Post Your Overclock!

[smarteyeball]

I haven't researched much into IB, but do they run warmer than Sandy chips do?

Yep

When I built this rig, I opted for the 2500K as a 'temporary' chip until IVY was released, but with the current heat issues and great performance as is - I might end up skipping IVY unless they do a revision.


Lately I've ignored my own advice and have been playing with the BCLK. 102.5 pushes me up to 4.726ghz and my 2133mhz modules seem quite happy @ 2190mhz.

They seem to have mhz headroom, but hate having the timings adjusted

 

[Britton30]

Hi Patrick, where ya been?
I suspect the higher RAM frequency will make up for a bit looser timing?

 

[Wishmaster]

From my limited experiences, most Intel chips seem to like more Bandwidth over tighter Timings anyway, so I wouldn't even worry about it.
2190Mhz is pretty fast

 

[smarteyeball]

Hi Patrick, where ya been?
I suspect the higher RAM frequency will make up for a bit looser timing?

I've been around

I just took a break from fiddling.

Raw mhz has definitely more impact than tighter timings.

From my limited experiences, most Intel chips seem to like more Bandwidth over tighter Timings anyway, so I wouldn't even worry about it.
2190Mhz is pretty fast

yeah, I've given up on timing tightening - There's no pay off. Raw mhz gives better bandwidth and lower latency.

I'm testing a bclk of 103.1 which is giving me 2200MHZ. I was running the DRAM at 1.62500 and VCCIO at 1.125 and got an idle BSOD at 2090Mz.

I've now upped the dram to 1.65 / 1.15 and adjusted the - offest a few notches. I might give it a proper stress later to see how it's affecting the PIC-E lanes (GPU) etc.

These are some 'realworld temps' atm (converting video) -.030 iirc






EDIT:

Nope - It did not like running that quick. BSOD. Time to dial and fiddle...

 

[smarteyeball]

*Sigh* back to the drawing board.

Edit:

Decided to start low, test properly and then increase as need be.

BCLK 102.5 / 4.695ghz / 2176mhz /-0.020 offset /1.65v DRAm and 1.5 VCCIO / CPU spectrum Enabled / LLC set to medium (%25)

10 pas of IBT maximum. Now priming.

If that is stable, then I'll do what I normally do -lower settings, don't bother 'stress testing' and cross fingers with real world usage

And if it fails Prime95 - I'll swear. Basically the tried and true method of overclocking :roflmao:

I'm getting the same bandwidth/latency @ 2175mhz that I was at 2200mz, give or take the odd Mb and .X ns

29K winsat and 43ns is quick in any mans language

 

[Dave76]

Nice OC, no swearing?

 

[smarteyeball]

No, there has been several bouts of swearing since then

P95 failure about an hour in. BSOD the next run. etc

Good old 'pass Ibt', yet fail Prime.

I'm dropping the BCLK each time. Now I'm trying using a + offset. I suspect there will be more swearing before the night is through.

I probably should be doing a memtest run but god damn that test bores me to tears...


One thing that's always had me curious about overclocking is:

"Rig has a decent and stable OC and all is fine". Then out of the blue it's suddenly not acceptable and things must be tinkered with to the point of frustration and the sometimes inevitable rollback to original stable overclock

 

[Britton30]

Computers are cranky. I've come back to a BSOD after a night's sleep...after a high pass with IBT. :huh:

 

[smarteyeball]

Computers are cranky. I've come back to a BSOD after a night's sleep...after a high pass with IBT. :huh:

Those are the absolute worst Dejection is the main emotion with frustration and annoyance mixed in.

I know the feeling all to well

I'm currently a couple of hours into a 4.75ghz / BCLK 101.1 / 2156Mhz / No LLC / + 0.010 offfset(?) + a few different tweaks and 'so far so good.

IBT max 10 didn't go above 70c and vcore topped at 1.308

Priming atm @ 1.288v with 59c across all cores.

I've been swear free for a few hours now

So far It's still quite an acceptable vcore for the clock speed with plenty of headroom.

I think this latest bout of tweaking is down to preparing for any potential SLI GTX 670 bottleneck eradication.

 

[Britton30]

Keep the world swear-free and enjoy. Why are we moved to fiddle-fart around with a stable system to "see if this works"?
1.288V is darned good for 4.75GHz, I can't seem to get mine to run more than 4.5.

 

[essenbe]

It's Gremlins, trust me!

 

[Wishmaster]

I played around with the BLCK setting myself a bit and didn't have much luck. At least not to the point it was worth doing. Ended up with better overall results just bumping the CPU multi.

Guessing IB is not really any better off in this regard than SB is.

Either way, impressive OCs Still, squeezing every mhz out of it to find the MAX is FUN!

 

[tw33k]

My chip needs an offset of +0.085 to run at 4.7GHz. LLC at level 1 (max) Ran 6 hours Prime with no errors so I'm happy. Funny thing is the temps were lower than they were when I was stressing with Aida64. Ambient temp was about the same. I've never seen that happen before. Usually temps under load with Prime are quite a bit higher than under Aida64

 

[smarteyeball]

Bah! Swears re-instated.



(The other two cores were still happily going)

Right there is proof for all those naysayers who reckon 1 or two hours of prime is enough and that it's "pointless to test for 8hrs+"

However I might bump the offset up a notch and 'real world test it. If no, or the very rare BSOD - it then gets the Pat F.I.S.E* seal approval.


(*F**k It, Stable Enough )

I played around with the BLCK setting myself a bit and didn't have much luck. At least not to the point it was worth doing. Ended up with better overall results just bumping the CPU multi.

Turns out, neither am I

 

[Wishmaster]

Yep. Couple hours just isnt enough. Ive had one fail at just over 9hrs before.

I think IBT is better though for a quick test. Just run 5 passes or so, and get a good idea in just a few minutes, rather than a few hours with Prime.
But, for the final stabilty test, I think Prime is better for a 12hr run.

 

[smarteyeball]

Yep. Couple hours just isnt enough. Ive had one fail at just over 9hrs before.

I think IBT is better though for a quick test. Just run 5 passes or so, and get a good idea in just a few minutes, rather than a few hours with Prime.
But, for the final stability test, I think Prime is better for a 12hr run.

Definitely - I usually run IBT for 5/10 runs high/max as a quick test and then switch to Prime95 for a 12hr+ run.

But then again, there will always be those who disagree and think 12hrs is ridiculous. Ultimately that's their choice - but it will never be the advice I give or condone when given by others.

When it comes to Overclocking and stability tests, personal experience and majority opinion will always trump 'the pointless crowd'.

 

[A Guy]

Would you say that while an OC might fail Prime95 after several hours, it would likely never be pushed that hard in day to day use? I'm not arguing your statement, but as one who has never OC (only a BCLK bump from 133MHz to 160MHz http://www.tomshardware.com/reviews/core-i5-750-efficiency,2500.html), I am curious if this isn't why others think such a long test seems unnecessary. I have certainly seen multiple instances of people reporting a failure after prolonged testing, even after 12 hours or more. A Guy

 

[smarteyeball]

Would you say that while an OC might fail Prime95 after several hours, it would likely never be pushed that hard in day to day use?

Definitely. Stress testing is completely artificial with the assumption that if it can handle 'worse case scenario', it can handle anything. That's also why it's a good idea to throw several types of realistic and unrealistic tests into the mix. Varying load, voltages involved, components etc.

I'm not arguing your statement, but as one who has never OC (only a BCLK bump from 133MHz to 160MHz Efficiency Explored: What's The Perfect Clock Rate For Your Core i5? : Hunting Down The Perfect Clock Speed For Core i5), I am curious if this isn't why others think such a long test seems unnecessary. I have certainly seen multiple instances of people reporting a failure after prolonged testing, even after 12 hours or more. A Guy

Everybody has their own approach and their own reasoning. I've also seen and experienced failures after prolonged testing. Also not everybody chases the same level of overclock.

Basically, no overclock can ever be considered fully stable no matter how it's stress tested. But prolonged testing does increase the chance of highlighting/uncovering an area or weakness where a likely failure is to occur.

Depending were and when it fails, can also help pinpoint what may the root cause (ie temps, particular voltage, Ram settings etc).

 

[A Guy]

Thanks Patrick

A Guy

 

[Britton30]

Would you say that while an OC might fail Prime95 after several hours, it would likely never be pushed that hard in day to day use? I'm not arguing your statement, but as one who has never OC (only a BCLK bump from 133MHz to 160MHz Efficiency Explored: What's The Perfect Clock Rate For Your Core i5? : Hunting Down The Perfect Clock Speed For Core i5), I am curious if this isn't why others think such a long test seems unnecessary. I have certainly seen multiple instances of people reporting a failure after prolonged testing, even after 12 hours or more. A Guy

Quite true as Patrick said. I think even the lightest stress test does more work than about 90% of day-to-day usage will do. Personally I've not seen any real world increase in performance with any overclock in CPU clocks or RAM fiddling. These things will get better scores in bench tests though.

So far I have seen quicker PC responses (apps opening, CCleaner faster, quicker boots, etc.) only with hardware upgrade to something which is "faster" than the previous. SSD takes the checkers on that with an upgrade to Intel over AMD a real close second. I even went from 8 to 16GB RAM and see no real difference except in program that use more RAM...graphics and Word come to mind. I understand CAD software or video transcoding (what's that?) also benefit with more RAM.