Discussion 100MHz that costs a lot

Mkazem

Commendable
Jul 18, 2016
60
0
1,540
2
For about two years I had my 6700K running at 4.6GHz, VCore 1.344 and 84C max temp (with Prime95)
with 4 memory sticks running at 3600MHz,16,16,36,2

Today I noticed that my setup is not really stable, so I decided to drop 100MHz

Now running at 4.5GHz, VCore 1.28 and 74C max temp with Prime95 with same memory speed and its all stable.

That extra 100MHz cost me a lot. lets do a simple calculation

Sacrifices:

4.6 to 4.5 i.e 2.17% reduction in frequency.

Gains:

85 => 75 i.e 11.76% reduction in temp. (therefor more speed)
1.344 => 1.280 i.e 4.76% reduction in voltage.
lower fan speed (less noise!)
lower wattage
and more stability :)

I didn't measure the watts carefully but I'm pretty sure it reduced about 10 watts under load.

So far I'm very happy with the results. just wanted to share this with you so if you are wasting energy for just 100MHz, you may consider reducing it by a notch :)
 

BringerOfTea

Reputable
Dec 17, 2015
1,299
18
5,665
124
my chip won't go lower than 1,315V on 4.5 GHZ

I used OCCT and 3Dmark to test for stability...

didn't exceed your temps though was about 72 C.

Low ambient 19-20 C... wool socks on wintertime gaming :D
 

Mkazem

Commendable
Jul 18, 2016
60
0
1,540
2
Mkazem,

For a more meaningful and valid comparison, there's a few more items of information needed:

Q: What is your ambient (room) temperature?

Q: Which version of Prime95?

Q: Which Torture Test?

Thanks,

CT 😎
Q: What is your ambient (room) temperature?

26C is my room temperature

Q: Which version of Prime95?

I think its the latest version: 29.6 Build 6

Q: Which Torture Test?

Custom
MinFFT size (in K) = 4, MaxFFT size (in K) = 8192
Memory to use (in MB) = 26000 (I have 32GB but didn't want to trash my SSD with page files)
Time to run each FFT size (in minutes) = 6

---

How ever this morning I saw that 1.280 volts didn't remain stable during the night. system was frozen.

So now its running @1.296 volts.

Max temps captured via HWMonitor after 6 hours

Core#0 : 76C
Core#1: 78C
Core#2: 72C
Core#3: 73C

So far its stable after 6 hours. I may leave this for another 6 hours.
 

Mkazem

Commendable
Jul 18, 2016
60
0
1,540
2
my chip won't go lower than 1,315V on 4.5 GHZ

I used OCCT and 3Dmark to test for stability...

didn't exceed your temps though was about 72 C.

Low ambient 19-20 C... wool socks on wintertime gaming :D
Yeah, my room is a little hot :)

for me it could get to windows @1.245v but went blue screen after torture test 💩
 

Bearmann

Admirable
I had almost the exact same experience with my i5-6600K. I had to drop it back to 4.5 GHz from 4.6GHz for stability around 2 years later. The chip degrades a bit over time and the built in headroom by the manufacturer has been used by the overclock.
 
Last edited:

CompuTronix

Splendid
Moderator
Q: Which version of Prime95?

I think its the latest version: 29.6 Build 6

Q: Which Torture Test?

Custom
MinFFT size (in K) = 4, MaxFFT size (in K) = 8192
Memory to use (in MB) = 26000 (I have 32GB but didn't want to trash my SSD with page files)
Time to run each FFT size (in minutes) = 6

---

How ever this morning I saw that 1.280 volts didn't remain stable during the night. system was frozen.

So now its running @1.296 volts.

Max temps captured via HWMonitor after 6 hours

Core#0 : 76C
Core#1: 78C
Core#2: 72C
Core#3: 73C

So far its stable after 6 hours. I may leave this for another 6 hours.
Mkazem,

The latest version of Prime95 is 29.4, which is probably what you ran.

Versions later than 26.6 run AVX code which can overheat your processor due to an overload of nearly 130%. Although you haven't mentioned your cooling solution, the custom P95 configuration you've chosen is a variation on the default test "Blend", which is a fluctuating workload well suited for testing memory stability. As your custom test configuration cycles through the lower FFT's sizes such as 4K, the CPU rather than memory is then primarily stressed during those test segments, which causes peak Core temperatures. As Small FFT's is a steady-state 100% TDP workload which yields steady Core temperatures, the small FFT's test selection is ideally suited for testing CPU thermal performance .

Even though your ambient temperature is 4°C above normal (22°C), since decreasing your overclock and Vcore, your Core temperatures are reasonable. However, in addition to negatively affecting Core temperatures, AVX has a negative affect on stability. If you run AVX workloads such as for rendering or transcoding, then you should configure "AVX offset" in BIOS for at least 3 (300MHz downclock), which will reduce your Core temperatures and increase stability without compromising your overclock for gaming performance.

If you don't run AVX apps, then there's no point in testing for AVX stability. You'll find that if you instead run Prime95 version 26.6 (non-AVX), you can revert back to your previous overclock settings while maintaining stability using lower Vcore with lower Core temperatures.

Prime95 v26.6 - http://www.mersenneforum.org/showthread.php?t=15504

Keep in mind that as per Intel's Datasheets, TDP and Thermal Specifications are validated "without AVX".

You might want to read Section 12 - Thermal Test 100% Workload: Intel Temperature Guide - https://forums.tomshardware.com/threads/intel-temperature-guide.1488337/

CT 😎
 
Were stability concerns about temps or bsods?

It's true prime95 regarding avx instructions in the later versions you shouldn't run. Temps are accelerated much above normal use.

Have you monitored temps doing something else?

I get the feeling everything was good (temp wise) all these years until you tried prime95 above 26.6 then thought there was a problem? Unless you actually had bsods?
 

Mkazem

Commendable
Jul 18, 2016
60
0
1,540
2
Were stability concerns about temps or bsods?

It's true prime95 regarding avx instructions in the later versions you shouldn't run. Temps are accelerated much above normal use.

Have you monitored temps doing something else?

I get the feeling everything was good (temp wise) all these years until you tried prime95 above 26.6 then thought there was a problem? Unless you actually had bsods?
not bsods but game freezing issues.

I play Startcraft 2, I had freezing issue in start craft 2, after 20 minutes it would freeze. at first i thought its the game issue

I don't have graphics card installed, so I'm using iGPU of 6700k. maybe the memory pressure caused the issue so i did 8 hours of Memtest86 (boot from USB), but it did not catch any errors.

I tried a lot of settings in the game to avoid freeze until i realized if i set affinity of Starcraft 2 process from 8 to only 4 cores, freezing issue would solve.

this was annoying, I had to set affinity every time and also something was wrong. I did prime95 test and realized my PC is not stable, I got bsods. (I did prime test about year ago and it was stable)

with 100MHz less, problems seems to be solved. now i don't have to set affinity and i can play at highest graphic settings with 8 cores.
 
Last edited:

ASK THE COMMUNITY