[Arm-netbook] EOMA server standard

[luke.leighton]

On Sat, Oct 27, 2012 at 12:41 PM, Dr. David Alan Gilbert
<dave at treblig.org> wrote:
> * luke.leighton (luke.leighton at gmail.com) wrote:
>> On Fri, Oct 26, 2012 at 5:52 PM, Baybal Ni <nikulinpi at gmail.com> wrote:
>> > On 26 October 2012 09:31, luke.leighton <luke.leighton at gmail.com> wrote:
>> >>  *that's* a good standard.  yes.  take the 8 USB wires.  you can
>> >> auto-negotiate all the way from USB1.1 right the way up to USB3.
>> >> that's _great_.  likewise with 10/100/1000 ethernet.  8 wires,
>> >> auto-negotiation.  likewise with SATA: exact same wires, you can
>> >> auto-negotiate all the way from 150kbit/sec right up to 6gbits/sec.
>> >
>> >
>> > Don't forget that autonegotiation works over any amount of wires. So
>> > it's possible to mandate that the backplane will ALWAYS have 8 lanes,
>> > and card can use either 8 or 4.
>>
>>  for PCIe?  my understanding is that PCIe can down-level negotiate
>> even to single-lane, and also detect the version of PCIe (1, 2, 3
>> etc.)
>>
>>  so again, yes, thank you baybal - PCIe is another good example.  hell
>> of a lot more wires, which is why i'm reluctant to add it to any
>> standard, but it's a good example.
>
> Is it really that many more

 yes! http://pinouts.ru/Slots/pci_express_pinout.shtml

 1x is 36 pins, and 4x is 64 pins!  it's a fixed overhead of appx 18
pins, and an average of appx 10 pins per lane.

> - one way around the networking stuff would
> be to put no networking over the connector, just put as many PCIe lanes
> as you can fit, and then the chassis can have a standard PCIe slot
> to put the users choice of networking card in.

 i can see the merits of that.  it's very tempting.

l.