5820k not OC'ing at all. What am I missing?

[cbliss1000]

Sigh. I've finished a brand new x99 rig, blissfully anticipating the 4.2-5 OC that seems so commonplace. One frustrating afternoon later and I have nothing to show for it - even modest 200-300mhz OCs are failing after an hour of stressing. Anything above that fails within 5 minutes in Intel Extreme Tuning Utility.

At stock everything works fine. I stress tested for 2 hours in Intel Extreme and 3 hours in AIDA64. Memtest passed with no errors after 8 hours. But... anything higher and it's unstable, regardless of what I do with vCore (1.3 was the highest I tried, no dice). The preset options don't work and I've confirmed I'm running the latest BIOS. I've tried with and without loading the mb's XMP settings for the RAM (by default the mobo is detecting a lower speed RAM), with and without the mobo's energy saving option, and with and without also adding a 1.9v input voltage. I followed the typical approach -- increasing multiplier, then increasing vCore once it fails -- but it once it became clear that nothing was working, I tried various combinations of the options above. No dice.

Temperatures have been fine (< 70c, except when I try high clocks -- even then, never > 80c).

Are there other approaches I can/should try? And what is most likely causing the issue - CPU, mobo, maybe PSU?

The build:

• i7 5820k
• ASRock Extreme4
• 4 x 4 Corsair Vengeance LPX CMK16GX4M4A2666C16
• be quiet! DARK ROCK PRO 3 Silentwings
• MSI R9 390
• Seasonic SS-660XP2

FWIW - and I think this is unrelated, but more info is good info, right? - I RMA'd the first Extreme4 I got, which was very erratic. I haven't experienced the same issues with its replacement.

Thoughts?

 

[Daemas]

what are you increasing the CPU voltage to? MOST cpus should be able to do 4.3-4.6 at 1.3v.

If you haven't gotten that high yet, I would just skip all the other lower clocks and set it for 1.3v and 4.3/4.4 GHz and see how it goes.

 

[cbliss1000]

what are you increasing the CPU voltage to? MOST cpus should be able to do 4.3-4.6 at 1.3v.

If you haven't gotten that high yet, I would just skip all the other lower clocks and set it for 1.3v and 4.3/4.4 GHz and see how it goes.

Yep, that was the first thing I did out of the box. 43/44 and a 1.3v vCore did not work. Neither does anything else except stock (I may have been able to get a 34 working with auto vCore, but I wouldn't have stress tested for long with such a small OC).

 

[cbliss1000]

(Deleted)

 

[Forsaken1]

You may be missing this



system agent offset set to .1 volts

cpu integrated faults VR set to disabled

cpu integrated faults VR efficiency mode set to disabled

cpu ratio set to all core

All core set to xx

CPU input voltage 1.9v-1.95v

 

[LoBoxJa]

Are you overclocking your memory at all? You might have to increase the power to the memory controller too

 

[LigTasm]

ASRock seems to have a bug with their X99 boards, I found on the last one I had I could not use any multipier above the highest turbo bin (for my 5930k it was x38). It would crash no matter what I did. However, it was perfectly happy at 125x36 for 4500mhz with only 1.330V. Try seeing if it crashes when you go one bin above the max turbo bin, and then try with a higher boot strap (don't forget this changes RAM speed and cache as well).

I've also seen recommendations to load the 4ghz profile and tweak that from people using the OC Formula board.

 

[cbliss1000]

You may be missing this



system agent offset set to .1 volts

cpu integrated faults VR set to disabled

cpu integrated faults VR efficiency mode set to disabled

cpu ratio set to all core

All core set to xx

CPU input voltage 1.9v-1.95v

Tried this at 40, no dice (and are you over on overclockers too? I posted there as well and received similarly specific advice). Thanks though.

ASRock seems to have a bug with their X99 boards, I found on the last one I had I could not use any multipier above the highest turbo bin (for my 5930k it was x38). It would crash no matter what I did. However, it was perfectly happy at 125x36 for 4500mhz with only 1.330V. Try seeing if it crashes when you go one bin above the max turbo bin, and then try with a higher boot strap (don't forget this changes RAM speed and cache as well).

I've also seen recommendations to load the 4ghz profile and tweak that from people using the OC Formula board.

I'll look into this. I didn't have any luck with the 4ghz profile - it crashed SUPER quickly - but I can experiment with higher speed straps. I'm not particularly experienced though, and don't really know what to adjust for the RAM - I haven't touched that at all.

Are you overclocking your memory at all? You might have to increase the power to the memory controller too

Nope - I haven't touched it all. I had the correct DRAM XMP profile loaded for a while but someone else told me to remove it and not play with RAM until I have a stable clock.

 

[LigTasm]

Tried this at 40, no dice (and are you over on overclockers too? I posted there as well and received similarly specific advice). Thanks though.

I'll look into this. I didn't have any luck with the 4ghz profile - it crashed SUPER quickly - but I can experiment with higher speed straps. I'm not particularly experienced though, and don't really know what to adjust for the RAM - I haven't touched that at all.

Nope - I haven't touched it all. I had the correct DRAM XMP profile loaded for a while but someone else told me to remove it and not play with RAM until I have a stable clock.

I couldn't even get into Windows at x40 multi, no matter how much voltage I tried. It definitely sounds like you have the same bios bug I was seeing. The CPU I was using was known good for 4.5ghz+ on other boards so I knew the motherboard was the problem.

 

[cbliss1000]

I couldn't even get into Windows at x40 multi, no matter how much voltage I tried. It definitely sounds like you have the same bios bug I was seeing. The CPU I was using was known good for 4.5ghz+ on other boards so I knew the motherboard was the problem.

Huh interesting, yeah that sounds very similar. So you fixed it by adjusting the base clock instead?

 

[LigTasm]

Huh interesting, yeah that sounds very similar. So you fixed it by adjusting the base clock instead?

Yessir. For an easy test, do 125x32 which is 4000mhz. Set cache manually to whatever it defaults to, 3400 or 3500 or whatever. Also adjust your RAM to its normal speed, or let it be as close to the default 2133 as possible. If you can get into windows like that then you are having the same problem I did.

 

[cbliss1000]

Yessir. For an easy test, do 125x32 which is 4000mhz. Set cache manually to whatever it defaults to, 3400 or 3500 or whatever. Also adjust your RAM to its normal speed, or let it be as close to the default 2133 as possible. If you can get into windows like that then you are having the same problem I did.

WELL WELL WELL this is very encouraging. Using the voltage settings recommended by forsaken1, and using the base clock OC idea from Ligtasm, I've found something that hasn't failed AIDA for the last hour (still running). Having seen dozens of these fail over the last few days I think it needs to run for another few hours before I claim victory, BUT this is the longest any of my OC attempts on this particular CPU/mobo have lasted (including ridiculously small attempts).

So... yeah. Fingers crossed. I'll post back regardless...

 

[cbliss1000]

Update: it crashed after ~3 hours or so. On the bright side, that's much better than what it was doing for virtually all my previous attempts, which would crash within 5-45 minutes (changing only vCore and multiplier).

Not sure what the next step is. Could continue fiddling with the base clock to see if there's something the CPU prefers.

Does anyone have a good test to determine whether this is a CPU issue or mobo issue (or something else?).

 

[LigTasm]

What voltage settings are you on? Also note on the ASRock boards LLC level 4 is the only one that actually does anything.

 

[Suprfire]

Try dropping the cache ratio or keeping it around 35x to 40x

 

[cbliss1000]

What voltage settings are you on? Also note on the ASRock boards LLC level 4 is the only one that actually does anything.

Trying again w/ the following

vCore -- 1.3v
input -- 1.93v
system agent offset -- .1 volts
cpu integrated faults VR -- disabled
cpu integrated faults VR efficiency -- disabled

Currently shooting for 4225 mhz
base clock - 132
multiplier - 32

I realized during my last run that the cache may have been set incorrectly, so I returned it to stock (3300 mhz).

Thanks for your help btw - you've been very helpful.

Try dropping the cache ratio or keeping it around 35x to 40x

Yeah, good shout. I'll try this if the current test fails. Other thoughts welcome. Thanks!

 

[cbliss1000]

OK, last and hopefully final update on this: I believe I have it working stably at the settings described above. OC'ing w/ the base clock and not the multiplier was definitely the key - keeping everything else constant and changing to a multiplier OC results in near-immediate instability while stress testing. With a base clock OC I just completed ~4 hours of AIDA64, no sweat.

Thanks again to everyone here for your help

 

[LigTasm]

OK, last and hopefully final update on this: I believe I have it working stably at the settings described above. OC'ing w/ the base clock and not the multiplier was definitely the key - keeping everything else constant and changing to a multiplier OC results in near-immediate instability while stress testing. With a base clock OC I just completed ~4 hours of AIDA64, no sweat.

Thanks again to everyone here for your help

Check the BIOS update page once in a while, I'm assuming they'll definitely fix it at some point. Glad you at least got it settled for now.