My first OC 8350 SUMOUT. Got some advice.
My first OC 8350 SUMOUT. Got some advice.
I'm working on keeping the system stable with an overclocked 8350. I've completed an hour of prime 95 with minimal issues, but once a memory test starts, the system shuts down due to SUMOUT errors. Currently, I'm using voltages of 1.36 for the CPU and 1.25 for the CPU Northbridge. All power-saving features are disabled, and the system runs quietly without any noise.
The multiplier is set to 21.5, the FSB is at 200 (stock), and CPU/NB and HT frequency is at 2200. I'm using the D.O.C.P. RAM profile for timing adjustments, as it helped stabilize the 1866 stable at stock speeds. This was necessary because the ASUS BIOS previously set CPU/NB and HT speeds to 2200 for my ballistix tactical 1866 to maintain stability.
To attempt an overclock, I've lowered the RAM frequency to 1600 but adjusted the timing to 8-8-8-24. All FTT small runs have remained stable for up to an hour without any rounding errors. The SUMOUT errors appear during blend and custom tests when using 75-90% of my RAM.
I've increased the Northbridge voltage by 4 volts, but this hasn't resolved the issue. I'm uncertain whether adjusting the RAM voltage further would help, as I haven't tested it yet. I'm at a loss and need advice. Should I record the Northbridge voltage readings or the HT voltage? Would increasing CPU voltage further work? I'm not sure if the problem lies with the RAM or the SUMOUT errors.
My system details are:
M5A99X evo 1.0
FX 8350 4.0GHz
Crucial ballistix tactical 1866 1.5V
1060 GTX
Corsair 850 bronze power supply.
Current temperatures:
Package 52C under load
Socket 59C
Idle package 11C!!
Idle Socket 35C
The errors consistently occur after 58 minutes, regardless of adjustments. No matter what I try, the RAM fails to stabilize. I'm stuck and would appreciate any suggestions.
Just another reason against using Prime95 on an AMD system—it was designed for Intel, its approach and testing methods. The first version is tailored exclusively for AMD, while the second offers more flexibility for adjustments. Prime95 should have been obsolete a long time back.
Just another reason against using Prime95 on an AMD system—it was designed for Intel, its approach and testing methods. The first version is tailored exclusively for AMD, while the second offers more flexibility for adjustments. Prime95 should have been obsolete a long time back.
For temperature and other measurements, refer to the BIOS settings. Benchmark and stability tests are more appropriate for the AMD platform.
I also have OCCT which appears to function more similarly to Prime with test halt options, allowing me to stop stresses if temperatures exceed limits. This is quite helpful. I'm wondering why I keep receiving sumout errors even though CPU stresses remain stable after four hours? It seems to be related to RAM, but I can't figure it out. My RAM is running at a clocked 1333MHz and is at 1866, yet it's failing to maintain stability with a multiplier of 21.5. I'm trying to find any voltage or configuration adjustments that could help stabilize the RAM. I have manual stock settings and tightened them from 1866 to 1600, but it still doesn't work. Voltages seem adequate for such a small OC. The CPU is at 1.38V, should I increase CPU/FB voltages from 1.25V? I have no idea what to do...
I changed the approach by adjusting the OC and voltage settings. Initially, I increased the voltage as much as needed to check cooling limits, then used multipliers to raise the frequency before lowering the voltage to find the optimal balance between performance and stability. Due to the sensitivity of FX chips to IMC, I rely on FSB only for final fine-tuning. Contrary to expectations, the highest frequency didn’t always yield the best overall results—it was slightly lower than the peak OC achieved. I mostly use Kingston RAM, which performs better with higher voltages up to 1.65V and appears more stable.
OCCT performed well overall. It turned out I can't run the chip at 4.5ghz in air, but I'm now testing it under stress at 4.4ghz with 1.75v for an 8-hour period. No errors were reported during a 30-minute test, regardless of whether it was small or large. The maximum temperature reached was 57°C, which is within the safe limit of 61°C at 4.5ghz (voltage of 1.875). I didn't want to push beyond that for regular use. Since I'm running at 1.75v and if this proves stable, I might increase the bclk to get closer to 4.5. I suspect I could reach a bclk between 202-203 or 4.48ghz.
The next question is whether I can achieve RAM speeds up to 1866 or if an error rate would appear at 1600. If not, I'll focus on tightening the timing to 8-8-8-24 @1600 or 7-7-7-21 @1333. I'll have to monitor the results. Frequency and HT are both 2400. I remember seeing 2200 with HT at 2600. I thought it would be better to match them, as adjusting bclk would be easier after that. I'll update you on what happens.
Thanks, Mike