The war on input lag revisited: USB ports (updated with links to video tweaks)

[Gamogo]

Edit: Previous input lag discussions and how to reduce/negate it:

Intial post: http://forum.ozhadou.net/index.php?topic=6600.msg124359#msg124359
Follow-up: http://forum.ozhadou.net/index.php?topic=6600.msg124379#msg124379

More input lag crap from me:

Windows defaults to a 125Hz sample rate for USB attached devices which while fine for the most part, falls far short of what USB attached devices are capable of. By upping the sample rate of USB ports, your PC is provided with a busier stream of updates from your attached devices, which also reduces any read delays (or input skipping outright) for inputs performed.

The most noticeable result of upping your USB sample rate is mouse movement. Input and cursor movement is much smoother and more precise. Due to the increased input polling of the device however, cursor movement will be around 25 - 30% slower, which will require a bump in the associated pointer speed to set things back to a familiar handling, and with it comes the improved accuracy and smoothness associated with a higher sample rate.

So how does this relate to Street Fighter (or PC fighting games in general)? Put simply - USB attached arcade sticks. I tinkered with the the USB sample rate of the port my stick is attached to (A Hori VLX) and found some interesting results. Firstly, I noticed a reduction in input drops and much greater consistency with tighter 1 and 2 frame links in SFIV. Secondly, I noticed the 'backup' input from plinked combos appearing much more consistently than before. For example, plinking c.mp~lp after E.Ryu's MK axe-kick felt not only more consistent, but I noticed multiple times if I flubbed it, the c.lp would more consistently come out to 'save' my link (this can happen if you perform the c.mp~lp input too soon after his MK axe kick).

I've noticed the PC version of SFIV has a few irritations (the way vsync works is particularly nasty) though I did find input registers to be noticeably sloppier than the Xbox which only made sense given the inherent variability the PC has as a games platform. Upping your USB port sampling rate seems to address this, at least it certainly does from where I am sitting.

There are a few resources online that detail how to alter your USB port sampling rate, though right out of the gate I'd recommend giving this a read:

http://www.ngohq.com/news/15043-how-to-increase-usb-sample-rate-in-windows-vista-7-a.html

Running the attached 'mouserate' application will give you a means to test whether your sample rate input changes have applied or not. The attached 'dimr' application and associated driver also have a setup utility that allows you to view and set the sample rate of all attached USB devices, including arcade sticks. I have mine jacked up to 1000Hz. That's almost an 8-fold increase in input sampling.

Some things to keep in mind:

- Applying this driver mandates that you DISABLE Windows driver signing. This allows you to install uncertified drivers and performance hacks, although it does enable the unsightly "Test Mode" dialogue in the bottom right corner of your screen. This can be removed with some additional tweaks however.

- You make these changes at your own risk! While this has worked nicely for me, it may not for you. I have not tested this on Windows 8, only 7.

- Mileage may vary. I've personally noticed an improvement in how inputs are registered. Input lag and delay of any kind irritates the unholy shit out of me so I'm always tinkering with ways to reduce it. Fighting games make a great acid test for this kind of stuff.

Those that give this a try please share your thoughts. Some discussions I had with some guys who are just as picky as I am about input lag who happen to play Super Turbo made me look into how this could work for SFIV on PC. They were looking at ways to reduce input lag on their outlandish arcade hardware emulators and stick hardware they had piggy-backing off a PC. USB port polling was among the areas looked at which helped in their case also.

[kyokugen]

Yeah, this was an issue in GGPO days, but I was always under the impression that Windows' drivers for 360 controllers always ensured that they were polled @ 1000Hz.

There isn't any explicit information out there on this, but I remember back in the day I 'felt' (derp) the difference between non-increased generic USB and a MadCatz stick.

I guess upping polling system wide doesn't really hurt either way. Just sharing my experience and whatevers.

[Niah]

this is a bloody great find. i never knew! should also make and sticky a thread on how you fixed that vsync problem so we don't have to scour through the depths of the pc thread to find your post

also, doesn't work in windows 8 and there's no info on the net about windows 8 either

[bbbenson]

Thanks Shane. Now if only you could remove your brain lag and thus improve your bottleneck, you'd win a few more matches.