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

Haswell Overclocking Guide [With Statistics] - Page 132

post #1311 of 19539
Quote:
Originally Posted by givmedew View Post

Just double checking but VCCIN, VRIN, and CPU Input Voltage all refer to the same thing correct?

Yessir
post #1312 of 19539
Quote:
Originally Posted by Big Texas View Post

From what I've noticed a non bsod reboot that just causes the screen to go black is VCCIN/VRIN related
i've upped the VCCIN from default 1.75V (AUTO in BIOS) to 1.77V. In CPU-Z it shows as 1.808 while running HandBrake. That must be because of the LLC is set to extreme. (CPU-Z 1.66 shows VCCIN as Vcore btw. It's difficult to find out the instantaneous VCCIN with this board with other utilities, even AISuite III.)

Anyways, I have few videos in a HandBrake queue. It should keep the busy the while time I'm at work. (TeamViewer is a lifesaver!).

I'll keep you guys posted about the progress.
post #1313 of 19539
Quote:
Originally Posted by Darkwizzie View Post

He didn't specify. if you don't specify it can mean anything. Even stressing is vague. How stable is stable enough? How hot is too hot? Linpack or Prime or Aida? Chess or encoding? If somebody is going to call others' results BS they better elaborate.

No offense but your graph lacks specifics. 90% of the entries do not specify length of whatever they ran and secondly lack any link to that person's post that you've accepted - I'm not going to go through 100+ pages trying to find it and see what exactly was shown - it's your thread and if you accepted it, that's fine. That is not the point.

Run XTU bench (as an example) on air @ 1.3v+ - show me your volts and temps? Don't run the stress test as that's close to useless when it comes to temps. Also when you run this - tell me what your RAM speed is; if I drop it to 1333Mhz from 2133Mhz my temps drop by over 10 degrees.

I've already considered that my loop has a problem but it doesn't; I can easily verify that by checking my GPU temps when I stress those. The CPU block.. re-seated that 5 times now and always with 1-2 degrees, can't see an issue there.
Just a 'puter
(18 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i7 6700K ASUS Maximus VIII Hero iGPU thanks to Nvidia 970 3.5GB Corsair Vengeance LPX CMK8GX4M2B3200C16 
Hard DriveHard DriveCoolingCooling
Samsung 950 PRO 512GB Crucial M4 256GB RX360 V3 Koolance CPU-380I 
CoolingCoolingOSMonitor
Koolance RP-452X2 Reservoir Koolance PMP-450 12V Variable Speed Pump Windows 10 Pro x64 benq XL2411Z 
KeyboardPowerCaseMouse
logitech G510 Seasonic X-760 Corsair 800D Mionix Castor 
AudioAudio
Creative Sound Blaster X-Fi Titanium HD DT 990 Premium 250Ohm 
  hide details  
Reply
Just a 'puter
(18 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i7 6700K ASUS Maximus VIII Hero iGPU thanks to Nvidia 970 3.5GB Corsair Vengeance LPX CMK8GX4M2B3200C16 
Hard DriveHard DriveCoolingCooling
Samsung 950 PRO 512GB Crucial M4 256GB RX360 V3 Koolance CPU-380I 
CoolingCoolingOSMonitor
Koolance RP-452X2 Reservoir Koolance PMP-450 12V Variable Speed Pump Windows 10 Pro x64 benq XL2411Z 
KeyboardPowerCaseMouse
logitech G510 Seasonic X-760 Corsair 800D Mionix Castor 
AudioAudio
Creative Sound Blaster X-Fi Titanium HD DT 990 Premium 250Ohm 
  hide details  
Reply
post #1314 of 19539
Is anyone experimenting with clocking certain cores faster than other cores? It seems like it is the same core that is failing in prime95. VCore fixes it (lots of vcore).
Silence
(20 items)
 
Secondary Rig
(9 items)
 
 
CPUMotherboardGraphicsGraphics
Intel i7-5820k  ASRock x99 extreme 4 AMD R9 290X AMD R9 290X 
RAMHard DriveCoolingCooling
ballistix ddr4 dual channel Samsung 850 EVO 500GB DT 5noz CPU Waterblock XSPC Razor Full Cover GPU Block 
CoolingCoolingCoolingCooling
XSPC Razor Full Cover GPU Block 4x120 Black Ice Xtreme GTX480 Swiftech MCP655 Danger Den Monsoon Pump/Reservoir 
CoolingOSMonitorKeyboard
Corsair Link Fan Controller Windows 8.1 Nixeus 2560x1440 Logitech G15 
PowerMouseMouse PadAudio
Kingwin PF-850 Ttesports Level10M Razer Vespula Sennheiser HD590 
CPUMotherboardGraphicsRAM
Intel i5-2500k gigabyte z68xp-ud4 Gigabyte GTX 970 G1 Gaming DDR3 1600 4x2GB 
Hard DriveCoolingOSMonitor
Western Digital Velociraptor 300GB w/ Intel RST... Corsair H100 Windows 8.1 1920x1200 HP 
Power
PC P+C 1000w 
  hide details  
Reply
Silence
(20 items)
 
Secondary Rig
(9 items)
 
 
CPUMotherboardGraphicsGraphics
Intel i7-5820k  ASRock x99 extreme 4 AMD R9 290X AMD R9 290X 
RAMHard DriveCoolingCooling
ballistix ddr4 dual channel Samsung 850 EVO 500GB DT 5noz CPU Waterblock XSPC Razor Full Cover GPU Block 
CoolingCoolingCoolingCooling
XSPC Razor Full Cover GPU Block 4x120 Black Ice Xtreme GTX480 Swiftech MCP655 Danger Den Monsoon Pump/Reservoir 
CoolingOSMonitorKeyboard
Corsair Link Fan Controller Windows 8.1 Nixeus 2560x1440 Logitech G15 
PowerMouseMouse PadAudio
Kingwin PF-850 Ttesports Level10M Razer Vespula Sennheiser HD590 
CPUMotherboardGraphicsRAM
Intel i5-2500k gigabyte z68xp-ud4 Gigabyte GTX 970 G1 Gaming DDR3 1600 4x2GB 
Hard DriveCoolingOSMonitor
Western Digital Velociraptor 300GB w/ Intel RST... Corsair H100 Windows 8.1 1920x1200 HP 
Power
PC P+C 1000w 
  hide details  
Reply
post #1315 of 19539
Quote:
Originally Posted by givmedew View Post

Is anyone experimenting with clocking certain cores faster than other cores? It seems like it is the same core that is failing in prime95. VCore fixes it (lots of vcore).

I have tried this but honestly my benchmark scores are lower.

for example I score higher at all cores at 4.7GHz than I do with two cores at 4.8GHz and two cores at 4.7GHz. the mpower max that im using is kinda difficult to adjust individual cores. im not sure i can do it from the bios, ive tried without success . i can however change individual cores through the xtu, but like i said i usually score lower so i have been keeping all cores at the same frequency. thats my experience with changing individual core speeds.
post #1316 of 19539
Quote:
Originally Posted by givmedew View Post

Is anyone experimenting with clocking certain cores faster than other cores? It seems like it is the same core that is failing in prime95. VCore fixes it (lots of vcore).

You can't control the speed of a specific core. Active cores all run at the same speed. You may be able to run different speeds based on how many cores are active (like turbo does) but you can't run Core 0 at a certain speed and Core 3 at something else.
post #1317 of 19539
Quote:
Originally Posted by Forceman View Post

Quote:
Originally Posted by givmedew View Post

Is anyone experimenting with clocking certain cores faster than other cores? It seems like it is the same core that is failing in prime95. VCore fixes it (lots of vcore).

You can't control the speed of a specific core. Active cores all run at the same speed. You may be able to run different speeds based on how many cores are active (like turbo does) but you can't run Core 0 at a certain speed and Core 3 at something else.

Yeh I noticed... the verbage was confusing in the manual. It calls it all core or per core. When I was OCing the phenomII chips you could go by each individual core and often needed to. With this it is just how many cores are active. So you could set a 45x multi for 2 cores and then set a 40x for 3 and a 35 for 4 if that is what you wanted to do but you can't single out a specific core and change it's multi.
Silence
(20 items)
 
Secondary Rig
(9 items)
 
 
CPUMotherboardGraphicsGraphics
Intel i7-5820k  ASRock x99 extreme 4 AMD R9 290X AMD R9 290X 
RAMHard DriveCoolingCooling
ballistix ddr4 dual channel Samsung 850 EVO 500GB DT 5noz CPU Waterblock XSPC Razor Full Cover GPU Block 
CoolingCoolingCoolingCooling
XSPC Razor Full Cover GPU Block 4x120 Black Ice Xtreme GTX480 Swiftech MCP655 Danger Den Monsoon Pump/Reservoir 
CoolingOSMonitorKeyboard
Corsair Link Fan Controller Windows 8.1 Nixeus 2560x1440 Logitech G15 
PowerMouseMouse PadAudio
Kingwin PF-850 Ttesports Level10M Razer Vespula Sennheiser HD590 
CPUMotherboardGraphicsRAM
Intel i5-2500k gigabyte z68xp-ud4 Gigabyte GTX 970 G1 Gaming DDR3 1600 4x2GB 
Hard DriveCoolingOSMonitor
Western Digital Velociraptor 300GB w/ Intel RST... Corsair H100 Windows 8.1 1920x1200 HP 
Power
PC P+C 1000w 
  hide details  
Reply
Silence
(20 items)
 
Secondary Rig
(9 items)
 
 
CPUMotherboardGraphicsGraphics
Intel i7-5820k  ASRock x99 extreme 4 AMD R9 290X AMD R9 290X 
RAMHard DriveCoolingCooling
ballistix ddr4 dual channel Samsung 850 EVO 500GB DT 5noz CPU Waterblock XSPC Razor Full Cover GPU Block 
CoolingCoolingCoolingCooling
XSPC Razor Full Cover GPU Block 4x120 Black Ice Xtreme GTX480 Swiftech MCP655 Danger Den Monsoon Pump/Reservoir 
CoolingOSMonitorKeyboard
Corsair Link Fan Controller Windows 8.1 Nixeus 2560x1440 Logitech G15 
PowerMouseMouse PadAudio
Kingwin PF-850 Ttesports Level10M Razer Vespula Sennheiser HD590 
CPUMotherboardGraphicsRAM
Intel i5-2500k gigabyte z68xp-ud4 Gigabyte GTX 970 G1 Gaming DDR3 1600 4x2GB 
Hard DriveCoolingOSMonitor
Western Digital Velociraptor 300GB w/ Intel RST... Corsair H100 Windows 8.1 1920x1200 HP 
Power
PC P+C 1000w 
  hide details  
Reply
post #1318 of 19539
Quote:
Originally Posted by Forceman View Post

You can't control the speed of a specific core. Active cores all run at the same speed. You may be able to run different speeds based on how many cores are active (like turbo does) but you can't run Core 0 at a certain speed and Core 3 at something else.
One question though. Say you set the multi like below for different core utilization.

1-core: 46x
2-cores: 45x
3 and 4 cores: 44x

How do you test if that is stable? If you stress test, all cores will be utilized and thus you'll always be limited to 44x.

Do you set the affinity in task manager? But it thinks 4770k has 8 cores. So you set the CPU affinity of the stress test to core 0 and 1 for 46x multi? You do this for all cores (0&1,2&3,4&5,6&7) separately?

I know I'm crazy to think about this when I cannot make 44x stable across all utilizations. I was just curious.
Edited by Anusha - 8/28/13 at 6:21pm
post #1319 of 19539
You'd have to use something like Prime95 That let's you specify the number of threads, and then use affinity to control which core was tested, like you showed. It would be an enormous pain, which is probably part of the reason why no one does it.
post #1320 of 19539
Quote:
Originally Posted by Forceman View Post

Prime95 took a lot more voltage than Aida for me as well. Aida is not particularly stressful.

I wouldn't worry about Prime, just test with x264 or chess or BF3 or something.

Just my two cents but I think a little more emphasis needs to be put on how long and what you are using to stress your system.

Reading all the posts on here I see so many people having trouble with BSOD's and stability but then it also appears like stress tests have only been run for short periods of time or even not at all and just gaming etc is used.

In my opinion the best test for stability is to use a combination of these different tests as well as regular use. It might take some time but in my experience it leads to less BSOD'/errors and time fiddling around to fix things in the long run.

And just because Prime or another testing software doesn't test a particular instruction set doesn't invalidate it as a stress testing/stability tool, however this is also why I recommend using multiple tools.

A lot of this has already been touched on before in the guide/thread and I don't claim to be an expert but this is my opinion only which is based on my experience overclocking several Sandybridge systems and now Haswell. I'm just hoping it may help people lead to increased stability. After I've set my vcore/clock multiplier/uncore etc etc I do the following.

Dial in clockspeed first with Memory at default speeds and uncore at default 35x.

1. Intel Burn Test - Run this on High. If this passes I know I'm atleast in the ball park of where I need to be.

2. Try running Cinebench and encoding something with Handrake - If these pass it's reassurance you're on the right track.

2. Aida64 - Run stability tests for FPU only and CPU/FPU/Cache/Memory for about 8 hours. You could substite the Intel XTU stability test here if you like. (Aida64 seems to be less stressful than Prime95 and Intel XTU Benchmark in my experience)

3. Prime 95/Intel XTU Benchmark - Try running Prime95 Blend for about 24 hours or Intel XTU Benchmark. (I used Prime). If your stable after this time you're pretty much good to go.

4. Play some games/Do what you normally do.

After this I would move on to upping the uncore and memory by repeating the above steps.

Remember to watch temps/throttling, but mainly voltage. Intel has stated before that it is Voltage that kills chips, not temps.

Watch your BSOD codes and remember that they can mean both too much voltage and not enough voltage. For example if you find you are constantly increasing vcore and still reaching instability with BSOD code 124, it may not be that you need more vcore, it could be you need to increase input voltage but will be able to run a lower vcore.

Haswell is a little bit tricker than Ivy and Sandy but if you put the time in at the start and are methodical about what you do, in my experience it leads to less instability down the road and less overall time wasted.

I estimate it took me about 6 days to get my Haswell stable at 4.5Ghz & 42x uncore with RAM at 2400Mhz XMP following the above. When overclocking Sandy it took less time due to less variables.

For my Haswell and Sandybridge Overclocks I followed the above and whilst it may be luck, I haven't run in to any stability problems/BSOD's during gaming and my everyday use after being able to pass those stability tests.

To be able to do the above beyond 4.6-4.7Ghz though (depending on the voltage you're pushing/temps you're seeing), unless you're running some extremely effective cooling like phase-change cooling you're probably going to need to delid.

Dont' rush in, be methodical, and you'll get results.

Just my thoughts.
Edited by combatant3219 - 8/28/13 at 9:36pm
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Intel CPUs
Overclock.net › Forums › Intel › Intel CPUs › Haswell Overclocking Guide [With Statistics]