7000x3D boosting data ready, please share your HWinfo! :)
7000x3D boosting data ready, please share your HWinfo! :)
Following the latest firmware release aimed at resolving 7000x3D Vcore problems, my peak boost increased noticeably. Earlier I was capped at 5.875GHz on the same Core 13, but now it reaches around 5.9GHz. The exact sample timing isn’t clear, but the image below captures HWinfo64 data. I’m interested in how others are experiencing maximum boosting across all 7000x3D units, particularly the 7900x3D. It seems the 7950x3D might have been grouped differently, though I suspect similar trends exist elsewhere. The average result here isn’t critical. Note: the highest stable Vcore is likely around 1.390V, possibly excluding the non-3D CCD section (hopefully). If the recent GN video mentioned stability after 1.3V on the 7800x3D, that aligns with what I see. This setup uses 'prefer cache', EXPO 6000 36-36-36-76, plus a 200MHz boost at -15mV PBO2.
I'm not sure about the X3D; I only have a 7700X. Based on what I'd guess, the voltage dropped slightly, which lets the CPU run cooler and allows it to operate at a higher speed.
My belief still stands, though reviewing past boost settings my peak Vcore was around 1.38V. It seems the overall Vcore performance is more important than differences between models. I’m not sure it affects CCD0 versus CCD1 much, as I question if the 7950x3D really pushes 1.39V into the 3D sensor.
The 7950x3D 8+0 setup reached a peak of 5.05GHz before the latest UEFI update. In a complete 16C configuration, it tops out at 5.25GHz compared to the 7800x3D's 5.05GHz. I also updated the name to clearly indicate the CPU and increased the Vcore settings since they weren't used before. I plan to try a full 16C configuration later.
Using Division 2 as the reference gave similar framerates across 8+8 and 8+0 setups. The 0+8 test wasn’t worth uploading because top boosts hit around 5.3GHz and didn’t work well. My control rates ranged from 180-210 fps on 3D modes, while the non-3D mode was roughly 160-180 fps.
These tests were performed on my NH-D15. The main changes included the cooler and its two 140mm fans (I also checked the bottom of the CPU and socket just to be safe). 8+0 and full 16c respectively. I can't upload the other screenshot for some reason. The 16c/32t model achieved 36,472 points with a 4.825-4.925GHz boost across all cores during the test. 1.015V Vcore for all cores. I also ran a 10-minute burn instead of one long run, which kept all cores between 5.85 and 5.875 and the results were around 18.2k. It will reach the thermal limit of 89°C throughout, which is normal given the 5800x3D model had similar performance.
For those with a 7800x3D, this advice applies if your Cinebench R20 is below 18,000. Consider reviewing your power plan and trying the 'balanced' setting instead of 'high performance.' This adjustment was observed in the 7950x3D when testing last night, especially after troubleshooting an issue with a friend's 7800x3D. The 'high performance' mode appears to enhance boosting but can cause the CPU temperature to drop, reducing its boost clock by around 200MHz. Your Cinebench score improved noticeably—reaching 400-500 points higher—and the average boost clock was also tracked during testing.
Balance remains the recommended profile. It is the sole option available in the Settings panel and should be configured as "Balanced." Unless you require advanced cooling methods like biffy cooling or direct die cooling for this Ryzen 7000 series processor, stick with this setting to avoid overheating issues. Excessive heat buildup can cause throttling, which may be evident in lower Cinebench scores.
I also tried this method because the screenshot doesn't reflect the actual device specs. My friend used his 7800x3D in high performance mode and saw his R23 drop below 18k. I observed the behavior you mentioned—high initial voltage spikes that likely overwhelmed the IHS/die, causing lower core clock readings. I might be able to verify if this resolved his problem, but I recreated a lower score by testing in both modes.