Dead Skylake 6700K - Was Motherboard Accomplice?

Hurin

2[H]4U
Joined
Oct 8, 2003
Messages
2,410
Hi All,

I had my brand new 6700K on an Asus Z170-WS running at 4.6GHz using Offset vcore mode. I was actually lowering voltage via offset mode: At the time of the CPU's demise, it was running offset -0.070 which was resulting in a vcore display of 1.328 in BIOS, an HWMonitor reading of 1.248v at idle, and a load value of 1.360v. Nothing scary at all. Temps while running RealBench were chilly at 55/59/53/55 (Corsair H115i).

Nothing else fancy in the BIOS. Just XMP enabled for my 3200MHz Corsairs, and pretty much everything else (the exception being offset vcore of -0.070) left at auto as recommended most places for Skylake.

I walked away from this system to run an 8 hour RealBench validation run (to confirm it was stable), and when I came back for a check-in around 90 minutes in. . . the screen was frozen, and yet I could still see HWMonitor reporting a vcore of. . .. wait for it. . . 1.984v.

one, point, nine, eight, four. w. t. f.

The CPU is toast. Won't boot into Windows. Even moved it into another computer and it misbehaves there too (while a swapped core i5 6500 works okay in mine).

So, my question: Would you chalk that 1.984v showing as vcore on the frozen screen as an anomaly due to the CPU dying at that moment. . . or rather would you look at it as the cause and perhaps suspect the motherboard? I have no reason to suspect the PSU. It has served me well in my Haswell rig and it's a Seasonic X-series 1250.

Just a dud CPU. . . or would you swap out the motherboard too? I love most things about this motherboard but I'm not sure I can trust it.

Also, any thoughts on how likely this was to have taken other components with it? I'm thinking vcore is pretty isolated, but would hate to have zapped my Titan X (old) or other components!

--H
 
I wouldn't try adjusting the CPU voltage offset again using that board.

Check for an updated BIOS before the next try.
 
Was using newest BIOS (0702). Since I put in this temporary i5-6500, I have put BIOS 0602 back on it.

I've just gotta return this motherboard. I don't trust it. And $300+ for a motherboard I can't trust is not acceptable. Which is a shame because I otherwise really liked it.

Meanwhile, I'll be testing my RAM and video card to make sure they weren't collateral damage.

--H
 
Man that is sketchy. I concur, replace that board if you can't verify it in confidence.
 
Offset might as well be called "Auto - Offset", no doubt there was some funky stuff going on... but I always go for Manual/Override mode.

It's unlikely that a high vcore would take other components with it
 
The motherboard and RAM appear to be fine (even at XMP 3200). Ran 16 hours of RealBench stress test successfully while using a loaned i5-6500. However, I'm still going to be returning this board. Might as well get something cheaper *and* trust it more. I hate to give the board a potentially bum rap. But I wouldn't feel safe putting any more CPUs in it. Nor am I willing to wait out another CPU RMA process in the future. I'm not the most patient person when it comes to that sorta stuff.
 
Its possible you did this accidentally.
High LLC settings can put ungodly voltages to the CPU, I harmed a 2500K accidentally like this.
Couple that with some of the more hardcore CPU tests, when you arent looking the toughest test starts and boosts voltage dramatically.
The amount of negative offset you used is nothing compared to how much LLC can add.

I use offset voltage as well with my 6700K but it can be dangerous during stability testing of high overclocks.
The system is still using automatic voltage selection (VID) which can use much higher voltages at high clocks.
This with high LLC and hardcore CPU tests might be what happened.
 
When I first got my Z170/6700K combo I tried overclocking using offset and found it to be very flaky. Even on stock settings I was getting up to a 1.4v in edge cases. I'd recommend everyone who's overclocking a Sky Lake processor use the manual vcore. The auto vcore function isn't reliable or safe when overclocking. I'm using a manual 1.3V and I tested it with a variety of loads to make sure the LLC setting was correct. As for anyone who thinks it will use a lot more power, the chip is still downclocking and using idle states, unless you're on a battery it will never make a difference.
 
Regarding LLC. . . at the time of its death, LLC was set to auto. But, two days prior I had experimented with it and while setting a manual vcore it behaved normally (slight vdroop up to level 5, slight boost under load at level 6). When using offset, LLC was causing vcore to go much higher than expected (compared to when vcore is manual) even at lower settings. Which is why I set LLC to auto which seemed to bring it back down to just doing a minor boost under load (almost identical to manual vcore). If I could have turned it off completely, I probably would have. . . but "auto" was the closest thing to "off" because even level 1 manually set was boosting.

I think we can all agree though that there should be some logic in the BIOS that says that no combination of vcore and LLC should allow for a vcore of 1.984v. =)

But yes, I was using offset mode because I liked the idea of throwing fewer volts at the chip when they aren't needed. Had I kept it on manual, perhaps there wouldn't have been the catalyst that caused (?) that vcore jump. Again, assuming that reading was accurate and was a cause rather than a symptom of a chip in its death throes.
 
The same reason I use offset as well, idle temps are way low!
I manually set LLC to prevent strangeness like you experienced.

On my mobo (Maximus Hero), low LLC "numbers" act as high LLC setting.
Seems to be backwards.
Another thing to check for.
 
The same reason I use offset as well, idle temps are way low!
I manually set LLC to prevent strangeness like you experienced.

On my mobo (Maximus Hero), low LLC "numbers" act as high LLC setting.
Seems to be backwards.
Another thing to check for.
Yes, LLC behavior seems to change dramatically when you go from manual vcore to offset vcore. I may underclock and undervolt my next CPU and test the relationships more thoroughly (which will hopefully be safe while undervolted).

I have returned the motherboard though. I think I'm going to spend about half as much and yet still get dual Intel NICs by going with the Gigabyte Z170X-UD5. Will miss the Asus BIOS though. It was awesome.
 
But I wouldn't feel safe putting any more CPUs in it. Nor am I willing to wait out another CPU RMA process in the future. I'm not the most patient person when it comes to that sorta stuff.

Did you really RMA your 6700K?
 
Did you really RMA your 6700K?
I think I see where this is going. . .

Let me answer your question with some questions. . .

--Given the bizarre pull of almost 2v to the vcore despite there being no such request set in the BIOS, do you know with absolute certainty that this chip wasn't defective (rare as that is)?

--Given that this CPU was running at 4.6GHz (overclocking, yes, but nowhere near an extreme) and well below what anyone considers "dangerous" vcore when this happened (1.36v, 6700Ks don't even bat an eye until 1.4 or even 1.42, some even go as high as 1.45 or 1.5), do you know with absolute certainty that the chip wasn't defective?

I realize certain risks are taken with overclocking. And had I pushed this CPU above safe voltage, acted outside the norms of what is widely expected of these chips, or otherwise abused it, you might have a point. But I didn't. Keep in mind, I'm not seeking to return this CPU because it isn't performing adequately ("wah! I wanted 5GHz on air!"). . . I'm returning it because it "popped" and died well within the norms of how it is commonly and routinely used.

So, yes, I am returning this CPU. I have an excellent relationship with the vendor from whom I purchased this. They know what is up and have no problem with it. One other question: Would it make you feel any better to know that I bought the OCing insurance too? If so, feel free to believe that.

Maybe some wear their "OC 4 Life" T-shirt to bed every night secure in the knowledge that if a brand new CPU they were validating ever died while running entirely safe vcore and clockspeed, they would be the right and honorable OCer and forfeit the expense of the week-old chip. I can only guess that they would then write themselves a haiku in their journal about how hardcore they are and praise their own astounding levels of personal integrity. I guess I'm just a terrible person in comparison.

The chip died at entirely reasonable voltage and clock speed and is an unlocked chip intended to have its clock speed tweaked (and thus, on most motherboards, its vcore as well due to "auto" vcore). So I'm not convinced it wasn't defective/weak from the start. My vendor has no problem with that. So, really, neither should anyone else.

--H

P.S. My apologies if I misread the intent behind that question. But I've seen this all before (though not directed at me) too many times on these forums. So if someone wants to sell indigant self-righteousness in this thread, I'm sorry, but I'm just going to laugh at them. Because I sincerely think it's misplaced here.
 
Make sure to use the manual vcore setting with the new chip so this doesn't happen again.
 
I think I see where this is going. . .

*Snip*

P.S. My apologies if I misread the intent behind that question. But I've seen this all before (though not directed at me) too many times on these forums. So if someone wants to sell indigant self-righteousness in this thread, I'm sorry, but I'm just going to laugh at them. Because I sincerely think it's misplaced here.

Hey man, no self-righteousness from me. I have zero issues with what you're doing. It is possible that that your CPU would have bricked itself running stock clocks and modified voltages. The issue is similar to the chicken and the egg predicament. Was it a defective CPU or did the Mobo kill it? The motherboard doesn't blindly apply voltage to the CPU (unless you use Manual mode). The voltage is pulled from the CPU's VID Table. Offset, as you know, applies the +- Offset specified agaisn't that VID table. Unless the mobo freaked out (should have a failsafe?) I don't see how that voltage was applied. Surely Intel doesn't have that high of voltage in the VID Table.

I would still let ASUS know what happened. What they do with that information is up to them. That info may help them in the future.
 
Hey man, no self-righteousness from me. I have zero issues with what you're doing. It is possible that that your CPU would have bricked itself running stock clocks and modified voltages. The issue is similar to the chicken and the egg predicament. Was it a defective CPU or did the Mobo kill it? The motherboard doesn't blindly apply voltage to the CPU (unless you use Manual mode). The voltage is pulled from the CPU's VID Table. Offset, as you know, applies the +- Offset specified agaisn't that VID table. Unless the mobo freaked out (should have a failsafe?) I don't see how that voltage was applied. Surely Intel doesn't have that high of voltage in the VID Table.

I would still let ASUS know what happened. What they do with that information is up to them. That info may help them in the future.
Cool. Sorry for the misunderstanding. Obviously, I'm a bit touchy. I've never killed a CPU before over a quarter century of doing this stuff. The shame! :unsure:

I will indeed document what happened with Asus when this is all sorted. Though, to really do it right, I would have had to be willing to put more CPUs at risk (ie, reproduce the failure).
 
i had a DOA 6700K and a few months back a 6600k that lasted about 2 days before it shat itself. Never had bad intel cpus before skylake
 
i had a DOA 6700K and a few months back a 6600k that lasted about 2 days before it shat itself. Never had bad intel cpus before skylake

This intrigues me...

Could you possibly give some further details on those Skylake deaths please?

I ask because, from what I've experienced over the years, I've been giving Intel CPUs some serious shit voltage and they've always taken it. Bar none. I mean I've killed many a graphics card, but never a cpu, no matter how hard I've tried. Mobos die first. My current Haswell 4790K is on 1.52v right now. Has been at least 1.5v for about a year, and my previous Sandy @5.2ghz @ 1.52v before that lasted 2 years and made it look easy, so...much appreciated, cheers!

Just to add, if it helps, I was recently was forced to move from my Asus z87 Hero to my backup PC's Asus z97 Pro because I straight up *fried* one of the RAM slots through over-voltage. That happened, but the damage to the CPU? oh yeah, it took it, nothing, like a Cat...

Cheers! :cool:
 
When I first got my Z170/6700K combo I tried overclocking using offset and found it to be very flaky. Even on stock settings I was getting up to a 1.4v in edge cases. I'd recommend everyone who's overclocking a Sky Lake processor use the manual vcore. The auto vcore function isn't reliable or safe when overclocking. I'm using a manual 1.3V and I tested it with a variety of loads to make sure the LLC setting was correct. As for anyone who thinks it will use a lot more power, the chip is still downclocking and using idle states, unless you're on a battery it will never make a difference.

If you could help me understand what I'm doing and how to do what you're doing, I would really appreciate that.

On my ASUS Sabertooth Z170, this is what I've set in the CPU voltage section:
CPU Core/Cache Voltage: Adaptive Mode (while in BIOS, it shows 1.168V next to it)
Offset Mode Sign: +
Additional Turbo Mode CPU Core Voltage: 1.332
Offset Voltage: Auto
Total Adaptive Mode CPU Core Voltage: 1.332V (this line is greyed out, it's a total not a setting)
 
This intrigues me...

Could you possibly give some further details on those Skylake deaths please?

I ask because, from what I've experienced over the years, I've been giving Intel CPUs some serious shit voltage and they've always taken it. Bar none. I mean I've killed many a graphics card, but never a cpu, no matter how hard I've tried. Mobos die first. My current Haswell 4790K is on 1.52v right now. Has been at least 1.5v for about a year, and my previous Sandy @5.2ghz @ 1.52v before that lasted 2 years and made it look easy, so...much appreciated, cheers!

Just to add, if it helps, I was recently was forced to move from my Asus z87 Hero to my backup PC's Asus z97 Pro because I straight up *fried* one of the RAM slots through over-voltage. That happened, but the damage to the CPU? oh yeah, it took it, nothing, like a Cat...

Cheers! :cool:
You are lucky.
My 2500K was subjected to 1.49V for 1/2hr on Prime95 by accident after a bios update gave me LLC (had no LLC before that).
This damaged it and changed its max overclock from 4.6/4.7GHz to 4.3GHz, both at 1.44V.
Water cooled with decent home made setup.

My Dad now has it with a lesser cooler and it has to run at 4.1GHz max.
 
I personally don't use offset modes for this very reason. I will set a voltage in the BIOS and then I will get an entirely different voltage in practice. No thanks. I want the voltage I specify, and that's it. Full manual it is.
 
You are lucky.
My 2500K was subjected to 1.49V for 1/2hr on Prime95 by accident after a bios update gave me LLC (had no LLC before that).
This damaged it and changed its max overclock from 4.6/4.7GHz to 4.3GHz, both at 1.44V.
Water cooled with decent home made setup.

My Dad now has it with a lesser cooler and it has to run at 4.1GHz max.

Isn't that end user abuse? You degraded the hell out of that chip. It's not Intels fault.
 
You did... read what I replied to?
And
You did... read what I said?

I dont think you got anything right.

I got the feeling from your post that it was the voltage in a bubble. You can kill a cpu instantly by feeding it like 1.9v but that's not what happened with yours. That was from the 1/2 hr of Prime95 that degraded it and not just the voltage.
 
6600k was installed on a Gigabyte ITX build. Installed OS, hadn't yet oc'd system as I was still setting up benches etc... Wouldn't post when I got home to work on it. RMA'd cpu and replacement was fine in the system, and eventually oc'd the replacement to 4.5ghz. Running fine for 6mos now about.

6700k was DOA. Wouldn't post on Asus Sabertooth board. I3 6100 worked in the board until the board shat itself; so I dunno, maybe the board coulda done something. Tried the 6700k in question on Asrock board, and no post. Replacement cpu works fine.

I also had a i3 6100 go back on me, but that coulda been from overclocking on an Asrock itx Z170 board with a bclk bios for non k cpus. In the almost 20 years I've been building systems, I cannot recall ever having a bad intel cpu, other than damage from use/misuse/oci'ing etc... I think I had 1-2 AMD cpus go bad, and were replaced under RMA. Run of bad luck or Skylake is more fragile I guess.
 
I got the feeling from your post that it was the voltage in a bubble. You can kill a cpu instantly by feeding it like 1.9v but that's not what happened with yours. That was from the 1/2 hr of Prime95 that degraded it and not just the voltage.
I didnt feed it 1.9V though...
Yes, I stated running prime for 1/2hr with 1.49V harmed it, not much deduction needed there.
I didnt blame Intel for anything, there was no need to say that.

My comment was for the guy I replied to, he had a certain amount of luck because he used higher voltage for a long time and he didnt suffer.
I had previously tested basic clock speed vs voltage and I was already at the max speed with 1.44V, voltage increase didnt help.
1.52V would have nailed my chip.

Maybe he will reply with his circumstances.
 
Last edited:
I didnt feed it 1.9V though...
Yes, I stated running prime for 1/2hr with 1.49V harmed it, not much deduction needed there.
I didnt blame Intel for anything, there was no need to say that.

My comment was for the guy I replied to, he had a certain amount of luck because he used higher voltage for a long time and he didnt suffer.
I had previously tested basic clock speed vs voltage and I was already at the max speed with 1.44V, voltage increase didnt help.
1.52V would have nailed my chip.

Maybe he will reply with his circumstances.

Again, its not that you used 1.49v, it's that you primed it for a 1/2 hour. That only leads to massive degradation. Sandybridge was a freaking voltage king, it sucked it up and spit out massive overclocks. I've had my old 2600K at over 1.53v pushing 5.2-5.3ghz on water, and it did not die. I didn't run 1/2 hours of Prime. At these voltages you have to be careful with the load you put on the chips. Benching progs utilize the chip maybe 40-50% at most so they hardly faze the chip. This chip now runs my nephews system at 4.8ghz with a DH14 aircooler.

It's not the voltage that kills it, its how you go about it. And fyi, I've run my Skylake to 1.49ish for a bench or two and all is fine. And this is a stock chip, not delidded. It still runs silly low volts for daily 24/7 clocks like 4.6ghz at 1.3v with igpu on.

I scored 7 921 in Fire Strike Ultra
 
I personally don't use offset modes for this very reason. I will set a voltage in the BIOS and then I will get an entirely different voltage in practice. No thanks. I want the voltage I specify, and that's it. Full manual it is.

It is sketchy unless you verify the voltage being applied.
Without a volt meter its easy to be mislead with Skylake unless you are aware of the problem.

Many comments about CPUID say it will give the correct voltage but the last few versions (possibly since the release of Skylake) only give VID which is much lower.
I only get the correct reading from the onboard sensors when using the Asus software that came with the motherboard.
I've stated that a few times in ongoing threads but I think it got lost among the noise.
The version of CPUID that came on the CD is also wrong.
(My mobo is Asus Z170 Maximus VIII Hero, I've also had an ASRock and MSI Z170 board)

Its really shit that its not simple to know the explicit voltage applied.
Almost like everyone cheaped out and gave us the simple VID instead.
The difference can be massive!
 
^^Yea, adaptive does throw off the voltage baseline from manual. I find switching to adaptive from manual raises the voltage baseline something like 4-5 hundredths of a volt and that is not something to lose track off. That said I use adaptive, but I rework my voltages accounting for the baseline changes.
 
^^Yea, adaptive does throw off the voltage baseline from manual. I find switching to adaptive from manual raises the voltage baseline something like 4-5 hundredths of a volt and that is not something to lose track off. That said I use adaptive, but I rework my voltages accounting for the baseline changes.
Adaptive uses VID and an offset.
I found for my chips and overclock it uses almost the right voltage under load, I have to add a tiny bit more.
If you can make it work for you it gives much lower idle voltage/power consumption. My machine is on 24/7 so this is worth the effort.

Thats on a 6600K and 6700K under water.
I dont exceed 1.44V.
For those that dare push higher, good luck.
 
Again, its not that you used 1.49v, it's that you primed it for a 1/2 hour. That only leads to massive degradation.
Again, I fucking know, I stated it in my original post ffs.
But unlike what you said it wasnt user abuse, it was an accident.
I also did not blame Intel, and I wasnt even replying to you so there was no need for your "intervention".
Fyi I dont use the Prime95 with AVX so it wasnt pushed anywhere near as hard as it could have.

Sandybridge was a freaking voltage king, it sucked it up and spit out massive overclocks. I've had my old 2600K at over 1.53v pushing 5.2-5.3ghz on water, and it did not die. I didn't run 1/2 hours of Prime. At these voltages you have to be careful with the load you put on the chips. Benching progs utilize the chip maybe 40-50% at most so they hardly faze the chip. This chip now runs my nephews system at 4.8ghz with a DH14 aircooler.
You missed the bit where I run mine 24/7.
Gltches occur when its left running that dont occur very often, once a week maybe.
Higher voltage was not recommended and would have been worse for my condition.

It's not the voltage that kills it, its how you go about it. And fyi, I've run my Skylake to 1.49ish for a bench or two and all is fine. And this is a stock chip, not delidded. It still runs silly low volts for daily 24/7 clocks like 4.6ghz at 1.3v with igpu on.
Dont make things up.
High voltage alone certainly does kill, silicon is voltage rated for a reason.
Both high voltage and high heat cause the most harm harm. And your run of luck on the silicon lottery.
 
Well, couldn't quite bring myself to leave Asus behind entirely. So I have a Z170-Deluxe on the way. Those considering the Z170-WS. . . it was a fantastic motherboard. But be cautious and carefully monitor vcore when using offset mode in case you see any jumps above safe voltage. Maybe mine was a fluke.
 
Personally I never set anything to "Auto". I do use Adaptive with a turbo offset though.
 
I didnt feed it 1.9V though...
Yes, I stated running prime for 1/2hr with 1.49V harmed it, not much deduction needed there.
I didnt blame Intel for anything, there was no need to say that.

My comment was for the guy I replied to, he had a certain amount of luck because he used higher voltage for a long time and he didnt suffer.
I had previously tested basic clock speed vs voltage and I was already at the max speed with 1.44V, voltage increase didnt help.
1.52V would have nailed my chip.

Maybe he will reply with his circumstances.

It's not luck, imho. If you know what you're doing these chips can take ~1.5v 24/7 on water easily from my experience. I can't talk beyond Haswell though.

The only CPU I've had degrade was an AMD Athlon XP Barton core something. I can't remember the exact circumstances, but it was under water with alot of voltage and just got steadily worse by about 100mhz over about 2 years.


Oh, and there was an Athlon XP I fried because I was using an external pump at the time which didn't turn on one day when I was out. Can't even remember what chip it was!

athlon_fried.jpg


The waterblock is scared where the hot cpu magic dust literally exploded out. :dead:
 
It's not luck, imho. If you know what you're doing these chips can take ~1.5v 24/7 on water easily from my experience. I can't talk beyond Haswell though.

What do you use to check Vcore?
I cant get anything to report the correct voltage except the tool that came with my motherboard, but its refresh rate is so darn slow.
Everything else seems to report VID.

What temps are you getting under hard testing?
With 1.392/1.408V (reported by Asus AI suite) I'm hitting 75C on Prime95 non AVX after the delid.
I cant imagine using 1.5V.

Cheers.

I either have bad luck with my chip being hot or my water cooler has a problem.

edit
Ah nvm, forgot we were talking about my old Sandybridge CPU.
The above is my relatively new 6700K.
 
Last edited:
Well, after a three week wait, my CPU is here and I'm back in the saddle. This time with an Asus Z170 Deluxe.

Having fried a CPU for the first time in two decades of doing this stuff, I'm incredibly apprehensive to OC this new one. It is currently at 4200 (set manually) and auto vcore (peaking momentarily at 1.360 via HWMonitor) while running RealBench. XMP has my RAM at 3200MHz. Temps are peaking momentarily at 59/61/51/56c in a 76f room.

So it looks like a good chip and a good mount of the Corsair H115i. I may never really know how good it is though since I'm unwilling to lose another CPU. So I'm going to keep this one at 4.6GHz or below and vcore manual at no more than 1.385v.

I might experiment with different LLC levels rather than keeping it at auto. But even that scares me at this point.

Heh, I've lost my nerve! I'm too old for this sh*t now. =P
 
Asus stated that up to 1.44V on water is ok as long as temps are kept in check.
You are very aware of your voltages now so playing with LLC will be safe.
But make sure the voltage readings you are using are not VID.

Whatever you are using, run it alongside CPUZ.
If the voltages match, you are monitoring VID, not true vcore.
This could be the reason why your previous chip was damaged.

If CPUZ is much lower (when under load), you have likely been reading the correct voltage.
The last chips damage was bad luck.
 
Tough break OP. The skylakes seem to have alot of issues. I know this is not a scientific poll, just forum lurching. Still, I am glad I got the o/c limited Xeon version of skylake. I am usually my own worst enemy.
 
Asus stated that up to 1.44V on water is ok as long as temps are kept in check.
You are very aware of your voltages now so playing with LLC will be safe.
But make sure the voltage readings you are using are not VID.

Whatever you are using, run it alongside CPUZ.
If the voltages match, you are monitoring VID, not true vcore.
This could be the reason why your previous chip was damaged.

If CPUZ is much lower (when under load), you have likely been reading the correct voltage.
The last chips damage was bad luck.
Just a couple of clarifications. . .

I was very careful to read vcore and not VID. I primarily used HWMonitor (experimented with some which I dropped because they were obviously reporting VID as vcore).

CPUZ and HWMonitor both report the same numbers for vcore (and VID, which is reported as different, appropriately).

I gradually came to understand that "VID" is what the CPU was declaring it needs, while vcore (in HWMonitor and CPUZ) was the motherboard's best guess at what it is actually providing.

I can say with 100% certainty that when I came in to the frozen screen and saw the bizarre 1.984v showing, it was definitely vcore. I wish I hadn't rushed so much to shut off the computer and had taken a photo. After all, the damage was done.

Anyways, this new chip seems to be a good one. I spent the morning documenting its LLC behavior with both manual and adaptive vcore settings. As expected, small boosting begins in both modes at LLC level 6. So I'm sticking with that. I did all this testing at manually set 4.2GHz and 1.32 vcore just to be safe. I will not be trying offset mode again.

I've already broken my pledge to not use Prime95 anymore by using it to quickly get a "worst case" load on the CPUs to test vcore as described immediately above. But for longer-term testing, I'm going with RealBench (though this was what was running when my last one died).

So far this new one, or this different motherboard just seem to be more reliable in the vcore department. I set it manually, and it sticks right around there (with a little droop or boost under load according to LLC setting). I seem to recall that the Z170-WS was more wobbly even at manual. But I'll have to check my notes.

As of now, I'm at 4.6, manual vcore of 1.3v, and it's stable through an hour of RealBench (but that seems too good to be true and I won't call it stable until it passes a few 8 hour runs). If it makes it to 2 hours, I may try 4.7GHz. But not if it requires more than 1.375v. I'm done pushing the envelope. :)
 
So it looks like a good chip and a good mount of the Corsair H115i. I may never really know how good it is though since I'm unwilling to lose another CPU. So I'm going to keep this one at 4.6GHz or below and vcore manual at no more than 1.385v.

I might experiment with different LLC levels rather than keeping it at auto. But even that scares me at this point.

Heh, I've lost my nerve! I'm too old for this sh*t now. =P


I run my Maximus Ranger with 6700K at 4.6 Ghz, Voltage settled in at 1.3 manually, actual is 1.31-1.34 and game stable, stress testing. Memory is 3000mhz XMP.
 

Attachments

  • 6700K.PNG
    6700K.PNG
    968 KB · Views: 55
Back
Top