You are not logged in.

#1 2014-01-09 00:26:33

frazer
Member
Registered: 2013-12-09
Posts: 60

[SOLVED] Pci-e lanes and availability

Hello Forum,

I have had a nagging question in my mind about pcie lanes.

Say I was to put a single lane USB3 interface card in a pciex1 slot, and say a Firewire interface in another pcie x1 slot and continued to populate available lanes/slots, how do I know or find out if a card has exclusive use of a lane please?  If all available lanes are exclusive, that answers my question.

In one example, I have two separate USB3 cards in 2 different pcie x1 slots of a motherboard, and would like to use each USB3 card to connect to two USB3 HDDs to the motherboard and copy files from one USB HDD to another.  Can I always assume that if there is a pcie x1 slot, it will not be shared with another lane?

I really hope that I have explained it properly, I just want to find the answer so that I can get the best performance out of any future card/motherboard setups.

Thank you in advance!

Last edited by frazer (2014-03-04 13:09:01)

Offline

#2 2014-01-09 00:34:46

ewaller
Administrator
From: Pasadena, CA
Registered: 2009-07-13
Posts: 19,802

Re: [SOLVED] Pci-e lanes and availability

PCI-e lanes are used exclusively by a single device.  High bandwidth interfaces may require multiple lanes, or may be able to operate in a degraded mode on a single lane.  But each lane is used for a dedicated purpose.
http://en.wikipedia.org/wiki/PCI_Express


Nothing is too wonderful to be true, if it be consistent with the laws of nature -- Michael Faraday
Sometimes it is the people no one can imagine anything of who do the things no one can imagine. -- Alan Turing
---
How to Ask Questions the Smart Way

Offline

#3 2014-01-11 00:03:35

frazer
Member
Registered: 2013-12-09
Posts: 60

Re: [SOLVED] Pci-e lanes and availability

Thank you ewaller for your reply.

I'm more concerned about the reality rather than the design ideals though as "real world" experience has shown me that a lot of times, things just 'don't add up'!

In practice with, for example, a USB 3 pci-e x1 interface card or, a sata 3 card, there is quite a significant CPU usage attributed to the use of the card, leading to a greatly decreased bandwidth in practice when compared to design.  I know that I am opening a can of worms here regarding semantics, and I am in no way trying to offend, instead I feel that I am missing a piece of the puzzle with regards to specifications and actualities.

For example, when I use a USB3 port on a/my motherboard, I will find a faster transfer rate (by far) than that of a pci-e x1 card, and with less CPU overhead.

The same proves true for a sata ii or iii card etc.  I know there may be discrepancies between USB3 speed and pci-e2 x1 bandwidth, but there doesn't seem to be any way of predicting the CPU overheads and the actualities of transfer speed without plugging an appropriate card into an appropriate lane and trying it....  This proves to be extremely frustrating when designing systems and then finding out the real world performance.

I just wish that anyone could expand on "rules of thumb" if there are any...

Maybe someone could advise on the advantages of 'motherboard' interfaces over pci-e lanes?

I suppose that we may end up talking about chipsets, but I really would love a 'rule of thumb' to go by.

Anything from the forum would be wonderful as it will allow me to look down different ways of configuring systems in the real world!

Many thanks for anyone's comments.

Offline

Board footer

Powered by FluxBB