5960X Overclocking - Complete Dog?

sk3tch

2[H]4U
Joined
Sep 5, 2008
Messages
3,341
It appears that I may be unlucky with my 5960X. No matter what I try, I can't seem to get it to OC. Can't even OC to 3.6GHz! I can, however, run at 3.5GHz with XMP on (3000MHz DDR4). I've used the tips from this thread, among others...only thing I can think of is maybe my PSU (currently: Seasonic X-1250 1250W). That's very unlikely - but I have a AX1500i on the way, anyway. I've pumped as high as 1.375v CPU/2.1 VCCIN/1.35V RAM - 3.8GHz crashed after about 6 minutes of ROG RealBench. Didn't bother going lower - no point in OCing with that voltage if it's for 300MHz (runs at stock voltage at 3.5GHz/3000MHz DDR4 XMP). Temps are fantastic. Just not a CPU that will comply to any OC. I just can't really believe it. Going to keep at it...but it has been a few days of try/reboot/test/reboot/re-try..

- Tried 45x100, 1.35v VCORE, 1.95v VCCIN, 1.35v DDR4 and went back 100MHz at a time - went as high as 1.375v VCORE and 2.1 VCCIN. Went as low as 3.7GHz and quit trying (may as well run stock at that point). Using ROG RealBench with 1 hour of success as a benchmark for a stable OC.
- Temps never an issue (except at higher voltage - just high then - in 90s) - so thinking it's just a dog CPU.
- Now at XMP (hence the BCLK) at stock speed for CPU.

Specs:
Intel i7-5960X CPU (Corsair H110 280mm Liquid CPU cooler), Corsair Vengeance LPX 16GB 3000MHz DDR4, Tri-SLI EVGA GTX 980 SC 4GB, Creative Sound Blaster ZxR audio, Samsung 850 Pro SSDs (256GB; 1TB), and a 2TB Western Digital Green WD20EARX 5400RPM HDD on an MSI X99S XPOWER AC motherboard inside of a Cooler Master CM Storm Trooper full tower – powered by a Seasonic X Series X-1250 (1250W; fully modular).

BIOs screenshots (with 3.5GHz/XMP - but the only things I'm not using are CPU core, VCCIN, DRAM CH_A/B, DRAM CH_C/D which I have altered previously for my OC attempts).


(click to enlarge)


(click to enlarge)

\
(click to enlarge)

\
(click to enlarge)


Any thoughts?
 
Last edited:
So XMP sets bclk to 125mhz? Because the 4th screenshot shows BCLK 125 MHz.
 
So XMP sets bclk to 125mhz? Because the 4th screenshot shows BCLK 125 MHz.


Yup. Did the same with my prior 5930K and 2666MHz DDR4 (EVGA X99 Micro board).

The screenshots are just to show my mobos options and what I've set (outside of what I outlined in OP).
 
Last edited:
- Tried 45x100, 1.35v VCORE, 1.95v VCCIN

1.35v is an awfully high starting point. I would start much lower and lock some settings down. An H110 @ that voltage will struggle to keep temps under control when really stressed.

1.25v VCORE, 1.95v VCCIN 40x100.
Memory @ 2133 at whatever voltage it calls for 1.35v maybe.
Lock the Ring Ratio @ 30.

Leave everything else at auto. Then start testing and see if you can get any stability. Work your way down the CPU Ratio if it is not stable.

Good luck.
 
1.35v is an awfully high starting point. I would start much lower and lock some settings down. An H110 @ that voltage will struggle to keep temps under control when really stressed.

1.25v VCORE, 1.95v VCCIN 40x100.
Memory @ 2133 at whatever voltage it calls for 1.35v maybe.
Lock the Ring Ratio @ 30.

Leave everything else at auto. Then start testing and see if you can get any stability. Work your way down the CPU Ratio if it is not stable.

Good luck.

Thanks! You may be onto something - perhaps too much voltage? I did use the ROG guide at first (so 1.30v). With 1.25v and 4.0GHz RealBench went 29 minutes before crash:


(click to enlarge)

Just very disheartening that even an extremely modest OC won't work. There has to be something obvious that I'm missing? I'm not looking for 4.5GHz or anything...just 3.9GHz or higher. :)

Off to try 3.9GHz with 1.25v.

EDIT: failed at 3.9GHz, 3.8GHz, 3.7GHz, and 3.6GHz. I have to be doing something wrong here, lol.

EDIT2: Going to re-start at 4GHz/1.25v (same as recommended) with AIDA64 System Stability Test. I have heard it's not nearly as tough as others - but at this point I just want to pass something! Ha.

EDIT3: 90 minutes passed with AIDA64 System Stability Test...so that's...something.


(click to enlarge)
 
Last edited:
Are you hitting the chips default TDP? I'm not sure if Haswell has the same options for turbo-boost, but I had to tweak with those settings to get my 3770k stable under Intel burn test.

With turbo disabled the chip would be mostly stable at 4.5 and 1.22v, but after maybe 10 minutes of IBT it would fail. Eventually I discovered that the chip was pegging against the default wattage limit around 77w. I was able to increase the power limit with Intel's Extreme Tuning Utility and the issue vanished. It took me a while to track that down because less-demanding stress tests would run fine.

Strangely, my motherboard locked the power controls unless turbo is enabled, so I switched it on, locked the speeds at 4.5 and increased the power limit. This was the only way I could get it IBT stable other than using the Intel tool in windows to increase the power limit after boot.
 
Are you hitting the chips default TDP? I'm not sure if Haswell has the same options for turbo-boost, but I had to tweak with those settings to get my 3770k stable under Intel burn test.

With turbo disabled the chip would be mostly stable at 4.5 and 1.22v, but after maybe 10 minutes of IBT it would fail. Eventually I discovered that the chip was pegging against the default wattage limit around 77w. I was able to increase the power limit with Intel's Extreme Tuning Utility and the issue vanished. It took me a while to track that down because less-demanding stress tests would run fine.

Strangely, my motherboard locked the power controls unless turbo is enabled, so I switched it on, locked the speeds at 4.5 and increased the power limit. This was the only way I could get it IBT stable other than using the Intel tool in windows to increase the power limit after boot.

Thanks for chiming in!

Are you talking about the Turbo Boost Power Max? Here's a screenshot of my system:


(click to enlarge)

I have turbo turned off in my BIOS settings (in OP) among some other power limiters being disabled. Is it only an XTU thing? If so, how do I make it "stick" and yet not affect everything else (I.E. do I have to set my multipliers in XTU to match what I have set in the mobo's BIOS? Seems like it's dicey to mix software and BIOS OC...).

Thanks!
 
Yep, thats the boost power max. From your screenshot, it looks like its already set to high enough values to not be an issue.

I believe any tweaks made in XTU will on-the-fly adjust whatever the bios set them to at boot. Tweaking in XTU worked for me because my bios was keeping a power cap even though turbo was disabled, and those power cap settings are locked in my bios unless turbo is enabled. XTU was showing the stock power limits when I first launched it and tweaking it there confirmed my suspicions.
 
Yep, thats the boost power max. From your screenshot, it looks like its already set to high enough values to not be an issue.

I believe any tweaks made in XTU will on-the-fly adjust whatever the bios set them to at boot. Tweaking in XTU worked for me because my bios was keeping a power cap even though turbo was disabled, and those power cap settings are locked in my bios unless turbo is enabled. XTU was showing the stock power limits when I first launched it and tweaking it there confirmed my suspicions.


Yeah, it didn't even let me edit it. That was just shown by default.

Stable after 8 hours of AIDA64 System Stability Test at 4GHz/1.25v VCORE/1.25V DDR/1.95v VCCIN.


(click to enlarge)

Just not sure how reliable that test is compared to the others. Stability testing tends to spark religious debates. :)
 
Last edited:
My 4670k @ 4400 was AIDA64 stable @ 1.232v. Prime95 27.9 stable @ 1.268v.
Arma 3 stable @ 1.28v.
 
You have retardedly high PCH 1.05 voltage and VCCIO, and I can't figure out why (both are above the max Intel specs). Set those back manually to 1.05 and somewhere south of 1.200V respectively and increase your SA voltage to 1.100-1.150V if you;re trying to find stability with high speed RAM. It looks like your BIOS might be a little wonky, I also have an MSI X99 board, my CPU is only a 5820k but I'm running at 4500mhz right now and my board did not auto-volt those things nearly as high. I'll try and grab a screenshot of it in a minute.

EDIT- here is some screens of my quick and dirty 4.5ghz, note the auto voltages my board has set in respect to yours (mine isn't a higher end OC'ing board, just the SLI plus so I don't have many of the settings you do though -also my CPU core voltage is only 1.295V under load, not sure why it shows 1.373V in BIOS):


MSI_SnapShot_01.png

MSI_SnapShot_00.png

MSI_SnapShot.png
 
Last edited:
You have retardedly high PCH 1.05 voltage and VCCIO, and I can't figure out why (both are above the max Intel specs). Set those back manually to 1.05 and somewhere south of 1.200V respectively and increase your SA voltage to 1.100-1.150V if you;re trying to find stability with high speed RAM. It looks like your BIOS might be a little wonky, I also have an MSI X99 board, my CPU is only a 5820k but I'm running at 4500mhz right now and my board did not auto-volt those things nearly as high. I'll try and grab a screenshot of it in a minute.

Thank you for taking the time to check it out! I will investigate that immediately. The screenshots were taken when I enabled XMP with my Corsair DDR4 3000MHz memory. I'm now back to standard 100 BCLK (right now: testing 4GHz/1.29v VCORE/1.35v DDR4/1.95 VCCIN with ROG RealBench). Upped the memory voltage after reading that it is likely the culprit (or potentially so).
 
Thank you for taking the time to check it out! I will investigate that immediately. The screenshots were taken when I enabled XMP with my Corsair DDR4 3000MHz memory. I'm now back to standard 100 BCLK (right now: testing 4GHz/1.29v VCORE/1.35v DDR4/1.95 VCCIN with ROG RealBench). Upped the memory voltage after reading that it is likely the culprit (or potentially so).

Yeah, definitely bump your SA voltage for 3000mhz RAM. I would try 1.150V first and see if it helps stability. I had to bump mine from .950V to 1.100V just for 2400mhz. Also be aware if you try to find your manual stable voltage and then switch to offset so it will lower the volts at idle, the MSI board has a big jump between 1.900V VCCIN and 2.000V, with 1.9 I would offset to +.075 and still not make the load volts I want, but moving to 2V I can do it with only 0.025 offset.
 
Yeah, definitely bump your SA voltage for 3000mhz RAM. I would try 1.150V first and see if it helps stability. I had to bump mine from .950V to 1.100V just for 2400mhz.

Thank you. Applied and rebooting now.

I noticed that you left turbo/EIST on and you're using dynamic mode for CPU ratio...mine is always fixed and that stuff is disabled. Doubt it matters much for benching/stability testing since we're pegging our CPUs - but that's something of note that is different, as well.

EDIT: won't even boot into Windows! Ha. I'll re-review. Wonder if my 3x GPUs is affecting things, as well?
 
Thank you. Applied and rebooting now.

I noticed that you left turbo/EIST on and you're using dynamic mode for CPU ratio...mine is always fixed and that stuff is disabled. Doubt it matters much for benching/stability testing since we're pegging our CPUs - but that's something of note that is different, as well.

EDIT: won't even boot into Windows! Ha. I'll re-review. Wonder if my 3x GPUs is affecting things, as well?

Not sure, but I do know quite a few boards struggle with 3000mhz RAM. I know Asus was quick to release new BIOS to fix the compatibility problems but I don't know about MSI (thats the reason I went with 2400, to keep the bus speed at 100mhz to save on headaches). You might also find some help over here: http://www.overclock.net/t/1511496/official-msi-x99s-motherboard-owners-club
 
Not sure, but I do know quite a few boards struggle with 3000mhz RAM. I know Asus was quick to release new BIOS to fix the compatibility problems but I don't know about MSI (thats the reason I went with 2400, to keep the bus speed at 100mhz to save on headaches). You might also find some help over here: http://www.overclock.net/t/1511496/official-msi-x99s-motherboard-owners-club

Yeah, I'm not even trying 3000MHz right now - just 2133 (or whatever the default is with these sticks...you only get 3000MHz with XMP or manual tuning). Bus is at 100. Thanks for the help. I'll hop over there, as well.
 
Last edited:
Yeah, I'm not even trying 3000MHz right now - just 2133. Thanks for the help. I'll hop over there, as well.

Have you tried just unplugging it and taking the battery out for 10-15 minutes and starting fresh? Sometimes just doing something simple like that will help.
 
Have you tried just unplugging it and taking the battery out for 10-15 minutes and starting fresh? Sometimes just doing something simple like that will help.

This is what I was thinking after reading this. Something is not right. I would start fresh and change ONLY your multiplier until it fails stress. Then add a little extra voltage. rinse and repeat.
 
Have you tried just unplugging it and taking the battery out for 10-15 minutes and starting fresh? Sometimes just doing something simple like that will help.

This is what I was thinking after reading this. Something is not right. I would start fresh and change ONLY your multiplier until it fails stress. Then add a little extra voltage. rinse and repeat.

Good suggestion, guys. I will try that next. Thanks!!

EDIT: first I reset all defaults in BIOS and then changed CPU ratio to 45, VCCIN to 2.0v, and CPU Core Voltage to 1.3v - leaving the rest standard (other than disabling 2nd NIC and onboard mobo audio) - even all of the energy saving crap. I'll work back from this. If this doesn't work - then I'll do the power pull/battery pull to do a complete refresh.
 
Last edited:
If you do decide to use XMP for the RAM, be aware that it may flip out and not be stable.

But then if you turn off XMP and set all the timings and voltages and speed settings to what XMP set it to, it will probably be stable.

XMP is basically trash for the most part and has been ever since it was introduced. The only thing it is good for is for getting a baseline of what to try.
 
If you do decide to use XMP for the RAM, be aware that it may flip out and not be stable.

But then if you turn off XMP and set all the timings and voltages and speed settings to what XMP set it to, it will probably be stable.

XMP is basically trash for the most part and has been ever since it was introduced. The only thing it is good for is for getting a baseline of what to try.

Thanks - XMP is next on my list (once I stabilize an OC - then you need to stabilize with XMP).

Amazingly - CMOS reset has led to the same results. 4.2GHz/1.3v VCORE/2.0v VCCIN/100 BCLK just failed ROG RealBench. Off to try 4.1GHz. :mad:
 
I can't help you OP, but this thread has been awesome for me. As a relative newcomer to OCing, the discussion has proven very enlightening.

Thanks [H] gurus! :D
 
sk3tch, I hope you are keeping the Ring Ratio @ 30. If you are not you are trying to OC it to the same frequency as the core which doesn't always work.

My 4670k will only do a Ring Ratio of 40 no matter what voltage I use. You need to do just the core, then move the ring ratio up until it starts crashing again. Once you get those 2 set then you can move on to memory.

That's just my experience with Haswell.
 
I can't help you OP, but this thread has been awesome for me. As a relative newcomer to OCing, the discussion has proven very enlightening.

Thanks [H] gurus! :D

Likewise! Love this place. Glad you can get something from this, too! I've been doing this since the Celeron 300A but the Haswell stuff is a whole "new" game. :)

sk3tch, I hope you are keeping the Ring Ratio @ 30. If you are not you are trying to OC it to the same frequency as the core which doesn't always work.

My 4670k will only do a Ring Ratio of 40 no matter what voltage I use. You need to do just the core, then move the ring ratio up until it starts crashing again. Once you get those 2 set then you can move on to memory.

That's just my experience with Haswell.

Thanks - I've been leaving that at "Auto" or hard-setting to 30.

Quick update as to where I am at:

FAIL FAIL FAIL FAIL

Back at stock now and running my stability tests to make sure it's not a software problem (i.e. ROG RealBench is at C:\ maybe that affects it - that type of dumb stuff). If I get an hour stable at stock, I am going to just presume that my CPU isn't fully stable at anything higher than 3.5Ghz. I can certainly boot into Windows and do most everything at 4.5GHz/1.3v/2.0v but I don't want to have that Russian Roulette wheel constantly spinning. :D
 
Leaving it on Auto makes it run @ whatever CPU ratio you have, that is why I was saying set it to 30.

OK. I'll try that now.

Latest news is even at stock (BIOS reset to defaults and just the mobo sound card and 2nd NIC disabled) it crashes! Could I have a bad CPU??? I'm using ROG RealBench on another PC and it's fine. I don't think I am doing it wrong. It's pretty braindead.


(click to enlarge)
 

Looks like that is what I should probably do. I will call NewEgg in the morning...because ideally I'd like to do a buy new and return so I'm not without a CPU.

Has anyone exchanged CPUs with NewEgg? Do they let you do a cross-ship kind of thing? I don't care if I have to "buy" one temporarily for 2-3 weeks or whatever.

Maybe I'll just buy a CPU now and straight up return the other one.

EDIT: that's right - no returns on the CPUs. That's fine. I think I might actually call Intel first tomorrow morning. If they can't do some kind of cross-ship then I will deal with NewEgg. Thanks all!!

(P.S. running Prime 95 28.5 Small FFTs right now and it has almost been an hour at stock speeds/voltages with no crash - will 99.9% RMA it, anyway - but now I'm just kind of fascinated with this chip).
 
Is it possible you have a bad stick of RAM screwing things up? I would think its very unlikely that you'd have a bad CPU and more likely that there are a few bad ram cells that are corrupting the stress test results.

It could certainly be the CPU, just statistically speaking there are a lot more bad ram sticks that make it through quality control than bad CPUs.

A bad CPU or glitchy memory controller on it would still cause errors in memtest, but the errors would be at random addresses with no pattern while bad ram cells will fail at the same address every time.
 
I doubt it is the CPU, I would suspect the board long before the CPU, especially with those ridiculous voltages being applied at auto. Have you tried with just one stick of RAM at a time?
 
It could be them memory or the board, using one stick and testing at stock should narrow it down pretty quickly. When I have had bad motherboards or memory the symptoms have been catastrophic, not failed stress tests. Random hard locks and BSOD's, he has not mentioned those signs at all, he just says failed. He said he was using prime95 with small fft's and it passed, I think that is the one that just loads on the CPU, I could be wrong.


I have never had a "bad" CPU, one that failed to work at advertised specs. I think we have have all had bad overclockers. Maybe he should get the Intel PTPP and just exchange it if he is not happy with it. http://click.intel.com/tuningplan/purchase-a-plan
 
Is it possible you have a bad stick of RAM screwing things up? I would think its very unlikely that you'd have a bad CPU and more likely that there are a few bad ram cells that are corrupting the stress test results.

It could certainly be the CPU, just statistically speaking there are a lot more bad ram sticks that make it through quality control than bad CPUs.

A bad CPU or glitchy memory controller on it would still cause errors in memtest, but the errors would be at random addresses with no pattern while bad ram cells will fail at the same address every time.

I ran Windows Memory Diagnostic and it did pass (did this a day or so ago - which I why I ruled it out). I'm still not 100% positive. Any other Windows 8.1-compatible tests I should run? My old favorite (memtest 86+) doesn't work with 8.1.

I doubt it is the CPU, I would suspect the board long before the CPU, especially with those ridiculous voltages being applied at auto. Have you tried with just one stick of RAM at a time?

I will try with one stick of RAM. Another thing is the GPUs - I've never had a problem before (I've run 4-way, 3-way, 2-way in many configs) but maybe that's somehow causing an issue? I have 4x 980s here and I swapped some out yesterday while I was resetting the CMOS battery (mostly due to cosmetics - my "GEFORCE GTX" lights aren't all the same arggh). No difference, yet.

It could be them memory or the board, using one stick and testing at stock should narrow it down pretty quickly. When I have had bad motherboards or memory the symptoms have been catastrophic, not failed stress tests. Random hard locks and BSOD's, he has not mentioned those signs at all, he just says failed. He said he was using prime95 with small fft's and it passed, I think that is the one that just loads on the CPU, I could be wrong.


I have never had a "bad" CPU, one that failed to work at advertised specs. I think we have have all had bad overclockers. Maybe he should get the Intel PTPP and just exchange it if he is not happy with it. http://click.intel.com/tuningplan/purchase-a-plan

On the Prime95 front - Prime95 28.5 Small FFTs over 12 hours - stable (stock frequencies):


(click to enlarge)

Good suggestions. I sign up for PTPP every time...unfortunately there's like a 30 day grace period until you can use it. So yeah, in about 25 days I could go that route. I am sure Intel would be cool on the phone if I called them now...but yeah, don't want to rip a CPU out just yet.

Right now I moved ROG RealBench (+ fresh files from the zip) to D:\ (instead of C:\) in case there's some permissions issue. Running 4.0GHz/1.25v/2.0v with everything else stock in this test. EDIT: that failed - so it isn't ROG RealBench's file location causing an issue. :( EDIT2: trying Prime95 28.5 Small FFTs with the 4GHz config. Maybe I have a bad stick of RAM or GPU causing RealBench to barf...

I'll test RAM sticks later today. I have MicroCenter in town so I can always go that route. Business trip tomorrow to Saturday so my play time is going to be limited soon. :)

In other news:
- Corsair AX1500i arrives today from NewEgg!!
- Ordered 2x Noctua NF-A14 iPPC-3000 PWM super fans for my Corsair H110 - yeah, they'll be LOUD!!! :) (if this proc is a dog - may as well take advantage of my PPTP coverage, tons and tons of headroom with AX1500i, and keep it cool - blow it out with 1.4v and see what I can get, ha).

Thanks, as always, all!!!
 
Last edited:
HCI memtest works in 8.1.

http://hcidesign.com/memtest/

It also tends to find problems a lot quicker than memtest86+ does.

Each instance of this test will only do 2GB, so you will have to run multiple copies at the same time.
 
THE PLOT THICKENS!

4.0 GHz / 1.25v VCORE / 2.0 VCCIN

8 hours of Prime95 28.5 Small FFTs = PASSED! (with the requisite high, high temps)


(click to enlarge)

That joins the AIDA64 Stability Test as two benches that have lasted 8+ hours at 4.0 GHz.

Running HCL MemTest right now - I wonder if it's my RAM??
 
Back
Top