F5F Stay Refreshed Power Users Overclocking AMD Ryzen 1600

AMD Ryzen 1600

AMD Ryzen 1600

R
RauloMenino
Member
224
10-06-2017, 12:34 PM
#1
I'm currently working on a blend test in prime95. I'm wondering what other users tend to overclock this chip to. They mentioned they usually cap at around 3.6 to 7ghz at 1.35v. My experience was different: after adjusting the settings to 3.7 and 1.3v or 1.35v, the CPU still wouldn't exceed 3.39ghz. Higher voltages or clock speeds forced me to lower RAM speeds, which seemed to be a trade-off. Eventually, I decided to try again. By this point, I had managed to keep RAM stable at 2800 but struggled to push the CPU beyond 3.39ghz. I reduced the voltage to 1.25 and increased the RAM speed, and it worked—Prime95 responded well. The question is: why does lowering voltage seem to unlock system potential and boost performance? Increasing voltage generally just speeds up wear or damage rather than improving efficiency. Would your thoughts be similar?

P.S. The system remains stable at 3.7ghz with 1.25v, though the temperature is still rising—around 68°C. I'm interested in how others handle similar temperatures and their conclusions. Check out the build: https://pcpartpicker.com/b/w3mqqs
R
RauloMenino
10-06-2017, 12:34 PM #1

I'm currently working on a blend test in prime95. I'm wondering what other users tend to overclock this chip to. They mentioned they usually cap at around 3.6 to 7ghz at 1.35v. My experience was different: after adjusting the settings to 3.7 and 1.3v or 1.35v, the CPU still wouldn't exceed 3.39ghz. Higher voltages or clock speeds forced me to lower RAM speeds, which seemed to be a trade-off. Eventually, I decided to try again. By this point, I had managed to keep RAM stable at 2800 but struggled to push the CPU beyond 3.39ghz. I reduced the voltage to 1.25 and increased the RAM speed, and it worked—Prime95 responded well. The question is: why does lowering voltage seem to unlock system potential and boost performance? Increasing voltage generally just speeds up wear or damage rather than improving efficiency. Would your thoughts be similar?

P.S. The system remains stable at 3.7ghz with 1.25v, though the temperature is still rising—around 68°C. I'm interested in how others handle similar temperatures and their conclusions. Check out the build: https://pcpartpicker.com/b/w3mqqs

B
BenTGreat
Senior Member
437
10-11-2017, 10:44 PM
#2
Do not execute any Prime95 version other than 26.6 unless you are certain your applications heavily rely on AVX encoding. Running newer versions could create excessive thermal and computational demands, making them unsuitable. Stick to Prime v26.6 only. Test for 15 to 24 hours after each adjustment to confirm stability. This ensures minimal risk of errors, data loss, or crashes due to instability and frequent bit errors.
B
BenTGreat
10-11-2017, 10:44 PM #2

Do not execute any Prime95 version other than 26.6 unless you are certain your applications heavily rely on AVX encoding. Running newer versions could create excessive thermal and computational demands, making them unsuitable. Stick to Prime v26.6 only. Test for 15 to 24 hours after each adjustment to confirm stability. This ensures minimal risk of errors, data loss, or crashes due to instability and frequent bit errors.

S
Streiyn
Posting Freak
768
10-12-2017, 05:16 AM
#3
Stop using Blend. It isn't suitable for checking CPU stability or thermal limits. While it mainly tests memory, it often runs with more memory than available, requiring virtual memory. This forces the hard drive to access data and the CPU to pause during that time, preventing a consistent workload. Therefore, no steady state is enforced.

You must always use Prime95 version 26.6, specifically version 26.6, and run the Small FFT option. Any other method is either ineffective or will produce incorrect results.
S
Streiyn
10-12-2017, 05:16 AM #3

Stop using Blend. It isn't suitable for checking CPU stability or thermal limits. While it mainly tests memory, it often runs with more memory than available, requiring virtual memory. This forces the hard drive to access data and the CPU to pause during that time, preventing a consistent workload. Therefore, no steady state is enforced.

You must always use Prime95 version 26.6, specifically version 26.6, and run the Small FFT option. Any other method is either ineffective or will produce incorrect results.

F
56
10-12-2017, 05:24 AM
#4
Interesting... all the sources I found suggest the "blend" setting is the most balanced choice. These reports also indicated that the blend evaluation was adequate enough to determine system stability for regular use. There were hints that the blend test might not be enough for heavy-duty tasks like 24-hour video encoding, but for someone mainly doing light gaming, it could work. I’m also worried about the prime95 test results. The maximum memory tested was 1500mhz, and the speed stayed steady at a constant rate. The CPU was kept at full capacity at 3.691ghz for 20 hours. Can you help clarify this? It seems like our findings might be conflicting. I think the "blend" option is more suited to CPU-heavy work and less so for memory-heavy tasks, which is why my result.
F
feral_blue_cat
10-12-2017, 05:24 AM #4

Interesting... all the sources I found suggest the "blend" setting is the most balanced choice. These reports also indicated that the blend evaluation was adequate enough to determine system stability for regular use. There were hints that the blend test might not be enough for heavy-duty tasks like 24-hour video encoding, but for someone mainly doing light gaming, it could work. I’m also worried about the prime95 test results. The maximum memory tested was 1500mhz, and the speed stayed steady at a constant rate. The CPU was kept at full capacity at 3.691ghz for 20 hours. Can you help clarify this? It seems like our findings might be conflicting. I think the "blend" option is more suited to CPU-heavy work and less so for memory-heavy tasks, which is why my result.

K
KleinerRaker
Member
58
10-12-2017, 05:31 AM
#5
I will definitely examine what you are saying.
Thank you.
K
KleinerRaker
10-12-2017, 05:31 AM #5

I will definitely examine what you are saying.
Thank you.

N
nathanspike
Member
171
10-12-2017, 07:14 AM
#6
Your conclusion may be incorrect, as the Prime95 source also supports a different perspective. This information comes from Computronix, one of possibly only a few individuals I have known who could clearly discuss the method with sound understanding.
N
nathanspike
10-12-2017, 07:14 AM #6

Your conclusion may be incorrect, as the Prime95 source also supports a different perspective. This information comes from Computronix, one of possibly only a few individuals I have known who could clearly discuss the method with sound understanding.

A
Amegahoney
Posting Freak
789
10-19-2017, 08:45 PM
#7
Here’s a revised version of your text:

Thank you for this helpful information. I read that the latest prime95 version is more optimized for Ryzen processors.
It seems many people mentioned that adjustments are needed, and each release focuses on different CPU features. This relates to the underlying architecture... "zen" please confirm if this makes sense.
I’ve been using version 26 and it’s performing well.
However, I wasn’t able to maintain the RAM at 1.35 volts at 2993 MHz. It’s stable at 3.7 GHz and 2800 MHz.
I’m thinking about increasing the voltage on the RAM in the hope of reaching 2993.
Do you know how RAM voltage and CPU speed are connected?
At some point, raising the voltage and MHz on the CPU can cause a voltage drop; how does this affect RAM voltage, timing, or speed?
A
Amegahoney
10-19-2017, 08:45 PM #7

Here’s a revised version of your text:

Thank you for this helpful information. I read that the latest prime95 version is more optimized for Ryzen processors.
It seems many people mentioned that adjustments are needed, and each release focuses on different CPU features. This relates to the underlying architecture... "zen" please confirm if this makes sense.
I’ve been using version 26 and it’s performing well.
However, I wasn’t able to maintain the RAM at 1.35 volts at 2993 MHz. It’s stable at 3.7 GHz and 2800 MHz.
I’m thinking about increasing the voltage on the RAM in the hope of reaching 2993.
Do you know how RAM voltage and CPU speed are connected?
At some point, raising the voltage and MHz on the CPU can cause a voltage drop; how does this affect RAM voltage, timing, or speed?

V
Vykor817
Member
214
10-19-2017, 11:41 PM
#8
Also, reset the BIOS settings I’m using at 3.7ghz and 2800mhz with cool n and c6 active.
All my power states are set to auto (need to verify they function properly).
My main concern is ensuring these power states stay above cool n quiet and c6 modes.
Any advice on overclocking the Asrock AB350M motherboard would be appreciated.
V
Vykor817
10-19-2017, 11:41 PM #8

Also, reset the BIOS settings I’m using at 3.7ghz and 2800mhz with cool n and c6 active.
All my power states are set to auto (need to verify they function properly).
My main concern is ensuring these power states stay above cool n quiet and c6 modes.
Any advice on overclocking the Asrock AB350M motherboard would be appreciated.

E
Emil1234
Member
55
10-20-2017, 01:42 AM
#9
You should avoid using any Prime95 version other than 26.6 unless you are certain your software relies heavily on AVX encoding. Otherwise, newer versions may not suit your needs since they could overload your system with unrealistic thermal and processing demands. Stick to Prime v26.6 only. After each adjustment of multiplier or voltage, let the program run for 15 minutes to confirm it operates within safe temperatures. For full stability, continue testing for 15 to 24 hours, or until you reach a point where errors and data loss become unlikely. Running Blend mode or additional tests is fine, but always start with the standard Prime Small FFT baseline. I suggest pausing every 15 minutes after changes and extending testing to 15–24 hours once stability appears. Many systems performed well up to 20–22 hours before failing, so exceeding this range may signal trouble. Ultimately, your setup will depend on your specific hardware.
E
Emil1234
10-20-2017, 01:42 AM #9

You should avoid using any Prime95 version other than 26.6 unless you are certain your software relies heavily on AVX encoding. Otherwise, newer versions may not suit your needs since they could overload your system with unrealistic thermal and processing demands. Stick to Prime v26.6 only. After each adjustment of multiplier or voltage, let the program run for 15 minutes to confirm it operates within safe temperatures. For full stability, continue testing for 15 to 24 hours, or until you reach a point where errors and data loss become unlikely. Running Blend mode or additional tests is fine, but always start with the standard Prime Small FFT baseline. I suggest pausing every 15 minutes after changes and extending testing to 15–24 hours once stability appears. Many systems performed well up to 20–22 hours before failing, so exceeding this range may signal trouble. Ultimately, your setup will depend on your specific hardware.