Prime is melting my 4770k, oh the humanity. Other options?

Welp, I had a blue screen on Civ 5 earlier. It probably didn't happen all of the other times I've played it because this is the first time we've had turns go by constantly (so it had to calculate AI actions constantly). This friend was a lot faster going through the turns lol.

Anyway, I know it'll probably be fine at 1.31vcore, but honestly I don't really know what to do with anymore, so I just utterly reset it to stock for now.

Also, for giggles I tried running Prime95 28.1 totally at stock. Was hitting around 83C. I'm wondering if I put my Kraken X40 on there wrong. Maybe I need to redo the paste and stuff. For now I'm just gonna leave it at stock so I don't have to worry about it. Might do a modest boost up to 4.3 or so eventually, but 4.5 is apparently unattainable for me without going to 1.31 vcore, and that's probably not safe.

Oh well, crap luck in silicon lottery ho! Not that my luck is better in much else.

Nothing wrong with removing your CPU cooler to check how effective your thermal paste application was!
 
Well what's interesting is that my MSI MPower actually tries to tell you to keep your vcore under 1.3. Out of all the companies, MSI is the one that's trying to keep you safe. Lol.
It's funny, they are one of few companies that protects users.

Awesome response Dan, thanks.

EDIT* Found this:

45 Nanometer 1st Generation Core i ..... 1.40 Vcore
32 Nanometer 2nd Generation Core i .... 1.35 Vcore
22 Nanometer 3rd Generation Core i ..... 1.30 Vcore
22 Nanometer 4th Generation Core i ..... 1.30 Vcore

http://www.tomshardware.com/answers/id-1694551/safe-voltage-temp-4770k.html#11533981

I suppose I'll stick with 4.7 @ 1.300V for now. May drop to 4.6 just to give myself some headroom as I'm on adaptive voltage.
Well my 45 nm E7200 arrived with a sticker: "never exceed 1.25V". Considering these CPUs are using much smaller manufacturing technology, they should be more sensitive.
1.3 V is for watercooling with custom loop. Real world 24/7 voltage should be bit lower, how much depends on each individual CPU. I'd look at news of Ivy degradation, there should be some statistic data, because HW uses the same process, it only added a VRM on a die thus HW is hotter.
OCnet "The Ivy Bridge Stable / Suicide Club" shows even they stay under 1.27, and under 88 C.

Nothing wrong with removing your CPU cooler to check how effective your thermal paste application was!
He would need to reapply thermal paste afterward to ensure perfect contact. (Also don't do that with Indigo xtreme. Things would become expensive really quickly.)
 
Nothing wrong with removing your CPU cooler to check how effective your thermal paste application was!

Well honestly I didn't even apply any, I just used what came already on the Kraken. Probably wasn't terribly good but whatever... I didn't have any to use. I had some Arctic Silver or something or other paste somewhere, but apparently someone made off with it or something. Whatever the application on the CPU was, it was whatever the Kraken makers thought would be a good distribution... >_>

I'm an impatient person...

Anyway I'm not sure when I'll get to overclocking this again, but chances are I'll have to just clock it down to 4.4 and something well under 1.3V. I'm really not sure if 1.31 vcore is okay or not, but no point in taking chances I guess. Well that being said, I think the only reason I took it off of stock was just for kicks...
 
Never ran Prime. Been running my system at 4.8 for quite a while and backed down last summer to 4.5 for daily use. I couldn't care less about Prime since my system has been rock solid at both 4.8 and 4.5 configurations running months on end without reboots. I only do gaming and browsing on my system so if that works, everything is fine.

Prime can suck it.
 
And this is with a 4770k cpu @ 4.8GHz?


Never ran Prime. Been running my system at 4.8 for quite a while and backed down last summer to 4.5 for daily use. I couldn't care less about Prime since my system has been rock solid at both 4.8 and 4.5 configurations running months on end without reboots. I only do gaming and browsing on my system so if that works, everything is fine.

Prime can suck it.
 
I bought 2 4770k's and with the first one (second one hasn't come yet) that im running 4.5ghz @ 1.2v (1.1931 reported) its going to 81c max, 78c average on core #0, all the other cores are a few degrees cooler (as of right now running prime95 v28.1 its core0:78,1:73,2:74,3:68). This is all a slow build from ~35-36c ramping up after a few minutes to said temps. CPU is cooler using an Corsair H50 set to intake with one fan.

Ill report back when I get my second 4770k
 
Major thanks for the tip on WHEA errors. After scanning event viewer, I did have a few. I gave my voltage a tiny bump (1.275 on my Ivy Bridge 3770K) and they're long gone no matter what I throw at it.
Things were always stable in the past, but who knows what kind of little screw ups might have been going on in the background?
 
And this is with a 4770k cpu @ 4.8GHz?

Probably not...

This is the issue exactly. OP is about v28.1 and unrealistic loads on a 4770K. People that don't have a Haswell may not have a proper conception of how ridiculous v28.1 acts. Again: why isn't it on the Mersenne site? I would not recommend 28.1 for a Haswell stability test. In particular newbies like me run it on their new Haswell with Auto or Adaptive voltage settings in BIOS and throttle the crap out of these specific processors and way overvolt them without realizing it for a while.
 
This thread has made me finally realize i need to de-lid my 4770k. Now if i de-lid this CPU, will i be able to get higher than 4.3ghz? I remember hearing something about a certain percentage of these chips not being able to go over a certain mark. I'm not sure if that's due to heat issues (which i wasn't having) or it just plain won't go over 4.3 no matter what i do. Leave it alone or de-lid? If i de-lid and it doesn't work i'm just gonna pay the 25 bucks for the Intel Protection Program and pray i don't get another crap chip.
 
I would never delid a cpu. Your warranty is gone once you do that. You can tell if the cpu will boot at a certain multiplier. But you can also use the block method. So you're not limited on how you can overclock it. You just have to find what works best for you. Then if you are running very fast ram that builds up heat also. Since it is all now controlled on the cpu.

I would just sale yours and buy another. You might get a decent one from Intel but I wouldn't count on it. Either get another one of buy one that has been shown to do it. But even custom loops are having a hard time dealing with the heat on these Haswells.
 
Last edited:
This thread has made me finally realize i need to de-lid my 4770k. Now if i de-lid this CPU, will i be able to get higher than 4.3ghz? I remember hearing something about a certain percentage of these chips not being able to go over a certain mark. I'm not sure if that's due to heat issues (which i wasn't having) or it just plain won't go over 4.3 no matter what i do. Leave it alone or de-lid? If i de-lid and it doesn't work i'm just gonna pay the 25 bucks for the Intel Protection Program and pray i don't get another crap chip.

These are important questions and I want to make sure I don't mislead you so I can't be saying I've tested 800 4770ks but from my limited personal experience trying to find one that can do 4.5+ temperature isn't the limiting factor necessarily to getting a higher clock at all. It's more likely that it just has to do with the quality of the lithography on the chip you get. Where it was in the cake, if that makes any sense. As far as I can tell, this is the widest varying and most negatively slanted unlocked chip Intel has ever made. Why do we like them anyway? Because a 100Mhz bump to core clock on a Haswell means more than a 100Mhz bump ever has before.

Long answer short: no, delidding is almost certainly not going to get you any higher than 4.3Ghz, especially considering temps aren't really a problem for you anyway. People generally only delid chips that are obviously good OCers and chips that are really bad for practice/HTPC low profile builds.

About the Intel Protection Plan. This seems ultra scammy to me. It's for DEAD procs that have been damaged by overclocking. If you are just unhappy with your chips OC potential that doesn't really fulfill the criteria. It's also quite difficult to actually kill one of these chips by OCing them. They have thermal shutdown and throttle very efficiently. I'd be concerned about hurting other components if I actually made it a goal to destroy a 4770K with voltage just so I could send it in under that Protection Plan.

I suspect you can get Intel to approve replacement RMAs ONE TIME PER PROC on underperforming 4770ks simply by stating that it does not meet your satisfaction. I did this with a 4770s and they swapped it for a new one without asking many questions at all. I simply told them it was throttling itself on AIDA 64 with turboboost turned off even, which was true. I dumped the new retail 4770s they sent me on eBay. Just tell them it runs way too hot and see what they do. After seeing them dump 75$ 4770ks on the retail program I no longer feel so good about paying $340 for the one I'm currently using.
 
I miss my 2700k I can tell you that. You really will not get much gains on a synthetic benchmark that was designed to test a gpu though. Perhaps it will help about 5% max but that would be really it. The gpu is really doing all the work.

No it is with 2600K at 4.8 Ghz. I even ran it at 5.1 Ghz with 1.6 volts but I thought it was too much and I only gained like 50 3dmarks lol.
 
For gaming I've seen very few real-world benefits of OC'ing beyond the normal realm that programs like MSI's Genie automatically do. The benefits tend to come out more with creative suite and the like.
 
A good 4770K wouldn't need that much voltage for 4.5GHz. That kind of voltage would get you 4.7+ if you can keep it cool enough.

Mine needs 1.4v manual to reach 4700. I guess I have a crappy chip then? System rebooted after a couple hours with only 1.38v so I bumped it up and it's been running for over 10 hours of prime now. Is 1.4v unsafe for long term use?
 
Mine needs 1.4v manual to reach 4700. I guess I have a crappy chip then? System rebooted after a couple hours with only 1.38v so I bumped it up and it's been running for over 10 hours of prime now. Is 1.4v unsafe for long term use?

I wouldn't run anything over 1.3v for long term use. As has already been stated that chip degradation has been seen in these chips with use of high voltage. If you can keep it at 4.5 with 1.3v or less I'd stick with that.
 
I wouldn't run anything over 1.3v for long term use. As has already been stated that chip degradation has been seen in these chips with use of high voltage. If you can keep it at 4.5 with 1.3v or less I'd stick with that.

I found this info but thought under water 1.4v might be ok. Guess I'll have to back off some then :( I left it priming while I'm at work so hopefully a 18 hour test won't kill it.

http://hardforum.com/showthread.php?t=1764817

I think 4600 took closer to 1.35v but I changed a few other settings too so I will have to go back and check. I was hoping to leave it at 1.4v and 4700 because it seems rock solid...Bummer. I got ahead of myself once I found a fairly stable 4500 OC and just moved on so I'm not sure the minimum voltage it will require yet.
 
I wouldn't run anything over 1.3v for long term use. As has already been stated that chip degradation has been seen in these chips with use of high voltage. If you can keep it at 4.5 with 1.3v or less I'd stick with that.

+1

I was running 4.7 @ 1.305 but have since dropped to 4.6 @ 1.275 just to avoid degrading CPU.

I'm not missing those 100MHz at all and my temps are 3-5C lower due to knocking VCore down by 0.030V

I found this info but thought under water 1.4v might be ok. Guess I'll have to back off some then :( I left it priming while I'm at work so hopefully a 18 hour test won't kill it.

http://hardforum.com/showthread.php?t=1764817

I think 4600 took closer to 1.35v but I changed a few other settings too so I will have to go back and check. I was hoping to leave it at 1.4v and 4700 because it seems rock solid...Bummer. I got ahead of myself once I found a fairly stable 4500 OC and just moved on so I'm not sure the minimum voltage it will require yet.

Being "Safe under water" applies more to the operating temps than the affect of the voltage. Even if your CPU is running at 70C max @ 1.400V under load you are still pumping 1.4V through your CPU.

As Dan touched on above, VCore in excess of 1.3V can lead to degradation of the CPU. You could always purchase the Intel PTPP to cover your ass for 3 years in the event that your OC nukes the CPU.
 
+1

I was running 4.7 @ 1.305 but have since dropped to 4.6 @ 1.275 just to avoid degrading CPU.

I'm not missing those 100MHz at all and my temps are 3-5C lower due to knocking VCore down by 0.030V



Being "Safe under water" applies more to the operating temps than the affect of the voltage. Even if your CPU is running at 70C max @ 1.400V under load you are still pumping 1.4V through your CPU.

As Dan touched on above, VCore in excess of 1.3V can lead to degradation of the CPU. You could always purchase the Intel PTPP to cover your ass for 3 years in the event that your OC nukes the CPU.

Well I know that ;) I was just hoping the concerns were more with heat on haswell and could be mitigated by deliding and running my crazy loop. It sounds like only 4400Mhz will be safe for me if I need to use under 1.3v, but I can work on getting 4500.
 
Well it was stable at 4700, ran all day anyways. I believe 4600 @ 1.35v got me a bluescreen but can't remember for sure.

I'm now testing 4500 @ 1.3 core 1.9 VID. I will see if I can lower voltage from there. small FFT has highest core temp at about 73c
 
Last edited:
VID != VCore

In short, VID is the voltage the CPU is asking for, VCore is what it's actually getting.

Setting 1.86 VCore would quickly result in a nuked CPU.

Exactly. People need to understand that the IVR on these things can call for a huge range of voltage. This range is based on the current clock speed and what the IVR things is needed for the current workload. Prime95 for example can call for 1.65v or something at 4.6GHz. This is why you'll see the thing overheat fast and why you have to lock the voltage down so it doesn't do that with that application.
 
Okay, I'm fighting the urge to post my own "are my temps okay thread?". . . so instead this thread seems appropriate since I'm obviously seeing the odd behavior of Prime95 (though v 27.9 not 28.1).

I'm only concerned because so many Corsair H100i users claim to have mounting issue on recent Asus motherboards. They've resorted to using washers to make the cooler mount more securely/evenly and swear up and down that dramatic decreases in temperatures have resulted. Personally, I suspect that a lot of them originally had the backplate oriented wrong (it only fits one way, which is documented NOWHERE in Corsair's instructions and their youtube guy even puts it on wrong in the H100 installation video).

But, it's nagging at me. Every time I see certain settings cause throttling, I wonder if maybe I need to go to the hardware store and get some nylon washers. But I don't like to do non-standard stuff like that for no reason. And I have an old case without a CPU cut-out. So re-mounting the backplate (again) requires complete system disassembly.

Here's what I'm seeing in Prime95. . . with some Intel Burn Test (IBT) and Intel Extreme Tweaking Utility (Intel XTU) thrown in for good measure as well. . .

System Specs
Intel i7 4770k
Asus Z87-Deluxe (Bios 1602)
Corair h100i w/ Stock Corsair Fans @ 1700RPM (max RPM only makes a couple degrees difference)
Antec Twelve Hundred v1 (no easy access to CPU backplate!)
Windows 7 (later will be 8.1)


============== Stock ==================
ambient 70
Hyperthreading On
vcore = auto (shows 1.069-1.072 under stress)
56c Maximum IBT
56c Prime95 8k FFT


============ Overclocked @ 4.1GHz, Default/Auto vcore ================
ambient 71f
Hyperthreading On
vcore = auto (shows 1.141 idle, 1.222 under stress)
Multiplier = 41 (4.1GHz)
CPU Cache Ratio = auto
81c Prime95 8k FFT


============ Overclocked @ 4.3GHz, Default/Auto vcore ================
ambient 71f
Hyperthreading On
vcore = auto (shows 1.32 under stress!)
Multiplier = 43 (4.3GHz)
CPU Cache Ratio = auto
99(+)c Prime95 8k FFT (throttling)


============ Overclocked @ 4.7GHz, 1.3 vcore ==============
ambient 71f
Hyperthreading Off (would throttle at 99+ if on)
vcore manual/locked 1.3
Multiplier = 47 (4.7GHz)
CPU Cache Ratio = 44 (4.4Hz) Cache
92c Standard IBT
91c Prime95 8k FFT
72c Intel XTU (5 Minutes)

Thanks!
 
Incidentally, I just don't trust an overclock to be stable until it has made it 24 hours while looping 3DMark and Prime95 simultaneously.

This issue with Prime95 artificially increasing volts and temps on Haswell though has me struggling to figure out a way to validate my overclock. I'm not comfortable revving my CPU up to 90c+ at regular intervals over a 24-hour period.

My solution: I'm running a custom Prime95 job where the min FFT size is 256. This seems to avoid the issue. Temps stay in the 60s and 70s with only a momentary single spike to 86. And it did still seem to weed out a few under-volted overclocks. Hopefully it's still a good stability test even without the 8k-128k FFT tests!
 
This thread has made me finally realize i need to de-lid my 4770k. Now if i de-lid this CPU, will i be able to get higher than 4.3ghz? I remember hearing something about a certain percentage of these chips not being able to go over a certain mark. I'm not sure if that's due to heat issues (which i wasn't having) or it just plain won't go over 4.3 no matter what i do. Leave it alone or de-lid? If i de-lid and it doesn't work i'm just gonna pay the 25 bucks for the Intel Protection Program and pray i don't get another crap chip.
Don't think the IPP covers de-lidding. Just FYI.
 
Once you delid you are on your own. That protection isn't for physical damage. So you pretty much flushed that warranty down the drain. But people seem to like buying delidded cpus here. So you could sale it to someone I'm sure of it.
 
After some playing with voltages I was able to get my system stable at 4.6Ghz with 1.235v, cache ratio at 41 with 1.13v, vrin at 1.9v. Temps hit the mid 90's using IBT 20 passes. Using AIDA64 temps hit mid 70's. I think I've pretty much hit the limit of my H80i. Though, now that I've seen the clocks I can get at low voltages I'm considering delidding this chip.
 
After some playing with voltages I was able to get my system stable at 4.6Ghz with 1.235v, cache ratio at 41 with 1.13v, vrin at 1.9v. Temps hit the mid 90's using IBT 20 passes. Using AIDA64 temps hit mid 70's. I think I've pretty much hit the limit of my H80i. Though, now that I've seen the clocks I can get at low voltages I'm considering delidding this chip.

Wow that's amazing man, good job. I'm at 1.285v and holding stable at only 4400mhz under prime 95 v28.1 for 8+ hours small fft. Pretty sure it will go 24 hours, but I had to install some updates so instead it did 2 8 hour passes. I think I can do 4500mhz but at around 1.32v under prime 95 v28.1.

Glad I tried the new version of prime 95 cause the old one showed the system to be stable, but under the new version I had a restart. It also passed IBT, however it was not stable under the new prime 95 v28.1 and I would get reboots after about 4 hours. My temps peak at around 92c on the hottest core after running the new version of prime, and that's with 2 480 radiators so your H80i must be one hell of a cooler, granted your running a lot less voltage. I'm now going to decrease voltages and see how low I can go, and try to be happy at 4400mhz.
 
Boo! Just had BSOD at hour 22 of my 4.7GHz overclock (4770K) torture test. 1.33v.

Was running Prime95 and looping the Unigine Valley demo at the same time.

I had set the FFT minimum size to 512 to try to avoid the worst of the artificial/erroneous voltage increases and heat. So I was only maxing at about 88c on core 3 (core 1 isn't quite as hot since it's busy running Valley instead of focusing on Prime95 exclusively).

I hope setting the FFT like that doesn't defeat the purpose of stress-testing. But I'm just not comfortable cooking my chip overnight with those tiny FFTs. So I'm gambling/hoping that instability will still be weeded out with the larger FFTs. So far, that seems to be the case.

I may have to start pulling back my multiplier to 4.5 or 4.6. I'm not comfortable increasing vcore and heat much (if any) more. Not with only my wee Corair H100i. ;)

Sigh. . . 22 hours. Wife called to tell me it had crashed. I left it at hour 20 to go to work. Grrrr.
 
Boo! Just had BSOD at hour 22 of my 4.7GHz overclock (4770K) torture test. 1.33v.

Was running Prime95 and looping the Unigine Valley demo at the same time.

I had set the FFT minimum size to 512 to try to avoid the worst of the artificial/erroneous voltage increases and heat. So I was only maxing at about 88c on core 3 (core 1 isn't quite as hot since it's busy running Valley instead of focusing on Prime95 exclusively).

I hope setting the FFT like that doesn't defeat the purpose of stress-testing. But I'm just not comfortable cooking my chip overnight with those tiny FFTs. So I'm gambling/hoping that instability will still be weeded out with the larger FFTs. So far, that seems to be the case.

I may have to start pulling back my multiplier to 4.5 or 4.6. I'm not comfortable increasing vcore and heat much (if any) more. Not with only my wee Corair H100i. ;)

Sigh. . . 22 hours. Wife called to tell me it had crashed. I left it at hour 20 to go to work. Grrrr.

I think I'm going to delid, then try 4500hmz again. Like you I'm just not comfortable leaving the CPU at 91c for 24 hours, although probably it should be ok. I anticipate a 12-15c drop if I delid, and use some liquid metal TIM.
 
I think I'm going to delid, then try 4500hmz again. Like you I'm just not comfortable leaving the CPU at 91c for 24 hours, although probably it should be ok. I anticipate a 12-15c drop if I delid, and use some liquid metal TIM.

Are you going to do the vice method? I brought one home with me that was sitting in my folk's garage for years the other day, along with a good steel mallet and I've been eyeing the vice trying to figure out how I'm going to mount it... or if I should.
 
Are you going to do the vice method? I brought one home with me that was sitting in my folk's garage for years the other day, along with a good steel mallet and I've been eyeing the vice trying to figure out how I'm going to mount it... or if I should.

IMHO the vice method looks dumb as hell. Why would you subject your chip to that much abuse?

The razor blade method or the hairdryer/vice method are the safest from what I've seen.

This is probably the least stressful delid I've seen. Hairdryer to loosen the paste and then offsets the chip in the vice and applies just enough pressure to break the seal.
 
That looks a little fake. If it was that hot how could he pick it up and then use his finger nail to remove ihs. I would have thought it would have been a little too hot to grab it and then do that.


IMHO the vice method looks dumb as hell. Why would you subject your chip to that much abuse?

The razor blade method or the hairdryer/vice method are the safest from what I've seen.

This is probably the least stressful delid I've seen. Hairdryer to loosen the paste and then offsets the chip in the vice and applies just enough pressure to break the seal.
 
Are you going to do the vice method? I brought one home with me that was sitting in my folk's garage for years the other day, along with a good steel mallet and I've been eyeing the vice trying to figure out how I'm going to mount it... or if I should.

Yes I think so :) One vice I have has smooth sides so I'll probably use that one to hold the IHS.
 
Finally passed 24 hour torture test of Prime95 (512k minimum FFT size) and Unigine Valley looping.

4.7GHz (HT Off)
vcore: 1.340
cache voltage: 1.300 (cache running at 4.4GHz)
input voltage (not vcore): 1.8

Temps never exceeded 88c on hottest core. But, again, I modified the FFT size to stay away from those killer ones. It would almost certainly throttle with adaptive voltage turned on if I let it do the 8k FFT or turn off all torture tests in AIDA except FPU.

But, for actual usage, I think I have it dialed in. I'm hoping those voltages are safe for a 4770K and Corsair H100i.

A fresh install of Windows 8.1, applications, games, and data seems to have gone off without a hitch with those settings. Pretty happy. But would like to now see if I can bring those voltages down a bit since I'm told that the last thing I did (raising input voltage) can actually allow vcore to be lowered. But, I just need to actually start using this computer!

--H
 
Last edited:
Back
Top