Comments by "Winnetou17" (@Winnetou17) on "" video.
-
2
-
1
-
@joeyvdm1 Oh wow, you really like this stuff :))
Well, maybe I heard/understood/remembered incorectly.
From what I remember (I also watch AdornedTV and CoreTeks?CoreTecks?) there was something that because you only have one CCX, the cache is now available for every core directly, without the need for accessing the other CCX or duplicating the data in the local cache.
The thing that might make our communication a bit harder here is that IPC is too broad of a term. Because it can be influeced by many things. Having a lower latency for... anything really, I see that as IPC increase. Don't know why companies wouldn't market it as such. Instead of saying "We got 5% lower latency, 5% better cache efficiency and 5% IPC", they will simply say "We got an average of 16% IPC increase". Because that's a bigger number, and very easy to understand: in average, everything will be 16% faster.
It's nerds like us who want to know how is that 16% made of, to better asses which workloads will be impacted more.
So, I still stand by my claim that IPC increase will encompass all the improvements, since that's the idea of IPC metric. Not "pure workload, no memory fetching or saving, no shared memory, no multicore communication" metric.
And, even if I'm wrong, I want to not be too hyped up. Seeing in some places that Zen 3 will have "only 10-15%" IPC increase, as if that's not great, only makes me sad.
Even with 10% IPC increase and absolutely nothing else improved, my quick mind maths tell me that this will be enough for Ryzen CPUs to match Intel in gaming (on average, on some games it will be better), while obliterating them in all other aspects. That's still something that I can't wait to see.
Cheers!
1