After the launch of the new Ryzen 3 with the 3300X in the lead and followed by its brother 3100, AMD had two different configurations for these processors even though they share a huge part of their specifications. This has caused suspicions about the movements for Zen 3 to soar and the data to be observed with a magnifying glass to discern what we can expect with the new processors. Is the Ryzen 3 3300X a prologue to what AMD has in store for us?
If we look at the specifications of both processors we will have a small dejavú between them. Both have 4 cores and 8 threads, both carry 2 MB of L2 and 16 MB of L3 , support PCIe 4.0 and have a TDP of 65 watts .
As expected, the largest of them has a higher frequency, as well as a price according to this difference in performance, but that would be the difference, or not?
Ryzen 3 3300X and 3100: two different core and cache configurations
By sharing the number of cores and threads, AMD has only enabled a single eight-core chiplet to bring these two processors to life, but this leaves two different variants at play that Lisa Su’s can use.
The difference between the two is simple: the Ryzen 3 3300X gets a quad core configuration within the same CCX, while the Ryzen 3 3100 does it from two CCXs with two cores each.
What are the advantages? It is very simple to understand. The Ryzen 3 3300X does not have to call the Infinity Fabric nor does it have to leave its CCX to import or export data, allowing this processor to have a unified and very constant latency thanks to its so-called 4 + 0 configuration, where the second CCX it is completely disabled.
In the Ryzen 3 3100 this does not happen in the same way, since its configuration is 2 + 2 and therefore it has both active CCX and must communicate through Infinity Fabric, which adds latency and penalizes performance. In addition, its cache design divides the 16 MB in two, so we will have 2 X 8 MB respectively.
Up to 50 nanoseconds difference between CCX and 4 more between cores
The measured data reflects a huge disparity in performance and especially latency, the main stumbling block of the Ryzen 3000 that have been dragging all Zen architectures. As you can see, the Ryzen 3 3300X has unified latencies between 32 and 33 nanoseconds , while the Ryzen 3 3100 ranges from 37nm between its four main threads and up to 90 nanoseconds if we have to access the other CCX.
This is almost three times more latency between the two, something that is also reflected in the access to RAM as is logical and that has an impact again on the final performance. Given the details and knowing that Zen 3 will work with improvements of this caliber, is this Ryzen 3 3300X a prologue?
AMD Zen 3, one step closer to unifying CCX, CCD and caches
What we know from the different leaks is that Zen 3 is going to leave the CCX scheme almost totally behind to leave single blocks that could well be more complex CCDs. That is, in Zen 3 a CCX will be a CCD because what AMD tries is precisely to end the latency problems and incidentally share the L3 among all the cores, as Intel currently does but with the advantages of its MCM system.
Given the data above and although Intel will have some advantage in caches for continuing with monolithic systems in its processors, the difference in performance should go down and therefore in tasks such as gaming they should be much more competitive, which added to the other improvements IPCs such as the implementation of new functions in the I / O die, front end and back end , as well as better frequencies, should put AMD above Intel, at least up to Rocket Lake-S and far ahead on servers.
In short, the Ryzen 3 3300X is effectively a prologue to the improvements that we can find in Zen 3, and due to this, after what we have said, it can only improve the result after the updates that the new architecture will bring in the Ryzen 4000.
