News AMD's EPYC Milan-X is Official: 3D V-Cache Brings Up To 768MB of L3 Cache, 64 Cores

[gg83]

Damn. I remember when my PIII motherboard that was the max amount of ram it could even take!!

Yeah, and now it's small enough to fit on the chip itself. It's crazy

 

[jeremyj_83]

They already did that with Jim Keller. And Sunny Cove which powers Alder Lake is the result. No need to wait 7 years, it's already here.

Sunny Cove was used in Ice Lake where as Golden Cove is in Alder Lake.

 

[Jimbojan]

Adding more L3 cache does not make the processor run faster, but costs more to make, especially not to drive up the stock price, it is a pure speculation. It is not clear how much is Microsoft or Face Book would buy from AMD. Intel is trying to reduce its chip pricing furiously, so that AMD is losing its margin, if there is still anything left for AMD; AMD will soon start losing money when AMD is forced to reduce its chip ASP by 30% or more. Intel is attacking the Desktop and mobile market now with Alder Lake with integrated graphics, look for Intel to get Sapphire Rapids for server and discrete graphic in December 2021, both AMD and NVDA are in trouble. I am sorry.

 

[spongiemaster]

According to Jim Keller, he designed the successor to Sunny Cove which is Golden Cove, not sunny cove.
https://www.techpowerup.com/259653/intel-sunny-cove-successor-significantly-bigger-jim-keller

Derp, you're right. It was Golden Cove, not Sunny Cove. The point still stands though.

 

[jeremyj_83]

Adding more L3 cache does not make the processor run faster, but costs more to make, especially not to drive up the stock price, it is a pure speculation. It is not clear how much is Microsoft or Face Book would buy from AMD. Intel is trying to reduce its chip pricing furiously, so that AMD is losing its margin, if there is still anything left for AMD; AMD will soon start losing money when AMD is forced to reduce its chip ASP by 30% or more. Intel is attacking the Desktop and mobile market now with Alder Lake with integrated graphics, look for Intel to get Sapphire Rapids for server and discrete graphic in December 2021, both AMD and NVDA are in trouble. I am sorry.

Run faster as in higher GHz that is correct add L3 won't help. Run faster as in it is able to process things faster that is a yes and no. Adding L3 will for sure help is memory bound applications as the cache acts as a buffer before having to go out to the main memory. If the information is found in the 500MB of cache it is quick to get. Azure has some benchmarks that they have posted that show great scaling in some applications with Milan-X vs Milan. https://www.servethehome.com/amd-milan-x-scaling-to-0-75gb-of-l3-cache-per-chip/

Right now Sapphire Rapids is slated to come into production Q1 2022 and Q2 2022 for ramp up. I haven't seen anything about Alder Lake going mobile. The CPU is pretty power hungry so they most likely have to do a few things to get it into a mobile power envelope.

 

[drtweak]

Yeah, and now it's small enough to fit on the chip itself. It's crazy

Right? I remember when 128MB was huge for a thumb drive lol