PS/2 limitation? CM Storm QF?

[mellohello]

So I caved and bought a CM Storm Quickfire.

I'm using the board over ps/2, and testing it in passmark keyboard test. I've noticed that the majority of keys are nkro with 2-4ms lag time. However, I notice that the altgr button on the right side of the keyboard causes exactly 5ms lag when depressed together with either right or left ctrl. Is there any particular reason why this might be? And also why is it all the keys to the right of the enter button (including the arrow cluster) suffer exactly 19ms lag when any two of them are pressed together? They're also 3kro not nkro, while if they are combined with keys from the "main" part of the keyboard they sometimes don't fully register. in other words if i press 3 arrows+some keys from the nkro keys, maybe only one or two of the arrow keys register.

Interestingly, in usb mode every single 6 key combination I can think of registers perfectly in other words the full board including the keys that are 3kro in ps/2 are 6kro. Furthermore in usb there's no 19ms lag between the keys on the right side of the board, and no 5ms lag on the ctrl+altgr either. Instead, every multiple key combination is 2-4ms.

I doubt either my new keyboard or my motherboard's ps/2 port are at fault, despite the cheapness of my motherboard. Is this actually a limitation of ps/2? I'm very curious to know and would be grateful to hear thoughts.

Thanks

[mellohello]

Sorry, I edited the post a fair bit to make it more accurate/readable. I'm still using ps/2, so the issues described aren't particularly bothersome, although I've had to remap my stepmania keys (rhythm based game) to the wasd cluster to avoid lag, since milliseconds matter in that game. I am however still curious . I've started to think it might be poor ps/2 implementation on my motherboard. Does anyone have an idea?

[Soarer]

Dunno. It does sounds at least partly like a limitation of PS/2. For the arrow keys etc. the sequence of codes sent can be quite long, particularly if numlock is on - even if it's a tenkeyless, probably your bios or windows is setting numlock on, it would be interesting to see if changing that made it quicker! Right control and right alt both have 'extended' codes - so two bytes instead of one for a press, and three instead of two for a release.

The 3kro effect could be related, since the long sequences have to fit in a buffer before sending to the PC. Clearly fewer will fit! That should only affect keys pressed or released at the same time though, rather than how many you can hold down.

How are you measuring the lag?

I'm guessing it's the Rapid you have? I still haven't figured out how they can do 1mS polling on low-speed USB. But if that's what you've got, it certainly seems to be working!

[mellohello]

wow, the numlock key was indeed set to on; hadn't at all considered the possibility! i just turned it off, though unfortunately it didn't reduce the lag.

i've also just noticed that the windows key(s) behaves just like the altgr key in terms of lag (5ms with a second key). strangely enough, when altgr is pressed with the windows key, the lag is 2-4ms, despite both of them being the lag instigators when combined with any other key. anyway it does seem to only affect the keys that have extended codes as you put it, so your explanation makes perfect sense. also thanks for stating that the 3kro keys can actually be held down simultaneously if pressed separately. i didn't even think to test that, and you are right of course.

the test i'm using is passmark keyboard test. by "lag" the software means how long it takes to register a second key after one has already been pressed. the bare minimum is 2ms, though i use a flat obect when pressing 2 keys at a time, to cut human error. the exact 19ms every time for the arrow cluster, etc seems very symmetrical. once again though your explanation makes sense i think! if those keys send a lot of code, then it stands to reason that usb is faster than ps/2. if this can be confirmed, i think it needs to be added to the list of pros that usb has over ps/2 .

yes, it's the rapid. it's growing on me! as for actual input lag, the software can't test that sadly. it only tests the time between 2 keypresses. so i personally can't verify whether usb or ps/2 is faster in terms of total input lag, and i can't verify whether 1000hz does anything either . if passmark is accurate though, ps/2 is pretty bad for arrow cluster gaming.

thank you for the very helpful response

edit: i've noticed that if i take the usb plug out and leave the ps/2 converter in, i can hotplug. it also comes up as "usb keyboard" under bios. i wonder since the multiple keypress lag on ps/2 should technically be 0ms (right?), and it should at least be different to usb, whether the keyboard isn't genuinely ps/2 because it's using a converter? is that even a thing? sorry to keep asking questions ^^"

[Soarer]

Ah, that's an interesting way to test the lag, I hadn't thought of that. namelos did some testing a while back that used extra hardware to press a key - not as easy to do as your test! He was more interested in comparing keyboards than checking each key. But I'm not sure how accurate your test can be, what limitations it has... the arrow keys in particular seem to have a higher lag than they should, compared to an alpha key, even allowing for the fact that a sequence of codes is sent (it should only be as long as the right alt code with numlock off, or double that with numlock on). Anyway, yes, the alpha keys are better for gaming than the arrow keys, over PS/2 :-)

Almost all the keys that weren't on the original XT/AT keyboards have extended codes. (Extended in this sense just means using more than single byte when pressed). Mainly that's the ones you've noticed already, and the / and Enter on a numpad. But not F11 or F12. All done for compatibility reasons that have now been completely irrelevant for many years!

I reckon you'd notice if the 1000Hz wasn't working - over USB the two presses would come in 8mS+ apart. Although, since USB can send two (or more) changes in a single report, it could be 0mS apart if you really pressed them at exactly the same time!

[mellohello]

ok, i'd now ditch ps/2 if i hadn't heard somewhere that usb hubs share bandwidth and two 1000hz high speed peripherals at the same time might cause slowdowns in both...

also 8ms, really? i guess the strangest thing in all of this then is that the software reports down to 1ms lag time on my original keyboard - the cheap 2kro dell that i've been using for the past 7 years. i've yet to see 1ms on the quickfire, usb or ps/2 lol.. and review sites seem to think 2ms is pretty good too. how on earth is my dell kb reporting such supposedly good results o.o. i remembered testing it way back when i first learned about polling, and thinking then that i must've been using the program wrongly. however i just got it out again and once again i'm getting occasional 1ms times. and it's definitely not 1000hz... something's very strange. this seems to be a commonly used program too.

ah it's getting late, i have to sleep x.x. too many typos before editing. thanks so much for the continued help though!

edit: ok god, after mega spamming the qf to death, i've seen occasional 1ms lag times over usb. definitely definitely nothing less than 2ms over ps/2 though. still don't understand why my cheap free membrane dell usb board gets 1ms so often though. it really can't be 1000hz. definitely bedtime now

[laffindude]

I am no expert but it sounds like to me that in each individual packet sent, the controller keeps track of when the keys are registered in 1ms resolution. Instead of sending out packets of key status in 8ms resolution.

Cheap membrane boards probably get more 1ms since they probably need less debounce time.

[Soarer]

The sharing bandwidth thing isn't a worry, it would only have an affect with devices that really use the bandwidth, e.g. hard disks, cameras, etc. For HID devices, even at their max rate, only a small part of the bandwidth is used and the protocol has a number of slots within each mS so it can allocate one for each device. Basically, if you keep all HID devices on one root hub and don't share it with high-bandwidth devices, it will be fine.

There could definitely be quirks using passmark for measuring lag, and it's probably better for PS/2 than USB. But for both, the debouncing in the keyboard can bunch key presses together, when the second press happens within the debounce time of the first. In that case, all it's really showing is how fast two events can be sent over the wire, which doesn't really tell much of the whole story about lag. Ironically, a better debounce algorithm might show a bigger lag (in passmark), which might explain the difference between the Dell and the QF.

[mellohello]

Once again, thanks so much for the input guys! It's been very helpful.

Yeah, I never found the usb bandwidth thing again after I saw it that one time. Thanks for clearing it up. And whatever the case, I'm not going back to the old keyboard. I guess in the end it doesn't really matter in terms of performance which to use, ps/2 or usb. I do feel a bit more knowledgeable though.

Thanks again

[Soarer]

You're welcome!

I wonder if you could do us a favour... it would be interesting to see if the trick for 1000Hz is apparent in the descriptors somewhere. There's easy to follow instructions on how to get them here.

USBlyzer also has a capture mode - if you're curious you might want to try that as well to see whether the two key event are being sent in the same report. It is a bit cryptic to use for that though!


edit: I'm still rather sceptical that they can actually do 1000Hz over low-speed USB without a driver. The standard says the minimum interval should be 10mS, whereas Windows implements a minimum of 8mS (125Hz) and actually increases the polling rate for anything that asks for between 8 and 16mS. However, maybe they aren't really claiming that - although it says 1000Hz polling, perhaps that actually just means scanning the keys at that rate.

[mellohello]

Sure! I'll do that straight away. Sorry, I've been away for a few days.

If I understand the program properly, it lists both my mouse and keyboard as 1ms (which I assume is 1000hz). What it also does is list my keyboard as usb (while using ps/2, just like my motherboard bios), and it also lists it on the same port as my mouse, even though the keyboard uses a ps/2 port, and even when I change the mouse to another port. I can change the mouse port to anywhere and the keyboard port will follow it.

Also, once again this keyboard continues to baffle. When I got home and switched on my computer today, my keyboard decided to stop taking 19ms breaks between the keys on the right hand side of the enter key. They're now taking 5ms like the altgr and windows keys. And they're now 8kro instead of 3?! This isn't the passmark software messing up, I genuinely can press 8 of those keys together and have them register now. Honestly don't know what's going on; nothing changed between then and now except that A. I haven't been using my keyboard and B. I've had the PC off at the power supply this weekend (which I don't normally do).

Anyway did you want a log file/any other information from USBlyzer?

Thanks