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

Haswell Overclocking Guide [With Statistics] - Page 564

post #5631 of 19539
Quote:
Originally Posted by Darkwizzie View Post

I'm not that familiar with custom loops but you must have a seriously nerfed loop then. People on air are reaching 1.2v as a starting point and I ended at 1.42v. I'll chart you, chanks for submitting!

He doesn't have enough rad space. 120.1 for every component in the system and one extra 120.1 is minimum. I.E. He really needs another 120 rad. Although, his money would better be invested in a 120.2 or 120.3 rad. It really just depends on his case requirements. To me, it sounds like he bought a Swiftech H22x or one of those and just added the GPU to the loop.
post #5632 of 19539
hyperthreading does add a lot of heat btw
if you want to try higher speeds and you are severely thermally limited, you can also disable hyperthreading if you're sure you don't have use cases for 8 virtual cores.... kind of a silly thing to do if you've bought a 4770k but it's an option (i personally wouldn't do it tongue.gif)

I've another question for you peeps -- if I want to define a custom turbo mode per core, it's annoying that you cannot define voltages per core. I.e. 1 core active, 4.8Ghz and say 1.45V, 2 cores active and 4.7Ghz and say 1.37V and 3 or 4 cores active, and say 4.6Ghz and 1.27V. I see a way to set the frequency like that but not the voltages. Thanks.
post #5633 of 19539
Quote:
Originally Posted by creos7 View Post

I've another question for you peeps -- if I want to define a custom turbo mode per core, it's annoying that you cannot define voltages per core. I.e. 1 core active, 4.8Ghz and say 1.45V, 2 cores active and 4.7Ghz and say 1.37V and 3 or 4 cores active, and say 4.6Ghz and 1.27V. I see a way to set the frequency like that but not the voltages. Thanks.

Adaptive?
post #5634 of 19539
Where is everyone checking to see there stock cpu voltage?
post #5635 of 19539
Quote:
Originally Posted by Doug2507 View Post

Adaptive?
Adapative doesn't quite work because the 'turbo boost' voltage you set is basically what happens under full load. For non-full load intel has some algo (i think you need to also have C states enabled possibly) but for boost it takes what you give it. The issue is full load for 1 core and full load for 2cores and full load for 3/4 cores will require different voltages. Adaptive only allows you to provide one.

EDIT: well it takes what you give it except for the special AVX (or whatever they are called) instructions where it actually takes your specified max + some offset, this is hardwired in the FIVR so it's not controllable, which is why you never stress under adaptive
Edited by creos7 - 11/21/13 at 10:25am
post #5636 of 19539
I got a horrible Costa Rica chip. I think I need to play around with alot more settings so that I can lower my voltage core. I am trying to get 4.5 ghz stable and I have not found an acceptable Vcore level yet - I was able to do so at 1.36 Vcore and incredibly high temps (60-70C for first round of Prime95 but 90C+ for 2nd round that stresses CPU more). I potentially can do 4.6 ghz at 1.42V with even higher temps.... Do I need to reapply my thermal paste or is that normal for high Haswell voltages with hyperthreading enabled for a water cooled loop?

I am running a 360 and 120 radiators in a custom loop for the 4770k and a 290X overvolted to 1.299 V (1.41V + vdroop). What kind of temperatures should I expect for around 1.325-1.42 Vcore range? I think I can get my Vcore lower with tweaking other voltage settings - hopefully for 4.5 ghz in the 1.31-1.34 range.

In my initial testing, I think I was unstable due too high stock voltage for my low power Samsung 1600 memory (should be 1.35V stock, motherboard was applying 1.5V). I was running the memory on my Ivy bridge motherboard at 1.45 V 1866 - it was quite unstable at higher voltages and the motherboard had some bug that higher frequencies wouldnt post. Should I be using similar voltages for a Z87 motherboard?
My System
(17 items)
 
  
CPUMotherboardGraphicsRAM
Ryzen R7 1700x Asrock X370 Taichi EVGA GTX 1080 Ti Founder Edition with Hybrid Co... 16 GB DDR4 G.skill Flare X 3200 
Hard DriveHard DriveHard DriveOptical Drive
512GB Samsung 840 Pro 3 TB Western Digital Black 3 TB Toshiba LG WH14NS40 Blu-ray Writer 
CoolingCoolingOSMonitor
NZXT Kraken X52 Gentle Typhoon AP-15 Windows 10 Samsung KU6300 40 inch 4k HDTV 
KeyboardPowerCaseMouse
Cooler Master MasterKeys Pro L RGB with Cherry ... EVGA Supernova 1000 P2 Phanteks Enthoo Luxe Tempered Glass Edition Logitech G900 Chaos Spectrum 
Mouse Pad
Razer Scarab 
  hide details  
Reply
My System
(17 items)
 
  
CPUMotherboardGraphicsRAM
Ryzen R7 1700x Asrock X370 Taichi EVGA GTX 1080 Ti Founder Edition with Hybrid Co... 16 GB DDR4 G.skill Flare X 3200 
Hard DriveHard DriveHard DriveOptical Drive
512GB Samsung 840 Pro 3 TB Western Digital Black 3 TB Toshiba LG WH14NS40 Blu-ray Writer 
CoolingCoolingOSMonitor
NZXT Kraken X52 Gentle Typhoon AP-15 Windows 10 Samsung KU6300 40 inch 4k HDTV 
KeyboardPowerCaseMouse
Cooler Master MasterKeys Pro L RGB with Cherry ... EVGA Supernova 1000 P2 Phanteks Enthoo Luxe Tempered Glass Edition Logitech G900 Chaos Spectrum 
Mouse Pad
Razer Scarab 
  hide details  
Reply
post #5637 of 19539
Jrcbandit, see below to give you an idea.
Quote:
Having power circuitry on-die adds heat. Haswell processors run hot when voltage levels are increased.

A very good air cooler is required for voltage levels above 1.15V.
1.20V-1.23V requires use of closed loop water coolers.
At 1.24V-1.275V dual or triple radiator water cooling solutions are advised.
This is assuming the processor will be run at full load for extended periods of time.

Using Vcore higher than 1.275V is not advised for 4 core 8 thread CPUs under full load as there are very few cooling solutions that can keep temps below thermal throttling point.

Obviously these statements are conservative and also apply only to the case of using synthetic stress tests. Real-world use patterns pretty much never stress the CPU that hard (unless you're doing some highly specalized stuff). But it should give you an idea of what to expect if you want to go down the path of standard Prime95 / IntelBurnTest testing.

You should read darkwizzie's guide again, he addresses those points!
Edited by creos7 - 11/21/13 at 11:00am
post #5638 of 19539
Quote:
Originally Posted by jrcbandit View Post

What kind of temperatures should I expect for around 1.325-1.42 Vcore range?

I can't hit 80deg with 28.1 and IBT running 1.35vcore and over 2.0vrin. Custom loop (see sig) but with delid.
Quote:
Originally Posted by creos7 View Post

Adapative doesn't quite work because the 'turbo boost' voltage you set is basically what happens under full load. For non-full load intel has some algo (i think you need to also have C states enabled possibly) but for boost it takes what you give it. The issue is full load for 1 core and full load for 2cores and full load for 3/4 cores will require different voltages. Adaptive only allows you to provide one.

EDIT: well it takes what you give it except for the special AVX (or whatever they are called) instructions where it actually takes your specified max + some offset, this is hardwired in the FIVR so it's not controllable, which is why you never stress under adaptive

Ah right, i thought with adaptive and power saving features it would automatically adjust anywhere up to a pre-set maximum dependant on load. Not really had a play about with it myself as i'm on manual most of the time so good to know.

Don't entirely agree with that last statement though, adaptive with non-AVX is a non issue. I get what you're saying though. thumb.gif
Edited by Doug2507 - 11/21/13 at 11:29am
post #5639 of 19539
Quote:
Originally Posted by creos7 View Post

Jrcbandit, see below to give you an idea.
Obviously these statements are conservative and also apply only to the case of using synthetic stress tests. Real-world use patterns pretty much never stress the CPU that hard (unless you're doing some highly specalized stuff). But it should give you an idea of what to expect if you want to go down the path of standard Prime95 / IntelBurnTest testing.

You should read darkwizzie's guide again, he addresses those points!

Good points, I will try x264 and Battlefield 3 as stress tests, any other games that are susceptible to overclocks? Seems like Prime is not the way to go for testing if you have a poor overclocker that requires 1.35V+.

I know that Vccin of 1.9V is safe, but my overclocks seem to require 2.0 or higher for stress testing 4.5 or 4.6 ghz. I wonder how good of idea is it to use 2.1...
My System
(17 items)
 
  
CPUMotherboardGraphicsRAM
Ryzen R7 1700x Asrock X370 Taichi EVGA GTX 1080 Ti Founder Edition with Hybrid Co... 16 GB DDR4 G.skill Flare X 3200 
Hard DriveHard DriveHard DriveOptical Drive
512GB Samsung 840 Pro 3 TB Western Digital Black 3 TB Toshiba LG WH14NS40 Blu-ray Writer 
CoolingCoolingOSMonitor
NZXT Kraken X52 Gentle Typhoon AP-15 Windows 10 Samsung KU6300 40 inch 4k HDTV 
KeyboardPowerCaseMouse
Cooler Master MasterKeys Pro L RGB with Cherry ... EVGA Supernova 1000 P2 Phanteks Enthoo Luxe Tempered Glass Edition Logitech G900 Chaos Spectrum 
Mouse Pad
Razer Scarab 
  hide details  
Reply
My System
(17 items)
 
  
CPUMotherboardGraphicsRAM
Ryzen R7 1700x Asrock X370 Taichi EVGA GTX 1080 Ti Founder Edition with Hybrid Co... 16 GB DDR4 G.skill Flare X 3200 
Hard DriveHard DriveHard DriveOptical Drive
512GB Samsung 840 Pro 3 TB Western Digital Black 3 TB Toshiba LG WH14NS40 Blu-ray Writer 
CoolingCoolingOSMonitor
NZXT Kraken X52 Gentle Typhoon AP-15 Windows 10 Samsung KU6300 40 inch 4k HDTV 
KeyboardPowerCaseMouse
Cooler Master MasterKeys Pro L RGB with Cherry ... EVGA Supernova 1000 P2 Phanteks Enthoo Luxe Tempered Glass Edition Logitech G900 Chaos Spectrum 
Mouse Pad
Razer Scarab 
  hide details  
Reply
post #5640 of 19539
Quote:
Originally Posted by Doug2507 View Post

Don't entirely agree with that last statement though, adaptive with non-AVX is a non issue. I get what you're saying though. thumb.gif
Oh ya, i meant you never stress with AVX under adapative. I'm not 100% familiar with the FIVR programming by Intel, but i do think the most significant ramp up in voltage occurs for AVX so you are right, as long as it's not AVX it is probably a negligible increase in voltage... smile.gif
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Intel CPUs
Overclock.net › Forums › Intel › Intel CPUs › Haswell Overclocking Guide [With Statistics]