and I think it's just HSUSB or maybe HSUSB + SDIO |
20:27.14 | eduprey | so that's one idea, maybe using a 3g dongle would be another, people have gotten huwei 3g dongles to do voice, SMS, etc |
20:28.22 | eduprey | http://code.google.com/p/datacard/ http://code.google.com/p/asterisk-chan-dongle/ for example |
20:29.16 | eduprey | basically just thinking of ways to get around certification issues to start with |
20:31.01 | A2Sheds | I'm thinking millions per year |
20:31.27 | eduprey | A2Sheds: ahh so yeah I'd be interested in owning one of those :) |
20:31.30 | A2Sheds | you build a community around the cards |
20:31.58 | A2Sheds | people can swap them when newer/bigger/better/faster cards come along |
20:31.58 | eduprey | though wouldn't EOMA-CF be smarter for smartphones? |
20:32.20 | A2Sheds | people can swap netbook parts as well |
20:33.02 | A2Sheds | bigger drives, different screen, you don't toss the old one you trade or sell to somebody who only wants/needs the older one |
20:33.49 | A2Sheds | eduprey: maybe |
20:34.59 | eduprey | A2Sheds: hardest thing is getting EOMA stuff into common retail channels |
20:35.10 | eduprey | but I guess there are manufacturers interested |
20:35.13 | A2Sheds | in the west |
20:35.29 | A2Sheds | it's different in India or China |
20:35.35 | eduprey | yeah definitely |
20:36.04 | A2Sheds | the west has been trained to only recognize brands |
20:36.38 | eduprey | so makes sense to start with asia etc, and when popularity gets high enough there businesses that sell online to people here will sell them and when enough get bought that way bigger retailers may be interested |
20:36.55 | eduprey | and of course our own direct ordering etc |
20:37.05 | A2Sheds | 2.5 billion is a pretty big market to start with IMHO :) |
20:38.41 | eduprey | I think in china / india the idea of taking your computer on a card to plug into whatever will be novel and attractive enough that people will do it regardless of the fact that it makes open software development easier :) |
20:39.21 | eduprey | but for some esp in education the fact that it does make it easier to get started making things could be of huge benefit |
20:39.28 | eduprey | and obviously low cost is a big factor |
20:40.54 | A2Sheds | I've had people tell me that it will be a pain to have to run down to the car to get the card out of the carPC to plug into the desktop when they forget |
20:41.35 | eduprey | heh, here people plug their iphone into their car to charge it wihtout forgeting their phone when they get out |
20:42.01 | A2Sheds | in their world $50 or $100 is no big deal for an extra card, but when $50 is a months rent it's another story |
20:42.23 | eduprey | heh yeah 1st world stuff |
20:43.06 | A2Sheds | I politely just suggest they travel outside their bubble more often |
20:43.46 | eduprey | A2Sheds: do we have any guesses as to what a card will actually cost direct order once it's into volume? |
20:43.51 | eduprey | for EOMA-68 A10 |
20:44.44 | A2Sheds | some of the contacts at the factory come back to work in a few hours |
20:45.10 | A2Sheds | Luke will have that info |
20:45.15 | eduprey | I heard rough numbers saying probably below raspberry pi pricing |
20:45.47 | A2Sheds | yeah, that's a good ballpark |
20:46.24 | A2Sheds | we'll trim the BOM to best price/qual |
20:47.38 | eduprey | it's funny when I see people excited at SCALE about XBMC running on raspberry pi .. considering this coming :) |
20:48.29 | A2Sheds | i never got excited about the Pi myself, it's like the panda, not designed to be integrated into products |
20:49.17 | A2Sheds | and lots of my old PC's are only worth $25 and they run linux |
20:50.13 | eduprey | there's certainly a place for a product like raspberry pi for education or whatever, I just think it'll be the minimal eoma68 host board :) |
20:51.18 | eduprey | and then that school's upgrade path is easier and kids can do all kinds of other projects using the same computer |
20:51.24 | A2Sheds | time will tell on the Pi, I see it as another way for broadcom to get free linux development under the guise of an open project that isn't really open for hardware |
20:52.11 | A2Sheds | and for the Pi to be useful, you need to add a display + keyboard/mouse |
20:52.30 | A2Sheds | now you're up to the cost of the OLPC XO |
20:53.05 | eduprey | something like an OLPC laptop with an EOMA68 socket in it would be awesome |
20:53.24 | A2Sheds | I know the devs there... |
20:53.33 | A2Sheds | Marvell gave them $5M |
20:53.43 | eduprey | wow |
20:53.54 | A2Sheds | so I don't expect much change unless we spin a marvell card |
20:54.16 | eduprey | even open projects are bought and sold, it seems |
20:55.18 | A2Sheds | http://arstechnica.com/open-source/news/2010/10/olpc-gets-56-million-from-marvell-to-build-android-tablet.ars |
20:56.28 | eduprey | ah, One Tablet per Child then? bizarre |
20:56.51 | A2Sheds | I'm waiting for M$ to want to be on eoma68 |
20:57.37 | eduprey | let them sell windows 8 for A10 :) i don't want it but some people wil |
20:57.38 | eduprey | l |
20:57.56 | A2Sheds | I'm waiting to see what broadcom has up their sleeve |
20:59.04 | A2Sheds | I'm suspicious of their ability to even ponder the thought of altruism |