New
#461
Ok for anyone interested, I just downloaded and tried IntelBurnTest(for the first time ever), and all tests, EXCEPT Maximum, it would jump to 4.9 often, but not stay there, on Maximum, it started out at 4900, but down clocked to between 4.0 and 4.4, so my system is stable, but the CPU down clocks to protect my system, I could turn of Intel Adaptive Thermal Monitor, but then chance frying my CPU, no way in heck I want to chance that, just to stress my PC to a point I will never reach in day to day stuff.
I did just use my day to day settings in BIOS for the test as it's getting late here, maybe tomorrow I'll try with enhanced turbo boost enabled, instead of set to auto, shut off C-State, set Vcore from adaptive to override, and CPU Ratio mode from Dynamic to Fixed, which will lock in my 4.9 OC(normally) and if it still starts to down clock, then I know it's the thermal monitoring kicking in even before the temps get above 72°C(that's where it stayed in IntelBurnTest at Maximum).
Like I said, it's a good thing as it protects the CPU, and I believe they started implementing it with the Skylake, so anyone with those or newer(Kaby and future ones) might know what to look at, and at least you shouldn't get a BSOD or system freeze at the worst time(unless benchmarking)
This morning I tried a few thing in BIOS that didn't work or help, so decided to try lowering Vcore as far as I could, and still us the 49x multiplier.
I was able to drop my Vcore from 1.450 to 1.420 and still be able to not only boot, but also load all my startup programs.
I just ran IBT and it peaked at 4.9GHz more often than it did when Vcore was set to 1.450, so my problem definitely is CPU self preservation, and here are the results for Very High:
Now to test with P95 and then OCCT, then try to see if CPU Mark and Catzilla remain stable(they are the hardest to please when it comes to the CPU)----------------------------
IntelBurnTest v2.54
Created by AgentGOD
----------------------------
Processor: Intel(R) Core(TM) i7-6700K CPU @ 4.00GHz
Clock Speed: 4.22 GHz
Active Physical Cores: 8
Total System Memory: 32731 MB
Stress Level: Very High (4096 MB)
Testing started on 18-Mar-17 06:06:51
Time (s) Speed (GFlops) Result
[06:08:11] 59.880 128.7185 3.414758e-002
[06:09:30] 60.121 128.2040 3.414758e-002
[06:10:50] 60.331 127.7577 3.414758e-002
[06:12:10] 60.461 127.4825 3.414758e-002
[06:13:30] 60.633 127.1200 3.414758e-002
[06:14:52] 62.393 123.5357 3.414758e-002
[06:16:11] 60.175 128.0890 3.414758e-002
[06:17:31] 60.058 128.3377 3.414758e-002
[06:18:50] 59.996 128.4710 3.414758e-002
[06:20:09] 60.006 128.4489 3.414758e-002
Testing ended on 18-Mar-17 06:20:10
Test Result: Success.
----------------------------
While running Prime95, I just noticed that it downclocks to 4.4GHz:
Which is exactly the setting MSI has chosen for OC Genie when a 6700K is installed(if on decides to use it):
System is stable so far, just the OC is not, I guess, I'll just have to live with the fact, that, under extremely heavy loads, it'll down clock, which means I have reached my limit on benchmarking when it comes to max CPU(for example Physics testing).
I do not want to mess with the thermal, current, & power limits in BIOS and cause a house fire
Very probable cause...
VDroop because as voltage drops, almost immediately after stress test start of course the OC multiplier goes down too:
What is VDroop?
vDroop - the difference between the idle and load voltage.. Generally, you get a higher voltage when idling than when at full load. That's because it works as a self-protection mechanism to prevent too much voltage surging into the chip and frying it, or not giving it enough power.
For example, if your chip needs can take a voltage range of 1.35v - 1.55v, you might find that your idle voltage is around 1.50v. When you put it under full load, it'd drop to 1.47v. That's vDroop.
The board gives it extra voltage at idle so that when the chip goes into full load it can handle the negative voltage drop without going below the minimum 1.35v.
Likewise, the voltage upshoot from from full load to idle is protected by vOffset. That's the difference between your idle voltage and the maximum the chip can handle. You might find that when switching from full load to idle the voltage goes over the 1.50v and then comes back down to settle at 1.50v. That's to give it some room up top away from the 1.55v maximum.
Images to help explain from this article on Anandtech.
With vOffset:
Without vOffset:
Last edited by Cliff S; 25 Mar 2017 at 02:10.
Man, I don't like it when a single graphic card takes up so many (3) PCI slots. I remember way back I bought two ATI 7970's. I asked the retailer - does the card take up 2 slots - yes. I needed slots 1 and 3 to run crossfire 16x. Returned them for Asus 770's and said to the retailer - what are you talking about - two slots? He said, well 2 1/2 slots.
What are you supposed to do with 1/2 slot?
MSI GeForce GTX 1080 Ti Gaming X Makes a Camera Appearance
would water blocks take more, or less space?
EKWB: TITAN X Pascal waterblocks compatible with GeForce GTX 1080 Ti | VideoCardz.com
EK Water Blocks, the world renowned premium liquid cooling manufacturer, is pleased to announce that the existing EK-FC Titan X Pascal series water blocks are compatible with the GeForce® GTX 1080 Ti, NVIDIA’s new flagship gaming GPU.
In order to simplify the search for suitable and compatible water block EK has already added NVIDIA® GeForce® GTX 1080 Ti graphics card to EK Cooling Configurator database.
Oh, by the way GPU-Z has an update to V. 1.18.0
Version History
v1.18.0 (March 15th, 2017)
- Added support for NVIDIA GeForce GTX 1080 Ti, Quadro P4000, P1000, P600, P400
- Added support for AMD Radeon RX 580, RX 570, RX 560, RX 550, HD 6430M
- Added support for Intel HD Graphics 500 (Apollo Lake Pentium N4200)
- Update check will now display the latest version number available
- Fixed error message display during BIOS upload
- Improved tray icon creation code when GPU-Z is launched during Windows startup
- Fixed clocks read as -1 on Pascal cards with overclocking disabled
- Invalid readings are no longer included in sensor average calculation (avoids -1.$ output)
- Addded support to display Boost clocks on Intel