GhislainG :
JackNaylorPE you probably meant Prime95 28.5 because 26.6 doesn't put that much stress on the CPU.
You can't do temperature tests at 50% load, there are just too many variables. Prime95 v26.6 provides the most stable, full load stress test that does not include AVX. You are referring to Prime95 v28.XX, which does include AVX. http://www.tomshardware.com/forum/id-1800828/intel-temp... This is article is written by another forum member and is meant to help with CPU temperature issues.
No..... while 28.5 is very capable of destroying your CPU because when AVX and other instruction sets are called upon, voltage spikes 0.10, 0.13 and higher. However, 26.6 still puts on a CPU load that is greater than you can possibly achieve using any application based benchmark.
P95 26.6 - like testing your new 5,000 pound towing capacity pickup truck pulling a 7,500 pound load up 6%, 2 mile long grades, at 4,500 feet elevation in 85F weather
P95 28.5 - like testing your new 5,000 pound towing capacity pickup truck pulling a 10,000 pound load up 12%, 7 mile grades, at 12,000 feet elevation in 105F weather
Neither is representative of your new 5,000 pound towing capacity pickup truck towing the 3,500 pound boat you actually own over the hills between your house and the local marina at seal level.
To use another analogy, P95 28.5 is like being the last guy left on a dodgeball game and the other team is using a pitching machine sending baseballs at 150 mph. P95 26.6 is like being the last guy left on a dodgeball game and the other team is using a pitching machine using baseballs sending baseballs at 120 mph. RoG RB is like you're still the only one left but they have 4 guys each sending tennis balls at 98 mph at the same time and they keep moving the machines
. you'll get hit a lot more but it's no where near as dangerous.
What is the point of testing w/ 26.6 when:
a) It creates temps that no combination of programs you will actually ever use are capable of producing . If you are thinking, I want the highest OC I can get w/o my hottest core going above 78C, you unnecessarily limit your OC.
b) Your have a modern CPU w/ modern instruction sets that allow you to complete certain tasks at speeds which older CPUs are incapable. Yet, you are "testing your OC for stability" using a tool that does not include these instruction sets.
c) You can be P95 26.6 stable for 2 hours and then crash using the individual applications in RoG Real Bench. You therefore wind up with a CPU and OC that can run benchmarks but not real applications. And unless you are a competitive overclocker, we buy our machines so that the can run real applications.
I hit temps in the mid - high 70s at 4.6 Ghz (1.375 volts... > 1.5v peak) under RoG RB. With 26.6 I can easily hit temps which could shorten chip life ... with 28.5, I could possibly kill the CPU.
I'm not sure what these mean. Also I didn't know that Hwinfo was a program and didnt use that, i used realtemp, with the results as follows:
Core 0: 61ºC
Core 1: 61ºC
Core 2: 64ºC
Core 3: 62ºC
Well the reason you use HWiNFO is it makes all other utilities obsolete. No need to keep 8 utilities for measuring this, that and the other thing .... sensors only is the mode you wanna use here.... tho the tool is extremely useful.
http://www.hwinfo.com/
As for the relative usefulness of testing tools any question of "is this high" can not be answered if everyone is using different rulers ... if one is in cm and one in inches, we aren't going to get the same numbers.... A game is the most useless of choices as the load various continuously throughout the game. So for example.... getting 65C walking alone thru the woods might be 75C during a big battle with lots of characters. A game also isn't going to be very much fun if it constantly does the same thing over and over again.... watching your toon gallop on a horse really, really fast for 2 hours doesn't show us that your machine is stable when other animations are initiated or lighting changes, shadows etc....that's P95.
Now that we both are using the same ruler (RoG Real Bench) using the same scale of measurement for comparison purposes I can conclude that you are hotter than I'd expect. I got temps similar to yours at 4.5 Ghz. At stock speeds I am at 46 - 52C. Remember also that the 4690k is reported by anandtech to run 5 - 10C cooler than my 4770k.
As for diagnosing your crash problem, again.... a single function synthetic benchmark is useless here. Again I will use car analogy, you hear a sound and bring the car in and the mechanic can't find it.... but did ya remember perhaps to tell him that the sound occured while braking while in a cloverleaf getting onto a major highway. If the mechanic drove it down the road, stopped, turned around came back he wouldn't find it.... that's the p95 test..... doing the same thing over and over again. Sometimes it takes a combination of factors to create the conditions which create an instability. By multitasking 4 different applications using Open CL, AVX, SSE2 in a random multitasking environment, you are testing not only the CPU but how the CPU interacts with the system in a wide variety if ways.
As previously indicated, I have used P95 to set TIM by putting on a moderate "shot in the dark" OC and then cycling up to the mid 80s an back down to room temperature finishing with a 2 hour run. On several occasions I have had that same "shot in the dark" OC fail under RoG Real Bench....have passed 8 hours of Intel XTU and failed under RoG Real Bench. The 8 minute benchmark test has saved me days of wasted time as I have uncovered instabilities with it in minutes..... anything with P95 under 2 hours is non conclusive.
So... where are we at .... your 61 - 64 temps is a very tight band.... sounds more like average numbers than max recorded. Another reason to use HWiNFO (4 columns - current, min, max avg) .... usually the spread is 10C from lowest to highest....It gives ya a better idea as to whether the peak was a little burp or a routine thing.
Here's what I'd suggest. Go into your BIOS and find these settings, right click on them to move these to your "favorites" page, .... don't change any settings as yet ....Once done, go to the tools tab, OC Profiles and save these settings to profile 1 and label it "stock"... Hit F10 (save and exit) ..... Go back into BIOS and now make the following changes (in bold).....most will already be there.
AI Overclock Tuner =
Auto
1-Core Ratio Limit =
42 (all others should automatically change with Sync all cores selected above)
Max. CPU Cache Ratio =
Auto
Min. CPU Cache Ratio =
Auto
Fully Manual Mode =
Disabled
Core Voltage =
Adaptive
Additional Turbo Mode CPU Core Voltage =
1.200
Core Cache Voltage =
Adaptive
Additional Turbo Mode CPU Cache Voltage =
Auto
Eventual CPU Input Voltage =
1.90
DRAM Voltage =
Auto
Again, go to the tools tab, and save this profile in the next slot and label it "4200", save and exit
Run RoG Real Bench again..... if it passes the 8 minutes w/ decent temps, run the stress test for 2 hours.... if it passes, set AI Overclock Tuner to XMP and retest.
If we get that far come back and we'll see where we go from there.
Next, while it didn't suggest it for your problem at hand, I did tell ya up top there that if you wanted to test for GPU issues and go GPU overclocks, the tool for watching GPU temps is Furmark. GPU-Z, MSI AFterburner and Unigine valley (among others is also useful). In the meantime, if ya wanna play with that
http://www.guru3d.com/articles_pages/gigabyte_geforce_gtx_970_g1_gaming_review,26.html
With AfterBurner we applied:
Temp Target 80 Degrees C
GPU clock +150 MHz
Power limiter 112%
Mem clock +500 MHz
Volatge + 87Mv
FAN RPM default
First save the default settings in the 1st "save slot" . Then Input the above settings into Afterburner (except memory) , and apply.... run..Unigine Valley check temps with Afterburner...adjust clock up or down by 10 based upon your success. When ya fail go back down and save the stable settings. Now start playing with memory in same manner. You will find that some games are just too unstable ....BF4 is one.
If ya get this far. we'll pick up at this point next time.
Facebook
Twitter
Instagram