Performance dropped unexpectedly in PassMark tests.
Performance dropped unexpectedly in PassMark tests.
Hey there! I'm seeing quite low performance in the Windows UI and DirectX 11 tests, which points to possible CPU issues. From what I understand, my single-threaded speed is pretty weak. Since I have a Ryzen processor, it seems RAM might be the bottleneck. I suspect latency problems are the main culprit. I read that dual-rank memory sticks can hurt performance, but I’m stuck with 32GB for now. Here are some screenshots of the relevant tests. I’ve reset my NVIDIA control panel and turned on only GSYNC plus background window frame rate limits. My Samsung RAPID is enabled on the 850 EVO boot drive—could that help? Key details: | MOBO | ASUS TUF Gaming x570 Plus | CPU | Ryzen 7 3800x – Clock capped at 4.2GHz (PBO off), undervolting to 1V | RAM | 2x 16GB Corsair Vengeance LPX – 3600MHz CL18 – version 3.34 (CMK32GX4M2D3600C18) | GPU | Zotac RTX 3060 Twin Edge OC – 12GB (Game Ready driver 512.15) | Monitor | 1x 1440p @144Hz Freesync with GSYNC forced | OS | Windows 10 Pro x64 build 19043.1586 | PSU | Sentey 725W Thanks for your support! CPU idle (apps running): Under load: GPU benchmarks: Memory benchmarks:
It doesn't excel in actual games but may struggle in controlled tests. The passmark score suggests issues, possibly linked to Vega 10 graphics, which also appears as the name for Vega 56/64.
Online scores suggest aiming for around 5000 points with a 3800x resolution, but you're falling short by over 200 points. Based on the data you've shared, it seems thermal throttling might be affecting performance, which is why lower voltage settings help. Key specs: 4.2GHz max @ 1.1V — score is close to expected range. PBO: Use default Ryzen Master configurations.
A score around 3% below the expected value isn't a cause for concern. The silicon lottery can sometimes deliver strong chips or underperformers. At stock speeds my 2600X is just slightly below the expected value, but a slight adjustment improves it to meet that standard. If your results are a few frames lower than you'd prefer (like 240 vs 265), it's not alarming. A drop from 75 to 200 would be more notable. Another question: where did you obtain your expected scores?
Did you turn off the voltage adjustments while using PBO? Just leave them on Auto and rerun the tests, making sure to clear all temporary files during benchmarks. The extremely low setting of 1V is causing your CPU to struggle, particularly with single-core performance—Ryzens tend to demand higher voltages, sometimes up to 1.5V for heavy single-core or multi-thread tasks. It seems you might be misunderstanding how voltage impacts Ryzens, especially since you're likely using an Intel platform that wouldn't run well at such low levels. Short summary: the issue is low voltage affecting performance.
Not that straightforward. As someone who's been using AMD for a lifetime and just getting into underclocking, this is really impressive. Usually it hits about 87°C during tests, but CoreTemp seems way off compared to Ryzen Master. Hopefully it gives us more clues. I'm starting to think maybe I should push my CPU further—seems like the cooler connection isn't doing a great job.
1. For undervolting, start with offsets rather than fixed voltage since static voltages can significantly impact multi-core performance and single-core efficiency on Ryzen processors. They often aim for voltages above 1.4V during light single-core tasks.
2. It seems you're dealing with overheating issues; without details about your cooler, it's best to check it first. For better monitoring, consider using HWiNFO64 with a pooling rate of 500-1000ms.