Maybe this is interesting for EOMA-68...
On Fri, Nov 13, 2015 at 7:50 PM, GaCuest gacuest@gmail.com wrote:
Maybe this is interesting for EOMA-68...
there's a practical/comfortable power guide of around 3.5 watts for any EOMA68 cpu card - 4.0 watts is slightly uncomfortable, and 5.0 watts is the hard external cut-off limit that will drop the power, stone dead.
i'm currently looking at a 64-bit quad-core ARM SoC that will have to be run at around 900mhz to stay within this limit. luckily, information's available on it so it's possible to use.
.... nvidia... yyeah, if someone can break through their barriers to entry, such that PCB CAD files and technical datasheets are made available, great, i can take a look.
l.
A pity it's got to be that low -- Atom Z2460 has a TDP of 4w... it's an SoC Atom from what I understand -- meaning no chipset, just add RAM, VReg, and a few interface bits.
But there's no way to keep the required external components less than a watt, I'm sure.
Worth noting, even VIA can't keep up with that. Closest is an Eden ULV 1.0GHz Esther core... it gets 3.5w TDP but that's before the chipset and everything else...
Oh well :(
Just a thought -- I forget -- are EOMA68 cards permitted to protrude from the slot any, the way PCMCIA/Cardbus cards almost always did? If so, it *may* be possible to wire up a small ventilation system -- fan with a bit of ducting is what I'm picturing, there's no room for a real heatsink in there, after all. But every little bit helps... ;)
On Fri, Nov 13, 2015 at 3:30 PM, Luke Kenneth Casson Leighton <lkcl@lkcl.net
wrote:
On Fri, Nov 13, 2015 at 7:50 PM, GaCuest gacuest@gmail.com wrote:
Maybe this is interesting for EOMA-68...
there's a practical/comfortable power guide of around 3.5 watts for any EOMA68 cpu card - 4.0 watts is slightly uncomfortable, and 5.0 watts is the hard external cut-off limit that will drop the power, stone dead.
i'm currently looking at a 64-bit quad-core ARM SoC that will have to be run at around 900mhz to stay within this limit. luckily, information's available on it so it's possible to use.
.... nvidia... yyeah, if someone can break through their barriers to entry, such that PCB CAD files and technical datasheets are made available, great, i can take a look.
l.
arm-netbook mailing list arm-netbook@lists.phcomp.co.uk http://lists.phcomp.co.uk/mailman/listinfo/arm-netbook Send large attachments to arm-netbook@files.phcomp.co.uk
On Fri, Nov 13, 2015 at 8:46 PM, Christopher Havel laserhawk64@gmail.com wrote:
A pity it's got to be that low -- Atom Z2460 has a TDP of 4w... it's an SoC Atom from what I understand -- meaning no chipset, just add RAM, VReg, and a few interface bits.
yep. intel's been told... but they keep taking the advantage of the geometry shrinks to increase speed, *not* reduce power.
so they're deliberately pissing in their own back yard, basically.
But there's no way to keep the required external components less than a watt, I'm sure.
32 bit 800mhz DDR3 (1.5v) RAM uses only around.... 300mA (0.3A), so it's doable.
Worth noting, even VIA can't keep up with that. Closest is an Eden ULV 1.0GHz Esther core... it gets 3.5w TDP but that's before the chipset and everything else...
yyep... and the eden is only in 45nm and isn't a very good design... *and* x86 has a performance/watt penalty inherent in its design, due to the CISC instruction set.... long story, i've written about this before.
Oh well :(
Just a thought -- I forget -- are EOMA68 cards permitted to protrude from the slot any, the way PCMCIA/Cardbus cards almost always did?
of course.
If so, it *may* be possible to wire up a small ventilation system -- fan with a bit of ducting is what I'm picturing, there's no room for a real heatsink in there, after all. But every little bit helps... ;)
graphite paper also helps.
l.
Maybe I'm being a little dense but I don't get the graphite paper reference/pun.
On Fri, Nov 13, 2015 at 3:54 PM, Luke Kenneth Casson Leighton <lkcl@lkcl.net
wrote:
On Fri, Nov 13, 2015 at 8:46 PM, Christopher Havel laserhawk64@gmail.com wrote:
A pity it's got to be that low -- Atom Z2460 has a TDP of 4w... it's an
SoC
Atom from what I understand -- meaning no chipset, just add RAM, VReg,
and a
few interface bits.
yep. intel's been told... but they keep taking the advantage of the geometry shrinks to increase speed, *not* reduce power.
so they're deliberately pissing in their own back yard, basically.
But there's no way to keep the required external components less than a watt, I'm sure.
32 bit 800mhz DDR3 (1.5v) RAM uses only around.... 300mA (0.3A), so it's doable.
Worth noting, even VIA can't keep up with that. Closest is an Eden ULV 1.0GHz Esther core... it gets 3.5w TDP but that's before the chipset and everything else...
yyep... and the eden is only in 45nm and isn't a very good design... *and* x86 has a performance/watt penalty inherent in its design, due to the CISC instruction set.... long story, i've written about this before.
Oh well :(
Just a thought -- I forget -- are EOMA68 cards permitted to protrude from the slot any, the way PCMCIA/Cardbus cards almost always did?
of course.
If so, it *may* be possible to wire up a small ventilation system -- fan with a
bit of
ducting is what I'm picturing, there's no room for a real heatsink in
there,
after all. But every little bit helps... ;)
graphite paper also helps.
l.
arm-netbook mailing list arm-netbook@lists.phcomp.co.uk http://lists.phcomp.co.uk/mailman/listinfo/arm-netbook Send large attachments to arm-netbook@files.phcomp.co.uk
On Fri, Nov 13, 2015 at 8:55 PM, Christopher Havel laserhawk64@gmail.com wrote:
Maybe I'm being a little dense but I don't get the graphite paper reference/pun.
reference. graphite is a heat conductor that's... i think it's something like over 100x more efficient than copper. graphite paper is used as a heat spreader in mobile phones. they simply lay it over the components and ensure it's also in good contact with the casework.
much less messy than thermal gel.
l.
Wax pad or cheap thermal gloop (please use at least the copper stuff, that silicone gunk looks like it came out the wrong end of a pigeon!) to the shell and blow some air across that shell.
My old ASUS 1000HE uses that exact method -- except that the shell is the keyboard underlay and palmrest... it's thin sheet metal... might be a half mm thick. Might. That and a 40x10 or 50x10 mm fan is all it ever needed.
Something like this barely needs a fan if at all, but (again) every little bit helps -- what I'm picturing is something like 30x7mm in an external housing that is both intake and exhaust -- just do the housing so that the air gets direction. I'll do up a drawing later, I'm cutting a bolt with a hacksaw right now and I'm almost done. Soon as I get the bolt off, fix the motor it's holding together, and put it all back together I'll draw something up, host it on Imgur, and throw a link up here. Shouldn't be too long now...
On Fri, Nov 13, 2015 at 3:59 PM, Luke Kenneth Casson Leighton <lkcl@lkcl.net
wrote:
On Fri, Nov 13, 2015 at 8:55 PM, Christopher Havel laserhawk64@gmail.com wrote:
Maybe I'm being a little dense but I don't get the graphite paper reference/pun.
reference. graphite is a heat conductor that's... i think it's something like over 100x more efficient than copper. graphite paper is used as a heat spreader in mobile phones. they simply lay it over the components and ensure it's also in good contact with the casework.
much less messy than thermal gel.
l.
arm-netbook mailing list arm-netbook@lists.phcomp.co.uk http://lists.phcomp.co.uk/mailman/listinfo/arm-netbook Send large attachments to arm-netbook@files.phcomp.co.uk
Here there are some documentation, if anyone are interested:
On Fri, Nov 13, 2015 at 9:35 PM, GaCuest gacuest@gmail.com wrote:
Here there are some documentation, if anyone are interested:
got it.
woo holy shit - max current (sustained) - 12 amps just on the cpu alone. at around 1.0 to 1.2 volts that's about 12-13 watts for the main cores. GPU's another 6 watts, just on its own.
that puts it into the "desktop / server" category. it would be a fantastic candidate for something like EOMA200.
l.
Direct link to photo of sketch. http://i.imgur.com/4ayahf5.jpg If the sketch isn't legible enough from that, I can do a proper scan...
Basic idea here is (as discussed earlier) an actively cooled (meaning fanned) Type-II size EOMA68 card.
On Fri, Nov 13, 2015 at 5:19 PM, Luke Kenneth Casson Leighton <lkcl@lkcl.net
wrote:
On Fri, Nov 13, 2015 at 9:35 PM, GaCuest gacuest@gmail.com wrote:
Here there are some documentation, if anyone are interested:
got it.
woo holy shit - max current (sustained) - 12 amps just on the cpu alone. at around 1.0 to 1.2 volts that's about 12-13 watts for the main cores. GPU's another 6 watts, just on its own.
that puts it into the "desktop / server" category. it would be a fantastic candidate for something like EOMA200.
l.
arm-netbook mailing list arm-netbook@lists.phcomp.co.uk http://lists.phcomp.co.uk/mailman/listinfo/arm-netbook Send large attachments to arm-netbook@files.phcomp.co.uk
not bad chris. type III is 8mm btw, that's a 10W limit.
On Fri, Nov 13, 2015 at 10:58 PM, Christopher Havel laserhawk64@gmail.com wrote:
Direct link to photo of sketch. If the sketch isn't legible enough from that, I can do a proper scan...
Basic idea here is (as discussed earlier) an actively cooled (meaning fanned) Type-II size EOMA68 card.
On Fri, Nov 13, 2015 at 5:19 PM, Luke Kenneth Casson Leighton lkcl@lkcl.net wrote:
On Fri, Nov 13, 2015 at 9:35 PM, GaCuest gacuest@gmail.com wrote:
Here there are some documentation, if anyone are interested:
got it.
woo holy shit - max current (sustained) - 12 amps just on the cpu alone. at around 1.0 to 1.2 volts that's about 12-13 watts for the main cores. GPU's another 6 watts, just on its own.
that puts it into the "desktop / server" category. it would be a fantastic candidate for something like EOMA200.
l.
arm-netbook mailing list arm-netbook@lists.phcomp.co.uk http://lists.phcomp.co.uk/mailman/listinfo/arm-netbook Send large attachments to arm-netbook@files.phcomp.co.uk
arm-netbook mailing list arm-netbook@lists.phcomp.co.uk http://lists.phcomp.co.uk/mailman/listinfo/arm-netbook Send large attachments to arm-netbook@files.phcomp.co.uk
I know, but IIRC the 8mm ones aren't going to be the first out of the gate, or even the third -- and an x86 offering would be, uhm, a very good idea IMO -- and the sooner, the better.
;)
On Fri, Nov 13, 2015 at 6:39 PM, Luke Kenneth Casson Leighton <lkcl@lkcl.net
wrote:
not bad chris. type III is 8mm btw, that's a 10W limit.
On Fri, Nov 13, 2015 at 10:58 PM, Christopher Havel laserhawk64@gmail.com wrote:
Direct link to photo of sketch. If the sketch isn't legible enough from that, I can do a proper scan...
Basic idea here is (as discussed earlier) an actively cooled (meaning fanned) Type-II size EOMA68 card.
On Fri, Nov 13, 2015 at 5:19 PM, Luke Kenneth Casson Leighton lkcl@lkcl.net wrote:
On Fri, Nov 13, 2015 at 9:35 PM, GaCuest gacuest@gmail.com wrote:
Here there are some documentation, if anyone are interested:
got it.
woo holy shit - max current (sustained) - 12 amps just on the cpu alone. at around 1.0 to 1.2 volts that's about 12-13 watts for the main cores. GPU's another 6 watts, just on its own.
that puts it into the "desktop / server" category. it would be a fantastic candidate for something like EOMA200.
l.
arm-netbook mailing list arm-netbook@lists.phcomp.co.uk http://lists.phcomp.co.uk/mailman/listinfo/arm-netbook Send large attachments to arm-netbook@files.phcomp.co.uk
arm-netbook mailing list arm-netbook@lists.phcomp.co.uk http://lists.phcomp.co.uk/mailman/listinfo/arm-netbook Send large attachments to arm-netbook@files.phcomp.co.uk
arm-netbook mailing list arm-netbook@lists.phcomp.co.uk http://lists.phcomp.co.uk/mailman/listinfo/arm-netbook Send large attachments to arm-netbook@files.phcomp.co.uk
On Fri, Nov 13, 2015 at 11:41 PM, Christopher Havel laserhawk64@gmail.com wrote:
I know, but IIRC the 8mm ones aren't going to be the first out of the gate, or even the third -- and an x86 offering would be, uhm, a very good idea IMO -- and the sooner, the better.
that's going to be down to intel, to get a decent SoC out their door that can fit in a 2.5W budget. there are other design factors involved as well, not least is that the... naive people at intel seem to think it's okay to put out 1,100 pin monsters with a 0.4mm pin pitch, then expect people to fork out thousands of dollars on protoytpe 10 to 12 layer PCBs @ only 1.2mm thick...
... and that's *just the prototypes* cost, when all the china-sourced SoCs make do with 6 layer (which costs around $600 for 5 samples), 4 layer (which can be had for around $500), and there are even china SoCs out there which can fit onto 2-layer PCBs, now.
then there's the cost of their offerings. the SoCs that go into the USB-PC dongles? look up the price on intel's web site: those are $32. $32 for fuck's sake! are their marketing team high or something?? there's *$5* quad-core 2ghz 64-bit ARM processors out there that can address up to 32 gigabytes of RAM, and they're trying to still pretend that the processor is the most important factor in the BOM of a product.
basically intel haven't got the faintest clue as to why they aren't even remotely in the market. and because they're not in the market, they haven't a chance to find *out* why they're not in the market.
they focussed so much on desktops, servers and laptops that they've had ARM and MIPS SoCs create an entire market that they're *never* going to get into if they keep up with their current "strategy". back in 2007 they even sold the PXA design because it was too embarrassingly good, it was making the intel atom look piss-poor by comparison in the performance/watt stakes.
anyone from intel - if you're reading this - for god's sake get a grip, contact me and i'll help you to spec out a decent SoC that will stand a chance in the china market. you've already found out why china fabless semiconductor companies can't work with you - that was another costly learning experience, wasn't it? when you're ready to listen, i'll be happy to walk you through what you need to do.
l.
arm-netbook@lists.phcomp.co.uk
-
Christopher Havel -
GaCuest -
Luke Kenneth Casson Leighton