F5F Stay Refreshed Power Users Overclocking GPU overclocked yet outcomes remain unchanged

GPU overclocked yet outcomes remain unchanged

GPU overclocked yet outcomes remain unchanged

C
CallyCallis
Member
53
04-05-2017, 02:17 AM
#1
I think I might have made a mistake, which is why I'm asking for help.
I own an EVGA 1070 ti FTW2 that I overclocked with Precision OC.
When I run the Superposition benchmark without the overclock enabled, the results are:
Graphics: 1835
Memory: 4006
Temp: 68 deg C
Utilization: 98%
Avg FPS: 107.2
Results: 14308

If I run the same test with the overclock active, the numbers change slightly:
Graphics: 1835
Memory: 4006
Temp: 69 deg C
Utilization: 98%
Avg FPS: 106.7
Results: 14266

It seems either the overclock isn't doing much, it's not working at all, or I'm misunderstanding the data. I contacted EVGA and they advised me to set the power target up to 130%.
I really need some advice here... thanks.
C
CallyCallis
04-05-2017, 02:17 AM #1

I think I might have made a mistake, which is why I'm asking for help.
I own an EVGA 1070 ti FTW2 that I overclocked with Precision OC.
When I run the Superposition benchmark without the overclock enabled, the results are:
Graphics: 1835
Memory: 4006
Temp: 68 deg C
Utilization: 98%
Avg FPS: 107.2
Results: 14308

If I run the same test with the overclock active, the numbers change slightly:
Graphics: 1835
Memory: 4006
Temp: 69 deg C
Utilization: 98%
Avg FPS: 106.7
Results: 14266

It seems either the overclock isn't doing much, it's not working at all, or I'm misunderstanding the data. I contacted EVGA and they advised me to set the power target up to 130%.
I really need some advice here... thanks.

J
Jarranield
Member
161
04-12-2017, 10:10 PM
#2
You didn't overclock your GPU. If you had, the graphics and memory clocks would be higher. Currently they are 1835 and 4006. Open Precision X, adjust the power slider to 120%, add 100mhz to the base clock, and run the test. Check for crashes, adjusting in small steps until stable. Return to 0 and reset the memory. Begin with a higher value like 300, test both at once. If needed, raise voltage and repeat.
J
Jarranield
04-12-2017, 10:10 PM #2

You didn't overclock your GPU. If you had, the graphics and memory clocks would be higher. Currently they are 1835 and 4006. Open Precision X, adjust the power slider to 120%, add 100mhz to the base clock, and run the test. Check for crashes, adjusting in small steps until stable. Return to 0 and reset the memory. Begin with a higher value like 300, test both at once. If needed, raise voltage and repeat.

X
xEIgoldo
Member
219
04-13-2017, 11:35 AM
#3
You didn't overclock your GPU. If you had, the graphics and memory clocks would be higher. Currently they're both at 1835 and 4006.

Open Precision X, adjust the power slider to 120%. Add 100mhz to the base clock and run the test. Check for crashes. Adjust in small steps of 5-10 until stable. Return to 0 and reset the memory. Begin with a higher value like 300. Test again, watching for artifacts. Increase by 25mhz intervals until you notice them. You should likely reach around 500mhz on the memory.

After that, revert your clock speed to what was stable earlier and test both together. If needed, raise the voltage and try again.
X
xEIgoldo
04-13-2017, 11:35 AM #3

You didn't overclock your GPU. If you had, the graphics and memory clocks would be higher. Currently they're both at 1835 and 4006.

Open Precision X, adjust the power slider to 120%. Add 100mhz to the base clock and run the test. Check for crashes. Adjust in small steps of 5-10 until stable. Return to 0 and reset the memory. Begin with a higher value like 300. Test again, watching for artifacts. Increase by 25mhz intervals until you notice them. You should likely reach around 500mhz on the memory.

After that, revert your clock speed to what was stable earlier and test both together. If needed, raise the voltage and try again.

H
hahahalo21
Junior Member
46
04-13-2017, 02:22 PM
#4
Thank you...thank you. I'll try that. Also, using the benchmark in high 1080p seems appropriate, rather than extreme or medium; is that the correct starting point?

Regarding the temperatures, the power target at 130% and the temperature target at 92°C—are they safe to maintain?
H
hahahalo21
04-13-2017, 02:22 PM #4

Thank you...thank you. I'll try that. Also, using the benchmark in high 1080p seems appropriate, rather than extreme or medium; is that the correct starting point?

Regarding the temperatures, the power target at 130% and the temperature target at 92°C—are they safe to maintain?

D
Dephunkpunk_2
Senior Member
484
04-14-2017, 09:34 AM
#5
Unexpectedly cold readings, ensure the temperate setting isn't at 73C or higher. Many default to 83C, which is quite surprising since it didn't hit either. Most Pascal GPUs perform well between 2000-2100Mhz. Remember, cooler temperatures can often improve consistency with Pascal, meaning higher power doesn't always help.
D
Dephunkpunk_2
04-14-2017, 09:34 AM #5

Unexpectedly cold readings, ensure the temperate setting isn't at 73C or higher. Many default to 83C, which is quite surprising since it didn't hit either. Most Pascal GPUs perform well between 2000-2100Mhz. Remember, cooler temperatures can often improve consistency with Pascal, meaning higher power doesn't always help.

N
Noctaflo
Member
118
04-14-2017, 09:57 AM
#6
GPU clock: +233 MHz
Mem clock: +548 MHz
Max Temp: 68 deg C, auto fan setting
Results: 11605 at 1080p high
I was wondering if adjusting these settings would help create a guard band. Thanks again for the guidance.
N
Noctaflo
04-14-2017, 09:57 AM #6

GPU clock: +233 MHz
Mem clock: +548 MHz
Max Temp: 68 deg C, auto fan setting
Results: 11605 at 1080p high
I was wondering if adjusting these settings would help create a guard band. Thanks again for the guidance.

T
Tristcuit_
Member
53
04-15-2017, 10:45 AM
#7
223 offers a great overclock potential. I don’t own an 1070 Ti, yet none of my Pascal cards could reach that speed. I’m not sure what you’re referring to by a guard band.
T
Tristcuit_
04-15-2017, 10:45 AM #7

223 offers a great overclock potential. I don’t own an 1070 Ti, yet none of my Pascal cards could reach that speed. I’m not sure what you’re referring to by a guard band.

Z
Zazzery
Member
69
04-18-2017, 12:15 PM
#8
Feelinfroggy777 shared some useful tips. The overclock worked well on my system, but other Pascal cards couldn’t reach that speed. I’m not sure what you meant by a guard band. Thanks for the advice; it really helped. The results were satisfying. Adjusting the memory clock to +10 gave stable performance with the GPU at +248 and memory at +548, avoiding crashes. Reducing the CPU clock to +223 and keeping memory at +548 also worked well. Guard band refers to setting a buffer between instability and optimal performance—found the GPU instability points and adjusted accordingly. I’m grateful for your guidance!
Z
Zazzery
04-18-2017, 12:15 PM #8

Feelinfroggy777 shared some useful tips. The overclock worked well on my system, but other Pascal cards couldn’t reach that speed. I’m not sure what you meant by a guard band. Thanks for the advice; it really helped. The results were satisfying. Adjusting the memory clock to +10 gave stable performance with the GPU at +248 and memory at +548, avoiding crashes. Reducing the CPU clock to +223 and keeping memory at +548 also worked well. Guard band refers to setting a buffer between instability and optimal performance—found the GPU instability points and adjusted accordingly. I’m grateful for your guidance!