cancel
Showing results for 
Search instead for 
Did you mean: 

Technical explaination of PCIE dipswitches?

segfaulted
Level 7
Curious feature of this board are the dipswitches that supposedly disable individual PCIe slots. Does anyone know how exactly they do that? Is it the card boots but is thrown in some kind of sleep or does it physically remove power from the device? I'd love to do PCIe hotswap for example but for reasons I can't explain I'm unable to get the board to do a rescan after powering a slot back up.

This is also an attempt to work around another problem I'm having with posting at all when populated with two 1080TIs. I can toggle one slot off and boot, toggle it back on and see the card has power (LED's are on), but again can't get linux to trigger a rescan. I have no problem doing this on other boards so I'm at a loss as to what is going on. Can't find any documentation on either the switches nor platform itself 😞

Any ideas?
4,455 Views
10 REPLIES 10

JustinThyme
Level 13
These switches remove power on one pin for trouble shooting purposes.
One should NEVER attempt a hot swap on PCIE slots.
The device in the slot will not work after post by just turing the switch back on. It has to go through the post process with the device installed and PCIE power on to be recognized. Thats all part of why positng in a destop take some time.

The TS process is to turn off slots and boot up simply to save time from having to remove them entirely. Any change in the state of the switches should be done with the machine powered down and PSU off.



“Two things are infinite: the universe and human stupidity, I'm not sure about the former” ~ Albert Einstein

JustinThyme wrote:
These switches remove power on one pin for trouble shooting purposes.
One should NEVER attempt a hot swap on PCIE slots.
The device in the slot will not work after post by just turing the switch back on. It has to go through the post process with the device installed and PCIE power on to be recognized. Thats all part of why positng in a destop take some time.

The TS process is to turn off slots and boot up simply to save time from having to remove them entirely. Any change in the state of the switches should be done with the machine powered down and PSU off.




I get that this isn't for the faint of heart but neither is half the things on the board (probe points for example). PCIe most certainly supports hotswap. Says nothing about the hardware itself being in a state to support such a thing though. Can you clarify? Kind of the reason for asking is it's not clear what those switches do. You can from at least in Linux, power some devices off entirely from the OS. Fun fact, you save battery power in the process.

JustinThyme
Level 13
Can you please point me to anything that says any PCIE device is hot swappable? Every single board manufactuere Ive ever read the manuals for 100% of the time say it many times over to power the system down prior to inserting or removing anything from the PCIE slots. SATA, some do support hot swap but not PCIE.
An OS doesnt power anything......
There is a huge difference between inserting someting to or removing from a live active bus and measuring voltage test points that are isolated.



“Two things are infinite: the universe and human stupidity, I'm not sure about the former” ~ Albert Einstein

JustinThyme wrote:
Can you please point me to anything that says any PCIE device is hot swappable? Every single board manufactuere Ive ever read the manuals for 100% of the time say it many times over to power the system down prior to inserting or removing anything from the PCIE slots. SATA, some do support hot swap but not PCIE.
An OS doesnt power anything......
There is a huge difference between inserting someting to or removing from a live active bus and measuring voltage test points that are isolated.


Manufactures do not, understandably, want end users simply removing hardware from a live system. Especially in consumer electronics since they cannot among other things, guarantee all hardware will do the right thing nor their own drivers. That isn't what I'm after though.

I'm asking what on a technical level (aka logical board), do the switches do because if they do kill the power to a device, physically removing it becomes much easier. What state the OS is in, is an entirely different discussion. I never said it powered the hardware, I said it can control some. And I was referring specifically to Linux (I have no idea if Windows can).

** To be absolutely clear for anyone casually reading this, I am not, in any way, suggesting people remove powered on hardware from a live system **

It is quite dangerous to be sticking anything metal on a board with power. The probe example was to say it's also an advanced thing your typical end user will not be doing . Then again if jtag headers, serial buses and poking pci addresses make you warm'n fuzzy inside, this is moot.

JustinThyme
Level 13
Well, you have your answer.......



“Two things are infinite: the universe and human stupidity, I'm not sure about the former” ~ Albert Einstein

JustinThyme wrote:
Well, you have your answer.......


What? Sorry I have no idea who you are nor if you work for Asus or not. Do you?

segfaulted wrote:
What? Sorry I have no idea who you are nor if you work for Asus or not. Do you?


I do not work for ASUS and as Menthol Stated they are not going to share schematics.

What I can tell you from the perspective of an electronics engineer is this is a single pole double throw dip switch so you have on and off of a single one line item be it power or signal. When switched to the off position the 5V standby power is gone. Logical deduction without having to trace down the entire board dictates that the switch interupts at least the 5V power but......being profieicnt in the field of electronics design and implementation, to remove the 5V+ from a circuit is not a good design. When switching DC the industry standard of switching the common is used.

With the being said there are but two things possible.
5V+ is being switched which is a poor design
DC common is being switched.

Take your pick, its still HIGHLY DISCOURAGED TO INSTALL AND REMOVE CARDS FROM A PCIE BUS WHILE THE POWER SUPPLY IS PROVIDING POWER AS IT IS NOT HOT SWAPABLE. PRIOR TO INSTALLING OR REMOVING ONE SHOULD ALWAYS HAVE THE MACHINE POWERED DOWN AND THE POWER SUPPLY EITHER SWITCHED OFF OR THE CORD REMOVED!!

Manufacturers dont push this practice simply to warn the faint of heart. They push it for two reason. One, it is likley to cause irreparable damage, the other is its a moot point as any thing in the PCIE slot must be in place at the time of post to be scanned by the UEFI and offered up for use by the OS.

Good luck with your efforts, you're going to need it. :cool:



“Two things are infinite: the universe and human stupidity, I'm not sure about the former” ~ Albert Einstein

JustinThyme wrote:
I do not work for ASUS and as Menthol Stated they are not going to share schematics.


OK? I mean, you keep saying that but I never asked for schematics .

JustinThyme wrote:


With the being said there are but two things possible.
5V+ is being switched which is a poor design
DC common is being switched.


.. and that would by why I'm asking.

JustinThyme wrote:

Manufacturers dont push this practice simply to warn the faint of heart. They push it for two reason. One, it is likley to cause irreparable damage, the other is its a moot point as any thing in the PCIE slot must be in place at the time of post to be scanned by the UEFI and offered up for use by the OS.


That may be the case with Windows and UEFI but I'm neither booting Windows nor with UEFI. As I said, Linux and yes rescanning is possible.

JustinThyme wrote:

Good luck with your efforts, you're going to need it. :cool:



Indeed. What I need - what I would like - however is less luck and a straight answer from Asus. For all I know the switch is tied to some IO pin on the chipset that drops the link and doesn't actually disconnect any physical bus lines. FWIW I have turned the slots off without an issue. It still baffles me why turning off slot 1 eliminates _all_ pcie errors (yes, they are recoverable but there is an overhead of doing that at thousands per second). This is either a defective board from day one or some design problem (RFI leaking in for example?) -- because it's fairly well known and reported elsewhere. Threads are typically shut down with "disable error reporting" or some such nonesense.

For the cost of this board and it's target audience, the lack of response from Asus at all is bewildering.

Menthol
Level 14
I doubt ASUS is going to share schematics of there hardware as they probably don't want to design other manufacturers hardware for them but I would think that one switch can only disable the function of an add on card and not fully disconnect every electrical contact, that would require a relay with as many contacts as electrical circuits where the switch controls power to the relay coils