[mantorker]

i have looking into tons of pdf and website,seems that the PCIE has lower latency than the USB, cuz USB use the Polling model to collect data from device. And PCIE can use DMA feature to directly writes data to sytem memory.So,here is what i'm thinking.

hardware : optical chip ---spi--->MCU--->spi(long distance implementation)---> FPGA--->PCIE
software: dev a specify driver collecting movement data ,than writes somewhere to tell the windows/linux the cursor is moving(this part I stil have to look deeper)

but ，are all these brain-expolde work worth doing？

 

[mott555]

There's already a dedicated low-latency mouse interface with hardware interrupt support which is better than USB. It's called PS/2.

 

[FlailScHLAMP]

no, mice don't need that bandwidth at all..

waste of interface space

 

[BinaryDemon]

Quote:

Originally Posted by mott555

There's already a dedicated low-latency mouse interface with hardware interrupt support which is better than USB. It's called PS/2.

I think it's the ps/2's slower polling rate that makes USB a better option for mice.

"However, USB mice have lower latencies than PS/2 mice because standard USB mice are polled at a default rate of 125 hertz while standard PS/2 mice send interrupts at a default rate of 100 hertz when they have data to send to the computer.[6][7] Also, USB mice do not cause the USB controller to interrupt the system when they have no status change to report according to the USB HID specification's default profile for mice.[8] Both PS/2 and USB allow the sample rate to be overridden, with PS/2 supporting a sampling rate of up to 200 hertz[1] and USB supporting a polling rate up to 1 kilohertz[6] as long as the mouse runs at full-speed USB speeds or higher." -https://en.wikipedia.org/wiki/PS/2_port

 

[ncck]

Quote:

Originally Posted by mott555

There's already a dedicated low-latency mouse interface with hardware interrupt support which is better than USB. It's called PS/2.

ps/2 feels better but you can't run it (without doing weird stuff?) past 125 or 250 hz? I forget the number, but if people could get ps/2 stable working with 500hz I'm sure people would consider it.. I use it for the keyboard!

 

[qsxcv]

usb is a mess, but on my pc i already get ~600us usb to display latency..
no point at all

 

[Rayleyne]

Quote:

Originally Posted by qsxcv

usb is a mess, but on my pc i already get ~600us usb to display latency..
no point at all

You do realize that, That small a timespace is litterally not detectable by the human brain right, So 600us is perfectly fine...

 

[akromatic]

better invest that brain power for thunderbolt3

 

[qsxcv]

Quote:

Originally Posted by Rayleyne

You do realize that, That small a timespace is litterally not detectable by the human brain right, So 600us is perfectly fine...

ya that's what i mean...
Quote:

Originally Posted by akromatic

better invest that brain power for thunderbolt3

i'd rather not have thicker cables.

 

[akromatic]

Quote:

Originally Posted by qsxcv

i'd rather not have thicker cables.

but you dont have to.

thunderbolt 3 > controller box > mice via thin cable

you can probly remove most of the circuitry out of the mice and have them in the controller box to have it lighter

 

[mantorker]

it's true that thunderbolt3 is an interface utlizing the PCIE bus, but cmpare to USB3.1 , collecting the spec of thunderbolt3 is much more diffcult, and will need an thunderbolt3 interface chip in the mouse to make it work, which is hard to find .

maybe this is alittle bit off topic:
http://www.neglectedpotential.com/2013/10/iphone-touchscreen-accuracy/

according to this article,touch screen may have a acc < 1mm, so maybe bulding a android-pc touch screen-control mouse movement is more considarable.

the whole touch screen of a android pad/phone to povide precise ratio movement to host pc according to the sceen dpi of host pc, and use a foot-pedal as mouse button to avoid complex gesture

 

[mantorker]

https://developer.apple.com/library/mac/documentation/HardwareDrivers/Conceptual/ThunderboltDevGuide/Basics01/Basics01.html

according to this info,
"Devices on a Thunderbolt bus exhibit higher latency than devices on internal slots-about 1.5 microseconds of round-trip latency per hop. This means that putting a device at the end of a Thunderbolt chain can add up to 9 microseconds of round-trip latency. Most devices and drivers handle this additional communication latency with no particular difficulty. In some cases, however, your driver or device firmware must be tweaked to tolerate that latency."

 

[Bucake]

Quote:

Originally Posted by qsxcv

usb is a mess, but on my pc i already get ~600us usb to display latency..
no point at all

do you happen to have some tips to lower latency a bit?
did you go for specific hardware, or is it more a matter of managing your OS / software right?

 

[mantorker]

i think qsxcv means his usb polling rate is about 2000hz, so the to-the-next-report latency is about 600us.

 

[qsxcv]

600us doesnt include polling-related latency stuff. iirc i measured from time of the usb poll/transaction to time of screen (crt) responding.
Quote:

Originally Posted by Bucake

do you happen to have some tips to lower latency a bit?
did you go for specific hardware, or is it more a matter of managing your OS / software right?

well
going from an lcd monitor to a crt makes a bigger difference in latency than 90% of the tweaks you can find

my only tips are to use common sense and ignore paranoid people

 

[ncck]

Quote:

Originally Posted by Bucake

do you happen to have some tips to lower latency a bit?
did you go for specific hardware, or is it more a matter of managing your OS / software right?

The only things I've seen that affect input lag is/are....

Maximum pre-rendered frames (most obvious)
Display Scaling (least obvious)
And disabling all high graphic settings / anti aliasing in the game you play, even doing something like turning texture or anisotropic filtering on will create small delays (may depend on game engines?) even if your framerate doesn't change it feels like there's added delay.. (not confirmed? I've just noticed comparing aliasing on and off if the fps stays the same it feels better off.. no fact to back this up_

Um.. other than that nothing much? I do set my nvidia GPU's from IRQ TO MSI; amd cards are default MSI.. I don't think that does anything but I always do it
Haven't seen anything else affect input lag personally unless you have some software or driver issues.. in reference to DPC latency or something no idea

 

[mantorker]

Quote:

Originally Posted by qsxcv

600us doesnt include polling-related latency stuff. iirc i measured from time of the usb poll/transaction to time of screen (crt) responding.
well
going from an lcd monitor to a crt makes a bigger difference in latency than 90% of the tweaks you can find

my only tips are to use common sense and ignore paranoid people

Ha HA ,the tech things we disscuss here mostly are invented by paranoid people.

 

[Artikbot]

PS/2 uses a 200Hz hardware polling rate. Some OSes limit it to 125Hz though.

Technically speaking, at the very ultra worst case scenario, on a 144Hz display you could get up to a single frame's worth of latency.

Since it uses interrupts to pass data (meaning it stops the loop and shoves the data in there as opposed to holding the data until the OS asks for it), I really don't see how USB is a better interface over PS/2 for inputs like a mouse or a keyboard.

 

[ncck]

Quote:

Originally Posted by Artikbot

PS/2 uses a 200Hz hardware polling rate. Some OSes limit it to 125Hz though.

Technically speaking, at the very ultra worst case scenario, on a 144Hz display you could get up to a single frame's worth of latency.

Since it uses interrupts to pass data (meaning it stops the loop and shoves the data in there as opposed to holding the data until the OS asks for it), I really don't see how USB is a better interface over PS/2 for inputs like a mouse or a keyboard.

I've used a mouse in ps/2, it's like a slideshow with how it stutters; keyboard is better.. mouse feels terrible

 

[Artikbot]

I suppose the problem is that the position commands are passed to the OS only at the interruption intervals, and if that isn't modified to run at 200Hz (I think all Windows releases use 125Hz default unless you change it), there are quite large steps between each interrupt.

I can certainly confirm that there was skipping with my G9 running at 125Hz, going to 250Hz was immensely smoother, 500 could have been placebo, and I really cannot tell the difference at all at 1000Hz.

I'd like to test my G9 on PS/2 set at 200Hz...