What does the Cross-Platform Rating in a Price performance comparison page of a CPU on PassMark mean? What kind of performance does it reflect? It seems the kind of performance it reflects is more like what Average CPU Mark does, which is basically multi-core performance. If so, what is the difference between these 2 numbers?

More specifically in my case, I am comparing these 3 CPUs to decide which one to buy: AMD Ryzen 7 3800X vs AMD Ryzen 9 3900X vs AMD Ryzen 9 3950X. The CPU Mark of the 3900X is lower than 3/2 of that of the 3800X, while the Cross-Platform Rating of the 3900X is higher than 3/2 of that of the 3800X. I am going after multi-core performance, so I am wondering whether the 3900X gives me 3/2 of multi-core performance compared to the 3800X.


From Passmarks own version log: https://www.passmark.com/products/performancetest/history.php

enter image description here

It's just that: a bunch of benchmarks cobbled together into a single score. Only the devs know how well these benchmarks scale to multiple threads, and how they weigh the individual results to condense them into a final score. So maybe don't focus too much on this.

Let's rather look at the CPUs themselves. A 3900X has a 50% higher core count compared to a 3800X. And it increases almost all CPU resources by at least 50%. Some shared resources like L3 cache even increase by 100%. If we stopped here, it would be safe to assume that the 3900X is 50% faster than the 3800X, at least in multi-threaded benchmarks with ideal scaling. With 2 notable exceptions:
1) Thermal design power is the same. This causes the 3900X to run at lower clock speeds when all cores are loaded heavily. Dropping the performance increase to a bit less than 50%, at least if the user does not overclock.
2) Memory bandwidth. While the 3900X has almost double the write memory bandwidth compared to the 3800X (due to 2 chiplets vs. 1 and limited bandwidth from chiplets to the IO-die), the read memory bandwidth is practically the same. This causes less than 50% performance advantage in benchmarks (and applications) that become memory bandwidth limited at higher thread counts.

In conclusion: The 3900X can be nearly 50% faster than the 3800X in multi-threaded benchmarks and applications. But as always, it depends on the application.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.