12-21-2014
07:08 AM
- last edited on
03-06-2024
02:41 AM
by
ROGBot
12-21-2014 11:34 AM
Quad5Ny wrote:
+800mV? That's a kinda high. Everything I've ever read has had it at +400mV.
SVID, VCCIN, FIVR-IN, VR Input, CPU Input Voltage or whatever you want to call it should be around 1.8-ish.
Just set LLC to 7-9 and leave the VCCIN alone unless your going to really be turning your voltage up more than +100mV (0.1V). -- So if your vCore is 1.2V than the default of 1.8V VCCIN is perfectly fine.
Remember VCCIN is the voltage your feeding to the CPU's integrated voltage regulator. There is a sweet spot (for converting that voltage to all the other ones) and you can totally burn the VR out if you give it too many volts.
12-21-2014 02:45 PM
12-21-2014 03:30 PM
NemesisChild wrote:
BF4 is a CPU intensive game. I can pass RealBench and other stress tests fine, but fail in BF4.
I can tell you that 1.210v for 4.6GHz is probably too low, unless you have an extremely good chip.
12-21-2014 09:34 PM
NemesisChild wrote:
BF4 is a CPU intensive game. I can pass RealBench and other stress tests fine, but fail in BF4.
I can tell you that 1.210v for 4.6GHz is probably too low, unless you have an extremely good chip.
Morais wrote:
What I do know is that from 4.5Ghz onwards the requiered vcore goes up exponencialy.
4.5Ghz needs 1.175v
4.6 GHz needs 1.220v
4.7Ghz requiered around 1.275v. 😛
12-22-2014 03:29 AM
12-22-2014 07:52 AM
12-22-2014 10:55 AM
Quad5Ny wrote:
4.7GHz @ 1.275V is a dam good overclock considering most Haswell/Devils Canyon CPU's need 1.3V+ to hit 4.7. 🙂
jab383 wrote:
I've seen very few 4790K that could run x46 stable at 1.2Vcore. The 124 BSOD code indicates that yours is close, because it ran and played BF4, but not solidly stable. Try adding a little more Vcore, to maybe 1.225, and repeating the tests. That would still be a better 4790K than mine, which takes 2.50Vcore to work at x46.
Note that Aida64 is a great monitor program, but its stress tests aren't very stringent. Realbench is better and can indicate a 24/7 stable profile good for most applications. A conclusive test for me is OCCT - the large data set test. A few minutes (3-5) there are enough to show good or bad.
Jeff
06-19-2015 03:37 PM
06-19-2015 07:37 PM
06-20-2015 02:02 AM
NemesisChild wrote:
Be thankful you can do 4.8GHz, you've undoubtedly hit the silicone lottery wall.