need help with 4.5 overclcok

[zan76]

hi guys need some help getting my 4.5 overclock stable LLC HIGH

so far i have run prime95 on

1.330

1.335

1.340

1.345 volts and failed with FATAL ERROR: Rounding was 0.4999871254, expected less than 0.4

so when i get this i stick the volts up a notch

any other setting i should change.

 

[sheroo]

Have a read through this it might help a bit. If you want to skip the text scroll down to the bios piccies.

http://hardforum.com/showthread.php?t=1578110

 

[zan76]

Right had a read thru that.....

So changed llc to ultra high and enabled pll overvoltage

Rerunning prime on 1.345 volts so far so good just over the hour Mark on prime.....

 

[sheroo]

Nice one - sounds good.

 

[Scoob]

Have a read through this it might help a bit. If you want to skip the text scroll down to the bios piccies.

http://hardforum.com/showthread.php?t=1578110

Useful information in that guide, I used it myself.

OP: I'd say that a bit more vCore should be all you need. LLC of High (which is the default) is fine, I only needed to up it to "Very High" (so one notch above default) to get 100% stable at 4.6.

Note: I do NOT change my CPU vCore directly, rather I use the offset method. For 4.6 I needed + 0.050v, for 4.5 I needed +0.040v.

So, for your 4.5 target on your board (which is basically the same as mine in many respects) I'd suggest:

vCore - unchanged, but enable Manual adjustment

vCore Offset - +0.040 shouldn't be far out

Everything else, bar multiplier, unchanged. I leave all idle state stuff enabled so the CPU idles at 1.6ghz - using the OFFSET method above will see the motherboard correctly lowering vCore based on the VID for the applied multi - in my case ~ 0.968v @ 1.6ghz idle. I see ~ 1.4v at 4.6ghz under HEAVY IBT load, slightly less when under Prime 95 load.

ALL chips of course can behave slightly differently, judging by what others are achieving my chip seems to be pretty much middle of the range.

Good luck.

Scoob.

 

[zan76]

How long should I do prime for....will I leave it for the 6 hours

 

[Scoob]

Right had a read thru that.....

So changed llc to ultra high and enabled pll overvoltage

Rerunning prime on 1.345 volts so far so good just over the hour Mark on prime.....

Hi,

I did not need to adjust either of those for 4.5 - you shouldn't really for just 4.5. I played around with such settings and while it gave me a degree of stability it also pulled more v's and got hotter than needed. As part of my own learning process I managed to refine my settings as in my post. Now my CPU runs COOLER at 4.6 than it did at 4.5.

The key is to adjust ONE thing at a time else you'll not know which gave the desired effect. Setting an excessive LLC can to a certain degree compensate for too low a vCore, but this can lead to instabilities when NOT at full load potentially, or just higher temps than need be. Even the linked document (unless it's been changed since I read it) says that LLC should not need to be changed for 4.5.

Again I'll add that all chips are different so possibly a more agressive LLC setting is needed for your particular chip. However I will suggest that if your current batch of tests seem stable you continue to refine the overclock to get it as optimised as possible.

Note: IBT (Intel Burn Test, their own testing tool) will push your system harder than Prime 95, it will usually show any instabilities within a matter of a couple of minutes. I've seen systems that will run prime for hours, but last only seconds under IBT. Equally, I've seen systems that are fine with P95 but have the odd CTD while gaming...getting the system IBT stable (just the standard test, takes a couple of minutes to run) saw all the issues go away.

Long-term testing using P95 still has it's place for sure, I still use it, but in my view IBT gets you 99% of the way to stability if not 100%. Be aware that IBT needs W7 Service Pack 1 installed for the new AVX instructions supported by Sandy Bridge - without this IBT will not push your CPU hard at all. From memory a STOCK 2500k should be seeing ~90 GFLOPS, at 4.5 this should be ~120 GFLOPS. If your values are approximately HALF this then you're either running a version of IBT older than 2.51 (older linkpack libs which do NOT use the new AVX instructions) or you're not Service packed.

Best of luck.

Scoob.

 

[zan76]

so what version of ibt should i get ...

 

[zan76]

run ibt 2.52 these are the results.

http://imgur.com/Zs0Xo

what should i do now ....

 

[zan76]

right ok i also tried it the other way you said offset mode 1.4v with +0.040v. llc still on ulra high and still with pll overvoltage enabled the results with ibt were ....

Processor: Intel® Core™ i5-2500K CPU @ 3.30GHz

Clock Speed: 3.33 GHz

Active Physical Cores: 4

Total System Memory: 8168 MB

Stress Level: Standard (1024 MB)

Testing started on 05/09/2011 16:34:58

Time (s) Speed (GFlops) Result

[16:35:11] 8.492 105.2597 3.061384e-002

[16:35:24] 8.574 104.2505 3.061384e-002

[16:35:37] 8.605 103.8746 3.061384e-002

[16:35:50] 8.415 106.2256 3.061384e-002

[16:36:03] 8.368 106.8190 3.061384e-002

Testing ended on 05/09/2011 16:36:03

Test Result: Success.

----------------------------

ok re done it still on offset mode now1.36v but lowerd this bit down to +0.020v instead off +0.040v volts under load were better than the one before.

Stress Level: Standard (1024 MB)

Testing started on 05/09/2011 16:43:52

Time (s) Speed (GFlops) Result

[16:44:05] 8.501 105.1513 3.061384e-002

[16:44:18] 8.358 106.9485 3.061384e-002

[16:44:31] 8.598 103.9607 3.061384e-002

[16:44:44] 8.568 104.3252 3.061384e-002

[16:44:56] 8.383 106.6268 3.061384e-002

Testing ended on 05/09/2011 16:44:56

Test Result: Success. This one seems pretty spot on to me .......

 

[Scoob]

Hi,

Looks like you've got some good stability there, I know the test doesn't take long but it IS really stressful.

One thing that does confuse me slightly however, you say you're using 1.4v with an additional offset of +0.040v...is 1.4v what your BIOS shows your vCore as default then? Mine, on a P8Z68-V Pro, is shown as ~1.28v WITHOUT me changing anything apart from enabling manual vCore control. I add an offset vCore of +0.050 for 4.6ghz and this equates to around ~1.360v under gaming type load, but ~1.4v under IBT load. It looks like our starting points are differnt OR possibly you didn't clear your prior manual vCore setting before switching to offset mode.

It is very much a process of refinement as overclock SB is quite different to my prior experience with my Q6600 and Q6700 processors.

My personal guide, i.e. rules I like to stick to when overclocking Sandy B. is to not exceed ~1.4v under heavy load - it's less under typical load, and ensure my temperatures do not exceed the low-70's in IBT, which equates to no more than the low-60's in intensive gaming. I think at the moment you're using both a manually set vCore and an additional offset - not quite what I'm doing.

I only have fairly modest cooling with my Antec Kuhler 620 closed loop. It copes very well at my current 4.6ghz but, when I was playing, it didn't cope so well at 5ghz, hitting the low 80's which is out of my comfort zone. Still, this was done during some rather warm weather.

I think you're making good progress, but some further refinements will see your OC stable and as cool and low-voltage as it can be.

Cheers,

Scoob.

 

[zan76]

before you gave me advise i had my volts on manual at 1.345 v but since ur advise i changed the volts offset mode .

when u say further refinements what else could i do ...

 

[Scoob]

before you gave me advise i had my volts on manual at 1.345 v but since ur advise i changed the volts offset mode .

when u say further refinements what else could i do ...

Hi,

I think it might be remembering your prior manual setting of the vCore. I generally restored everything to default when I was changing how I overclocked.

I'm not saying what you've done is wrong in any way, it's just slightly different to how I did it. Remember that there may also be difference between how our (albeit similar) boards do things. My understanding is that the Z68 Pro's and P67 Pro's from ASUS as very very similar in regards to BIOS settings.

For me it was very much a learning process, trying things different ways & saving profiles when I hit a sucessful overclock. For example I have a couple of "4.5ghz" profiles on my machine, each achieving a similar result but done slightly differently. This is how I learnt.

My first 1155 mobo was a cheap one and had VERY limited options and was not a good overclocker...I got to 4.5 without too much issue but the board really couldn't control the vCore very well so it ran hotter than it should. Moving to the Z68 from ASUS I achieve 4.5 more readily, at a lower vCore and a correspondingly lower load temperature.

I think you're doing well, I'd recommend saving your successful overclock to a profile and trying things another way (after restoring defaults) if you're inclined to. This way you can always easily record each step in your OC process as a step away from Default.

To be honest, with Sandy B. and this particular board my FIRST OC was easy. Set multi of 45, changed nothing else & reboot! Ok, system wasn't stable under load but I got to windows and could do undemanding tasks right away. Benching and stressing allowed me to refine things, i.e. a bit more vCore via offset mode, bump up the LLC to get a stable 4.6 etc. etc.

Just keep an eye on your temps, I personally would be uncomfortable at anything over the mid-70's in IBT, but remember that IBT really does push your CPU harder than anything else, so day-to-day activities and gaming will be a good 10c below this level - again, in my experience.

I'm a little OCD when it comes to getting my OC perfect in my mind, whereas a friend of mine just adds an obscene amount of vCore and leaves it at that! Lol. Being on bespoke water, you can maybe be a bit more blaze regarding vCore as temperatures remain good...but that's not quite how I do it.

Part of the fun is finding out what works for you

Btw: can you confirm that your CPU is idling to 1.6ghz and that your vCore drops accordingly? In my experience (limited, I don't buy lots of new kit or review it!) setting vCore directly stops it reducing when the CPU idles, setting it via offset however allows it to drop at idle - which is great for your temps when you're just doing desktop stuff.

Sorry, I know sometimes this stuff can feel like information overload...I've tried to keep it simple in my own experimenting by changing just one thing at a time and restoring from defaults when trying a different approach.

Cheers,

Scoob.

 

[SmoulderingCircuits]

Don't hesitate to use the highest possible LLC your board offers, the settings are dramatised to sound severe when they aren't

 

[zan76]

Hi,

I think it might be remembering your prior manual setting of the vCore. I generally restored everything to default when I was changing how I overclocked.

I'm not saying what you've done is wrong in any way, it's just slightly different to how I did it. Remember that there may also be difference between how our (albeit similar) boards do things. My understanding is that the Z68 Pro's and P67 Pro's from ASUS as very very similar in regards to BIOS settings.

For me it was very much a learning process, trying things different ways & saving profiles when I hit a sucessful overclock. For example I have a couple of "4.5ghz" profiles on my machine, each achieving a similar result but done slightly differently. This is how I learnt.

My first 1155 mobo was a cheap one and had VERY limited options and was not a good overclocker...I got to 4.5 without too much issue but the board really couldn't control the vCore very well so it ran hotter than it should. Moving to the Z68 from ASUS I achieve 4.5 more readily, at a lower vCore and a correspondingly lower load temperature.

I think you're doing well, I'd recommend saving your successful overclock to a profile and trying things another way (after restoring defaults) if you're inclined to. This way you can always easily record each step in your OC process as a step away from Default.

To be honest, with Sandy B. and this particular board my FIRST OC was easy. Set multi of 45, changed nothing else & reboot! Ok, system wasn't stable under load but I got to windows and could do undemanding tasks right away. Benching and stressing allowed me to refine things, i.e. a bit more vCore via offset mode, bump up the LLC to get a stable 4.6 etc. etc.

Just keep an eye on your temps, I personally would be uncomfortable at anything over the mid-70's in IBT, but remember that IBT really does push your CPU harder than anything else, so day-to-day activities and gaming will be a good 10c below this level - again, in my experience.

I'm a little OCD when it comes to getting my OC perfect in my mind, whereas a friend of mine just adds an obscene amount of vCore and leaves it at that! Lol. Being on bespoke water, you can maybe be a bit more blaze regarding vCore as temperatures remain good...but that's not quite how I do it.

Part of the fun is finding out what works for you

Btw: can you confirm that your CPU is idling to 1.6ghz and that your vCore drops accordingly? In my experience (limited, I don't buy lots of new kit or review it!) setting vCore directly stops it reducing when the CPU idles, setting it via offset however allows it to drop at idle - which is great for your temps when you're just doing desktop stuff.

Sorry, I know sometimes this stuff can feel like information overload...I've tried to keep it simple in my own experimenting by changing just one thing at a time and restoring from defaults when trying a different approach.

Cheers,

Scoob.

hi scoob no it's on a consant 4.5

yea everthing feel's a bit like overload

so were should i go from to get it perfect

 

[Scoob]

Don't hesitate to use the highest possible LLC your board offers, the settings are dramatised to sound severe when they aren't

Yes, but only if it's needed. At 4.5 I didn't need to increase it, and at 4.6 just one notch up - though remember default is "high", one up is "Very High" which as The Protocol says is somewhat misleading! Given the same vCore bumping up the LLC will give higher vCore under load (that's sorta the point) but I feel it's important to ensure that the underlying vCore (via offset in my case) is correct else you might find that things work perfectly when stress testing, but you get odd instabilities at partial load - i.e. the more typical scenario when gaming.

If, for example, I were to set a higher LLC on my overclock now, my vCore under load would be far higher than it need be. I may then be able to reduce my underlying vCore (again via offset) slightly and still stress test perfectly. Though I know (from testing) that I'd then start getting stability issues doing other stuff when not at full load.

Setting a setting to max without going through the process of understanding what each increment does isn't the best approach in my view. Different motherboards do vary so much in this though. For example on my MSI I had to set the highest LLC to get anywhere, but there were only two settings on that board and it added sooo much vCore as to be hitting 1.5v just for 4.5 - it was stable yes, but that's way too much vCore under load for just 4.5. Additionally I could not lower the underlying vCore enough to keep load Vcore sensible while maintaining stability at partial load.

As mine and zan1976 have a similar board, I'd hope our results can be compared more directly - though of course there still will be a degree of variance between us most likely.

Cheers,

Scoob.

 

[SmoulderingCircuits]

Yes, but only if it's needed. At 4.5 I didn't need to increase it, and at 4.6 just one notch up - though remember default is "high", one up is "Very High" which as The Protocol says is somewhat misleading! Given the same vCore bumping up the LLC will give higher vCore under load (that's sorta the point) but I feel it's important to ensure that the underlying vCore (via offset in my case) is correct else you might find that things work perfectly when stress testing, but you get odd instabilities at partial load - i.e. the more typical scenario when gaming.

If, for example, I were to set a higher LLC on my overclock now, my vCore under load would be far higher than it need be. I may then be able to reduce my underlying vCore (again via offset) slightly and still stress test perfectly. Though I know (from testing) that I'd then start getting stability issues doing other stuff when not at full load.

Setting a setting to max without going through the process of understanding what each increment does isn't the best approach in my view. Different motherboards do vary so much in this though. For example on my MSI I had to set the highest LLC to get anywhere, but there were only two settings on that board and it added sooo much vCore as to be hitting 1.5v just for 4.5 - it was stable yes, but that's way too much vCore under load for just 4.5. Additionally I could not lower the underlying vCore enough to keep load Vcore sensible while maintaining stability at partial load.

As mine and zan1976 have a similar board, I'd hope our results can be compared more directly - though of course there still will be a degree of variance between us most likely.

Cheers,

Scoob.

Use LLC on highest to probe around for a stable voltage then lower the LLC and up the Vcore till you reel in a sustainable voltage config

 

[zan76]

right iam lost now complete overload ibt say stable with what have it at

llc high

offset mode 1.36v with +00.15

my mouse is acting funny keeps freezin for a few seconds now and then .

 

[Scoob]

hi scoob no it's on a consant 4.5

yea everthing feel's a bit like overload

so were should i go from to get it perfect

Hi, yeah sorry - a lot of information there

You're doing well, though I think the combination of a manually set vCore AND an offset will complicate things a little. Also, why have you chosen to be at a constant 4.5? On the older core2 (like my Q6600) this was the accepted way of remaining stable, due to the nature of FSB overclocking. With Sandy B. we're just adjusting the multi so it's simpler. With offset mode your board should (assuming it's like mine) adjust the vCore based on the multi being used, i.e. when at idle (16x multi for 1.6ghz) it should reduce vCore accordingly based on your CPU's VID at that multi) It's a really neat way of not running at full throttle all the time if you like. Also, there are concerns that giving a CPU it's full-load vCore of say 1.400v when it's idling and only wants say 1.000v is not a good thing.

It's true that instabilities can occur if your motherboards VRMs do not respond quickly enough when the clocks ramp up, but to be fair our levels of OC (4.5/4.6) ain't pushing things hard enough for that to be an issue just yet.

Personally if I were you I'd SAVE a profile for 4.5ghz, then I'd restore defaults and start from scratch just overclocking using offset mode, leaving all other setting bar the multi the same. This way you will see how the chip behaves at idle & clocks down to keep things cooler. Stressing will allow you to test your OC of course.

Really, my first stable OC on my ASUS board consisted of the following changes - done in steps of course at the time as I learned what worked:

i) Set multi to 45

ii) Set vCore offset to +0.040

Done.

That was it. Simple as that!

Ok, to get to 4.6 required a little more work for my chip, I needed to bump the LLC up one level.

So, from stock 3.3ghz to my overclock of 4.6 I just did the following (again, in steps initially)

i) Set multi to 46

ii) Set vCore offset to +0.050

iii) Set LLC to "Very High" from the default of "High"

Done.

Stress (Prime 95 and IBT) and game stable, while keeping within my desired temperature limits and vCore peak levels.

When you start changing too many things at once it can really get confusing, for example if you've overridden both vCore and offset vCore what do you change when things aren't stable? In theory it shouldn't matter surely as both just add up to equal your vCore right? Well no, in my experience it doesn't work that way...the offset vCore only seems to apply under load. For example IF I were to set my vCore to 1.4v and offset to 0 (i.e. not using it) then I might see 1.36v vCore at load AND idle. However leaving it all "auto" sees me idle at ~0.968 vCore and load at 1.32v and maybe NOT fully stable. Bump the offset value a little to +0.050v and we still see ~0.968v vCore at idle, but a healthier ~1.4v vCore under heavy load...with less at lighter loads - system stable!

Again, all I'm telling you is what I've experienced based on this board (and the old MSI but that was a bit crap) and my own experimenting and refining of my OC.

The more you can learn of how your board responds to the settings you change, the more reliable your OC will be and the easier it will be to push things further in the future if you so desire.

Another essay from me there, sorry dude, hope it's of use!

Cheers,

Scoob.

 

[SmoulderingCircuits]

right iam lost now complete overload ibt say stable with what have it at

llc high

offset mode 1.36v with +00.15

my mouse is acting funny keeps freezin for a few seconds now and then .

If your mouse is freezing the system is not stable, and from my experience too high a CPU/NB causes that...