Speedstep and the impact on mouse pointer performance

[BrightCandle]

This is a bit of an odd one. For a while I have this feeling the mouse was lagging, that it was not going where I put it as quickly as it should do. I wasn't sure why that was the case, it just didn't feel right. I notice the difference between 120hz and 60hz really keenly (there is a bug there in NVidia's drivers for example with multiple monitors where both render the mouse at 60hz when a second monitor is added regardless of its hz setting, really annoying!). I certainly felt the mouse was wrong. In the past I have noticed that the mouse movement speed is impacted by CPU clock speed changes, when machines downclock to 1.2Ghz and such to save power when I move the mouse it feels like its slower, actually I think its just latency.

I decided yesterday to go on a system wide driver update, haven't done one in months. Got a new bios as well and the new bios messed up the overclocking on the board. So I got a beta bios from the company in response to my support ticket. In the process however I found that the bios that didn't overclock properly did not have a laggy mouse, which was very odd. The beta bios as soon as I over clocked reintroduced the mouse lag.

Despite having c states and speedstep and such on with the beta driver the clock speed never drops below what I have set. However the voltage does change quite a bit (using dynamic VID to save power consumption for 24.7 overclock) dependent on load. When I turn speedstep off the voltage stops varying and the mouse stops lagging as badly as it did.

Its a curious effect, I wonder if anyone else has noticed this. I guess its quite subtle since I don't see many people complaining about it but to me its really obvious. I can tell whether SLI is on or off just by moving my mouse for example, it has a noticeable impact on the performance of moving windows and the mouse around the screen.

 

[SPBHM]

I never noticed any problem, my mouse is extremely consistent, no matter the CPU clock.

have you tried just playing with the power settings on windows, forcing it to 100% as minimum clock?

 

[Eug]

It took me a year and a half to figure out that the problems I was having with Netflix HD stuttering were directly a result of AMD's Cool'n'Quiet.

I just couldn't figure out why a 2.9 GHz triple-core AMD Athlon II X3 435 couldn't play a Netflix HD stream smoothly (even when paired with two different GPUs, one nVidia, one AMD/ATI, both with with HD video decode acceleration).

Then I changed one BIOS setting to turn off Cool'n'Quiet and the problem immediately disappeared almost completely.

http://forums.anandtech.com/showthread.php?p=34919065#post34919065

How does this relate to the OP's problem? Not sure it is related at all, but I too suspect it is a bug related to the software implementation and its inability to properly deal with Cool'n'Quiet, as opposed to a CPU max performance issue.

Plus, I suspect if the CPU had no Cool'n'Quiet but cut out a full quarter of the performance from this 2.9 GHz triple-core CPU, it'd still have the horsepower to deal with the HD video even if purely software decoded. (I never went past 65% CPU usage, yet the video would stutter.)

 

[Ben90]

I notice the difference between 120hz and 60hz really keenly (there is a bug there in NVidia's drivers for example with multiple monitors where both render the mouse at 60hz when a second monitor is added regardless of its hz setting, really annoying!).

I can tell whether SLI is on or off just by moving my mouse for example, it has a noticeable impact on the performance of moving windows and the mouse around the screen.

Seems like your drivers are jacked up to me. I know computer bugs can be hard to replicate sometimes, but I've never had a problem with my mouse and multiple monitors at refresh rates 120+ on Nvidia hardware. Your post just screams SLI issues. Having a downclocked CPU is probably only exemplifying the problem to be noticable and not the actual source of it.

I don't know your operating system, but if its 7 try turning off Aero see if you can reproduce the problem in classic mode. I remember having some accute problems in Aero that very few people on 120Hz screens would notice.

*edit*
And to address the person above me, Cool and Quiet has always had performance problems. Intel's solution has worked with a very minimal performance impact since the Core 2 days.

 

[mikeymikec]

Re: AMD CnQ:
http://www.anandtech.com/show/2775/2
I'm not 100% sure how to read this. The last conclusion I drew when reading it today is that the AII's CnQ implementation is broken, possibly through lack of C1E support... I think.

I'm wondering whether both issues are down to board implementation issues of SpeedStep / CnQ, though I think the OP is overclocking, which I would get rid of for the sake of testing for this issue.

 

[BrightCandle]

My drivers would have to be jacked across Windows 8 and Windows 7 across 3 installs and 10 or so different driver revisions and 2 different motherboards as well. Its not like this is a recent problem, I wrote about it nearly a year ago first, I just only found out the cause.

Yes I am running 100% minimum clock speed, which is why the CPU always stays at 4400.

It is only really when overclocking, but any amount of overclocking will do it. I wonder if its something to do with the way the system forces all cores to the same clock speed.

 

[mikeymikec]

It is only really when overclocking, but any amount of overclocking will do it. I wonder if its something to do with the way the system forces all cores to the same clock speed.

IMO, yes, board implementation issue. I once had a board that if I specified the same RAM timings as the board would use automatically, memtest86+ would show memory faults. Set it back to the auto settings, no problem.

 

[BrightCandle]

IMO, yes, board implementation issue. I once had a board that if I specified the same RAM timings as the board would use automatically, memtest86+ would show memory faults. Set it back to the auto settings, no problem.

I have had it across 2 boards however, on the Asus Pro and again on the Asrock extreme 9. I can no longer confirm if turning off speedstep fixes it on the Asus because it committed suicide some months ago but having had the same problem on the Asrock it does seem likely that its not board specific.

 

[blackened23]

Sounds like a motherboard USB driver issue to me. Are you using the intel USB 2.0 port? I have never had such an issue (using Asus mobos). I have had weird issues pop up when not installing correct intel USB drivers, but aside from that - mouse movement has always been super smooth.

 

[tweakboy]

There is no lag for me I use C1E and Speedstep ...... thx gl

Something is up with your rig if your laggy delayed mouse etc... gl

 

[roob]

i've had the same issues when using the keyboard. with speedstep active, moving the current line up or down multiple times via the arrow keys would be noticeably jerky. You would also see the cpu go from 800-1200mhz to ~2 ghz while moving the caret. Ever since I've disabled speedstep and installed throttlestop it's been fine.

imo, these sort of gui edge cases are impossible to prevent with speedstep. by the time the cpu enters higer power states, the animation will have already been in progress, making the difference more jarring (laggy -> smooth). then you stop scrolling or whatever, cpu downclocks and the same happens again.

though in my case it's most likely due to the intel graphics, I would think using a dedicated gpu this shouldn't be an issue.

 

[Sheep221]

This is a bit of an odd one. For a while I have this feeling the mouse was lagging, that it was not going where I put it as quickly as it should do. I wasn't sure why that was the case, it just didn't feel right. I notice the difference between 120hz and 60hz really keenly (there is a bug there in NVidia's drivers for example with multiple monitors where both render the mouse at 60hz when a second monitor is added regardless of its hz setting, really annoying!). I certainly felt the mouse was wrong. In the past I have noticed that the mouse movement speed is impacted by CPU clock speed changes, when machines downclock to 1.2Ghz and such to save power when I move the mouse it feels like its slower, actually I think its just latency.

I decided yesterday to go on a system wide driver update, haven't done one in months. Got a new bios as well and the new bios messed up the overclocking on the board. So I got a beta bios from the company in response to my support ticket. In the process however I found that the bios that didn't overclock properly did not have a laggy mouse, which was very odd. The beta bios as soon as I over clocked reintroduced the mouse lag.

Despite having c states and speedstep and such on with the beta driver the clock speed never drops below what I have set. However the voltage does change quite a bit (using dynamic VID to save power consumption for 24.7 overclock) dependent on load. When I turn speedstep off the voltage stops varying and the mouse stops lagging as badly as it did.

Its a curious effect, I wonder if anyone else has noticed this. I guess its quite subtle since I don't see many people complaining about it but to me its really obvious. I can tell whether SLI is on or off just by moving my mouse for example, it has a noticeable impact on the performance of moving windows and the mouse around the screen.

Overclocking, speedstep, frequency lock and undervolting, isn't that enough?

Your system only works just under some conditions. This for sure has to cause problem with data processing between CPU and Chipset which controls such a tasks.

 

[BrightCandle]

Sounds like a motherboard USB driver issue to me. Are you using the intel USB 2.0 port? I have never had such an issue (using Asus mobos). I have had weird issues pop up when not installing correct intel USB drivers, but aside from that - mouse movement has always been super smooth.

Yes using the standard USB 2.0 headers provided by Intel on my board. Although why that would be impacted by power saving on the CPU I don't know you'll have to explain your logic to me there. I notice a bundle of little delay problems associated with power saving with Intel, I just guess most people aren't that sensitive to these small transitions and don't have the monitor to make it obvious. I was one of the few people that said that crossfire was awful 20 months ago on release of the 7970 and it took most people well over a year and the framerate tests to come to the same conclusion. I am just very sensitive to this stuff. The fact you don't see it unfortunately doesn't mean its not there. Could be something else in my setup but I have it ever since I got a x79 and I have seen it on an Asus and Asrock board, on 7970's and 680 graphics cards in single and SLI/Xfire. I am pretty certain this is a CPU power saving issue when overclocking.

I first noticed this when I was using the machine after a bios upgrade and benchmarking for stability. With prime95 running in the background the mouse was more responsive than when it wasn't. But since the clockspeed wasn't varying that this was very weird. That has no impact on the GPU at all, it doesn't change clocks just with prime 95 in the background, its a pure CPU related.

Removing speedstep seems to have improved it but even now testing against prime95 in the background its still not right, so I guess the c states need to go as well. Power saving doesn't look quite as seamless as Intel would have us believe as far as I can tell. Its potentially quite subtle depending on your monitor and mouse (most people have totally rubbish gamers mice).

 

[blackened23]

A little over-sensitive there buddy? Either everyone that doesn't experience it is crazy, or everyone does experience it, or there is something specific to your mouse, mouse driver, or motherboard causing the issue. I was just throwing a suggestion out there: try a different USB port, different mouse or updated intel USB drivers. I get weird Usb issues when I use my ASMedia USB ports on my asus board, but absolutely zero issues with the intel ports. In fact I have had my mouse pointer go awry when using non intel ports, but this is very specific to your motherboard and how the USB implementation is done. Or even your mouse drivers. Razer mice are absolute garbage in this respect, I have since quit using them for strange issues such as that.

Whatever, though, man. I was just throwing that suggestion out there for you. Best of luck getting your stuff fixed. /peaces the hell out of this thread

 

[NTMBK]

It's all in your head.

 

[mikeymikec]

Yes using the standard USB 2.0 headers provided by Intel on my board. Although why that would be impacted by power saving on the CPU

I seen it before with AM3 CnQ resulting in significantly lower USB transfer performance (something like 10MB/sec less IIRC, it was about three years ago I think). No special circumstances either, clock frequencies, default settings, etc.

Not saying that your problem is or isn't caused by xyz, just responding to that point in particular.

 

[soccerballtux]

Re: AMD CnQ:
http://www.anandtech.com/show/2775/2
I'm not 100% sure how to read this. The last conclusion I drew when reading it today is that the AII's CnQ implementation is broken, possibly through lack of C1E support... I think.

I'm wondering whether both issues are down to board implementation issues of SpeedStep / CnQ, though I think the OP is overclocking, which I would get rid of for the sake of testing for this issue.

I use PhenomMSRTweaker to get around the CnQ implementation completely. Only 2 pstates, super slow and full speed are all you usually need. Maybe diff for netflix, you can make more states if you want but meh.