Help overclock my 6600k
Help overclock my 6600k
I just configured it for auto voltage at 4.6ghz, and it’s functioning perfectly. There are no high voltage spikes, and the voltage adjusts when not needed.
The voltage variation is displayed in real time by CPUID HWMonitor, as shown in the example. The "Auto" mode usually works too broadly and isn’t very helpful.
In truth, the voltage doesn’t drop completely when not in use. It reduces under load to cut power consumption because current also increases. Therefore, the voltage peaks when idle and drops when under load when it’s most required.
It’s better to manually adjust the LLC (CPU Load Line Calibration) setting in BIOS to "High" for a Gigabyte motherboard or set it to 4 or 5 for an Asus board (the scale ranges from 1 to 7). This helps control voltage fluctuations and can further lower the CPU VCore voltage.
At 4.6 GHz, CPUID HWMonitor (free to download) indicates a voltage range of 1.344 (under load) to 1.356 (idling)—two consecutive values at 12 mV. I manually set the CPU VCore to 1.355 V in BIOS. This results in a more efficient voltage use and may allow even lower CPU VCore requirements.
Yes, I might have had several programs checking the voltage and it was showing incorrect readings. After only running cpuid and intelburn test at high voltage, the actual voltage was 1.354v under load and 1.2v at idle. I prefer auto voltage because it saves power when the voltage is stable; however, this motherboard seems to handle it well.
grimsin :
yes I am, I may have had more then one program monitoring the voltage and it was giving me false voltage reading, just checked now with only cpuid running and intelburn test on very high, voltage under load was 1.354v, 1.2v at idle.
I like auto voltage as i down the voltage when not it isnt needed. I have overclock many computers and always used manual voltage, but this motherboard seems to monitor it well.
Hi grimsin,
It should be the other way around: 1.356 V at idle and 1.2xx (like 1.284) under load. The motherboard reduces the voltage under load because the current is also much higher under load. And as the power (W) equals (=) the voltage (V) multiplied (x) by the current (A), it increases too, and so does the heat and therefore the temp. of course. When the motherboard reduces the voltage (by an amount called vDroop), the power goes down as well as the heat and the temp. to prevent damage to the CPU. It is a self-protection mechanism.
What goes down when not needed is the frequency (like from 4.6 GHz dropping down to 800 MHz) thanks to the speedstep technology (provided you haven't disabled it).
The problem is that with OC, this automatic voltage reduction under load can starve the CPU of the required voltage when it needs it the most: precisely under load. It is a source of instability (crashes, blue screens, errors in tests, etc.) and requires to raise the CPU VCore to avoid the starving (which also increases the temp.). Tuning LLC (CPU VCore load Line Calibration) is a way to counteract this and increase the stability of the OC. I explained LLC in at least one other post on a couple of threads of this forum.
it isn't a reverse situation
with auto voltage it reduces the voltage (I have speed step turned on so it also lowers the frequency when it's not needed, and raises it when necessary).
And I'm completely crash-free.
Hi Grimsin,
I understand your point, there’s definitely something unusual with your setup. However, a false reading like the one you mentioned earlier is still possible. As long as you don’t physically measure the real VCore with a multimeter at the output capacitors, you could try using "only" CPUID HWMonitor (free download) to check and look under "Voltages" at the very first line starting with "CPU VCORE" while running a serious test like Prime95 (latest version) on its default blend test (4 cores).
Here’s what you might want to note:
1. What value do you get by checking CPU VCore directly in BIOS (not through any app)?
2. What is the CPUID HWMonitor reading for "Package" power (W) in the "Powers" section of the "Intel Core i5 6600K" category?
3. After running Prime95, what are the three numbers in the columns of CPUID HWMonitor at the right of CPU VCORE (Value, Min, Max)?
4. Lastly, can Prime95’s 4 workers (cores) remain green after two hours of testing? And what is the CPUID HWMonitor value for "Package" power (W) in the same section?
My CPU is running smoothly at 4.6 GHz with VCore manually set to 1.355 V in BIOS (stable since 1.350 V) and my current CPUID HWMonitor reading is: Value=1.356 V, Min=1.344 V, Max=1.356 V (which is essentially idle). When I start Cinebench, it instantly drops to 1.344 V for protection reasons. This is the normal and expected behavior, which makes sense given the motherboard’s protection system for the CPU.
You can turn off this protection in BIOS (I did it by disabling all options except "package C state limit," which can’t be fully disabled) and it’s sometimes done for overclocking to boost stability by gaining more control (see: http://forums.bit-tech.net/showthread.php?t=291605). But it shouldn’t work in the opposite direction (as far as I know). I didn’t disable it because I prioritize CPU protection over easier overclocking. It’s up to you, of course ;-)
I understand all this information, as I've mentioned before I've overclocked several rigs back to my Pentium 4 single core 3.0ghz CPU.
I monitor voltages and temperatures using cpuid.
In the BIOS it reads 1.2'sh volts, which matches what I mentioned in my previous post—just checked now with only cpuid running and the Intel burn test.
I don't use Prime95; I stick with Intel burn test. It passes 20 runs at the high setting, which works well for me.
I understand all this information, right? As I mentioned before, I've overclocked several rigs back to my Pentium 4 single-core 3.0GHz CPU. I monitor voltages and temperatures using cpuid. In the BIOS, it shows 1.2 volts, which matches what I reported in my previous post—just checked with cpuid running and Intel Burn Test on high load, where voltage under load was 1.354 volts at idle. I don’t use Prime95; I rely on Intel Burn Test. It passed 20 runs at high settings, which is acceptable for me.
Hi Grimsin,
Yes, you might be able to <<know all this>>, but other forum members (readers) might not. They could miss the fact that your setup is unique and unusual, which might confuse them since their motherboards wouldn’t behave the same way. It’s worth noting how unusual this is (you started by explaining it).
Considering this, a part of the motherboard could be faulty. Fortunately, not every board increases BIOS VCore from 1.2V to 1.354V under load—that’s a 12.8% rise. If you set VCore to 1.355V in BIOS, your board would go up to 1.52V, which is near the upper limit Intel suggests. Running at 1.52V would only be stable briefly due to heat and power limits. I reached a CPU package temperature of 85°C at 4.7GHz after two hours with VCore manually set to 1.40V. This seems typical for most users.
Assuming we’re clear on this, I suspect a motherboard issue is likely because it doesn’t make sense otherwise. In the Tweak-Town 6600K test (http://www.tweaktown.com/guides/7481/twe...ndex6.html), where VCore is set to 1.50V, your board would push it up by 12.8%, reaching 1.692V and potentially damaging the CPU. Their graph shows a big drop from idle to load voltages, especially at 4.8GHz (from 1.325V to 1.284V). They measured these directly with a multimeter, not via CPUID.
Your situation isn’t a major problem if you understand the behavior. Most people would need to set VCore around 1.40V for ~1.35V under load, but you’d just need to set it to 1.2V for ~1.35V. It’s unusual, but if it works, it’s fine.
It’s similar to seeing a car’s speedometer show 100 km/h when it’s actually around 95 km/h. In your case, the odometer would show 85 km/h when it’s likely closer to 95 km/h. It works differently for each person. If you’re aware of this “other way around,” it’s okay. But if you ever sell it, let the buyer know—otherwise, an overclock could push VCore to 1.692V and destroy the CPU. Personally, I’d prefer replacing a bad board, whether I sell or not. For the car, fixing the odometer before driving would be wise, especially if you’re being “the other way around.” That could be risky!
Just returned from work.
Checked the bios to verify the auto voltage, it was 1.36v... Not sure what was happening before...
Adjusted to 1.35v for 4.5ghz and set LLC to level 5, but it wouldn't boot. Then changed it to 1.355v and it worked, returned to the bios, set voltage back to 1.35v, and added "cpu current capability" at 130% – it booted and passed Intel burn tests. It also runs cooler than with auto enabled. Now I'm trying for 4.7ghz, lol.