6850K, 4.5Ghz Manual @ 1.32 Vcore Safe for 24/7?

[3cl1ps3]

I've been keeping my current CPU for a long time at 4.4 GHz at 1.29 Vcore but in the end I pushed for more and tried 4.5 which worked OK, fully stable at 1.334, but for gaming and in cases when it is not running at utilisation above 70-80% at 4.5GHz I am keeping it at 1.324. Gaming for example works great. My motherboard is a Rampage V Edition 10 with 32GB of RAM at 3.2 GHz run in XMP mode. CPU Cache voltage is at 1.15 (reaching 1.16 under load) at 36 multiplier (31 standard if I remember right). CPU input voltage (VCCIO) is manually set to not go above 1.8 and is mostly staying at 1.792 even at full and long-term load. Stable and working OK.

I was feeling comfortable having the CPU run at 1.29 (below 1.3 Vcore), but now since I am at fully manual mode and to be honest having some in moments hard time to properly fine tune it, I feel a little cautious about staying at 1.324 at all times. Anyone running their 6850K (Broadwell-E) chip at 4.5GHz 24/7 in manual mode? If so, at what settings? Anyone using adaptive voltage (Vcore) and if so, is it safe at load? It is not that I couldn't try adaptive mode but first do not really have the time to test and am also concerned the motherboard irrespective to the settings will push even higher voltages under load. I am running a 1080Ti factory overclocked edition and when gaming a few days ago at maximum game settings the PC rebooted and upon booting I will hear three or four motherboard beeps (1 long and 3 short) which from checking meant "no VGA" found. I turned the PC off the grid, waited for a while, and turned it back on, resulting in a proper operation but until I turned the PC off completely again, before Windows lock screen (and even in Windows for seconds) I can see a square blue artifact on the very top of the monitor which made me think the card is gone. So these fine-tuning re/sets/starts sometimes are not safe at all and may lead to damage to any components, especially around the edge of the silicon's sweet spots.

My CPU LLC (load-line calibration level) is 7 out of 11 if I remember (but Asus BIOS maximum recommended one for overclocking).

So, any suggestions? Will 24/7 running this CPU at almost 1.33 vcore (all cores synched) be OK? Shall I go for adaptive mode instead and is it recommended from personal experience? Don't get me wrong, most of the time computer is operating at 1.2GHz browsing, watching stuff, doing different work with real loads relatively rarely (10-15% of the time). So far the CPU never throttled or reached even 90 degrees Celsius, even once, including the CPU package. Most of the time CPU and CPU core temperatures stay well below 70-72 under load (like in their high 30's), up to maximum 80 for short time. But the cores run at almost 1.33, even at 1.2GHz and no load whatsoever...

Thanks for any suggestions.

 

[3cl1ps3]

After some tweaking today trying to reach lowest possible stable voltage settings, a few clock_watchdog and other BSODs I managed to get my CPU stable using dynamic mode. It is only a bit higher than the only manual mode but my current vcore across the cores now is 0.7/0.8V, randomly reaching 1.something and 1.33+ at full load. More or less all settings are similar to those at full manual mode and I am glad Asus gave us some fine control also in dynamic mode. Will see how it goes but for now I've got what I wanted.

 

[3cl1ps3]

Thanks everyone, this is why in the end I started this thread - there is not too much information around about adaptive mode and I already read quite a lot about dead chips on auto settings. darket, I will check system agent voltage although to be honest I think it must be within the limits specified. My RAM is Corsair and I am running it using its XMP profile @3200MHz applied in BIOS. Latest BIOS 2101.

Before going adaptive, I've read an Asus blog (their own) on overclocking Broadwell-E and they recommended adaptive mode for "any normal overclocking". I am not doing liquid nitrogen experiments so consider myself in this group. They also recommended not using offset mode and setting cache voltage manually (admitting the MoBos will mess things around), which I did to 1.15 for my 36 value. I can see it goes to 1.14 when idle to 1.66 under load. The RAM voltage - 1.35 default - also goes to about 1.66 under full load. Not really sure if this is too much but this is what the motherboard pushes to it. My BIOS setting is 1.35.

From what I know and read before what in first place kills Broadwell-E is cache voltage, especially if left on auto or otherwise uncontrolled (above 1.20) since this range of CPUs has a very sensitive/weak memory controller. Since going adaptive I've been only gradually pushing the entire system little by little looking at each single setting/temp as it goes and so far all of them seem to have been normal.

I can see 4.5GHz for my 6850K is already around its limits as a chip as both heat and these few-seconds only temperature spikes some 10+ degrees higher are quite scary and much more frequent. From what I read most 6850K can hardly reach 4.4 at much higher voltage with many left below. 4.5 and 4.6 when these can be reached at all already require some crazy Vcore with custom loops and the performance for the complications involved is not worth it (at least for me). I will keep this system for as long as possible and eventually use it as a second workstation/gaming machine as upgrading to another still 14 nM Intel CPU will be frankly a waste of money. I was considering AMD for a newer machine but there are still many complaints about them being unable to overclock enough, many motherboard and RAM issues, and last but no least, I always use discrete audio cards being an audiophile and currently B-series AMD motherboards (all) have problems with some of them - for example Xonar Essence STX II which I have. Asus apparently tried with different drivers but it didn't work across the range and now they officially state on their website that no AMD B-series motherboards, including the high-end ones, are supported.

Not sure however how many years Intel will need to even go to their first gen 10nM and if it will be worth it as much as my Broadwell-E with its Spectre and Meltdown microcode disaster. Glad I still get motherboard updates as otherwise I will need to amend system32 files and edit the registries.

 

[3cl1ps3]

Core voltage with broadwell-e isn't the issue , what degrades and kills these chips is Xmp , and specifically the SA (system agent voltage).

For some reason X99 boards love to overvolt the crap out of the sytem agent voltage when either overclocking the ram or using Xmp ,there is no need for this to be higher than 1 volt and generally in the range of 0.8 - 1.0 v for memory overclock tuning, anything higher does NOTHING for stability. I cannot stress enough to NOT leave this set to auto and if using Xmp check the voltage settings after you have applied the xmp profile.

For work engagements I didn't have time to look at VCCSA before. It was moving in between 1.23 and 1.24 and since the adaptive mode doesn't allow manual setting (it requires full manual mode) I used negative offset, initially of 0.3, then of 0.2 as with 0.3 the VCCSA went to low 0.8-something. With 0.2 offset it is now at 0.848-0.856. I am not sure if this will be too low, especially under load and as some oveclocking BE guides advise for 1.2+ VCCSA if memory speed is above 3GHz (will test), but you were right that X99 boards seem to be overvolting the VCCSA. From what I explored meanwhile, 1.3V is the absolute maximum (Asus themselves are not that extreme but advise not to go too far away from 1.30).

Thank you for the advice and I will reduce the offset if for any reason under load I get crashes.