Hi,
[email protected] wrote:
>
> On Jun 5, 2011, at 9:38 PM, Adam Schrotenboer wrote:
>
>> On 6/5/11 3:54 PM, [email protected] wrote:
>>> Ok, the chip is from 3Q'10 and PCI-E 3.0 is from November 2010, but
>>> 1.0a is a bit silly too.
>>
>> Blame Intel and their silly market segmentation. xref Intel's DMI bus
>> for more laughs.
>
> Agreed.
Although Intel often irritate me with their cripled parts, it's not that
simple, the Atom E6xx is a very low power part, and more and faster PCIe
ports will use more power....
> Gigabit ethernet isn't necessary for the routing part, but many people
> like to pretend Linux bridging is the same as switching and expect
> it to go at line rate. It does make me wonder if the on-board can go
> at line rate themselves, given that the processor has 4 x1 lanes of
> PCIe and the chipset has only 1 Gig-E. Are the 4 x1 lanes not taken
> up by the two mini card and the two full-sized PCIe headers?
Notice the two 4 ports IDT PCIe switches on the net6501 picture.... Yes,
the just 4 PCIe ports on the Atom E6xx made it a challange to give
enough bandwidth to everything, so basically:
1 PCIe port to PCIe slot #1
1 PCIe port to Gigabit Ethernet controller #1
Gigabit Ethernet controller #2
PCIe slot #2
1 PCIe port to Gigabit Ethernet controller #3
Gigabit Ethernet controller #4
Mini-PCIe slot #1
1 PCIe port to PCH (SATA & USB)
That is in my opinion a good balance without it getting too expensive.
PCIe slot #1 had full bandwidth to ex. a upcomming 4 ports gigabit
ethernet controller, and if needing full gigabit ethernet bandwidth use
ethernet controller #1 & #3 or #2 and #4.
And yes, Mini-PCIe slot #2 do not have PCIe, only USB and SATA, but
that's based on the assumption that most people will use one of the
Mini-PCIe slots for a mSATA SSD anyway.
Best Regards,
Soren Kristensen
CEO & Chief Engineer
Soekris Engineering, Inc.
_______________________________________________
Soekris-tech mailing list
[email protected]
http://lists.soekris.com/mailman/listinfo/soekris-tech
