BCLK overclocking

[GirlFromYonder]

I am gonna over clock a locked cpu because overclocking. I have the gigabyte z170-HD3P on the F2 version and a stock cooler. I am thinking to use Intel XTU. How much should i oc, how should and oc and most importantly, am i good to start ocing with my bios version and cooler????

 

 

THANKS

[Benji_w]

Use IBT AVX, it's easier to find instability also when you use BLCK it will overclock your other components like RAM etc so you'll have to increase in increments, BLCK overclocking isn't easy if you've never done it before because of this.

[reptileexperts]

Never a good idea to overclock on a stock cooler especially if you live in a warm environment, will just add to the heat.

 

BCKL overclocking can be hard to pin down, because as you raise the multiplier from 100, it will effect most of the other components in your system which may or may not handle overclocking well. This includes your PCI-E lanes and Memory. I would always recommend increasing in increments of .25 at most each time, stress it for a little while, then increase again. Using XTU makes it quick, but it also uses resources so in benching it is a better idea to find your clock stability point in XTU, then hard wire it into your bios. You should definitely do more research into the topic before attempting though.

[GirlFromYonder]

5 minutes ago, reptileexperts said:

Never a good idea to overclock on a stock cooler especially if you live in a warm environment, will just add to the heat.

 

BCKL overclocking can be hard to pin down, because as you raise the multiplier from 100, it will effect most of the other components in your system which may or may not handle overclocking well. This includes your PCI-E lanes and Memory. I would always recommend increasing in increments of .25 at most each time, stress it for a little while, then increase again. Using XTU makes it quick, but it also uses resources so in benching it is a better idea to find your clock stability point in XTU, then hard wire it into your bios. You should definitely do more research into the topic before attempting though.

but i read on overclockers that on skylake cpu clock speeds are independent of PCI-E clock speeds thus i can easily overclock. 

[reptileexperts]

hmm unless format has changed for LG1151, BCKL has always effected the PCI-E lanes. I haven't heard of that but I haven't done BCKL overclocking since my i7 4710HQ which DID effect everything. You can check your GPU specifications in XTU and see if it changes when you change the BCKL, either way, you might be required to downclock your RAM in order to maintain stability.

[MageTank]

10 minutes ago, DeezNoNos said:

I am gonna over clock a locked cpu because overclocking. I have the gigabyte z170-HD3P on the F2 version and a stock cooler. I am thinking to use Intel XTU. How much should i oc, how should and oc and most importantly, am i good to start ocing with my bios version and cooler????

 

 

THANKS

These people did not do their research. Skylake non-K overclocking disables AVX, so you can run even Prime95 on your stock cooler without heat being much of an issue. I own the Core i5 6600T, and took mine to 4.5ghz on stock cooler and hit 90C under 48k FFT load. This was at 1.39V (trying to get the CPU as hot as physically possible in my ITX Azza Z case). If you use a moderate overclock like 4.4ghz and only 1.3ish volts, you will not get anywhere near as hot.

 

Let's address the multiple elephants in the room. Non-K overclocking is not for everyone, and here's why. 

 

#1. Disables AVX

#2. Disables iGPU

#3. Cache speed is reduced by 75%

#4. Disables C-States

#5. Loss of thermal sensors

 

If all of this is okay with you, then feel free to proceed. With Skylake, BCLK is no longer tied to PCIE. Not only that you are given much more fine grain control (0.1 increment changes using intel's internal clock gen, or 0.06 increments using vendors external clock gen) meaning you can dial in more finely tuned overclocks. It is true that overclocking with BCLK will adjust your memory speed, but you can simply change that strap to something else to compensate. As long as you set your memory speed near its rated XMP speed or below, it will be just fine. If you want to set it higher, apply more vDIMM and pay attention to VCCIO/SA voltages. Try to keep VCCIO/SA under 1.15v each. You should not need more than that to keep memory stable.

 

As for the stock cooler, here are my results using one: 

 

 

Here are the exact same settings, using my ID Cooling VC45 ITX cooler:

 

 

16C difference in max temps. Worth every penny. Please note, this test was performed using absurdly high Vcore for such an OC, and was done in an Azza Z ITX case. In a better case, with less Vcore, your temps will not be as high as mine. 

[GirlFromYonder]

9 minutes ago, MageTank said:

These people did not do their research. Skylake non-K overclocking disables AVX, so you can run even Prime95 on your stock cooler without heat being much of an issue. I own the Core i5 6600T, and took mine to 4.5ghz on stock cooler and hit 90C under 48k FFT load. This was at 1.39V (trying to get the CPU as hot as physically possible in my ITX Azza Z case). If you use a moderate overclock like 4.4ghz and only 1.3ish volts, you will not get anywhere near as hot.

 

Let's address the multiple elephants in the room. Non-K overclocking is not for everyone, and here's why. 

 

#1. Disables AVX

#2. Disables iGPU

#3. Cache speed is reduced by 75%

#4. Disables C-States

#5. Loss of thermal sensors

 

If all of this is okay with you, then feel free to proceed. With Skylake, BCLK is no longer tied to PCIE. Not only that you are given much more fine grain control (0.1 increment changes using intel's internal clock gen, or 0.06 increments using vendors external clock gen) meaning you can dial in more finely tuned overclocks. It is true that overclocking with BCLK will adjust your memory speed, but you can simply change that strap to something else to compensate. As long as you set your memory speed near its rated XMP speed or below, it will be just fine. If you want to set it higher, apply more vDIMM and pay attention to VCCIO/SA voltages. Try to keep VCCIO/SA under 1.15v each. You should not need more than that to keep memory stable.

 

As for the stock cooler, here are my results using one: 

 

 

Here are the exact same settings, using my ID Cooling VC45 ITX cooler:

 

 

16C difference in max temps. Worth every penny. Please note, this test was performed using absurdly high Vcore for such an OC, and was done in an Azza Z ITX case. In a better case, with less Vcore, your temps will not be as high as mine. 

If i lose thermal sensors then how will i know the temps. How did u know the temps

[chiwas]

1 minute ago, DeezNoNos said:

If i lose thermal sensors then how will i know the temps. How did u know the temps

You can still monitor CPU package temp, but not each core temp.

[GirlFromYonder]

Just now, chiwas said:

You can still monitor CPU package temp, but not each core temp.

ooh thats cool then

[MageTank]

Just now, DeezNoNos said:

If i lose thermal sensors then how will i know the temps. How did u know the temps

Use HwInfo64, and look for PECI, or the Platform Environment Control Interface sensor. It is extremely accurate for telling you your CPU package temps (which you can also view, listed as CPU). While you will not be able to know the individual core temps, you will be able to see how hot the total package is, which is enough to determine if your cooling is sufficient. In my screenshots above, look for CPU and PECI temps, to see what I am referring to. PECI is actually highlighted in the first Screenshot.

 

I hope this helps you out. You can get pretty far on that stock cooler, but I would suggest upgrading coolers later on if you plan on going even higher. As long as your temps remain under 90C during 48k FFT Prime95, you are perfectly fine with anything you can throw at it. 

[GirlFromYonder]

1 minute ago, MageTank said:

Use HwInfo64, and look for PECI, or the Platform Environment Control Interface sensor. It is extremely accurate for telling you your CPU package temps (which you can also view, listed as CPU). While you will not be able to know the individual core temps, you will be able to see how hot the total package is, which is enough to determine if your cooling is sufficient. In my screenshots above, look for CPU and PECI temps, to see what I am referring to. PECI is actually highlighted in the first Screenshot.

 

I hope this helps you out. You can get pretty far on that stock cooler, but I would suggest upgrading coolers later on if you plan on going even higher. As long as your temps remain under 90C during 48k FFT Prime95, you are perfectly fine with anything you can throw at it. 

Understood so all is good. I can get more GHzzzzzzzzzzz

[chiwas]

This guide is a good starting point.

[MageTank]

20 minutes ago, chiwas said:

This guide is a good starting point.

No idea why that guide says "avoid high memory clocks". My ram is running at 3500mhz CL14-14-14-28-CR1 just fine. Also, 1 hour of 1344k won't tell you vcore is stable. That will take several hours of both 1344k and 2688k to be 100% certain. Aside from that, it's a decent guide. 

[Benji_w]

Just now, MageTank said:

No idea why that guide says "avoid high memory clocks". My ram is running at 3500mhz CL14-14-14-28-CR1 just fine. Also, 1 hour of 1344k won't tell you vcore is stable. That will take several hours of both 1344k and 2688k to be 100% certain. Aside from that, it's a decent guide. 

Shame you disable AVX when overclocking locked chips, IBT AVX will tell you VCORE is unstable in mins sometimes seconds.

[chiwas]

3 minutes ago, MageTank said:

No idea why that guide says "avoid high memory clocks". My ram is running at 3500mhz CL14-14-14-28-CR1 just fine. Also, 1 hour of 1344k won't tell you vcore is stable. That will take several hours of both 1344k and 2688k to be 100% certain. Aside from that, it's a decent guide. 

Same, Got my 2666MHz sticks running at 3k without any issues

[MageTank]

2 minutes ago, Benji_w said:

Shame you disable AVX when overclocking locked chips, IBT AVX will tell you VCORE is unstable in mins sometimes seconds.

Again, no it won't, lol. Intel's AVX is no different than Prime95 or Linpack (They actually use a custom Linpack version in their burn test). It's less stressful than the aforementioned ways of finding unstable vcores. It takes hours of stress to determine if something is 100% stable. You cannot know this in minutes or seconds, because the time it takes to test all of the variables cannot be done in such a limited window of time. 

 

I've seen plenty of people pass IBT, but fail the moment they start folding GROMACS or running Prime95. 

[Benji_w]

1 minute ago, MageTank said:

Again, no it won't, lol. Intel's AVX is no different than Prime95 or Linpack (They actually use a custom Linpack version in their burn test). It's less stressful than the aforementioned ways of finding unstable vcores. It takes hours of stress to determine if something is 100% stable. You cannot know this in minutes or seconds, because the time it takes to test all of the variables cannot be done in such a limited window of time. 

 

I've seen plenty of people pass IBT, but fail the moment they start folding GROMACS or running Prime95. 

Then that's usually something else unstable like memory, IBT AVX isn't Intel made it was made by someone else for the overclocking community and passing it on Xtreme stress mode on maximum will give you 90-100% stability, it's not easy to pass at all. 

[MageTank]

1 hour ago, Benji_w said:

Then that's usually something else unstable like memory, IBT AVX isn't Intel made it was made by someone else for the overclocking community and passing it on Xtreme stress mode on maximum will give you 90-100% stability, it's not easy to pass at all. 

No, it's not memory. I already know IBT isn't intel made. I just said it used a custom version of Linpack. 

 

Again, I can show you an OC that will pass IBT just fine, yet fail Prime95/GROMACS. Do you know why? IBT's version of Linpack, even the "Xtreme Stress Mode" is very tame in comparison. There are several factors that come into play to determine whether or not vcore is stable. Including its application at different load levels of CPU stress, and even memory usage. Contrary to popular belief, overclocking your ram adds heat to your CPU package. Going from 2133mhz to 3200mhz added roughly 5C of heat to my CPU package under full stress. Stressing both ram and CPU at the same time is enough to shut down what is previously seen as a "stable" OC in IBT.

 

I use a version of Linpack far more brutal than IBT, if you want to test it for yourself. However, even that is not enough for me to test stability in an hour or minutes. I use it to test pseudostable OC increments. Once it fails Linpack, i dial back some, then do a deep Prime95 run (8 hours+). Don't know if I am allowed to post the file here randomly, so PM me if you want to test it for yourself. 

 

EDIT: Proof to my claims.

 

 

 

Real overclockers appreciate the burning smell. IBT just isn't stressful enough.

[Benji_w]

2 hours ago, MageTank said:

No, it's not memory. I already know IBT isn't intel made. I just said it used a custom version of Linpack. 

 

Again, I can show you an OC that will pass IBT just fine, yet fail Prime95/GROMACS. Do you know why? IBT's version of Linpack, even the "Xtreme Stress Mode" is very tame in comparison. There are several factors that come into play to determine whether or not vcore is stable. Including its application at different load levels of CPU stress, and even memory usage. Contrary to popular belief, overclocking your ram adds heat to your CPU package. Going from 2133mhz to 3200mhz added roughly 5C of heat to my CPU package under full stress. Stressing both ram and CPU at the same time is enough to shut down what is previously seen as a "stable" OC in IBT.

 

I use a version of Linpack far more brutal than IBT, if you want to test it for yourself. However, even that is not enough for me to test stability in an hour or minutes. I use it to test pseudostable OC increments. Once it fails Linpack, i dial back some, then do a deep Prime95 run (8 hours+). Don't know if I am allowed to post the file here randomly, so PM me if you want to test it for yourself. 

 

EDIT: Proof to my claims.

 

 

 

 

Real overclockers appreciate the burning smell. IBT just isn't stressful enough.

Could you give me your linpack then so I can test it, because I'm not going to ever use prime for 24hours because I'm not a moron. Also speed doesn't add heat, voltage does, And I'm yet to see my RAM add any form of heat to any of my chips because this intel and my 5ghz 8350 didn't suffer from me overclocking my RAM at all.

[MageTank]

6 hours ago, Benji_w said:

Could you give me your linpack then so I can test it, because I'm not going to ever use prime for 24hours because I'm not a moron. Also speed doesn't add heat, voltage does, And I'm yet to see my RAM add any form of heat to any of my chips because this intel and my 5ghz 8350 didn't suffer from me overclocking my RAM at all.

What? Speed does not add heat? What world do you live in? While voltage causes far more heat, clock speed in and of itself also adds heat. This is basic level overclocking knowledge. Also, since when did using Prime95 for 24 hours make someone a moron? I run Prime95 for 27 hours to finish every single FFT length. It does not make me a moron. The only morons are the people that actually believe Prime95 kills CPU's.

 

As for the part about memory speed and heat added to the CPU, go test it again. I assure you, the additional VCCIO/SA voltages supplied to the IMC from overclocking memory will add more heat to the total CPU package. I am willing to bet on it. As for the version of Linpack i use, here you go:

 

LinX_IntelMlib.zip