No CO and undervolt issues are affecting your PC (7950X3D) even after updating the ASUS BIOS.
No CO and undervolt issues are affecting your PC (7950X3D) even after updating the ASUS BIOS.
Received my 7950X3D in May according to the build specified in the signature. The motherboard arrived with the April BIOS 2204, and everything functioned properly at first. At EXPO, I ran tests at 6000CL30 with CO at -20°C; most cores stayed stable except a few CPU-heavy benchmarks. In July, I updated to BIOS 2403 and it resolved the issue—EXPO now runs smoothly at 6000, though even a -10 crash still causes Windows to freeze on startup and reboot. It’s puzzling. Could the new BIOSes be problematic (perhaps due to Zen5 compatibility)? Or did a recent Windows 10 update reduce stability?
You're experiencing the same problem on your 7800x3D with the latest BIOS on an Asrock X670E board. If you try any PBO system, it runs at around 3fps and sometimes freezes completely if it doesn't restart. You've already tested temperatures from -20 to -10°C at 75°C and haven't found a solution. While you're unsure about the exact rule of thumb for 7000 series chips (+200MHz and SOC up to 1.15V), your RAM issue hasn't been resolved either.
I don't believe the silicon is the problem, since my CPU functioned properly up to CO-30 in the previous April 24 BIOS update... It's only six months old, so degradation likely hasn't occurred yet. I'm confident they may have made an error with voltage settings for Zen5-compatible BIOSes. The fact that Expo worked then but not now is another indication.
I just installed Windows 11 24H2 last night and uplift seems to be working. The CMOS was cleared this morning, and I plan to try again tonight after work. My goal is to reach at least CO-15 or -10, but I’m really unsure if I’ll hit 5GHz that way. This version might help, though I haven’t tested games or Cinebench R23 yet.
I observed core parking issues and game failures switching to CCD0. After updates, I realized PBO might need enabling or Curve Optimizer settings could be faulty. Fixed it using vcache-try from cocafe. It turns out the initial frequency/vcache CCD setting is controlled by a registry entry that changes dynamically. The Curve Optimizer itself has long been problematic—Core8 and Core9 are better with -8, while others range from -10 to -14. Vcache CCD supports CO values between -12 and -28. Identifying the unstable core is tough because WHEA doesn’t flag failures on this generation, making it hard to pinpoint the cause. Working with the 5000 series was significantly simpler.
To control CCD affinity you can use Process Lasso, which performs well. What seems unusual is how CO levels changed across BIOS versions: -20 in April, -10 in July, and now 0 after the September update. I hope the next release will require a positive CO again—what a surprise!
I didn’t mean to create tension, but why did you quickly switch to the newest bios even though your old ones were fine? Were there any problems? Updates are usually necessary only when fixing something, and from what I understand, all recent bios mainly add support for 9000 series CPUs.
I developed a routine of refreshing the BIOS on AM4 regularly. It never caused problems with my X570 Gigabyte BIOS over the years on a 3900X and later models. While the April BIOS functioned mostly well, CMR wasn't stable, and the one-minute delay was quite noticeable. I hoped for smoother updates, but it didn’t materialize, and EXPO became unusable. With a now-unstable BIOS, I was hesitant to proceed with an update.