But there is a ton of architecture changes and we don't know how the chiplets without all that IO affects AMD's ability to clock. Also even if people don't like him adores video after the EPYC reveal when talking about Yields. Intel can really wratchet up the clocks because the produce so many chips. The 8086k is a perfect example of the extreme end of that (and the 8400 the other end having specs that let it be a catch all for the lowest common denominator). A lot of AMD's limits on clocks (at least to 200-300mhz) has to do with volumes being low enough that there aren't enough chips to bin much higher. The increased count even if sales are equal of chiplets AMD will be producing should give them some wiggle room to offer models with clocks higher than what the process shrink with no other changes would normally allow.Shrink ratio of just under 0.7, clocks up between 12%-13% and power down approximately 10% (ranging between 20% and 5% depending on workload - c.f. Ananadtech with Techreport - could be driver induced variance?).
So without any architecture changes, and assuming no changes in bottlenecks/ceiling sources, AMD should be able to get Zen boost clocks up from 4.0 GHz (1800X) to 4.5 GHz while dropping power from from nominal TDP of 95W to 85W. Sounds alright to me - if not quite the paradigm shift that some were maybe hoping for.
We know there is serious changes. Twice as wide fpu units as huge change for a starter. Beefed up all over to feed it. 64 cores on server side. It's going to wreck Intel solutions to pieces at least for some workloads. Obliterate it. That's a certain. We don't know the cache and memory latency figures. And that's what is going to be interesting especially for gamingShrink ratio of just under 0.7, clocks up between 12%-13% and power down approximately 10% (ranging between 20% and 5% depending on workload - c.f. Ananadtech with Techreport - could be driver induced variance?).
So without any architecture changes, and assuming no changes in bottlenecks/ceiling sources, AMD should be able to get Zen boost clocks up from 4.0 GHz (1800X) to 4.5 GHz while dropping power from from nominal TDP of 95W to 85W. Sounds alright to me - if not quite the paradigm shift that some were maybe hoping for.
that is nothing newWhat making old GPU architecture on a new production node and then overclocking it does have in common with builing new CPU architecture on the same production node???? 7 nm technology is just a "building brick" and you can make anything with any parameters you wish using it.
Yes, I know there are serious changes folks - but we've no barometer for them in this context.
AMD have said 4x the FP performance for EYPC Zen3 - we know half of that is from core count, half from FP width. Does that mean clocks stay the same so the changes are iso-power? To me that is an extrapolation too far as we know many applications don't use AVX - so boost clocks for the most part shouldn't be driven by the FP width change.
What's your rough expectations for eg. fmax on the 8c model?I keep having to post this slide:
This means that running at the same power, you either get double the cores, or more clocks. Not both at the same time. For peak FP throughput, double the cores matters more than improved clocks, therefore peak throughput is 4x.
IMHO the big question I'm really interested in is clocks when running with =< 8 cores. I'd bet there is substantial improvement, but the exact amount will determine whether they beat Intel in all gaming or not.
Which correlates very nicely with the 7nm GPU and even adding a few more % points increase due to the increased memory bandwidth.I keep having to post this slide:
This means that running at the same power, you either get double the cores, or more clocks. Not both at the same time. For peak FP throughput, double the cores matters more than improved clocks, therefore peak throughput is 4x.
IMHO the big question I'm really interested in is clocks when running with =< 8 cores. I'd bet there is substantial improvement, but the exact amount will determine whether they beat Intel in all gaming or not.
Which correlates very nicely with the 7nm GPU and even adding a few more % points increase due to the increased memory bandwidth.
I can't understand the statements that this GPU example portends poorly for Zen2.
They seem to think that a die shrunk GPU has a correlation to what to expect from a CPU designed from the start for 7nm.
Not going to read the 1900 replies. But I care for core performance increases. They can put 120 cores in it if they want but the Zen needs to improve on core performance per clock or the 9900K will continue to eat their lunch.
When Does AMD launch a DDR5 platform? Is that how Ryzen 3000 gets to 16 cores with only 2 memory channels?
That is great news. For AMD's sake in gaming, I hope it is an 8/16 core CPU. I think if the 2000 series had 10 cores and was clocked at 4GHz, it could probably put up the same horsepower as a 9900 in multithreaded processing.well Test was shown that 8/16 ES Ryzen 3gen is almost equal too 9900K in Cinebench r15
https://i.redd.it/q65gpw17wf921.png
Eh I think that is mostly played out. There are some workloads that you should have a certain amount of bandwidth per core. But those aren't the markets AMD is aiming with the client platform. I don't see memory bandwidth holding AMD back from releasing a 16c CPU.When Does AMD launch a DDR5 platform? Is that how Ryzen 3000 gets to 16 cores with only 2 memory channels?
I am the only one seeing absolutely no problem running desktop (really desktop) workloads with 16C/32T pretty much without bottlenecks?
GDC
March 18-22, 2019
San Francisco, CA
"Join the AMD ISV Game Engineering team for an introduction to the AMD Ryzen™ family of processors followed by advanced optimization topics. Learn about the Ryzen line up of processors, profiling tools and techniques to understand optimization opportunities, and get a glimpse of the next generation of "Zen 2" x86 core architecture."
https://schedule.gdconf.com/session...software-optimization-presented-by-amd/864865
I feel lonely on "May 1st" release bandwagon
Is that really when AMD will launch Matisse? As in, on store shelves?
I assume they want to "make some noise" on the company 50th birthday :>