Slower Clock, faster CPU?

[siravarice]

So I was running my Q8200 @ 3.2GHz but started getting some crashes when I was running games after a while. Did some scans for viruses and what not and ran CCleaner and was still getting them. Turned the overclock down to 3GHz and now it seems to be actually running faster and doesn't crash anymore. Is it possible that it was originally an unstable overclock even though in stress tests it was fine?

Check sig for other components.

 

[86sixsix]

Are you running windows?

I suppose it's possible. If something like the 'System' process on windows falls down and gets locked into a loop waiting for some hardware to respond or something... Or the NT kernel or whatever... It'd make the machine feel sluggish but not necessarily fall over...

I guess a simple test would be to run the stress tests at the higher clocks in safe mode. Then 1 by 1, start adding back in all the daft services called up.

See what's in your event logs. That might shed some light on it. Perhaps something svchost has called went a bit squiffy...

Perhaps at really silly clock speeds you could get soft ram errors and whatnot. Are you running any posh ram, ECC or whatever?

Were you getting application crashes or more serious problems with blue-screens and the like?

If it was a software issue like so called malware or viruses I'd imagine you'd get the errors regardless of clock speeds...

I'm new on this forum and one thing I have seen is that if a benchmark/stress test completes without errors they class the overclock is stable. This is something I don't agree with. People are doing things like 8 hour stress tests. What's 8 hours when they might want to a 24 hour gaming marathon or whatever...

Don't forget - when you're running these tests... Only little bits of your machine's going to be working. Moving a mouse and generating interrupts or playing music and playing havoc with DMA and things can make a machine fall down...

In a nutshell... if you're getting errors - it's unstable. Regardless of clock speeds. A stable machine is one that produces no errors. Ever.

 

[siravarice]

No just some Kingston 800 ram. It's not overclocked though. It wasnt blue screens or anything, it would literally freeze and do nothing and you would have to restart. Running windows 7 pro 64bit. I know it was originally a bug with windows 7 but i thought they had fixed it by now. Anyway, everything seems fine now, haven't had a problem since putting it down to 3GHz. If I start getting the problem again then I will look into it more deeply.

 

[86sixsix]

Just noticed 'i ran ccleaner'...of course you're in windows... When it froze could you CTRL-ALT-DEL to reboot or did you have to prod the reset switch?

I've just put together a couple of boxes running 7 64 ultimate...but with AMD/nvidia gubbins in... Be good for us to compare faults/flaws...

If you get bored of 3Gig... perhaps have a look in system information (run msinfo32 as admin) and have a look in summary/hardware resources/conflictssharing...

Try and get that list as short as possible. Things like soundcards sharing interrupts and stuff can flatten a windows machine... Ideally the most 'responsive' system won't have anything in that list. That'd make it feel faster clock for clock. And possibly help with any stability issues hiding in the background.

Can you for example run the machine and do your benches and whatnot at 3.2gig with only a chip,harddrive,ram and a mouse plugged in? and with all the extra motherboard things like sound/firewire/usb/etc etc switched off? This takes stress away from your systemboard as it doesn't have to poll lots of oddbod hardware.

A hard-lock (just freezing, no errors, no bluescreen) tends to be memory related...or perhaps you're pushing a mosfet or 2 over the edge heat-wise...

If it locks while playing a sound and the sound loops like a stuck vinyl records...or moving the mouse or pressing keys...then it's IRQ/address/DMA/IO related...

Best of luck with the 7 bud... hope you get on with it better than I am! (I'm solaris till i die...)

 

[86sixsix]

Have you got a link to the 'original bug in 7' you mentioned? I've never seen anything on technet...

And perhaps try running slacker latency on the ram at 3.2gig...see what happens then...

 

[siravarice]

No I don't I'm afraid, this was one of the first bugs to be encountered and I was pretty sure they fixed it. The computer would freeze and I could do nothing except reboot. It's been fine for ages but just started doing it all of a sudden, when I get more time I will see if I can loosen the timings on the RAM.

 

[86sixsix]

What build was the bug in? (And what's the build you're on now?) - I'll see if I can replicate it on an intel box...

If it was bob-on for ages at 3.2 and suddenly gone freaky then i'd bet a device driver or a naughty system .dll is the cause of your woes...

Try giving your sticks something really lazy like 6's 18 2T and see how you get on... maybe it'll cut the controller some slack... I shouldn't ask but I trust you're at the right voltages ;-)

 

[siravarice]

Oh yes very much so, just the standard voltages for ram, and the what was the perfect voltage for the overclock. I have no idea how to find out what build revision it is though lol

 

[86sixsix]

For the version information - run "slmgr -dlv" from an admin command prompt... 6.1.xxxx.xxxxx

I had a look through the 7 knowledge base and couldn't find a single mention of a hard lock that was down to the OS...

As I've said...if it was right at 3.2 for ages and then gone mad all of a sudden (and you haven't changed a single piece of hardware or changed anything in the BIOS) then it's down to either:

a) A background process (Run a 24hour stress test in safe-mode)

b) A new driver, either a poorly coded one or one that's made a device change its resources (Have a look at your APIC tables and see what's going on regarding IRQs and whatnot) - try removing all unnessary devices as I've said earlier

c) A couple of tracks or mosfets on your board or other bit of hardware getting a bit old and tired

Let me know your build and I'll try to replicate the problem... I've got a system that's close to yours... But with Corsair RAM...

 

[biscuitboy]

I seem to have a Black screen problem in 7 64bit too..... Hmmm, Stable and stress tested, still randomly if left Idle the screen goes black and Will not resume or anything....

I'm not so sure they have resolved the issue tbh.

 

[86sixsix]

All this points to flaky display drivers causing soft ram errors... Uninstall the 'time zone' security update... Install the latest 'windows defender' definitions... Then disable the defender service (don't just turn it off)... Disable realtime AV/malware/spyware/whatever... Disable anything remotely power-management related... Enhanced halt states, speedsteps, CoolnQuiet, Screensavers, turning off the display, adaptive suspend, Set windows to 100min and 100max as far as CPU goes in powermanagement... Don't let windows enter 'away mode'... Uninstall your display drivers, switch to the "Standard VGA" driver as provided on the windows dvd...don't let windows update pick a newer version... and stress test in safe mode.

At least that'd differentiate between a software / hardware hiccup...

Perhaps S1 and S3 states are coming into play at silly times? I must level with you... windows isn't my thing... but these are the 1st thing's i'd try.

Do the op and biscuitboy get the problems in safe mode? And what build of 7 is everyone on? I run 7600.16385 but really can't handle it's crappy performance scaling...

The op didn't mention a black screen so I doubt these issues are related...

A good test for biscuitboy's blackscreen-ness would be to do my suggestions as per my last post - then switch off, cold boot and let it sit at the login screen without even touching the keyboard... see if you've gotten the Winlogon shell registry key problem... maybe, maybe not...

...another testing step...not likely but still...

Run in safe mode, create a new Admin-rights user, cold boot...and test again in safe mode. Perhaps something's mangled in your user profiles? don't trust ccleaner and the like... they seem to be a bit over-zealous in what they delete...

 

[siravarice]

Just realised that all you need to do is open up command prompt and it says what build revision it is at the top line. It's build 6.1.7600. Although the full build number is 6.1.7600.16385

 

[86sixsix]

Indeed it does...but it's the build (end bit) that's important...

Have either of you done any testing with the Standard VGA driver in safe mode?

I trust everyone's done the basic stuff, restoring every clock you've got to stock, tapping num-lock to be sure it really has locked, making sure windows is up to date, re-ran directx setup, chkdsk'd everything, unplugged everything they don't need, not let ccleaner and whatnot remove everything it found, disabled all the twaddle services (indexing, defender, firewall (unless it's the only one you've got lol), media sharing and what have you), turned off any realtime rootkit and av stuff, tried when logged on as a *new* user, made sure nothing freaky's coming back from secedit or sfc, got rid of all device sharing, closed all their local user's processes...

I shouldn't ask about all that rubbish...I'm new here and don't want to upset anyone but again someone says 'my machine is stable but it locks up now and again' - surely it's one or the other lol...but sometimes it's easy to overlook the basic stuff...I do dumb stuff all the time lol

It's just the only time I've seen hard locks in windows machines were due to soft ram errors, cr*p code in display drivers, or having keys missing or with the wrong permissions in the registry...

I'd love you help everyone but without seeing eventlogs and minidumps or whatever windows calls them these days I'm just guessing...

 

[86sixsix]

That's about all I can think of windows wise...

...i trust people've got all the cke strengths, clock drives/dram drives and terminations set properly?

 

[siravarice]

Not to that extent, but it's not something I'm worried about really. It's running faster and smoother now that it ever has. I updated my drivers for my graphics card (they were quite out of date) and everything is up to date. I perform regular scans and what not, run CCleaner, do disk checks and so on. I'm pretty sure it's not the motherboard, it has always been very good to me and my case has always been pretty cool. Voltages are what they need to be and not over the top, I lowered them as much as possible before the overclock was unstable, then raised them a bit until i felt everything was working great. Planning some upgrades soon anyway so if the whole thing does go kaput, I will just bring the upgrade forward!

 

[86sixsix]

^Those are just some of the basics...

You've done the right thing by trying some new display drivers (195.81 is the most stable for me as far as my nvidia bits go)... As an example of something you may have fixed, version 182.50 didn't drop the grfx cards out out of 'performance 3d' mode and as a result they used to pull more amps than they needed and eventually took a teensy tiny bit out of my 3.3v rails on the boards...

Just don't always trust ccleaner - unless you *really* know what all the classid's and controls you're frying actually do... In which case - you need to teach me your kung-fu...

Most of the stuff I'm into is (what's the term?) 'more low-level' than windows lets you get into... but I hope I've thrown some at least quite useful hints people's way...

OFFTOPIC: And if you ever go on a mission with your 260... be sure to let me know the results... The most I get stably out of mine at stock.v is 700/1490/1275... ...I'm sure there's more to get out of the ram on these things...

Best of luck...

 

[siravarice]

Hmm, I could probably give it a go. My 260 does still run very cool considering the huge overclock.