Discussion in 'Overclocking & Hardware' started by ledhead900, Feb 18, 2019.
UEFI Pics Here - Click to view UEFI config
1.36v is roughly accurate for that speed and CPU.
Test it and see if its stable.
1.29v (under load it would reach 1.33v and would happily sit in windows and run X4 Foundations but it did crash eventually when trying to boot up GTA V, so I knew right away it has potential but needs fine-tuning.
I lowered it to try to conserve heat
This thing loves getting into the 75c under load, I am not sure if the block needs to be re-mounted I used the pea method for thermal compound (arctic silver 5) I am not sure what the baseline for heat should be under no voltage adjustment as it appears to sit into the high 60s even on stock settings though this is Sandy-B we are talking about and it does have more cores than my old one.
Currently, I have both the H100iv2 fans ramped up into the 2500rpm and the core package is at 45-48c at idle (Vcore 1.35v ish)
If you have any for-thought advice on what to mess with or some rough values to throw at it let me know. I made a new baseline profile in bios for testing will take some pics for you guys.
'OP Updated with additional info and new clock speed'
For some reason tho the multiplier refused to take 100%, If I set the multiplier to 47, restart back into the bios it reads 4700mhz, I boot into windows and it will read 4.6ghz in CPUz and HWmonitor.
I then restart back into the bios and it now reads 4600mhz, although the multiplier has not changed. It seems to do this for values over 46, Turbo On/Off seems to make no difference.
Could be a Bios firmware bug or could be some sort of voltage or current protection I have not yet adjusted.
Had another crack at this with the Vcore ramped up too 1.44v, I am still unsure if the temps are in normal ranges for this CPU, or If should take the water block off and apply a slightly larger pea size drop of arctic silver 5, possibly even de-lid and re-do the compound on the die.
However, when running OCCT to test for stability between 2-5min in I get that dreaded quack quack sound and an error is detected (tho it is not telling me which core) something tells me I have a bit of Vdroop going on as monitoring software indicates a vcore of 1.43vi sh under load when clearly the BIOS is set too 1.47v and I don't know many volts this CPU will take before it explodes a little scared to push it up into 1.5v however every step I increase seems to increase the load voltage under operation in the monitoring software and I Know if Vcore would sit on 1.45-1.46 it it should happily be stable at 4.7/4.8ghz as all I did to get this stable leaving the Vcore in bios at 1.47v was drop the core back down to 4600mhz and now it passed OCCT for 60mins without error.
So below we have what is going on, I can get 4700mhz stable enough to bench but as you can also see the CPU voltage is not hitting the required voltage to become stable and I am trying to figure out what settings in BIOS I need to tweak to get this to use most of the Vcore set in bios under load without dropping down the voltage ( I already have all the power saving steps disabled)
Again the cooler used is an H100iv2 only this time the fan duty is ramped up too 70%
I think the problem above is Vdrop and LLC is not accounting for enough at 90% LLC. 4.6ghz at Bios Vcore of 1.46v and it was pretty stable and happy to run for 50mins no errors using OCCT - with large data set, however attempting 4.7ghz OCCT detected an error usually within 2-6min into the run.
Update 1 - 12:58pm
Vdroop under load is bouncing around 1.425v-1.44v (notes: may need to increase CPU LLC to 100%
14min into the 50min OCCT run @ 4.7ghz longest I have seen w/o (quack quack quack) so far albeit Bios Vcore was pushed a little higher than I would like for 24/7 use, Monitoring reports a pretty severe Vdroop I suspect still as under load it is only doing 1.428-1.440v That is with CPU LLC at 90%
Side thoughts -
This Xeon seems to yoyo the temps shockingly
I am going to re-mount the block with a tad bit more thermal compound this time and at the same time replace a cooling fan which seems to have a bearing issue and rattles. (might need to de-lid the Xeon????)
Update 2 - 1:14pm
20mins into OCCT stress test, longest I have seen yet, Temps are a bit of an issue tho-
Running on water with H100iV2 @ a fixed 75% duty cycle and pump speed at max.
Starting to think I need to do something about the thermal contact with this CPU.
Update 3 - 1:42pm
OCCT 50min test completed w/o errors.
Vcore is in a range I would want it to be but Vdroop I think is a pretty big issue for this MOBO or the BIOS revision one of the two.
Temps are a little crazy for my liking but such is life with Sandy-B Tj max of 91c under normal operation it should be fairly safe but will try to lower voltage anyway, appears de-lid is kinda pointless because they solder these buggers on although I have read cases where the solder was not evenly distributed causing fluctuation issues between one side of the die and the other I have no way to knowing if this causing the cores to have such varied degrees of temp difference but I am not about to take the lid of too find out as solder is better than paste.
Seems to be stable for roughly 60mins of abuse will just have to play games and use as normal if anything is all good will start to lower Vcore in increments of 0.05 and run a 20min test.