Overclock.net banner
4401 - 4420 of 4920 Posts

·
Facepalm
Joined
·
10,019 Posts
Discussion Starter · #4,401 ·
I tried increasing voltage didn't fix, atm I use LLC6 with 1.20v in BIOS resulting in 1.14v-1.15v in full load, you think I should increase to LLC7? But I would like the same load voltage so I would have to reduce the vcore in BIOS a bit to have the same 1.14v at load, definitely reducing ring ratio and increasing CPU vcore didn't fix. I have a big stuttering when the error occur while in game but I can still play fine for a long time before the 2nd WHEA error. It doesn't crash the game or anything, just a stutter.
No you don't decrease the voltage. Just increase the LLC and keep the voltage the same. And you probably didn't increase the vcore enough to remove the error. If you're at the edge, sometimes it requires a substantial increase. So it's better to just raise the LLC and keep the voltage constant.

You can't have your cake and eat it too. If you want to remove the parity errors, you have to pay the entrance fee. Or lower the overclock.
 

·
Registered
Joined
·
650 Posts
I tried increasing voltage didn't fix, atm I use LLC6 with 1.20v in BIOS resulting in 1.14v-1.15v in full load, you think I should increase to LLC7? But I would like the same load voltage so I would have to reduce the vcore in BIOS a bit to have the same 1.14v at load, definitely reducing ring ratio and increasing CPU vcore didn't fix. I have a big stuttering when the error occur while in game but I can still play fine for a long time before the 2nd WHEA error. It doesn't crash the game or anything, just a stutter.
That's very low voltage. Are you setting that as manual override? What is the CPU clock speed?
 
  • Rep+
Reactions: fray_bentos

·
Registered
Joined
·
1,220 Posts
No you don't decrease the voltage. Just increase the LLC and keep the voltage the same. And you probably didn't increase the vcore enough to remove the error. If you're at the edge, sometimes it requires a substantial increase. So it's better to just raise the LLC and keep the voltage constant.

You can't have your cake and eat it too. If you want to remove the parity errors, you have to pay the entrance fee. Or lower the overclock.
But my CPU stable overclock for 5.0/5.0Ghz and 4.7Ghz cache is 1.17v, I increased it to 1.20 no success, and then increase to 1.25v and reduced cache at the same time to 43x, no success again, I mean it's almost 0.1v increase and still getting errors.
 

·
Registered
Joined
·
1,220 Posts

·
Facepalm
Joined
·
10,019 Posts
Discussion Starter · #4,405 ·

·
Registered
Joined
·
1,220 Posts
Try raising VCCSA or VCCIO (or both)
Or maybe you simply aren't as stable as you think.
Hmm I will try that, I mean the OC is stable through Prime95 112k (RAM/CPU), RealBench 2.56 and Aida64, also RAM OC stable through all the stress tests.
 

·
Registered
Joined
·
351 Posts
Overclocking my 63 score CPU. Help if possible. I just re applied thermal compound and moves my h100i v2 rad to the front as an intake rather than at the top as an exhaust. I'll include a link to my spreadsheet detailing the things I have tried. As well as pics of the f steps and what not.

As you can see from my spreadsheet so farm my best cinebench r23 score has been on auto everything. I mostly use this beast for gaming and x265 encoding.

Spreadsheet, my 5.2 vid idle is at the bottom of the spreadsheet

Currently I am at 1.26 vcore bios, and cores are set (2x50)(6x49)(10x48)(10xA) , though I don't think I ever see 2 of them at 5.0ghz, not sure how that thing works. In the spreadsheet you can see my stable voltage at 4.9, might be better now that I redid the cooling.

Looking for some help and suggestions.
This is an underclock compared to stock. Stock should hit 49x10.
 

·
Registered
Joined
·
351 Posts

·
Registered
Joined
·
405 Posts
Hi,

Do you know if rog strix z490-g gaming BIOS has the octvb menu?
 

·
Registered
Joined
·
1,220 Posts
This is the 10900K thread.
Lol I know but the other thread is dead and 10900k has the same problem sometimes, so why not ask here for help :)
 

·
Optimal Pessimist
Joined
·
3,174 Posts
Lol I know but the other thread is dead and 10900k has the same problem sometimes, so why not ask here for help :)
It is more like a 10700k. 😉
 

·
Registered
Joined
·
3,810 Posts
I'm still having a hard time getting advanced adaptive to work properly lol.

The difference between light and heavy threaded workloads voltage wise is way too big and no matter what I change in terms of LLC levels or AC/DC it will not get any closer together. On x52 sync all core with x46 cache I run AC/DC "1" and LLC4 with -0.150 adaptive curve at x52 and the rest of the curve "Auto".
  • This gives me a BIOS voltage of 1.308v.
  • Windows idle 1.344v.
  • Prime95 Small FFT AVX On 20 threads 1.372v.
  • Prime95 Small FFT AVX On 5 threads 1.384v.
  • Prime95 Small FFT AVX Off 20 threads 1.344v.
  • Prime95 Small FFT AVX Off 5 threads 1.328v? (This one is weird, should be higher)
  • Gaming anywhere from 1.356v in heavy games to 1.308v in light stuff like RuneScape or just browsing which is way too low and unstable and causes random freezes, BSOD's and sometimes even random hard reboots. I need a minimum of 1.334v for 5.2 non-AVX to be stable.

When I set the same values through XTU, so Auto curve with -0.150 on x52 with the same AC/DC and LLC values and sync all cores x52 then it stays at 1.344-1.356v in any test. AVX doesn't overshoot 0.04v, light threaded loads don't drop like a brick, it just.. works. No crashes no hard reboots.. can even run higher cache freq stable. Problem is I want a BIOS level OC, not XTU / Windows.

Power limits are on "water cooled" in the BIOS a.k.a. 4096w and infinite time, CPU is direct die liquid metal cooled under water, worst-case Prime95 Small FFT AVX On 20 threads temps high 70's low 80's at 340-355w power draw. Only very short bursts to test cause I only wanna see how voltage acts, not fry the CPU.
 

·
Registered
Joined
·
650 Posts
@Imprezzion What is the "curve" on your motherboard? Is it the CPU internal VID? Increasing AC LL will give the CPU more voltage. I don't know how AC LL values work on your board but on my Asus XIII I am currently running AC LL 0.08 for 51x sync all cores. I would try to leave your curve on auto, set adaptive voltage to auto, try LLC 5 and adjust the AC LL for voltage stability. Do you have Speedstep turned off? Your idle should drop under 1.000v with adaptive + Speedstep/Speedshift.
 

·
Registered
Joined
·
592 Posts
I had issues getting 52x to work as well. As soon as I went down to 51x, where the VID table actually makes sense, it worked perfectly.
 

·
Registered
Joined
·
351 Posts
Yeah 4.9 x10 at 1.265 is stable but hits 100c so it gets a lower score in Aida compared to everything on auto.
There is something seriously wrong with your cooler/mount if that's the case. What is you actual Vcore reported in HWiNFO64 under heavy load? I suspect you are talking about set BIOS voltage and have an aggressive LLC setting meaning your load voltage is much higher than 1.265 V. If not, and 1.265 V is your actual load voltage, then there is something not right on the mechanical/cooling side.
 
  • Rep+
Reactions: Nizzen

·
Optimal Pessimist
Joined
·
3,174 Posts
@Imprezzion What is the "curve" on your motherboard? Is it the CPU internal VID? Increasing AC LL will give the CPU more voltage. I don't know how AC LL values work on your board but on my Asus XIII I am currently running AC LL 0.08 for 51x sync all cores. I would try to leave your curve on auto, set adaptive voltage to auto, try LLC 5 and adjust the AC LL for voltage stability. Do you have Speedstep turned off? Your idle should drop under 1.000v with adaptive + Speedstep/Speedshift.
MSI is like Gigabyte and AC/DC loadline units are in 1/100 mohm. So MSI ACLL = 1 = .01 mohm = minimum you can have.

I take it @Imprezzion voltage is higher than the BIOS because it is reports socket sense and not die sense .

Intel provides a boost for AVX high loads because it needs more than non-avx loads. So I expect the results are explained by AVX boost increase and decrease on high loads due to the LLC. But that Non-AVX anomaly IDK. Do you know what mohm LLC4 loadline corresponds to on this board?

For XTU, are you sure it is not dropping into fixed voltage mode when you make changes? Sure you are in adaptive mode and have speedstep or spendthrift enabled? Never did trust XTU.

EDIT: just tried small FFT noavx with 20 threads vs 5 threads on mine and the 20 thread is at a lower voltage than the 5 thread as you'd expect.
 

·
Registered
Joined
·
3,810 Posts
MSI is like Gigabyte and AC/DC loadline units are in 1/100 mohm. So MSI ACLL = 1 = .01 mohm = minimum you can have.

I take it @Imprezzion voltage is higher than the BIOS because it is reports socket sense and not die sense .

Intel provides a boost for AVX high loads because it needs more than non-avx loads. So I expect the results are explained by AVX boost increase and decrease on high loads due to the LLC. But that Non-AVX anomaly IDK. Do you know what mohm LLC4 loadline corresponds to on this board?

For XTU, are you sure it is not dropping into fixed voltage mode when you make changes? Sure you are in adaptive mode and have speedstep or spendthrift enabled? Never did trust XTU.

EDIT: just tried small FFT noavx with 20 threads vs 5 threads on mine and the 20 thread is at a lower voltage than the 5 thread as you'd expect.
XTU stays on V/F curve it just skips the AVX boost all together and doesn't add any AVX voltage.

I have speed shift + speed step enabled. It runs at about 0.766v idle @ 800Mhz.

The advanced offset option on MSI is indeed CPU internal VID curve which you can offset in the BIOS per multiplier (x8 x16 x25 asf.)

It's weird how the BIOS shows 1.308v but it normally shows 1.334v in windows (if I force 100% CPU idle) but it does drop to 1.308v under certain non AVX loads. The downside is that I know it will handle AVX at 1.344v. That is what I need for it to be 100% stable in any load when using Fixed voltage. There is no real way to make it run 1.344v under load on any adaptive mode using BIOS. It always goes all over the place with the AC/DC and AVX voltages.

I mean, if I set it up so that 1.344v is the lowest observed voltage it will go as high as 1.408v under AVX loads (like BFV). If I set it so that 1.344v is the highest load observed voltage the lowest voltages observed are 1.272v ish in single core non AVX loads which is nowhere near stable. So I basically always have to give it way more voltage then it actually needs due to all the random added voltages.

It's not that big of a deal as even 1.408v is fine for my cooling (barely).

I am now at 5.1/4.7 advanced offset -0.160, AC/DC 1, LLC4. It's better, but not great. I know from testing with fixed voltages I need 1.290v to be 100% stable in any load including AVX. I set it so that minimal non AVX voltage is 1.289v. the highest I've seen it go with a all core AVX load is 1.332v which is.. acceptable. And at least no random idle crashes now even on 47 cache..
 

·
Optimal Pessimist
Joined
·
3,174 Posts
I believe MSI doesn't report die-sense core voltage, it reports socket sense which will be higher than the diew-sense voltage, and hence the reported vcore will be higher than the actual vcore. This difference of about 40 mv is about the right ballpark. It drops under load because of droop from the LLC by Vcore = BIOS set votlage - LLC*Current for non-avx load. AVX loads are a bit of a mystery to me. Not a simple boost I think.

There is this though. A light AVX load like 1344/1344 FFT will result in a larger voltage because of the AVX boost and lower current, than a heavy AVX load because of the larger current draw and hence a much larger Vdroop.

EDIT: Do you have TVB Voltage optimization enabled? That will drop the vcore more.
 

·
Registered
Joined
·
3,810 Posts
I just wish there was a way to just have speed step + speed shift and adaptive voltages but with a load voltage that is always what I set it to be.

This can be done with XTU but I don't want software to control it as it relies on the host OS and thus more software running, more overhead and more chance of it not applying properly or not working in VM's / remote desktops and such.

I mean, I could just run fixed voltage and idle at the full voltage but why would I want the extra temperature and power consumption when watching some YouTube or just listening to some music while doing something else or even when playing stuff like RuneScape or whatever.

I'll just stay on 5.1/4.7 for a while as it hasn't done anything weird this evening at least and voltages and temps seem to be kind of reasonable with a minimum reported of 1.289v (RuneScape + Chrome / YouTube) and a maximum of 1.322v (BFV Multiplayer with DXR + Ray Tracing Ultra so using AVX).

It's a shame as I can get away with running 5.3 all core with 48 cache at 1.390v fixed voltage but there's no way I can do that with adaptive as it will overshoot to the moon to like 1.448v in some AVX loads and while I can cool 1.390v AVX doing the same on 1.448v is a whole nother problem.. setting it so that it loads AVX at 1.390-1.404v means the non-AVX voltage is as low as 1.356v which is nowhere near stable and causes loads of internal / L0 errors or even hard crashes.. it's almost like this specific CPU needs as much voltage for non AVX workloads as it needs for AVX. I did test this shortly with fixed voltages at 5.3 and for AVX I need 1.390v and for non AVX I need a minimum of 1.372v to be remotely stable. Same goes for 5.2. AVX is 1.344v, non AVX is 1.330v.

I do kinda wanna switch to a Z490/590 Apex only the price just isn't justifiable with how small the real world gains are over the Ace.
 
4401 - 4420 of 4920 Posts
Top