Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
I've noticed the same long-term stability problem. Sometimes it takes many hours, sometimes only 1 or 2, but it inevitably seems to BSOD. I have also observed this at Optimized Defaults with memory at XMP 2133. I'm wondering if it's some motherboard components overheating, although I can't do much better than the 200mm fan currently over the motherboard.
How does my results compare with your guys?
For those of you with long-term stability problems, have you bumped up your VRIN +0.01v and re-tested? That's what I needed to get long-term stability on mine. Well that and System Agent/VCCIO-D/VCCIO-A +0.1v (even without RAM overclocking this helped). Give it a try, I am quite curious.
We have similar results. Mine needed 1.21v vcore and 1.15v uncore at 36x to get P95 stability at 4.3Ghz. Beyond that has been difficult. 4.4Ghz takes 1.26v and VCCSA +0.10v and VRIN 1.81v for P95 stability. 4.5Gh is still not stable even at 1.32v but I will keep testing.
I delidded the original cpu in the first post!
Results in one word? Surprising. In short, there is a ~10c drop in temps on longterm averages and ~20c drop in spikes! I used Arctic Silver 5 on both the die and the Corsair H110 block. The MAX spiking temp is now 72-75c (one core goes to 81) using 1.375v/ 4.6ghz....with fans running at 750rpm (quiet mode). The cores bounce 50-70 instead of 70-90. Before I'd never see anything lower than 60 in the spikes. Not seeing 90-100c spikes is a relief.
I used the vice method (bought a $20 "Bessey" at Lowes) and honestly it was easier than I thought. Simply lock it in place, get a wood block on the edge, and do some medium strength taps with a hammer and it comes off...careful where you do it because the CPU might go flying...i put a pillow in front. I would NOT recommend using a razor because the VRAMs are really close the edge...you probably going to cut them unless you're extremely steady handed. Overall, the chip is really sturdy and I doubt you'll damage it even if you throw it across the room. The hardest part that took hours was cleaning the black stuff off..the stuff was like 1mil thick. After trying different chemicals, I recommend mineral spirits + plastic tooth picks!
I will be trying EK Supremacy direct die CPU block + cold water next.
Side note, I noticed that increasing CPU freq increases heat in a big way. 4.5ghz at 1.375v is a lot cooler than 4.6 at 1.375v. Also, increasing cache past 39 seems to cause the system to freeze and no memory dump/ BSOD. It will work for a while, but eventually lock up. Clock watch dog BSODs are related to RAM... you must increase CPU voltages related to RAM, decrease your ram clock, and/or increase your RAM voltage. I did benchmarks with higher cache and noticed 0 difference....but it keeps right to have core/cache be the same speed.
I recommend you go get a vice at Lowes and delid...results may vary, but a good chip would be one that runs on the lowest voltage. I would ignore the stock heat because how tight the intel heat sink to the die is will vary chip to chip. In the posts above, a Costa Rica chip seemed to gave good temps, but it required a lot of voltage to be stable. I suspect if I delidded it, it would more like ~10c temp drops or less. If you have a hot chip, I really think you should behead the sucker.
Btw, My chip is from Malay.
If that is stable for 24+ hours on AIDA64, that is a nice chip. Dozens of chips and 1.25v for 4.5ghz was the best I could find that was stable. Of course, many boot at that voltage, but basically none are stable.
You could probably do 4.6ghz at 1.35v and temps should be doable if you got a decent water cooler. Delid it!
For those of you with long-term stability problems, have you bumped up your VRIN +0.01v and re-tested? That's what I needed to get long-term stability on mine. Well that and System Agent/VCCIO-D/VCCIO-A +0.1v (even without RAM overclocking this helped). Give it a try, I am quite curious.
I delidded the original cpu in the first post!
Results in one word? Surprising. In short, there is a ~10c drop in temps on longterm averages and ~20c drop in spikes! I used Arctic Silver 5 on both the die and the Corsair H110 block. The MAX spiking temp is now 72-75c (one core goes to 81) using 1.375v/ 4.6ghz....with fans running at 750rpm (quiet mode). The cores bounce 50-70 instead of 70-90. Before I'd never see anything lower than 60 in the spikes. Not seeing 90-100c spikes is a relief.
NICE.
Add CLU TIM for max cooling performance from your direct die mount.
CLP should be better, CLU works better with non-lapped IHS because it can fill the gaps a bit better, but CLP is a better TIM for IHS. Phobya also make a liquid metal for about 1/2 the cost of CLP, supposed to be pretty similar.
I delidded the original cpu in the first post!
Results in one word? Surprising. In short, there is a ~10c drop in temps on longterm averages and ~20c drop in spikes! I used Arctic Silver 5 on both the die and the Corsair H110 block. The MAX spiking temp is now 72-75c (one core goes to 81) using 1.375v/ 4.6ghz....with fans running at 750rpm (quiet mode). The cores bounce 50-70 instead of 70-90. Before I'd never see anything lower than 60 in the spikes. Not seeing 90-100c spikes is a relief.
I used the vice method (bought a $20 "Bessey" at Lowes) and honestly it was easier than I thought. Simply lock it in place, get a wood block on the edge, and do some medium strength taps with a hammer and it comes off...careful where you do it because the CPU might go flying...i put a pillow in front. I would NOT recommend using a razor because the VRAMs are really close the edge...you probably going to cut them unless you're extremely steady handed. Overall, the chip is really sturdy and I doubt you'll damage it even if you throw it across the room. The hardest part that took hours was cleaning the black stuff off..the stuff was like 1mil thick. After trying different chemicals, I recommend mineral spirits + plastic tooth picks!
I will be trying EK Supremacy direct die CPU block + cold water next.
Side note, I noticed that increasing CPU freq increases heat in a big way. 4.5ghz at 1.375v is a lot cooler than 4.6 at 1.375v. Also, increasing cache past 39 seems to cause the system to freeze and no memory dump/ BSOD. It will work for a while, but eventually lock up. Clock watch dog BSODs are related to RAM... you must increase CPU voltages related to RAM, decrease your ram clock, and/or increase your RAM voltage. I did benchmarks with higher cache and noticed 0 difference....but it keeps right to have core/cache be the same speed.
I recommend you go get a vice at Lowes and delid...results may vary, but a good chip would be one that runs on the lowest voltage. I would ignore the stock heat because how tight the intel heat sink to the die is will vary chip to chip. In the posts above, a Costa Rica chip seemed to gave good temps, but it required a lot of voltage to be stable. I suspect if I delidded it, it would more like ~10c temp drops or less. If you have a hot chip, I really think you should behead the sucker.
I noticed your post what you said you must increase CPU voltages related to RAM is very true but at same time there other voltage you can play with that also bring temps back down . as for going straight die with the EK Supremacy is a bad idea . great block btw but you lose surface area, I tried it this week and what happens is you cpu take the heat from the silicon vs is you have the heat sink on it distributes the heat to the water block instead od you cpu doing all itself . what your problem is your using artic silver between the die and heatsink , try some Coollaboratory Liquid Ultra between and then on top usin some mx2 or mx4 that will come with the ek block . also straight die take alot of screw adjustments to the block to make all the cpu pins make contact you get alot of 55 error until you get right . Hope i saved you time
you got a nice 4770k there 45c average/ 61c spike with 1.375v/4.6ghz very nice actually if you get time try reliding with better thermal paste you see temps drop , why were not seeing this in the forum is cause most people straight do die and or happy with temps . but again great chip
Dude, can you PLEASE try to type like you actually speak English? Sentence structure, punctuation etc..It really makes it hard to understand what you are trying to say when you just type a bunch of random words together..
Has anyone's chip degraded yet? Reason I ask is because I cannot do my 4.3GHz overclock anymore with 1.220v. I was stable 12+ hours Prime95 but after going through various boards, I've concluded that it's slowly degrading. Plus, the IMC is horrible - barely can do 2200MHz.
Has anyone's chip degraded yet? Reason I ask is because I cannot do my 4.3GHz overclock anymore with 1.220v. I was stable 12+ hours Prime95 but after going through various boards, I've concluded that it's slowly degrading. Plus, the IMC is horrible - barely can do 2200MHz.
My 4770k sucks. I've been kicking the tires on this thing for over a week and a half and I can't get very far with it. I'm moving from a 2500k @ 4.5ghz. First, to level set a bit:
4770k running on an Asus Z87-Pro with the 1205 BIOS
16gig (4x4) Crucial Ballistix Tactical 1600mhz
Air cooling right now, CoolerMaster 212+ in push / pull
I did the initial test, 46x multiplier with 1.2v manual vcore. It starts loading Windows but then either reboots or blue screens before the login screen appears. I tried increasing the CPU Input Voltage to 1.9v - 2.0v but that didn't help. I tried increasing the CPU cache voltage to 1.2v, but that didn't help either. SVID is disabled.
Next test I tried was 45x at 1.2v. I can get to the Windows login screen but it either bluescreens or reboots before the Desktop appears. Tried the same CPU Input / CPU cache voltage adjustments with no success.
I repeated this until I got down to 42x, which seems like it might be stable at 1.2v. Honestly, 4.2ghz @ 1.2v seems pretty crappy to me. With such a low overclock I'm not sure its even worth it over leaving it set at 39x with mutlicore enhancement turned on.
I've tried tinkering with CPU LLC and the Max CPU cache multiplier but neither of those seemed to have any substantial effect.
Am I missing something?
Thought I was stable at x44 with 1.180v but apparently not. After using the system for a week normally I got a random BSOD (WHEA) while encoding video with FFmpeg and playing games at the same time.
I have to admit I only ran Prime95 for 2 hours before I settled with 1.180v. Upped it to 1.195 and still BSOD after ~6 hours Prime95 Blend. Running 1.205v now and hope it will last for 12+ hours.
Would you guys consider 4.4Ghz at 1.205v a dud or still above average?
Checked the BSOD and its a 0x00000124. A quick google search suggests that its memory related, but it happens even if I don't change anything memory related and just run at the JDEC values (1333mhz). I went ahead and swapped my 4x4gig 1.5v sticks for 2x8gig 1.35v sticks. All are Crucial Ballistix Tactical 1600mhz sticks. The 1.35v are low profile.
For what its worth, it didn't seem to make any difference.
I've tested three Haswell chips so far; two of them needed 1.25V for 10min of OCCT stability at 4.4Ghz and one needed 1.21V. I'd say your chip is probably a little above average.
Checked the BSOD and its a 0x00000124. A quick google search suggests that its memory related, but it happens even if I don't change anything memory related and just run at the JDEC values (1333mhz). I went ahead and swapped my 4x4gig 1.5v sticks for 2x8gig 1.35v sticks. All are Crucial Ballistix Tactical 1600mhz sticks. The 1.35v are low profile.
For what its worth, it didn't seem to make any difference.
I repeated this until I got down to 42x, which seems like it might be stable at 1.2v. Honestly, 4.2ghz @ 1.2v seems pretty crappy to me. With such a low overclock I'm not sure its even worth it over leaving it set at 39x with mutlicore enhancement turned on.
I just got up and was happily surprised it made it through a whole night of Prime:
http://imageshack.com/a/img29/9972/gd9w.png
What I don't get get is that HWiNFO64 is not showing the vcore I set in BIOS, CPU-Z would should the correct one. It's set to 1.205v in BIOS now, HWiNFO64 showed it as 1.216-1.232v fluctuating all the time.
Temperature spikes of 81°C are worrying though, this is on a custom water loop. My GTX Titan in the same loop is never exceeding 36°C under load. Maybe bad mounting or its really just the heat transfer problem and I need to delid. On a Corsair H80i I would get around 90°C at this voltages, used one before I went for custom water. Maximum normal load temps when encoding videos are around 58°C. I'm using FFmpeg a lot which is AVX enabled so I think I still to manual voltage.
My 4770K will *always* BSOD with 0x124 when the multi is too high and/or the voltage is too low. In Event Viewer, it will also say "Cache Hierarchy Error", but I have concluded it does not refer to the L3 cache, so increasing Cache Voltage makes no difference. It simply means "The Overclock is unstable, do something about it".
Pretty similar to mine. It does 4.3 GHz @ 1.218V. @ 4.2 GHz, I could probably reduce the voltage to around 1.2V.
Honestly the 4770K is one hell of a CPU even at stock, so if you don't feel the need for it, just leave at stock and enjoy awesome power savings at idle
Which Prime95 test did you run? I've found "Small FFT" the best both for stress testing for temps and stability. Before de-lid, mine would spike at 86-87C with 8K FFT's at 4.3 GHz / 1.218V. After de-lid, I'm at 80C @ 4.4 GHz / 1.24V.