Need Advice, having issue w/ overclock I5-3570k on gigabyte board

[stang725]

Can't seem to OC my memory above stock XMP profile of 933Mhz (1866).

Also want some feedback on if I should even attempt lower vcore. I don't even see an option to adjust BCLK in the BIOS, can adjust in WINOS..

I'm kind of a newb to Intel chips, by day I make ARM processors

Build:
Gigabyte z77x UDH5
I5-3570k
212 EVO
Corsair Dom Platinum 1866 2x 4GB (8GB)
Corsair 600T w, stock fans

Running at 4.5Ghz w/ vcore set manually in bios at 1.2 (showing 1.17 at full prime 95 load)
Max average temp at 24hr full load 66C, ambient temp ~20C

Some non-default BIOS settings:
PWM mode to high-perf
Voltage response fast
Turbo is on for 4.5
all C settings on
manual vcore to 1.2

So here is a capture of where I am at:

 

[Adroid]

For one thing, your memory is running at 1866. CPU shows 933, multiply that x 2 to get the speed you are actually running.

For another thing, I am a bit jealous of your chip if it's really running 4.5ghz at that low voltage, low temperature, and stable. I would check your windows event logger under custom => administrative => and see if you are getting lots of WHEA errors.

 

[stang725]

Yeah, I understand the 2x for the memory.... I meant each channel is running at the 933. I have seen people talking about going to a BCLK of ~104 on these to get a decent OC with only relaxing the timing to 10-10-10-27 using stock 1.5V. I couldn't boot at 1.6V @ 104 w/ 44x multiplier using vcore of even 2.7.

After reading a lot of forums, I am pretty happy with the voltage. No WHEA errors over the last week. Have done well over 24 hour of Prime Blend.

Being an ARM process engineer, I have some ideas about where in the process you might get lucky and get a wafer(chip) that gets a perfect FEOL run (where the gates sit that perform the crunching)... It may sound kind of bad, but the industry often sends out wafers that have numerous issues with their structures (just because they still meet the spec limits and will function as designed. May not overclock well or even if they do it may require giving them more juice... but will run as marketed. I have seen some pretty flawed wafers leave the fab, but at the same time, there are the 1 in 200 wafers that hit our exact targets at every step of the process and you land a couple of particles on the wafer here and there during the critical steps that may block a photo, etch, or a film step and you will be down to well under 80% yield of the wafer in terms of usable chips. The real reason as to why some are better than others out in the hands of consumers is we can't afford to throw away the marginal ones that still meet spec limits... especially true for Intel, where they prob have roughly ~500 chips per wafer. Not to say we send out bad chips by any means, just when working w/ anything below 45nm.... especially true at 22nm, its really hard to hit target consistently at those dimensions and keep defects under control.

 

[Adroid]

Yeah, I understand the 2x for the memory.... I meant each channel is running at the 933. I have seen people talking about going to a BCLK of ~104 on these to get a decent OC with only relaxing the timing to 10-10-10-27 using stock 1.5V. I couldn't boot at 1.6V @ 104 w/ 44x multiplier using vcore of even 2.7.

After reading a lot of forums, I am pretty happy with the voltage. No WHEA errors over the last week. Have done well over 24 hour of Prime Blend.

Being an ARM process engineer, I have some ideas about where in the process you might get lucky and get a wafer(chip) that gets a perfect FEOL run (where the gates sit that perform the crunching)... It may sound kind of bad, but the industry often sends out wafers that have numerous issues with their structures (just because they still meet the spec limits and will function as designed. May not overclock well or even if they do it may require giving them more juice... but will run as marketed. I have seen some pretty flawed wafers leave the fab, but at the same time, there are the 1 in 200 wafers that hit our exact targets at every step of the process and you land a couple of particles on the wafer here and there during the critical steps that may block a photo, etch, or a film step and you will be down to well under 80% yield of the wafer in terms of usable chips. The real reason as to why some are better than others out in the hands of consumers is we can't afford to throw away the marginal ones that still meet spec limits... especially true for Intel, where they prob have roughly ~500 chips per wafer. Not to say we send out bad chips by any means, just when working w/ anything below 45nm.... especially true at 22nm, its really hard to hit target consistently at those dimensions and keep defects under control.

Welp, your chip is a beauty then. Mine gets Whea errors @ 4.4 ghz and 1.265 voltage. Not to mention, it got to 86C during prime95 largeFFTs

I have the 212 evo, which isn't the greatest, but for my chip I had to settle for 4.3ghz and still a very fast CPU. Overall im happy with it.

Also just FYI, Intel does actually bin their CPUs, anything with a K next to the name is "above average" binning - although I agree you have a very small chance to get a chip as good as yours.

Seems like I never land a good chip... maybe next build