Overclocking (Main Thread)

Well, I finally threw in the towel—I’ve reverted everything back to stock settings, though I did leave the HPC (High Performance mode) toggled ON just to prevent those annoying frequency fluctuations under load...

It seems like hitting that overclock target isn't happening, and frankly, my desire for total silence has evaporated😁 My dual-fan CPU cooler is barely spinning at 600rpm, yet to push this processor to 4.5GHz, I'd need a staggering 4.45v. Even then, temperatures in OCCT climb to the 60-65°C range, and once you factor in that Intel IBT-AVX setting we discussed, it hits 70°C—which is starting to feel like the danger zone...

And I have to wonder—is there even any real-world benefit to successfully pushing this to 4.5GHz? Since it already hits 4.2GHz on stock using Turbo Boost, are we actually going to notice a tangible difference in daily use? Or am I just chasing ghosts? Besides, I've realized the downside of manual voltage tweaking: the power draw stays constant. With "Cool & Quiet" enabled, it just lowers the frequency, whereas on stock settings, the voltage drops down to a much more reasonable 0.9v during idle.

Once you land on a stable OC, you can switch back to Cool and Quiet.

The actual difference in clock speeds really depends on which software you're running. I’d suggest aiming for a 4.2-4.3 GHz overclock; you should be able to hit that at noticeably lower voltages. Otherwise, those processors become a bit of a nightmare to cool once you push past 1.40v.

I've got an 8350 running at 4.3GHz on 1.33V... should be solid enough.

Apparently, I have far too much time on my hands—again—so I decided to dive back into testing. It really just refuses to hit 4.5GHz. Seriously. Stock voltage? Nothing. Pushing it to 1.45v? Still nothing. I’ve played around with LLC settings and various other tweaks, but I have absolutely no desire to push past 1.45v... why bother risking it? Even at 4.3GHz with stock voltage, I'm hitting errors in Prime95 within ten minutes. However, at 4.2GHz using stock voltage—which, mind you, drops as low as 1.32v during the actual stress test while set to Auto in the BIOS—it stays rock stable.

I also wanted to ask—after some digging and searching online—I stumbled upon this link where someone suggests avoiding HPC and disabling APM entirely, yet the advice here seems to suggest the exact opposite.http://www.ronwoods.us/2014/02/what-...ion-power.htmlAny thoughts on this?

Also->


Is it normal for Windows to report 4.0GHz when it's actually running at 4.2? And is it standard for it to show 4 cores when there are actually 8 present?

If you don't enable HPM, you're going to run into those fluctuations. Your best bet is to dig through the various overclocking threads. The thing is, there isn't one official OC guide out there, so everyone ends up coming up with their own unique fixes. It varies wildly depending on the vendor and the specific motherboard you're using. These figures are what you'll see once you're inside Windows.
BTW, it looks like a certain feature is still toggled on, which is why your clock speed is dropping to 1400MHz. You might want to reach out to Overclock_freak; he’s a pro clocker and a massive Gigabyte enthusiast—he practically knows those boards inside and out.

Well, I went ahead and set HPC to off while keeping APM on—this guy in the link provided a pretty solid, well-reasoned argument for why we should stick to that configuration, and honestly, the results speak for themselves. I’m seeing a legitimate 5°C difference in my benchmarks: HPC off/APM on sits at 55°C, whereas HPC on/APM off jumps up to 60°C.

But now I find myself wondering... in any actual, real-world scenario, would the hardware's ability to throttle down its clock speed even be noticeable? This guy argues it won't be—that it happens within a millisecond and is essentially irrelevant.
With these specific settings:
During Prime95 Small FFT tests, my frequency is bouncing all over the place between 3600 and 4200 MHz—Hardware Monitor likely doesn't have a fast enough refresh rate to catch everything, so it probably just displays the lowest dip it sees.
Then, during the Prime95 Blend test, the frequency stays locked right at 4200 MHz. So, it's entirely possible that HPC isn't actually playing a role here; it might just come down to the application itself. It seems the computer simply identifies an opportunity to drop the clock speed for the user, and it happens so seamlessly that nobody even notices.

Frequency shouldn't be fluctuating if you're trying to overclock. It’s like testing stability under a full load, only to have the system downclock or undervolt because some specific feature kicks in. Besides, plenty of people have moved away from using Prime95 lately—especially those on Intel platforms (there's a rumor that Haswell can cause damage, and Skylake might just crash entirely). Are you getting through IBT okay? What kind of heavy lifting are you planning to do with this build?

I finished running the 3-4x IBTA standard—it passed. I’m about to kick off 20 passes at the "very high" setting, just like you suggested, since you mentioned those results are actually reliable. I'll be cross-referencing everything with Prime95, OCCT, and IBT to be sure.

Hmm, I don't know... GTA 5?

There is also this one thing that keeps bothering me during the overclocking process—specifically when I try manual voltage settings in the BIOS. When I set it manually, the voltage refuses to drop during idle. Even though Cool & Quiet drops the clock down to 1.4GHz, that manual voltage stays stuck right where I put it. Meanwhile, if I leave it on Auto, the system drops to 1.4GHz and the voltage dips down to 0.9V perfectly fine. Is this just the inevitable fate of anyone attempting an overclock, or am I doing something wrong here?...

Most tutorials suggest that C&C doesn't play nice with overclocking settings, though some people claim you can just toggle it on "once you've found a stable overclock" and run it without any issues 🤷. Personally (usually when working with different processors), I've never bothered turning C&C on after an overclock.

Back again

I’ve been wondering—is there actually any game or specific workload, or even just general PC use, that can heat up my CPU and motherboard as intensely as running Prime95 Small FFTs (that CPU-heavy stress test) or an Intel burn test?

My overclock itself is perfectly stable; the issue is strictly the motherboard—and I mean the board itself, not some minor setting, I've double-checked this a thousand times. When I run Prime95 Small FFTs, once things start getting hot, the VRMs throttle down to 1400MHz for a few seconds every single minute.

When I switch over to Large FFTs, everything stays fine; temperatures stay about 10°C lower, and there’s absolutely no throttling.

So, the question is: does this kind of thermal struggle only happen during these extreme, artificial stress tests, or is it possible that in "real life"—especially once summer hits—I might run into this throttling while I'm just playing games?

Hey, please reach out—I really need to hear from someone regarding this. 😁

Your VRM is losing speed, which means the CPU is throttling because of some specific feature—either the temperatures are too high or the voltages are just too much for your motherboard to handle. Because of that, your overclock isn't actually stable according to that stress test if the system can't even maintain its full clock speed. I saw that photo you posted over on the PC forums where you applied enough thermal paste to cover twenty processors; maybe that was part of the issue 😬
.
Regarding stress tests, some of them are honestly unrealistic for real-world usage, and some are supposedly even harmful (certain versions of Prime95 for Haswell if they run for too long), so you should define your own criteria for what counts as a "stress test." It’s easy to check if you're throttling during actual gaming by using something like HWMonitor or checking your in-game clock speeds via HWInfo

The board is throttling—or so says some guy I chatted with on an overseas forum. I can never quite recall which specific site it was, but apparently, the Giga UD3 (my current board) suffers from a hard-coded BIOS flaw that causes the VRMs to throttle prematurely due to temperature spikes, if I remember correctly.

I mean, look, the VRM doesn't have its own clock speed, but it heats up in direct correlation with the processor, right? When the CPU is under a heavy load, it draws more power, and since that current flows straight through the VRM, it naturally runs hotter.

So, I finally made the jump to the green side 🙂—solved my overheating issues and picked up an i7-2600
.
For the same amount of cash, I'm getting similar performance while pulling about 100-150W less under load
.
It’s a non-K processor, so I have to ask: am I actually left with any overclocking options here? All I've managed to do is bump the multiplier to 39, which leaves me sitting at 3.9GHz. There are options for up to 42, but nothing seems to stick; even though the BIOS claims 4200MHz, Windows keeps reporting 3900MHz
.




Since the frequency just stays locked at whatever value I set the multiplier to, perhaps a better route would be bumping the Turbo Boost multiplier? I have an unlimited range of numbers there too, yet the absolute ceiling I can hit is 3.9GHz across all four cores
.
A few questions: is 3.9GHz my hard limit if I don't mess with the BCLK?

Turbo seems to shut itself off once I hit around 55°C... is there any way to raise that thermal threshold? Perhaps some power-saving feature or a specific power limit is throttling things? Are these settings on this page looking alright to you
?


Also, if overclocking is out of the question, maybe I could at least try underclocking. But the BIOS won't let me drop the voltage any lower—or maybe I'm just missing it. This is everything I've got to work with
.

You can't overclock the non-K models. If I recall correctly, being able to bump up the multiplier was just a lucky break because they enabled it for those processors later on. You should honestly be glad you have that option at all, considering how Intel eventually locked those features down on newer chips via microcode updates through BIOS or Windows updates.

I’d steer clear of messing with the BCLK if I were you; you're more likely to cause headaches than see any gains. They aren't really designed for that kind of tinkering—it's not like working with Intel Skylake CPUs or AMD systems.

There's no way to configure the Turbo Boost.

I'm not entirely sure why you're hitting a wall with undervolting; it might be an issue with your chipset.

Is there any way to actually OC an i5-6400

Yes

Can WE still Overclock Non-K Intel Skylake CPUs?

Timothy Kim60 said:Yeah,

Can WE still Overclock Non-K Intel Skylake CPUs?

Woohoo. Thanks, I actually didn't know about that situation with Windows 10, which might actually be a bit of a headache for me then.

Basically, I need to find someone who actually knows their stuff...
I've got an i5 8400 overclocked, running TurboBoost up to 4.0... I'm pretty sure I'm going to need some serious cooling for this. If anyone can recommend something high-quality (but quiet!) and budget-friendly, let me know. Ideally, I'd love to find it at Best Buy since that's where I'm building the rest of this rig right now.

You aren't going to push past a 40x multiplier or a 102MHz BCLK, so I guess you won't really need anything fancy for cooling; maybe just an LC-CC-120.