The Ryzen 1700 is causing issues with overclocking.
The Ryzen 1700 is causing issues with overclocking.
Hi,
In another discussion I mentioned my attempts to identify the highest overclock I could achieve on a Ryzen 1700. By adjusting the voltage to 1.35 I reached a maximum of 4.2, and every time I increased the speed I performed a stress test without any issues.
The problem arose when I ran a CPU benchmark; my score dropped significantly. At a stock voltage of 1.25 I achieved a score of 1558 on Cinebench, but at 4.2 it fell to 500. When I lowered it back to 3.6, the score remained around 500.
I discovered that reverting the voltage to the stock level of 1.25 and running the benchmark at 3.6 restored my score to about 1558.
I can push the CPU up to 4.2 for stress testing, but it doesn’t score well. Going beyond the stock voltage causes instability. It’s puzzling because this processor is designed to run smoothly at 3.7 with default voltages, yet overclocking to 3.7 with standard settings causes the PC to shut down.
I can keep the CPU at 4.2 and pass stress tests, but performance drops at any higher voltage. The temperatures stay within normal ranges (under 40°C), and during stress testing at 1.35 I saw the voltages reported as 1.34 max.
Well, to ensure a solid stable OC for 24 hours, you'll be able to tell if it's reliable. Some might think it works just fine for two hours, but I prefer certainty. With two extra cores, your score of 3.9 on 1.35 is quite close (considering all chips vary slightly). A 3.8 result isn't too bad either—it gives you a more accurate picture of your performance. You can build from there.
From the beginning, I took a practical route. Starting at 1.325 with stock, I ran it for 24 hours without any problems. I got 3.75 at 1.325 stable again after 24 hours. Now I'm at 3.9 with 1.35 stable. I won't push further; my wraith max cooler works, but I prefer keeping temps below 80°C, which is around 77°C right now.
cinebench isn't meant for stress testing; it's for benchmarking. Use prime95 or AIDA to push your CPU to its limits. Your temperatures at 40°C are too low for real stress testing. That means you're not really pushing it. Unless you're running a very intense water loop, you won't reach those low temps during genuine stress tests.
Roland Of Gilead:
cinebench isn't for stress testing—it's just benchmarking. Use prime95 or AIDA to push your CPU to its limits. Your temperatures at 40°C aren't high enough for real stress testing; that means you're not really pushing it. Unless you're running a very intense water loop, you won't reach those low temps during genuine stress tests.
I've tried stress testing with prime95 and benchmarking with cinebench. The CPU reaches 45°C in the blend test. With a h100i v2 Corsair cooler, performance drops significantly. If I increase the voltage, the scores drop even more. Consistently getting low scores at 3.6 or 4.2 is concerning. I've noticed the CPU temps have been unusually cold, with the highest ever being 45°C.
aha, the issue is clear. You should run it on small FFTs. This will only challenge your CPU and cache. Please report back with temperatures. It would be surprising if it passes Prime 95 at 4.1/4.2, but keep an eye on the temps—they’ll rise significantly.
That’s the plan. Push it under full load with high temperatures, then you’ll have a stable output.
the h100 will definitely maintain low temperatures, though not at 45 full load Prime with small ffts. You might want to examine LLC and load line calibration—it functions better at higher output but requires understanding its application and caution. It helps remove vdroop and can improve OC stability. However, setting it too high risks overvoltage. Proceed with care.
Roland Of Gilead would keep temperatures low, though not at 45 full load Prime with small ffts. Consider examining LLC and load line calibration; it functions better at higher OC levels but requires careful understanding. It can remove vdroop and improve stability, though setting it too high risks overvoltage. Proceed with caution.
He tested running prime95 on small ffts at 3.6 stock voltages. CPU temperatures reached 45°C. Increasing the voltage to 4.1 with 1.35V caused a similar peak at 45°C. Benchmarking at 3.6V gave 1558, while at 4.1V with 1.35V it dropped to 500.
Raising voltages beyond default causes the CPU to lock at higher settings, but logging shows it reports 4.1. Running prime95 at 4.1 instead of 3.6 would likely increase temperature. At 4.1GHz the max voltage reached was 1.344.
He found LLC in the BIOS, set to Auto with low, medium, high, and extreme options.
I launched Ryzen Master and noticed my cores were capped at 1.55ghz. Checked the documentation and believe it's due to a Ryzen issue. I'll check that source for more information.
Yes, it would definitely raise the temperatures. It might be a bug, but I haven’t felt it myself. When voltage goes up, heat increases too. It’s just physics. I’m puzzled about what you’ve suggested. One possibility is adjusting the windows power profile—make sure you choose a high-performance setting or use the Ryzen profile so your CPU stays at full performance. Let me know what you think. I’m curious about the solution.
Roland Of Gilead provided some insights on optimizing performance settings. They mentioned that maintaining lower temperatures is important, but not at 45 full load with Prime using small FFTs. They suggested exploring LLC and load line calibration for higher output, noting it can reduce voltage droop and improve stability. However, they warned against setting it too high, as it might cause overvoltage. They shared personal testing results: running Prime 95 with small FFTs at 3.6 stock voltages caused CPU temperatures to peak at 45°C. Increasing the voltage to 4.1 with a 1.35V setting reduced the maximum temperature to 45°C, but benchmarking at 4.1 with 1.35V gave a lower value of 500. They observed that raising voltages beyond defaults often locks performance and prevents further increases. When logged in, it displayed 4.1 instead of 3.6, which they believed would increase temperature. They found LLC settings in the BIOS with options for low, medium, high, or extreme, recommending medium or high for a 4.1 OC, avoiding extreme. They advised starting with medium and noted that it helps maintain voltage consistency during full load or idle, though some adjustment is needed. Extreme was strongly discouraged due to the risk of severe overvoltage.