Hello everyone, it’s been a month since I got my hands on the new 7950X3D, and I’ve been watching almost all the posts and YouTube videos about how to overclocking and undervolting this beast.
As far as I know, many people struggle to achieve a all core negative 30 in the Optimal Curve or an average of 30 in per core setting. However, I just throw the numbers like setting CCD0 all to -40, CCD1 all to -30, and added a boost clock value of 150+ making the highest clock 5900.
Here is the detailed data in ryzen master:
[https://imgur.com/a/iU7gpk5](https://imgur.com/a/iU7gpk5)
Luckily, after I’ve been running uninterrupted stress tests for 2 weeks, not a single crash has happened. Does this prove that I’m lucky and got a golden sample CPU?
If not, how can one determine if a CPU is a golden sample?
Perhaps I could push it further to find this CPU’s limit, but I’m too lazy to do so because it’s time-consuming, and the additional performance gains are minimal. ..
5 Comments
How many watts is it pulling on full load?
I’m not sure it’s running over stock.
Good undervolts are just 1 indicator of a Golden Sample.
Yes, it is a strong one because it indicates good silicon.
Generally speaking you can only be sure once you are able to keep that behavior on substantial overclocks too.
F.e. My current gs, a ryzen 7 5700x, can do -30 on all cores on oc and outside of it, while I am able to keep a performance gain until 175w ppt.
A similar Ryzen 7 5700x I have can do -20 on all cores at low voltages. The performance gain stops at 130w ppt tho and the undervolt becomes unstable. That cpu can only do -10 when at 130w ppt now.
Thats a very extreme example since generally the differences arent that big but it happens. So even tjo it might look like its a crazy good cpu, it may show its weaknesses later on.
I dont recommend pushing that cpu any further just to know whether it really is a golden sample.
In the end what counts is how it performs in your current configuration. And since in your current config you can easily do -40, you can be very very happy with that. Theres absolutely no reason to overclock it too much just to see whether it is a genuine golden sample when the chance of it being one is pretty high already looking at your current undervolt
> If not, how can one determine if a CPU is a golden sample?
Compare your results with the top HWBot scores: https://hwbot.org/hardware/processor/ryzen_9_7950x3d/
I can’t even do 102 bclk or -5 on my 2ed CCD. 1st one is great. 2ed is terrible
Don’t use frequency offsets with AM5 CPUs, they break CPPC which means your actual day-to-day performance will be much worse. If you want to go to 5.9 on the frequency cores you need to use ECLK instead.