AMD issues driver ups for future Zen 4 and Zen 4C CPUs.
AMD Preps for Zen 4: Different Types of Cores Now Supported in Linux : Read more
AMD Preps for Zen 4: Different Types of Cores Now Supported in Linux : Read more
The scheduled forum maintenance has now been completed. If you spot any issues, please report them here in this thread. Thank you!
Zen 4 vs 4c has nothing to do with 3D cache as far as I'm aware. AMD describes Zen 4c as being optimized for power efficiency and density, so roughly analogous to "little" cores like Gracemont cores in Alder Lake.Pretty sure this confirms what we already know. 2 types of Zen4 core ccd parts. Those with 3D cache and those without.
Changes to error correction to address the higher cache parts makes sense.
Changes to drivers and/or updated die temp sensor location makes sense with 3DV parts.
Nothing I see points to anything beyond that. Not sure about server parts but based on the preso and brief look at the transistor changes on the package, the desktop plan is different.
"..Two types of cores in one.." or something along those lines. They're going to modify power and clocks on-the-fly in conjunction with an updated scheduler.
Zen 4 vs 4c has nothing to do with 3D cache as far as I'm aware. AMD describes Zen 4c as being optimized for power efficiency and density, so roughly analogous to "little" cores like Gracemont cores in Alder Lake.
So you think Zen 4c will have 3D cache and Zen 4 won't, or vice versa? Where have you seen AMD mention "density optimized cache" with respect to Zen 4c?I respectfully disagree..
4c is "cloud enhanced". Of course they haven't yet provided the deets but I read that to be designed for specific, repetitive tasks requiring less clock and power per thread.
"Density optimized cache" ie..stacked and power enhancements are stated as 4c features but who knows..
They don't have a big/small, big/little, big/bigger arrangement taped yet and we won't see it in Zen4, 4c, 4xxx.
They are, quite simply, using an updated scheduler and fancy up or down-clocking to execute threads more efficiently. Pretty slick as an interim response to Intel plans.