Overclock.net banner
1 - 8 of 8 Posts

·
Registered
Joined
·
26 Posts
Discussion Starter · #1 · (Edited)
Hello!

with my 5900x i am observing an 8-9K Temperature Delta at full load.
see screenshot.
Curve Settings are all in range of -21-30 with exception of core0 at -16. So i dont suspect it to be accountable.
Cooler is an ekwb supremacy evo with about 85l/h flow rate.
is it worth considering a repaste with thinner TIM Layer or could it be a soldering issue from Packaging?

edit: i included boost testing results in screenshot.

2517867
 

·
Registered
Joined
·
200 Posts
Most of the 5900X have that delta, more or less.
Havent found an explanation but they have.
Mine is 3-4 grades.
 

·
Registered
Joined
·
5,969 Posts
well, for one, ccx 2 has only 4 active cores instead of 8 so probably less localized heat. Secondly, the scheduler typically favors ccx 1 cores so they tend to carry heavier load.
 

·
Registered
Joined
·
2,909 Posts

·
Iconoclast
Joined
·
31,434 Posts
CCD temp differentials in this range are not uncommon, even for perfectly lapped parts, irrespective of TIM. You might be able to improve the differential slightly, but the ultimate cause is usually that the first CCD is just hotter.

Higher clocking dies tend to be leakier and slightly more power hungry even at the same clocks/voltages, and since it will be boosting higher anyway, it's very likely for the better clocking die to be running hotter. This isn't universally true and can be strongly influenced by one's custom curves, but is a general trend among these parts.

If you want to know if there is an actual problem cooling one of the CCDs, you need to equalize their power consumption/output. Test with a fixed voltage at fixed clocks and compare power, then reduce the clock of the hotter CCD until the cores have roughly the same power as the cooler CCD--temperature itself will influence power consumption, but it should not be meaningful if the delta is not more extreme than what you're seeing. If there is still more than a few C temperature differential, then some sort of cooling imbalance may well be the cause.

i included boost testing results in screenshot.
Look at the power figures for each core. More power means more heat, all other things being equal.
 

·
Registered
Joined
·
26 Posts
Discussion Starter · #6 ·
CCD temp differentials in this range are not uncommon, even for perfectly lapped parts, irrespective of TIM. You might be able to improve the differential slightly, but the ultimate cause is usually that the first CCD is just hotter.

Higher clocking dies tend to be leakier and slightly more power hungry even at the same clocks/voltages, and since it will be boosting higher anyway, it's very likely for the better clocking die to be running hotter. This isn't universally true and can be strongly influenced by one's custom curves, but is a general trend among these parts.

If you want to know if there is an actual problem cooling one of the CCDs, you need to equalize their power consumption/output. Test with a fixed voltage at fixed clocks and compare power, then reduce the clock of the hotter CCD until the cores have roughly the same power as the cooler CCD--temperature itself will influence power consumption, but it should not be meaningful if the delta is not more extreme than what you're seeing. If there is still more than a few C temperature differential, then some sort of cooling imbalance may well be the cause.



Look at the power figures for each core. More power means more heat, all other things being equal.
Thanks for the input.
i did a boost test with slightly cooler water at about 23-24C

CTR BOOST TESTER RESULTS (test version)
CORE / FREQUENCY / VID / POWER / TEMP
C01F 4929V 1.458W 16.42T 59.46
C02F 4922V 1.447W 15.95T 59.23
C03F 4914V 1.451W 16.12T 58.15
C04F 4939V 1.473W 17.47T 63.15
C05F 4898V 1.458W 16.6T 61.54
C06F 4900V 1.432W 15.8T 61.8
C07F 4824V 1.492W 12.39T 52.34
C08F 4899V 1.493W 12.27T 53.27
C09F 4860V 1.495W 12.42T 54.74
C10F 4900V 1.489W 12.49T 53.04
C11F 4818V 1.474W 11.9T 54.39
C12F 4835V 1.473W 12.2T 55.49

compared to about 32C coolant

CTR BOOST TESTER RESULTS (test version)
CORE / FREQUENCY / VID / POWER / TEMP
C01F 4884V 1.436W 16.06T 65.02
C02F 4851V 1.431W 15.92T 66.19
C03F 4852V 1.441W 16.48T 65.67
C04F 4894V 1.422W 15.87T 66.94
C05F 4824V 1.418W 16.02T 66.2
C06F 4836V 1.401W 15.12T 66.82
C07F 4772V 1.493W 12.26T 59.18
C08F 4848V 1.49W 12.21T 59.22
C09F 4800V 1.486W 12.08T 60.9
C10F 4850V 1.49W 12.32T 60.13
C11F 4750V 1.447W 11.28T 59.39
C12F 4773V 1.452W 11.95T 61.07

the higher cppc cores on ccx1 run at less voltage with roughly same clock but higher wattage. that is quite an odd picture; but possibly normal on those cpus.
i will try a repaste tomorrow and have a second look at it.
 

·
Overclock the World
Joined
·
3,123 Posts
Thanks for the input.
i did a boost test with slightly cooler water at about 23-24C

CTR BOOST TESTER RESULTS (test version)
CORE / FREQUENCY / VID / POWER / TEMP
C01F 4929V 1.458W 16.42T 59.46
C02F 4922V 1.447W 15.95T 59.23
C03F 4914V 1.451W 16.12T 58.15
C04F 4939V 1.473W 17.47T 63.15
C05F 4898V 1.458W 16.6T 61.54
C06F 4900V 1.432W 15.8T 61.8
C07F 4824V 1.492W 12.39T 52.34
C08F 4899V 1.493W 12.27T 53.27
C09F 4860V 1.495W 12.42T 54.74
C10F 4900V 1.489W 12.49T 53.04
C11F 4818V 1.474W 11.9T 54.39
C12F 4835V 1.473W 12.2T 55.49

compared to about 32C coolant

CTR BOOST TESTER RESULTS (test version)
CORE / FREQUENCY / VID / POWER / TEMP
C01F 4884V 1.436W 16.06T 65.02
C02F 4851V 1.431W 15.92T 66.19
C03F 4852V 1.441W 16.48T 65.67
C04F 4894V 1.422W 15.87T 66.94
C05F 4824V 1.418W 16.02T 66.2
C06F 4836V 1.401W 15.12T 66.82
C07F 4772V 1.493W 12.26T 59.18
C08F 4848V 1.49W 12.21T 59.22
C09F 4800V 1.486W 12.08T 60.9
C10F 4850V 1.49W 12.32T 60.13
C11F 4750V 1.447W 11.28T 59.39
C12F 4773V 1.452W 11.95T 61.07

the higher cppc cores on ccx1 run at less voltage with roughly same clock but higher wattage. that is quite an odd picture; but possibly normal on those cpus.
i will try a repaste tomorrow and have a second look at it.
Core 7,8,9,10 frequency throttle by reaching 1.5v VID
More negative CO helps bad cores ~ as they are nearly never "bad" in reaching the frequency. But rather overvolted
2518241

1 - 2 - 3
-18,-16,-14,
4 - 5 - 6
-20,-22,-13
(my setup)

 

·
Registered
Joined
·
26 Posts
Discussion Starter · #8 ·
Core 7,8,9,10 frequency throttle by reaching 1.5v VID
More negative CO helps bad cores ~ as they are nearly never "bad" in reaching the frequency. But rather overvolted
View attachment 2518241
1 - 2 - 3
-18,-16,-14,
4 - 5 - 6
-20,-22,-13
Boost +Core01234567891011
CPPC181177168181164173147156143160139152
150CO
-14​
-24​
-26​
-21​
-27​
-27​
-24​
-28​
-30​
-27​
-28​
-27​
VID
1,432​
1,437​
1,456​
1,423​
1,431​
1,4​
1,49​
1,489​
1,486​
1,492​
1,449​
1,452​
Boost MHz
4920​
4850​
4864​
4894​
4807​
4810​
4774​
4850​
4803​
4868​
4752​
4774​

Hello!

these are current CO settings. ccx2 is as tuned down as it gets. tested stable corecycler up to prime95 sse/avx2 moderate/heavy and y-cruncher kageri 30min per core.
i was wondering why VID goes that way up on ccx2 even with those negative offsets.
Possibly PBO trying to sqeeze out the bad cores to max Frequency. With TIM repaste they merely go above 50-53C. ccx1 is still hotter at 56-61C while boost testing.

edit: table formatting is a bit off from C&P
 
1 - 8 of 8 Posts
Top