CPU overclocking results confusing me

[AverageLeo]

Did 5 tests with i7-7700k and CineBench, got this results:

1: 4.2 1.000v 2285 temp 82 = This is the default stats of the CPU.
2. 4.9 1.250v 2461 temp 87
3. 4.9 1.300v 2438 temp 96
4. 4.7 1.200v 2341 temp 83
5. 4.8 1.250v 2289 temp 89

 

What the heck am I looking at and what you experts think I should do/check?

Thanks

[Jurrunio]

thermal throttling...

[AverageLeo]

2 minutes ago, Jurrunio said:

thermal throttling...

which means? 

 

[DoctorNick]

2 minutes ago, Jurrunio said:

thermal throttling...

This and get a new cooler - maybe even consider delidding the CPU - if 1.000v is default voltage on 4.2ghz in Cinebench, then congrats, you should be able to OC pretty well with a decent cpu cooler and a good ol' delid.

[AverageLeo]

Just now, DoctorNick said:

This and get a new cooler - maybe even consider delidding the CPU - if 1.000v is default voltage on 4.2ghz in Cinebench, then congrats, you should be able to OC pretty well with a decent cpu cooler and a good ol' delid.

I use kraken x62 AIO which supposed to be great, maybe I didn't attached it well?

And about the thermal throttling, it means I have bad airflow in my system?

[Mateyyy]

3 minutes ago, AverageLeo said:

which means? 

At a certain temperature threshold (95-100C), the CPU will start lowering its speed in an attempt to lower its temperature for safety, so despite you setting a particular frequency, like 4.9GHz, it's probably actually running at something lower during full load.

Check clocks with CPU-Z, and temps with HWiNFO or HWMonitor.

[Mister Woof]

12 minutes ago, AverageLeo said:

Did 5 tests with i7-7700k and CineBench, got this results:

1: 4.2 1.000v 2285 temp 82 = This is the default stats of the CPU.
2. 4.9 1.250v 2461 temp 87
3. 4.9 1.300v 2438 temp 96
4. 4.7 1.200v 2341 temp 83
5. 4.8 1.250v 2289 temp 89

 

What the heck am I looking at and what you experts think I should do/check?

Thanks

well for one you'd want to do multiple runs on each setting to get an average, not 5 total with different settings.

two, what does temp mean? is that average, max, what? 

 

 

[AverageLeo]

1 minute ago, Mateyyy said:

At a certain temperature threshold (95-100C), the CPU will start lowering its speed in an attempt to lower its temperature for safety, so despite you setting a particular frequency, like 4.9GHz, it's probably actually running at something lower during full load.

Check clocks with CPU-Z, and temps with HWiNFO or HWMonitor.

Got it. So nothing to do here?

1 minute ago, Mister Woof said:

well for one you'd want to do multiple runs on each setting to get an average, not 5 total with different settings.

two, what does temp mean? is that average, max, what? 

 

 

Thanks for the tip,

and it's max temp.

[Mateyyy]

Just now, AverageLeo said:

Got it. So nothing to do here?

Well you should probably try going back to stock settings and starting again, but little by little this time in adjusting the clock speed. I'd leave the voltage on auto, increase the multiplier until it's not stable, then increase the voltage accordingly. Meanwhile make sure your temperatures are in check and that you're not running into any power constraints, if your PSU is too weak.

[AverageLeo]

1 minute ago, Mateyyy said:

Well you should probably try going back to stock settings and starting again, but little by little this time in adjusting the clock speed. I'd leave the voltage on auto, increase the multiplier until it's not stable, then increase the voltage accordingly. Meanwhile make sure your temperatures are in check and that you're not running into any power constraints, if your PSU is too weak.

PSU is strong, Corsair RMx 750.

is this temp "4.2 1.000v 2285 temp 82" is normal for default settings benchmark?

cuz if it's too hot maybe I have cooling problems.

 

 

[Mateyyy]

9 minutes ago, AverageLeo said:

PSU is strong, Corsair RMx 750.

is this temp "4.2 1.000v 2285 temp 82" is normal for default settings benchmark?

cuz if it's too hot maybe I have cooling problems.

82C sounds fairly high imo for a 7700K at stock speeds with a 280mm AIO, while running Cinebench, not ridiculously high either though. 

Make sure the fans attached to the radiator have ample airflow around them (assuming you've set them up as exhausts in the top), and try increasing the pump speed too and see if it makes any difference. I wouldn't worry about a bad mount unless idle temps are also very high, like 60-70C+.

[Mister Woof]

23 minutes ago, AverageLeo said:

Got it. So nothing to do here?

Thanks for the tip,

and it's max temp.

If indeed that's max temp, i doubt any of those results hit thermal CPU throttling (maybe #3, but def none of the results in the 80s).  

 

There won't be any performance hit from the CPU side until it it starts throttling at 100c.

 

It looks like from best to worst you have a 7.x% performance increase and a 16% clock speed increase. 

 

Seems a bit low, (my 8700k gets ~15% better scores with a 16% increase in frequency), but I don't think it's thermal on the CPU side (if those are indeed max temps).

 

Maybe your VRM is throttling. What motherboard and do you have any cooling over the VRM components? What is your VRM temperature?

[AverageLeo]

59 minutes ago, Mister Woof said:

If indeed that's max temp, i doubt any of those results hit thermal CPU throttling (maybe #3, but def none of the results in the 80s).  

 

There won't be any performance hit from the CPU side until it it starts throttling at 100c.

 

It looks like from best to worst you have a 7.x% performance increase and a 16% clock speed increase. 

 

Seems a bit low, (my 8700k gets ~15% better scores with a 16% increase in frequency), but I don't think it's thermal on the CPU side (if those are indeed max temps).

 

Maybe your VRM is throttling. What motherboard and do you have any cooling over the VRM components? What is your VRM temperature?

Not sure what you're talking about  

My motherboard is Asus PRIME Z-270A.

[Mister Woof]

56 minutes ago, AverageLeo said:

Not sure what you're talking about  

My motherboard is Asus PRIME Z-270A.

 

Intel CPUs will operate at full power up until tjmax, which is 100c. They generally won't throttle until they hit that.

 

Quote

1: 4.2 1.000v 2285 temp 82 = This is the default stats of the CPU.
2. 4.9 1.250v 2461 temp 87
3. 4.9 1.300v 2438 temp 96
4. 4.7 1.200v 2341 temp 83
5. 4.8 1.250v 2289 temp 89

 

Your best score was 2461 @ 4.9ghz, and your worse was 2285 at 4.2ghz. That's a 7.7% difference in score and a 16.6% difference in clock speed. 

 

VRM is the power delivery capability of the motherboard. There are components on the motherboard that will generate heat, and they themselves will throttle if overheaded, won't provide enough power to the CPU to perform.

 

#2 and #3 are close in score, with the difference being 50mv and 9c. It's possible #3 scored worse because it may be thermal throttled between sensor polling rates. 

 

#1, #2, and #5 are the values that make me think your VRM is throttling at higher voltages, not your CPU. In those scenarios, CPU is well within TJmax, and shouldn't throttle.

 

However, the #5 result had nearly the same performance as the #1 result, despite having 14% more clock speed. Since neither of them thermal throttled, how can this be explained? The difference is that the #1 result only was 1.0v, whereas the #5 result was 1.25v.

 

Additionally, #4 got better results than 5, even though neither thermal throttled and it had less clock speed.

 

That doesn't completely explain everything, because you were able to get better values with #2 and #3, and those also had higher voltage values.

 

But I am pretty sure so far your CPU itself is not thermal throttling (with the exception of #3). Presuming your data is accurate.

 

Of course, your data might be hard to judge, because you only have one run per setting. You need to do multiple runs with very tightly controlled variables. I can run cinebench r20 three times in a row and get as high as 4010 and as low as 3920 with no apparent changes in anything and with temps well under control. Need an average.

 

To be sure you aren't thermal throttling or power limit throttling or something, download XTU and run a stress test. See what it reports. Although I don't think you are thermal throttling with those temps, I could be wrong or your sensors could be wrong.

 

 

[AverageLeo]

5 hours ago, Mister Woof said:

 

Intel CPUs will operate at full power up until tjmax, which is 100c. They generally won't throttle until they hit that.

 

 

Your best score was 2461 @ 4.9ghz, and your worse was 2285 at 4.2ghz. That's a 7.7% difference in score and a 16.6% difference in clock speed. 

 

VRM is the power delivery capability of the motherboard. There are components on the motherboard that will generate heat, and they themselves will throttle if overheaded, won't provide enough power to the CPU to perform.

 

#2 and #3 are close in score, with the difference being 50mv and 9c. It's possible #3 scored worse because it may be thermal throttled between sensor polling rates. 

 

#1, #2, and #5 are the values that make me think your VRM is throttling at higher voltages, not your CPU. In those scenarios, CPU is well within TJmax, and shouldn't throttle.

 

However, the #5 result had nearly the same performance as the #1 result, despite having 14% more clock speed. Since neither of them thermal throttled, how can this be explained? The difference is that the #1 result only was 1.0v, whereas the #5 result was 1.25v.

 

Additionally, #4 got better results than 5, even though neither thermal throttled and it had less clock speed.

 

That doesn't completely explain everything, because you were able to get better values with #2 and #3, and those also had higher voltage values.

 

But I am pretty sure so far your CPU itself is not thermal throttling (with the exception of #3). Presuming your data is accurate.

 

Of course, your data might be hard to judge, because you only have one run per setting. You need to do multiple runs with very tightly controlled variables. I can run cinebench r20 three times in a row and get as high as 4010 and as low as 3920 with no apparent changes in anything and with temps well under control. Need an average.

 

To be sure you aren't thermal throttling or power limit throttling or something, download XTU and run a stress test. See what it reports. Although I don't think you are thermal throttling with those temps, I could be wrong or your sensors could be wrong.

 

 

thanks a lot for the reply, appreciated.

I disabled XMP in bios which apparently "overclocked" my ram even tho I was at max speed for the motherboard (max ram speed 3000mhz, max motherboard support 2200mhz) and played a bit with the stats.

new results here:

1: 4.2 1.000v 2128 max temp 75
   4.2 1.000v 2159 max temp 76
2. 4.8 1.250v 2402 max temp 84
    4.8 1.250v 2290 max temp 88
3. 4.9 4.4 1.275v 2458 max temp 79
    4.9 4.4 1.280v 2457 max temp 80 crash
4. 4.8 4.4 1.275v 2460 max temp 80

Any suggestion?

XTU doesn't report any throttling or errors, survived stress tests and benchmarks.

 

[Mister Woof]

1 minute ago, AverageLeo said:

thanks a lot for the reply, appreciated.

I disabled XMP in bios which apparently "overclocked" my ram even tho I was at max speed for the motherboard (max ram speed 3000mhz, max motherboard support 2200mhz) and played a bit with the stats.

new results here:

1: 4.2 1.000v 2128 max temp 75
   4.2 1.000v 2159 max temp 76
2. 4.8 1.250v 2402 max temp 84
    4.8 1.250v 2290 max temp 88
3. 4.9 4.4 1.275v 2458 max temp 79
    4.9 4.4 1.280v 2457 max temp 80 crash
4. 4.8 4.4 1.275v 2460 max temp 80

Any suggestion?

XTU doesn't report any throttling or errors, survived stress tests and benchmarks.

 

i'd say get your ram to the right speed, xmp profile on and manually set dram voltage to 1.35v. run something like A64 or Realbench to test for stability and keep an eye on temps and voltages.

 

Also, check to see what others are getting for those clock speeds. compare results.