AMD RyZen CPU Architecture for 2017

They said at the ryzen presentation that they will not do a "Tick" architecture that all ryzen CPUs will be "Tocks" not minor changes.
Tick is same process but substantially different architecture. Tock is same architecture but new process.
"Zen+" is neither of those, it's the equivalent of Intel's new "Optimization"/"Refresh" steps. Probably just some low-hanging fruit (because why not?). If it comes with 12LP (formerly "14nm+" in slides) we'll be looking at modest gains in CPUs, just like the Skylake->Kabylake transition.
Zen 2 is indeed a Tock, since it comes in 7nm and they're promising significant changes in the uarch.
 
Tick is same process but substantially different architecture. Tock is same architecture but new process.
"Zen+" is neither of those, it's the equivalent of Intel's new "Optimization"/"Refresh" steps. Probably just some low-hanging fruit (because why not?). If it comes with 12LP (formerly "14nm+" in slides) we'll be looking at modest gains in CPUs, just like the Skylake->Kabylake transition.
Zen 2 is indeed a Tock, since it comes in 7nm and they're promising significant changes in the uarch.
I just quoted what AMD said. I think zen+ would more that marginal gains(only working in the memory optimization side they can get important gains) and I hope for betters clocks as well.
 
I just quoted what AMD said. I think zen+ would more that marginal gains(only working in the memory optimization side they can get important gains) and I hope for betters clocks as well.

Zen+ isn't a thing.

It appeared on the first roadmaps waaay back but was quickly renamed to Zen 2. There are no architectural changes expected from Pinnacle Ridge, it's pretty much Devil's Canyon for Ryzen.
 
Zen+ isn't a thing.
Not a major change, but improved steppings would seem likely. Tweaks to power distribution and various errata encountered. Even a new clock domain for Infinity would be practical. Might solve the complicated memory scaling issues and still perform inline with the originals. Process improvements would be expected as well.
 
Feb 2018 :oops:
That should be not just taped out but samples already well into qualification by now right?

From the link
AMD managed to achieve 19% sequential growth in second-quarter 2017 revenues and expects the amount to grow further by 23% in the third quarter.
First statement of affect/sales of Ryzen I've seen
 
Feb 2018 :oops:
That should be not just taped out but samples already well into qualification by now right?

Yes. It's important to note that the change from 14nm to 12nm process is not very large, and this makes the work needed to switch to the new process not all that hard. Don't expect miracles from this. The next big advance probably comes from Zen2 on 7nm.
 
True but any improvement on an already competitive chip this quickly is very welcome, the rework should enable some low hanging fruit to be fixed/improved.
If they can get in a couple of hundred extra Mhz base clock &/or raise that pretty hard OC limit it would have to be a big worry for Intel.
 
Last edited:
More like improve the subtimings of the memories and the IF dependencies on it crutos has shown 2 digit gains only working on the memory.
 
True but any improvement on an already competitive chip this quickly is very welcome, the rework should enable some low hanging fruit to be fixed/improved.
If they can get in a couple of hundred extra Mhz base clock &/or raise that pretty hard OC limit it would have to be a big worry for Intel.

Quite competitive against Kaby Lake, yes—against Coffee Lake, it might be a different story. The new process is supposed to bring an extra 10% in performance, so +300/400MHz should be possible, plus perhaps memory controller tweaks, which might also help. I don't think Intel will lose too much sleep over this, however.
 
Back
Top