Ryzen 5000 - 3 Features in Zen 3 That Will Make It Worth It!

[Gav]

Let’s get started with AMD. Some people at the company teased an announcement which was then announced for the announcement of their upcoming next-gen CPU and GPU’s. Yes, I’m making fun of their marketing. Anyways, we got the next-gen Ryzen announcement on the 8th of October and Today we’re talking about the big changes from Zen 2 to Zen 3.

So,Next-gen Ryzen, it would be called Ryzen 4000 like expected but rumors are popping up with a possible change to 5000. It’s just a name though, so who cares. That next-gen family of CPU’s will sport Zen 3 cores. We’ve known through some EPYC documents that this architecture would feature some changes to the cache. Instead of 16mb of L3 cache for groups of 4 cores, or CCX’s, Zen 3 would have a unified 32MB block of L3 cache for all 8 cores on each chipset.

 

[Marzipan]

I wonder why AMD would consider skipping the 4000 name, unless it's from backlash for their 4000 series APU's or something?

 

[Entz]

I really hope that is true. Having APUs/Mobile a "tier" above was silly and had to throw OEMs fits. You should be able to look at the number and figure out the generation. People would be like "why would I get a 3700 over a 4600 the number is higher on the latter"...

 

[Marzipan]

indeed. wonder what's up their sleeve.

 

[Bond007]

4 is considered unlucky in China (I believe), so maybe that is why.

I am all on board if they align cpu and APU naming though.

 

[CMetaphor]

Wait, back to combined L3? Does that mean the overly complicated OCing will go away? To this day I still don't think I understand all of the CCX "attributes". Never understood what was wrong with locking all the cores together - for OC purposes - and getting the maximum OC to the point where the weakest core is stable. It still just makes more sense to me. One core gets almost the maximum of it's potential, and the rest are OCd and comfortable.

... Maybe it's just me?

 

[Bond007]

Wait, back to combined L3? Does that mean the overly complicated OCing will go away? To this day I still don't think I understand all of the CCX "attributes". Never understood what was wrong with locking all the cores together - for OC purposes - and getting the maximum OC to the point where the weakest core is stable. It still just makes more sense to me. One core gets almost the maximum of it's potential, and the rest are OCd and comfortable.

... Maybe it's just me?

I suspect OCing will be very similar.

if you do tasks that don’t take advantage of all the cores,then you would be better off with fewer faster cores.

 

[CMetaphor]

if you do tasks that don’t take advantage of all the cores,then you would be better off with fewer faster cores.

That much is pretty obvious...? Dynamic "turbo" OCs are already everywhere in abundancento boost the clocks of a few cores when not all are being loaded. Although from what I've seen most people disable it when they go for Max OC

 

[Marzipan]

I wonder if they could offset cache for higher clock speeds and how that would effect end performance.