SetFSB Game Timing Problem? (G60JX)

First, thanks to anyone that takes the time to read and respond to this post. I really appreciate it.

I recently got an Asus G60JX and of course, I had to overclock it. Since SetFSB is the only thing that works on my Core i5 430m, I bought a copy. It overclocks well, I can max out the FSB @ 170 mhz which results in a 2.9-3.2 ghz overclock, depending on how Turbo Boost wants to manage my multipliers.

However, in Assassin's Creed Brother and Crysis 2, the overclock has recently been causing everything to go in 'hyper mode': if you've ever taken the frame limiter off an emulated game, you know what I mean. It's running at 60 FPS but everything is super fast: jumping, sounds, dialogue overlapping, characters walk too fast, etc. Please see this thread for another user's example:

Setfsb, MAJOR problems with it - Overclockers Forums
http://forum.notebookreview.com/asu...395-setfsb-ocing-g73jw-cpu-2.html#post7148617

My question is, how do I stop this? Is there any way to force the game's timing back into place without messing up the performance?

If anyone knows a fix, please let me know. Thanks.


EDIT: SOLVED, thanks to UncleWebb! Anyone that overclocks with SetFSB with ANY computer should read this!

 

I noticed this in Half-life 2 and other Valve games, dialogue and even some scripted events would become out of sync.

 

Did you figure out a solution? It seems to happen a lot in Source games (I didn't see it in Left 4 Dead), but also AC Brotherhood and Crysis 2.

 

Nope, no solution at all. my solution was to turn off my overclock.

Also, in my Battlefield games, overclocking the cpu resulted in severe stuttering and lag issues. Once again, returning to stock clocks resolved the issue.

 

I have a somewhat related problem, but different platform(G73JH). When I use SetFSB, my G700 Logitech mouse actually slows down. I have to double the DPI in order to get the performance back to what it was prior to using SetFSB. I've just stopped using it since I did not see any performance increase in the games I was playing. The only time I did see improvement was in benchmark programs like 3DMark.

 

Damn, that sucks. I wonder if overclocking somehow messes with Turbo Boost and the way it changes multipliers.

Can you turn SetFSB on and off to check if it's 'throttling' your mouse's response in real time? Or does it take a while to show the effects?

 

As soon as I set the FSB the mouse slows down. The only way to get it back to "normal" is to reboot the PC.

 

I've come across this myself, and the best explanation I've been given is this. With the i series cpu's, when you use SetFSB, it messes with the "timing" of Windows. Meaning the internals are out of sync. What I mean by that is not so much the clockspeed of the components, but the latency, I think.

Instead of the videocard and audio card for example running at a 1 to 1 ratio, they get out of whack and get the "yipps" so to speak.

This may be a poor explanation, but this is the best way I can translate what is in my head to typed words. Perhaps someone else can better explain it.

 

That's really weird, but maybe Keith's theory below could explain it.

I understand completely. Basically what you're saying is that when Windows starts up, your CPU is 'timed' a certain way. And then changing the FSB dynamically in Windows messes things up. This makes sense in a way, but then I look at Power4Gear and its Turbo/Extreme Turbo settings; when it raises the FSB in Windows, it's doing what I'm doing with SetFSB. I uninstalled Power4Gear, but I wonder if it also causes game timing problems when it raises the FSB.

One thing that's odd is that when I set the FSB to 141 like Power4Gear does for Extreme Turbo (turbo = 138 FSB), games don't SEEM to be sped up. I've kept my FSB below 150 and so far I've noticed no game timing problems.

This means either:

A. when Power4Gear raises the FSB (to 138/141(, the effects of game timing speed-up are without direct comparison/measurement, but they ARE happening, but it isn't as perceptible as the higher FSB settings or

B. The game timing problems only happen after a certain FSB, perhaps the games only have a certain threshold of timing correction and setting the FSB too high causes them to 'lose sync'.

It's really hard for me to play Crysis 2 and compare 133 FSB vs 141 FSB and notice a difference without recording it, but 160-170 FSB is noticeably sped up. Anyone know an objective way to determine whether A or B is true?

After my above post, I tried a new Google search and found a really old, interesting thread:

FSB overclocking on AMD2600+

itseme:
At 170MHz the System runs stabile but in Games like Unreal2 it seems to have timing-problems so i play at 12.5x168MHz (timing-problems means that the Game stutters some times)

That's interesting that the FSB is stable and runs faster (in theory this would give higher performance) yet it causes game timing problems in Unreal 2 similar to what you had in Battlefield (stuttering etc). Which looks like it means that at a certain FSB, games actually take a performance hit due to timing issues.

So this might mean there's a certain 'game-stable FSB' that gives a performance boost without timing issues, sort of how I theorized above that there's a 'threshold' for games.

 

Power 4 Gear or SetFSB for that matter will not have any efect on the timing up to a certain point. Once you get beyond a certain point, 141, 142 143, etc. that's when the timings get messed up.

 

Ok, so one of those theories is correct then. Have you tested it with SetFSB to see where the threshold lies?

 

Not officially.

But I can tell you it seems like once you go beyond where Power 4 Gear overclocks to, that's when things seem to start getting ot of whack.

 

Hmm ok. And perhaps going past Power4Gear's 141 FSB doesn't give huge gains in games. I know it gives huge gains in benchmarks and everyday computing, though.

 

The problem you guys are running into is a new Windows 7 "feature".

To accurately measure time in a Windows PC, software uses the QueryPerformanceCounter function. Many applications, especially games, depend on this Windows function to accurately calculate FPS as well as for many other timing purposes in game including sound.

The problem is that for many computers, this function in Windows 7 is now based on the clock speed of the processor. If you overclock in the bios and then boot up, this timer is calibrated and will work 100% correctly. If you use SetFSB or a similar program and you change the bus speed from within Windows, you have now just screwed up this very important timing mechanism. The number of applications that can choke after you do this is surprising, especially games.

I wrote a program called WinTimerTester to test for this problem.

WinTimerTester 1.1
http://www.mediafire.com/?xzo9n84d8lze9nb
http://www.sendspace.com/file/xadvhe

What it does is it runs two different clocks in your CPU at the same time and compares them to make sure they are both running at the same speed. It compares the Windows GetTickCount function to the QueryPerformanceCounter function and within approximately 60 to 100 seconds, these two timers should be running at a perfect 1.0000 : 1 ratio.

If you overclock your CPU with SetFSB and WinTimerTester reports that these timers are not running at a 1.0000 : 1 ratio, that shows that you have this bug and you have also overclocked this important timer within your computer which can screw up a variety of programs in a variety of ways.

Luckily there is a solution to this Windows bug. Microsoft actually considers this a feature but for anyone that uses SetFSB on a regular basis, it's a big bug.

To fix this problem, open up a command window and type in this:

bcdedit /set useplatformclock true

You will have to exit the command window and reboot before this setting can take affect. What this does is it changes Windows so the QueryPerformanceCounter function will be based on a fixed counter that is not influenced by SetFSB. That one line of code should correct this problem and your games will run fine when using SetFSB.

If you ever want to go back to the original buggered up timer that Windows uses then open up a command window and type in this.

bcdedit /deletevalue useplatformclock

Once again, reboot and things will be back to the default Windows 7 timer. These two images should show you when your internal timers are broken and what WinTimerTester will report when your internal timer is fixed. This only applies to Windows 7. There is a similar fix for Windows XP so PM me if you need it.

http://img202.imageshack.us/img202/3778/timerbroken.png

http://img96.imageshack.us/img96/8417/timerfix.png

Let me know if this finally solves your problem. My friend some_one over on XtremeSystems told me about this little trick.

Disgustipated: On a different note. It would be a good idea to run ThrottleStop and increase your TDP/TDC settings to ensure that you get full turbo boost when you are overclocking with SetFSB. If you don't increase TDP/TDC then your bus speed can go up but your multiplier might go down when fully loaded so you won't get the full benefit of your overclock. Send me a PM if you need to know more about how to properly test for full turbo boost at full load.

One last thing. Many 3D gaming benchmarks don't correct for this problem so you can't trust their results when using SetFSB if your laptop has this timing issue. I believe the 3DMark series corrects for this problem and can be trusted.

 

WOW. I've been using Throttle Stop and it's amazing, but to have the creator come and totally figure out what's causing my problem with a *different* program AND offer a solution - just wow.

THANK YOU, Unclewebb. I'm going to test and see if indeed this does fix the timing program, if so, SetFSB overclocking for games is a GO!

I'll be sending you a PM to learn about testing for full turbo boost.

Again, thank you. You might have just solved a game timing issue for not just G60JX owners, but for all gaming laptop owners limited to SetFSB.

Update! Just finished testing and changing settings - Unclewebb's done it! Works perfectly - no more timing issues, the games play perfectly but with a nice boost in performance thanks to my 3.1 Ghz overclock (stock turbo boost only goes to 2.5 ghz).

Stable so far in small FFT mode with Prime95 @ 165 mhz for the FSB. RAM is @ 660 Mhz.

 

You're welcome Disgustipated. I guess you've got nothing to be disgusted about now.
You'll be too busy checking out all the games that were screwed up before.

On my desktop computer, this was a bug in Windows XP, Microsoft fixed it in Vista and then reintroduced it as a new feature in Windows 7 so they could be ACPI compliant.

IBM and other manufacturers know about this problem so if anyone is afraid to try this solution, you can read all about it here.

http://www-947.ibm.com/support/entry/portal/docdisplay?brand=5000008&lndocid=MIGR-5084072

Copy and Paste this info to other forums so users can find out about this easy and permanent fix.

For Windows XP, you need to add this to the boot.ini file.

/usepmtimer

 

Thanks UncleWebb. I posted this on NeoGAF and in the Gaming section of NotebookReview. This is going to help a lot of people.

 

Yeah, thanks! This worked for me as well. My mouse doesn't slow down any more after SetFSB is activated. Performance has increased slightly too.

One thing though, when I run your program after the "fix" to check the ratios, I always get .998x:1 wether I use setfsb or not. This is better than the 1.05xx:1 I would get prior to the fix when using setfsb. Should it be a perfect 1.0000:1?

 

When you boot up, before you use SetFSB, what does WinTimerTester say after you let it run for about 100 seconds? After that amount of time, it's usually a perfect 1.0000 : 1 ratio for me.

A tiny difference like 0.998x means there is still some error in your high performance timer but at least that shouldn't cause any noticeable problems in games.

The problem might not be noticeable after a slight bus speed adjustment but the problem is still there. Try adjusting the base clock from 133 MHz to 135 MHz and WinTimerTester should show you that you are screwing up your internal clocks.

 

So in other words even the built in overclock apps from Asus and Msi for example will have this bug too, but not be as obvious. Thank you sir.

 

http://forum.notebookreview.com/7364805-post14.html

The images in post #14 showed this problem when I was using Asus TurboV on a desktop board. If you are using Windows 7, any application that you use to adjust your bus speed from within Windows will likely screw up the QueryPerformanceCounter function that many games depend on for the accurate measurement of time. Some games and applications know about this and will correct for it so you won't ever notice if there is a slight timing error. If software doesn't check and correct for this bug, you can end up with parts of games running at variable speeds and it might also cause timing problems when playing sound files. Any software that assumes that function is 100% accurate will not function correctly.

I don't own an Asus laptop so if you're curious, do some testing with WinTimerTester and post your results. If you run WinTimerTester for at least 60 seconds, it should have no trouble detecting a bus speed adjustment of as little as 1 MHz. If you boot up at 133 MHz and you increase that to 134 MHz with SetFSB, WinTimerTester will show the ratio 1.0075 : 1 (134/133)

If it shows 1.0000, your laptop doesn't have this bug or you fixed it with the bcdedit /set useplatformclock true function.

I've seen some boards where if you boot up at one frequency and then use SetFSB to go higher and then change your mind and go back to the original bus speed, you might end up with a slight amount of error in the timing. This is a sign that SetFSB doesn't always set the bus speed to the exact same value that the bios sets the bus speed to.

 

Okay, I tested again. This time actually doing it right. I wasn't letting the prior test run long enough. I had the "bug", but now, usually between 60-110 seconds, the test shows a 1.0000:1 ratio. It usually starts at .996x:1, but then rises to the perfect 1:1 ratio. Thanks unclewebb!

 

I just heard from front360 and this timing problem can also happen when using the Asus Twin Turbo feature.

If anyone uses Twin Turbo, can you download WinTimerTester and run a test with that while using Twin Turbo and then post a screen shot. Thanks.

WinTimerTester 1.1
WinTimerTester_1.1.zip

Windows timers that are running out of sync can cause problems for a variety of games and programs that depend on these timers for high precision timing. Follow the first post in this thread if you want to fix that problem.

 

Unclewebb yes it does, the timing fix you created also fixes the problem with TwinTurbo mode. When active without the fix it continues to raise the ratio as per my video (SIG) I show this happening.

It appears all of the issues that we faced struggling to understand why TwinTurbo ran so poorly was also fixed. I notice a positive difference in games and software using Twin Turbo (which raises the PLL in SetFSB from 139-153 on a 720QM) which I did not see before applying the DOS fix.