AMD Epyc 7452: Rome CPU with 32 cores and 2.35 GHz spotted

Published by

Click here to post a comment for AMD Epyc 7452: Rome CPU with 32 cores and 2.35 GHz spotted on our message forum
https://forums.guru3d.com/data/avatars/m/218/218363.jpg
I'll run Crysis just fine 😉
https://forums.guru3d.com/data/avatars/m/267/267153.jpg
no it will not .)
https://forums.guru3d.com/data/avatars/m/106/106401.jpg
I am wondering how well the TR 64C will perform with PBO active, Also would be like to see YouTuber's commenting about the CPU performance increase since first Rizen [Intel CPU core cout increas, price drops etc]and how did it affect them regarding the Video Editing/Rendering times,
https://forums.guru3d.com/data/avatars/m/250/250418.jpg
"When AMD will introduce the second Epyc generation called Rome with up to 64 cores is still unclear. Most recently, AMD had called only the third quarter of 2019 for the market launch. Originally the server CPUs should appear before the desktop counterparts, but this has apparently changed." Original plan was dumb, as they need to stock up on that 1% best chips for these. Although +7nm is already on trials and going to mass production by 2020, 7nm is still expensive and yields are not perfect. Thankfully Zen2 chiplets are small enough.
https://forums.guru3d.com/data/avatars/m/106/106401.jpg
Can AMD make the Zen2.0 Chiplets on GF12nm? [in Theory], Lets say 4 8c chiplets + IO chiplet on TR4? So they can make 32C EPYC/TR Zen2.0 for cheap?
https://forums.guru3d.com/data/avatars/m/270/270008.jpg
HWgeek:

Can AMD make the Zen2.0 Chiplets on GF12nm? [in Theory], Lets say 4 8c chiplets + IO chiplet on TR4? So they can make 32C EPYC/TR Zen2.0 for cheap?
No they can't without redoing the design, new masks etc. This would cost a ton of cash so no this will not happen nor is it a good idea on so many levels. The more they make on 7nm the cheaper it will get as TSMC works out the kinks. AMD's CPU chiplets are really small so they will eventually get very good yields so lower costs. Currently AMD's IO die and chipset are actually the same layout, exactly the same size chip to help with economies of scale manufacturing the both dies at GloFlo. This is where they saved money and it was indigenous to make the IO die and the chipset the same. I expect on Zen3 this IO die/chipset to move to TSMC as well and be made on the "old" 7nm process so we no longer need a fan to cool the chipset and it will also lower the IO dies power needs from 15w.
https://forums.guru3d.com/data/avatars/m/246/246171.jpg
A bit weird how this is CPU is an all-around upgrade, yet the part number is lower. I still haven't fully figured out the Epyc number scheme. Anyway, if we are to take these results seriously, that suggests a 15% IPC gain, which is pretty much what AMD has been touting this whole time.
https://forums.guru3d.com/data/avatars/m/270/270008.jpg
schmidtbag:

A bit weird how this is CPU is an all-around upgrade, yet the part number is lower. I still haven't fully figured out the Epyc number scheme. Anyway, if we are to take these results seriously, that suggests a 15% IPC gain, which is pretty much what AMD has been touting this whole time.
I think what you are going to see is the higher the core counts the larger AMD's advantage. AMD's 7nm chips use so much less power they will be able to clock the chips higher than Intel parts to fit into typical server thermal envelopes.
https://forums.guru3d.com/data/avatars/m/106/106401.jpg
schmidtbag:

A bit weird how this is CPU is an all-around upgrade, yet the part number is lower. I still haven't fully figured out the Epyc number scheme. Anyway, if we are to take these results seriously, that suggests a 15% IPC gain, which is pretty much what AMD has been touting this whole time.
Epyc 7XX1 = Zen 1.0 Epyc 7XX2 = Zen 2.0
https://forums.guru3d.com/data/avatars/m/106/106401.jpg
Rome launches on August 7th.