One of the few videos i didn't watch. I think it may be a job for tomorrow, when no one is around to watch me get frustrated
which i often do
-- EDIT ---------
Ok i got bored so decided to have a mess with under voltage. All of these tests ran a Large FFT cycle for 30 minutes
No Undervoltage - Max temp: 68
Attempt 1 - Max temp: 61 drop of 7'C
Attempt 2 - Max temp: 60 drop of 8'C
Attempt 3 - Max temp: 59 drop of 9'C
Test 3 was able to run the test but blue screened when downloading HL2-Lost Coast to try a game benchmark. I'm currently running a test with 1.02V to see if that will run stable and keep the same temps as Test 3. I will continue to test this tomorrow, and if the Test 4 passes 1 hour. i will leave it to run over night, so another 8 or 10 hours. So at the moment I know i have a solid drop of 0.12V which isn't bad.
Those temps are with all fans, and pump running on the lowest it can. We will see what I can get to tomorrow. The suggestion seems to be working I think. And i may see a drop when i rotate my CPU waterblock to the recommended orientation. Might not be for a while if this continues to be 24/7 stable. Hopefully it will still be stable when the max clock hits its 3.7GHz clock when i load up single cores.
I also read that the voltage thing my MB was doing, which auto dropped the voltage when idle can be adjusted, So i may be able to decrease the voltage for when idle and set it to the stable for max load. Not sure how that works though
-- EDIT ----------
Test 4 results are in.
Attempt 4 - Max temp: 59 drop of 9'C
So i got the same temp as with 1V, and that test was after 1 hour. i ran lost coast benchmark maxed out and got a max temp of 51'C which was nice to see instead of a normal game temp of 63'C
One thing i will note though is that the 59'C max was only a temp spike. Whilst watching the stress test the actually temp sat between 56, and 57'C. May decrease slightly if i do up the fans a little but im relatively happy with these temps now. I'm going to run a long test tomorrow now instead so i can watch a movie xD
Hopefully the system is 100% stable now, and will try some games tomorrow, most likely Dota 2 to make sure its in-game stable. If it is then i wont touch the voltage anymore as there is no point trying to lower it as the temps wont decrease much more now. Time to see what the idle temps are like while i sleep now
-- EDIT ----------
Ok everyone, I'm awake and have just finished a 6 hour large FFT cycle at 1.02V. So here are my results for you all
Final Test
As you can see the extra 5 hours did push the max temp up by 4'C however i believe this was again a spike with the average temp sitting at 56-57'C. After upping my rad fans slightly you can see above the temp it sits at is 54'C, this does fluctuate to 55'C every so often but it seems that the average temp whilst running with the fans slightly fast is 54-55'C which is a huge drop of 10'C off the average before undervolting.
Over night the room did get rather warm, with an ambient temp of around 24'C
Idle temps sat as follows; (EDIT: Left is night, Right is the current lowest during the day)
Ambient: 24'C 22'C
Core #0: 30'C 27'C
Core #1: 32'C 27'C
Core #2: 30'C 26'C
Core #3: 26'C 23'C
so as you can see, changing the voltage hasn't effected my idle temps, as the minimum temp was either 2'C or 1'C warmer then ambient, this i can put down to heat from the pump, plus heat from the GPU's passing through the rad aswell
I may look into the offset voltage settings, see if i can find a way to set the voltage when not idle to 1.02V and see how low i can get the idle voltage to sit. From having a fast read before going to sleep last night i understand that the offset is relative too the voltage the system sets which is based off the VID voltage set by the CPU.
Not sure if the VID voltage should be messed with or not, and if i can get my voltages working as i wanted to. I will have to have a dig around and see what i can pull up
-- EDIT ---------
I have had a look round and here is what i have found. I will post the link to the source if requested but i will break it down for anyone that may want to know it aswell. I will only be looking at the negative offset but the positive seems very similar.
Why use the offset over fixed you may ask?
Well the answer is simple. When using a fixed voltage as i have been, the voltage to the CPU stays the same when under load, and when idle. Now whilst under load the voltage may have decreased. When idle the voltage may have actually increased causing idle temps to increase along with idle power consumption. Using offset allows the voltage supplied to the CPU to be fluctuated providing lower voltages when the system is idle and providing more when under load.
So using the offset in the bios is linked to the CPU's in-built VID voltage setting. The setting which is used to dynamically increase the voltage for the CPU when higher clocks are reached. In this case the maximum VID for my processor is 1.231V, this can be seen using CoreTemp whilst running a Prime95 stress test.
This is the voltage which the offset will work from. In this case i want to undervoltage so i want to set my offset to "-". As my stable clock voltage is 1.02V my offset value would be a huge 0.211V. This is calculate as the difference between the VID and the maximum voltage you want
Offset = Desired Voltage MINUS VID
-0.211 = 1.02 - 1.231
Using this offset value the maximum Vcore would be set to around the 1.02V however as you can see below, my VID when idle is rather a lot lower then when it is stressed
Giving me a lowest VID of 1.0708V sometimes getting lower to 1.0654V. With this in mind we now need to factor in the offset to find the voltage the CPU will sit at when idle, use the lowest VID you see to calculate the lowest possible voltage
Idle Voltage = VID MINUS Offset
0.8544 = 1.0654 - 0.211
This voltage is extremely low, as with my motherboard the lowest fixed voltage you can set is 0.8V. Bear this in mind. If you do set the offset to -0.211 the CPU may not have enough voltage when idling to keep running. Therefore setting this may cause system crashes and un-stability. If this is the case then lowering the offset is required until the system is stable when idle. This will increase the Vcore when stressing however but if like me your system is mainly idle, you will have lower power usage, and lower temps while the system is sitting around. Great for when power saving is a factor.
Hopefully that is all correct. If not someone please correct me and i will change it as needed
-- EDIT --------
i just put the newest HWmonitor on my system and it actually shows all my voltages and the VID min and max for my processor. I take it this is accurate as it is whats used in OCCT. I will put a 20 minute short test on OCCT to see what that pics up the lowest and highest voltages as
Results
Ok so HWmonitor says i have a lowest VID of 1.066V and a highest of 1.261V
That one thing that interests me however is that whilst running the OCCT test the Vcore was as low as 1.016V and highest of 1.032V. So a margin of +-0.008V from the fixed voltage
a strange thing is that whilst OCCT also uses HWmonitor for temps, its maximum temperatures were lower on 3 cores, and higher on the 4th. This i think is very strange but ooo well
