Overclock.net › Forums › Intel › Intel CPUs › Kaby Lake Overclocking Guide [With Statistics]
New Posts  All Forums:Forum Nav:

Kaby Lake Overclocking Guide [With Statistics] - Page 210

post #2091 of 2872
+rep
post #2092 of 2872
Quote:
Originally Posted by skingun View Post

If the CPU could potentially request this, and referring to note 7 in the table, how could it not be safe? Please excuse me ignorance.
As I stated in an earlier post quoted from Asus O/C guide "Now, if you happen to be the type of user that spends more time running Prime95 than using a PC for other tasks, then we advise you reduce the maximum Vcore. “By how much?”, you ask. Well, you’re on your own for that. Remember, it’s current that degrades or kills a CPU. Be mindful of how much load you’re placing on the chip long-term and act accordingly. There’s nothing worse than pushing insane levels of current through the die and then moaning when the there’s degradation"
post #2093 of 2872
While I cannot speak for others, my personal situation does not utilise heavy AVX loads. I just game and browse the internet.
post #2094 of 2872
Quote:
Originally Posted by garyd9 View Post

Exactly... the VID is set in the CPU. It's set by Intel as a "stable" voltage for the given frequency. No, it can't be changed. We're on the same page with this.

So if, for a given frequency of 4400, my i7 has a VID of 1.17 and yours has a VID of 1.16, then wouldn't that indicate that Intel, when they set the VID's, felt that your chip required less voltage for 4400 when compared to mine?

The reason for using 4400 is that it's much easier to cause all 4 cores to go to 4400 than it is to get 1 core to 4500 (when using default intel settings for the chip.)
This doesn't agree with my own testing even while I type this. I have HWInfo open (which reports a VID for each of the 4 cores) and the only "stress" on the machine is me moving the browser window around. That causes all 4 cores to pop to a 44x multiplier, and VID on all 4 of them to 1.170.

I stop moving the browser window, and all 4 cores drop to a 8x multiplier (800MHz) and the VID drops to 0.800 for each of them. (There are occasional spikes for single cores as Windows does stuff in the background.)

Now I start 8 threads of linpack (via linx.exe) and the cores jump back to 44x (4400 MHz) with a VID of 1.170. Linpack starts by getting the memory buffers populated... during that time, my temps jump to 50C. After a few seconds, linpack starts actually processing AVX instructions and my temps jump to ~80C, but the frequency (multiplier) and VID are still the same (4400 and 1.170.) Obviously, while processing the AVX instructions, the load is much greater, but the VID isn't.

I'll attach a screen snippet (capture.png) showing the pattern. In the "Clock" graph you can see that the core jumps to and stays at 4400. As well, the VID jumps to and stays at 1.17 (or 1.169), but the temperature graph shows obvious differences of load...

I don't think you are doing the testing correctly. Do you clear HWiNFO64 in between testing, also what you are testing might be the same load for voltage.

Here is two tests to prove my point correctly. VID Prime95 Value 1.239v Maximum 1.277v RealBench Value 1.176 Maximum 1.220v.


post #2095 of 2872
Quote:
Originally Posted by scracy View Post

As I stated in an earlier post quoted from Asus O/C guide "Now, if you happen to be the type of user that spends more time running Prime95 than using a PC for other tasks, then we advise you reduce the maximum Vcore. “By how much?”, you ask. Well, you’re on your own for that. Remember, it’s current that degrades or kills a CPU. Be mindful of how much load you’re placing on the chip long-term and act accordingly. There’s nothing worse than pushing insane levels of current through the die and then moaning when the there’s degradation"

I've lost count of how many times I have read that guide wink.gif
post #2096 of 2872
Quote:
Originally Posted by skingun View Post

The Intel datasheet for the 7th Gen Intel® Processor Family for S Platforms suggests a maximum operating voltage of 1.52 for a 7700k. See Table 7-2 pg.144

http://www.intel.com/content/www/us/en/processors/core/7th-gen-core-family-desktop-s-processor-lines-datasheet-vol-1.html

Unless I am missing something, provide temps are in check and the measurement of vcore is 'real', it is safe to use up to 1.52V thumb.gif

that 1.52V max vcore requires that other conditions are met (see the foot notes). These are current(amperage/TDP), and temperature, etc. ... so, I'm not sayin you shouldn't run 1.52V, just be aware that the value is not to be taken alone when thinking 24/7 and durability.

Edit - the VID request bin depends on frequency AND load, not only on frequency and when a CPU experiences load change the VID request can pull a higher value from the table momentarily.
Edited by Jpmboy - 4/22/17 at 5:14pm
x99
(28 items)
 
x79
(21 items)
 
z170/z270
(27 items)
 
CPUCPUMotherboardGraphics
i7 5960X i7 6950X Asus Rampage V Extreme (STD and model-10) 2 GTX TitanX w/ EK Blocks & Backplates 
GraphicsGraphicsRAMHard Drive
Nvidia Tesla C-class 2 GTX Titan X Pascal (water cooled) GSkill 3200c14 TZs 8x8GB @ 3400c13 Intel 750 NVMe 400GB 
Hard DriveHard DriveHard DriveOptical Drive
Plextor 256GB M6e M.2 2x Plextor SSD 256 Raid 0 (Win 7) Samsung NVMe 950 Pro M.2 Plextor 810 
CoolingCoolingCoolingCooling
Aquacomputer GiGant 1680 (4x420) external Rad XSPC RX360 Koolance 380i Laing DDC-1T (2) 
CoolingOSOSOS
Koolance EXC-800 Chiller Windows 10 x64  Windows 7 Pro Linux Mint 
MonitorMonitorKeyboardPower
ASUS Swift 1440P 144Hz Seiki 50" 4K monitor (720P-2160i) 240-30Hz Das Keyboard Model S Pro Corsair AX1500i 
CaseMouseAudioOther
MicroCool Banchetto 101 Steelseries Rival Ultimate Ears TF10 Koolance Pump/Fan Controller 
CPUMotherboardGraphicsRAM
i7 4960X @ 4.7 Asus Rampage IV Black Edition R9 295x2 w/ Koolance Block G Skill 2400 c9 (@ 2133c8) 
Hard DriveHard DriveHard DriveHard Drive
Plextor 512 SSD VelociRaptor 600GB WD Black 2T 4 Terabyte NAS 
Optical DriveOptical DriveCoolingCooling
Asus BluRay R/W Lite On  AquaComputer 720XT Koolance 380i 
CoolingOSMonitorKeyboard
Liang DCC Aux Pump Win10x64 Samsung U28D590 4K/60 (primary) SteelSeries G6 (Cherry Red) 
PowerCaseMouseAudio
Corsair AX1200 Sliverstone TJ09 G700 on-board 
Audio
Ultimate Ears TF10s 
CPUCPUCPUMotherboard
i5 6600K i7 7700K i3 7350 ASUS Max iX APEX, Asrock MOCF, M8Impact 
GraphicsGraphicsGraphicsRAM
ASUS GTX 1080 Turbo ASUS GTX 960 Mini EVGA GTX 550Ti G.Skill 3600c15 
RAMHard DriveHard DriveHard Drive
G.Skill 4266c19 Samsung SM951 WD VR 300 Intel 750 NVMe U.2 
Hard DriveHard DriveOptical DriveCooling
Mushkin 512 Intel 600 NVMe Raid0 Plextor DVD ROM Custom Water (Koolance 380i / XSPC 360v2) 
CoolingOSOSOS
Corsair H80iV2 (M8I) Windows 10 Pro Window 7 Linux Mint 
MonitorMonitorKeyboardPower
Seiki 4K AOC 1080P Ducky Cherry Black Silverstone 500W SFX form factor/ Seasonic 760W 
CaseMouseAudio
Bench Logitech G700s Ultimate Ears TF10s 
  hide details  
Reply
x99
(28 items)
 
x79
(21 items)
 
z170/z270
(27 items)
 
CPUCPUMotherboardGraphics
i7 5960X i7 6950X Asus Rampage V Extreme (STD and model-10) 2 GTX TitanX w/ EK Blocks & Backplates 
GraphicsGraphicsRAMHard Drive
Nvidia Tesla C-class 2 GTX Titan X Pascal (water cooled) GSkill 3200c14 TZs 8x8GB @ 3400c13 Intel 750 NVMe 400GB 
Hard DriveHard DriveHard DriveOptical Drive
Plextor 256GB M6e M.2 2x Plextor SSD 256 Raid 0 (Win 7) Samsung NVMe 950 Pro M.2 Plextor 810 
CoolingCoolingCoolingCooling
Aquacomputer GiGant 1680 (4x420) external Rad XSPC RX360 Koolance 380i Laing DDC-1T (2) 
CoolingOSOSOS
Koolance EXC-800 Chiller Windows 10 x64  Windows 7 Pro Linux Mint 
MonitorMonitorKeyboardPower
ASUS Swift 1440P 144Hz Seiki 50" 4K monitor (720P-2160i) 240-30Hz Das Keyboard Model S Pro Corsair AX1500i 
CaseMouseAudioOther
MicroCool Banchetto 101 Steelseries Rival Ultimate Ears TF10 Koolance Pump/Fan Controller 
CPUMotherboardGraphicsRAM
i7 4960X @ 4.7 Asus Rampage IV Black Edition R9 295x2 w/ Koolance Block G Skill 2400 c9 (@ 2133c8) 
Hard DriveHard DriveHard DriveHard Drive
Plextor 512 SSD VelociRaptor 600GB WD Black 2T 4 Terabyte NAS 
Optical DriveOptical DriveCoolingCooling
Asus BluRay R/W Lite On  AquaComputer 720XT Koolance 380i 
CoolingOSMonitorKeyboard
Liang DCC Aux Pump Win10x64 Samsung U28D590 4K/60 (primary) SteelSeries G6 (Cherry Red) 
PowerCaseMouseAudio
Corsair AX1200 Sliverstone TJ09 G700 on-board 
Audio
Ultimate Ears TF10s 
CPUCPUCPUMotherboard
i5 6600K i7 7700K i3 7350 ASUS Max iX APEX, Asrock MOCF, M8Impact 
GraphicsGraphicsGraphicsRAM
ASUS GTX 1080 Turbo ASUS GTX 960 Mini EVGA GTX 550Ti G.Skill 3600c15 
RAMHard DriveHard DriveHard Drive
G.Skill 4266c19 Samsung SM951 WD VR 300 Intel 750 NVMe U.2 
Hard DriveHard DriveOptical DriveCooling
Mushkin 512 Intel 600 NVMe Raid0 Plextor DVD ROM Custom Water (Koolance 380i / XSPC 360v2) 
CoolingOSOSOS
Corsair H80iV2 (M8I) Windows 10 Pro Window 7 Linux Mint 
MonitorMonitorKeyboardPower
Seiki 4K AOC 1080P Ducky Cherry Black Silverstone 500W SFX form factor/ Seasonic 760W 
CaseMouseAudio
Bench Logitech G700s Ultimate Ears TF10s 
  hide details  
Reply
post #2097 of 2872
Not an entry, just testng a chip. 1.4V in BIOS LLC5. wink.gif


Edited by rt123 - 4/22/17 at 6:37pm
post #2098 of 2872
Quote:
Originally Posted by wingman99 View Post

I don't think you are doing the testing correctly. Do you clear HWiNFO64 in between testing...
Not always unless I'm doing a long running test. For the screen shots I provided earlier in the thread, I just eyeballed the "current" column. It doesn't invalidate that the VID is going to 1.17 no matter what load is on the processor iff all 4 cores are running at 4400. Once the cores slow down, the VID drops.
Quote:
Originally Posted by wingman99 View Post

..., also what you are testing might be the same load for voltage.
Are you saying that moving a 'Chrome' browser window around the desktop on an otherwise idle machine uses the same load as running 8 threads of linpack? They both showed the same VID of 1.17...

Perhaps defining "load" might be a good idea. My definition for the purpose of this discussion is "how hard the processor is having to work at a given frequency." I usually guess the difference of two loads by comparing the "CPU Package Power" parameter in HWInfo (assuming both tests cause all cores to use a 44x multiplier.)

So, using linpack as an example again, on my processor, when linpack is loading up memory, 4 cores are at 4400, VCore is at 1.12, VID is at 1.17, and CPU power is at only 40 to 42 watts. When linpack is done loading up memory and actually starts chewing data with AVX instructions, the core speeds, vcore, and VID are still the same, but the CPU's package power fluctuates between 91 and 95 watts.

Obviously, there's a massively different load on the processor between the two.

(Oh, and when the machine is idle, moving the chrome browser window around causes the power to jump to 21W, with 4x cores at 4400, VID at 1.17, and VCore at...1.12)
Quote:
Originally Posted by wingman99 View Post

Here is two tests to prove my point correctly. VID Prime95 Value 1.239v Maximum 1.277v RealBench Value 1.176 Maximum 1.220v.
...
I don't doubt that you are seeing something different than I am. smile.gif The massive difference is that I'm running everything with stock intel settings (and even disabling the Asus crap that throws extra voltage at the CPU.) Perhaps if you were to record/save your current BIOS settings, reset them to defaults, turn off anything on your board that auto-OC's or tries to be "helpful" with voltages, and try again you might see results similar to mine with the VID staying the same when the cores are all at max (4400.)

I'm NOT saying that you're wrong and I'm right. However, I'm trying to demonstrate with the least common denominator of stock intel settings, while your contrasting that with overclocked settings. We might as well be speaking different languages..

It'd be considerably easier for you to match my current settings (which are just stock defaults for intel) than for me to try and match your settings (which would likely be impossible as we don't even have the same motherboard.)

Otherwise, you and I comparing things doesn't make much sense.

Remember, please, that my original suggestion was to publish VID's for loaded cores with STOCK INTEL settings.
    
CPUMotherboardGraphicsRAM
Intel i7-7700K Asus ROG Maximus IX Hero Gigabyte GeForce GTX 1070 G1 Gaming 4x 8GB Vengeance LPX Series 
Hard DriveCoolingOSMonitor
1TB Samsung EVO 960 NVME M.2 EKWB based custom loop: Supremacy EVO, SE360 R... Windows 10 Pro x64 Asus PB328Q 32" WQHD 
MonitorKeyboardPowerCase
Asus 25" 16:10 (forgot the model number) Ducky Shine 5 EVGA SuperNOVA 650 G2  Corsair Crystal Series 460X RGB 
MouseMouse Pad
Logitech G900 Corsair MM300 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
Intel i7-7700K Asus ROG Maximus IX Hero Gigabyte GeForce GTX 1070 G1 Gaming 4x 8GB Vengeance LPX Series 
Hard DriveCoolingOSMonitor
1TB Samsung EVO 960 NVME M.2 EKWB based custom loop: Supremacy EVO, SE360 R... Windows 10 Pro x64 Asus PB328Q 32" WQHD 
MonitorKeyboardPowerCase
Asus 25" 16:10 (forgot the model number) Ducky Shine 5 EVGA SuperNOVA 650 G2  Corsair Crystal Series 460X RGB 
MouseMouse Pad
Logitech G900 Corsair MM300 
  hide details  
Reply
post #2099 of 2872
Quote:
Originally Posted by garyd9 View Post

Not always unless I'm doing a long running test. For the screen shots I provided earlier in the thread, I just eyeballed the "current" column. It doesn't invalidate that the VID is going to 1.17 no matter what load is on the processor iff all 4 cores are running at 4400. Once the cores slow down, the VID drops.
Are you saying that moving a 'Chrome' browser window around the desktop on an otherwise idle machine uses the same load as running 8 threads of linpack? They both showed the same VID of 1.17...

Perhaps defining "load" might be a good idea. My definition for the purpose of this discussion is "how hard the processor is having to work at a given frequency." I usually guess the difference of two loads by comparing the "CPU Package Power" parameter in HWInfo (assuming both tests cause all cores to use a 44x multiplier.)

So, using linpack as an example again, on my processor, when linpack is loading up memory, 4 cores are at 4400, VCore is at 1.12, VID is at 1.17, and CPU power is at only 40 to 42 watts. When linpack is done loading up memory and actually starts chewing data with AVX instructions, the core speeds, vcore, and VID are still the same, but the CPU's package power fluctuates between 91 and 95 watts.

Obviously, there's a massively different load on the processor between the two.

(Oh, and when the machine is idle, moving the chrome browser window around causes the power to jump to 21W, with 4x cores at 4400, VID at 1.17, and VCore at...1.12)
I don't doubt that you are seeing something different than I am. smile.gif The massive difference is that I'm running everything with stock intel settings (and even disabling the Asus crap that throws extra voltage at the CPU.) Perhaps if you were to record/save your current BIOS settings, reset them to defaults, turn off anything on your board that auto-OC's or tries to be "helpful" with voltages, and try again you might see results similar to mine with the VID staying the same when the cores are all at max (4400.)

I'm NOT saying that you're wrong and I'm right. However, I'm trying to demonstrate with the least common denominator of stock intel settings, while your contrasting that with overclocked settings. We might as well be speaking different languages..

It'd be considerably easier for you to match my current settings (which are just stock defaults for intel) than for me to try and match your settings (which would likely be impossible as we don't even have the same motherboard.)

Otherwise, you and I comparing things doesn't make much sense.

Remember, please, that my original suggestion was to publish VID's for loaded cores with STOCK INTEL settings.
Why don't you do a test like I did and show your results using Prime95 and Realbench so you can see like I have done already many times.

If I test like you I have the same results moving windows around causing a full load for the VID,why do you think the multiplier goes up to 4.4GHz doing that. That kind of test is meaningless.
Edited by wingman99 - 4/22/17 at 7:50pm
post #2100 of 2872
Quote:
Originally Posted by wingman99 View Post

Why don't you do a test like I did and show your results using Prime95 and Realbench so you can see like I have done already many times.

If I test like you I have the same results moving windows around causing a full load for the VID,why do you think the multiplier goes up to 4.4GHz doing that. That kind of test is meaningless.
And comparing an overclocked processor to a stock processor isn't?

The ROG test had the lower VID for you, correct? Here's ROG "Stress" test set at the default 4GB running for a few minutes. I changed the graphing to show Core0 clock, Core0 VID and the CPU Package power.

The significant point here is that while ROG Stress was running, the package power never went over 61 watts, and.... wait for it... the VID was constant at 1.170 (+/- 0.001) the entire time that the cores were at 4400. (I stopped the test before taking the screen snippet. ROG stress makes user interface interactions jumpy for me.)



Here's a few minutes of linpack. Cores at 4400, VID constant at 1.17 (+/- 0.001), and package power going between around 41W and ~95W depending on what linpack is doing. (This is setting linpack to use 16GB of memory for a problem size of 46290.)




Does that satisfy you that I'm not clueless? Being that I performed the test you asked for (as close as I'm able to with what I have installed), how about you do the same for me? Load up intel stock defaults for your processor, run any stress test (ROG bench, prime95, whatever) and see what happens with VID while the cores are all at 4400. (If they aren't at 4400, you aren't stock.)
    
CPUMotherboardGraphicsRAM
Intel i7-7700K Asus ROG Maximus IX Hero Gigabyte GeForce GTX 1070 G1 Gaming 4x 8GB Vengeance LPX Series 
Hard DriveCoolingOSMonitor
1TB Samsung EVO 960 NVME M.2 EKWB based custom loop: Supremacy EVO, SE360 R... Windows 10 Pro x64 Asus PB328Q 32" WQHD 
MonitorKeyboardPowerCase
Asus 25" 16:10 (forgot the model number) Ducky Shine 5 EVGA SuperNOVA 650 G2  Corsair Crystal Series 460X RGB 
MouseMouse Pad
Logitech G900 Corsair MM300 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
Intel i7-7700K Asus ROG Maximus IX Hero Gigabyte GeForce GTX 1070 G1 Gaming 4x 8GB Vengeance LPX Series 
Hard DriveCoolingOSMonitor
1TB Samsung EVO 960 NVME M.2 EKWB based custom loop: Supremacy EVO, SE360 R... Windows 10 Pro x64 Asus PB328Q 32" WQHD 
MonitorKeyboardPowerCase
Asus 25" 16:10 (forgot the model number) Ducky Shine 5 EVGA SuperNOVA 650 G2  Corsair Crystal Series 460X RGB 
MouseMouse Pad
Logitech G900 Corsair MM300 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Intel CPUs
Overclock.net › Forums › Intel › Intel CPUs › Kaby Lake Overclocking Guide [With Statistics]