AMD Zen4/Phoenix CPUs May Feature Hybrid Architecture with Performance and Efficiency Cores
Click here to post a comment for AMD Zen4/Phoenix CPUs May Feature Hybrid Architecture with Performance and Efficiency Cores on our message forum
fantaskarsef
So in the future we have a CCD which performs normal, and a downclocked CCD? So they're putting the worse / cut down chiplets on those CPUs?
With X3D, do we then get a 3D cache CCD and a bad chiplet on the side, much like a budget 7900X3D right now?
And yeah, way to go, just do what the competition does.
H83
I hope not, I'm not a fan of this approach for desktop Cpus.
Embra
Nooooo!! Haha .... Just kidding.
Horus-Anhur
Now AMD and Intel are going to enter a race to see who can cram more e-cores into a CPU, and then brag about the highest cinebench score.
https://media4.giphy.com/media/5WUsLZLkZlMdO/giphy.gif?cid=ecf05e471on7pfzm18depfp1n2ee4los5t1ikl66xq5d178u&rid=giphy.gif&ct=g
Undying
Craming more e-cores into the cpu does not increase gaming performance. I hope Amd keep the v-cache cpus for the gaming crowd and cinebench players can have their hybrids.
WhiteLightning
Moderator
makes sense . So people can turn them off again because their game is unplayable because of them LOL
Picolete
That's entirely post 2012 Microsoft's fault, and their scheduler; multi chiplet or big little CPUs work better on linux
schmidtbag
The only valid reason to complain about this is because MS does a terrible job at writing schedulers, and Intel has already got a lot of the dirty work done. Conceptually, this is a smart move. To my understanding, AMD already uses fewer transistors per core than Intel's P cores and achieves better performance-per-watt. If they make E-cores that actually function as E-cores (Intel does too much heavy lifting with their E-cores) and they only offer up to 4 of them per package, then I think it'll be totally fine. These cores would be small enough that they might just fit on the I/O die.
Reddoguk
I call BS on this one. Why? because i'm 100% sure that Intel would of trade marked the hell out of it. I bet they have patents for everything they do.
tsunami231
I not shocked by this, but then again there obvious issue with game not working right cause hiting e cores. and if game isnt patch ( in the case of old games) the OS needs to be smart enough to not put the load on ecore till AFTER pcores are loaded. if the game is old enough it should not matter as the ecores are world better then cores on my 6700k if 6700k dont have issue with those game ecore should not either. atlest till it newer games. that seem to be release unoptimised for pc these day and general in bad condition.
This is issue from what i been seeing it almost as bad if not worse then the HT needing to to be disable for game to work right. though i dont remember the HT issue cause massive stutter like as been seen in recent new game when they hitting those ecores.
imo Hybrid is "good" idea the problem is MS scheduler is still bad. maybe it will be fixed in windows 14?, by that time we will of had 3 more revamped "ui" that suck? 🙄
user1
they've been doing that for a while already:D,
this is probably for supporting zen4c alongside zen4 cores, could make sense for an apu, where you want to save die space, though it is also possible that amd intends to release a zen4 +zen4c dual ccd cpu, which would increase the number of cores beyond 16, however both possibilities depend on the assumption that the chiplets/ccx used on bergamo have a higher core density, which is hard to say , since bergamo is 128 cores, but 128 doesn't divide into 12(chiplets). so either amd is increasing the number of chiplets for more cores , or decreasing them to 8 chiplets and doubling the number of cores per ccx to 16.
big.LITTLE/ non-symmetric designs aren't exclusive to intel, arm chips have been doing it for like a decade.
JamesSneed
Its likely an older chiplet tech for the efficiency core just like Zen 4c. The idea is to shove more cores into the same space. I would expect the full efficiency chiplet to be 16 cores. Considering typical desktop workloads, you usually don't need more than 16 high performance threads which an 8-core CPU will cover. Once past that you are going for parallelism where a weaker 16 core part is going to be stranger than a stronger 8-core part. We have seen this work for ARM and Intel already so its a pretty clear path. This will be cool for the X3D parts as well.
JamesSneed
I expect them to take a Zen3 core to use for efficiency cores. Just build it on 5/4nm vs 7nm. Zen4 cores have 58% more transistors than Zen3, so it seems like a logical and easy move.
Pryme
Is a good arquitecture for mobile, like laptops. For gaming, or some other workloads on desktops, I dont see being relevant and it is a really good ideia if we had desktop chips with only performance cores.
This war for Cinebench points need to end. My CPU have half the Cinebench points of a 5900x, that diference is going to be materialized on games? Nope. Intel and AMD need to meditate better sometimes on some products.
Agent-A01
That sounds like a typical AMD response. But no performance plus efficient cores have been a thing for many years.
alanm
Not encountered any issues in any game so far. Are ppl up to date with their info re e-cores? I know some complaints came up after Alderlake release, but would have thought they've been addressed by now. Links to gaming performance issues since Intel 13th gen released?
cucaulay malkin
Don't need to rely on e-cores for games if p-cores do the job just fine. E-cores are there to help with MT applications, and that very much includes games that use more than 6c/12t, which these days is pretty much the majority of them.
You find a 6/12 cpu in the top10 of the gaming cpu chart
https://www.techpowerup.com/review/amd-ryzen-9-7950x3d/20.html
13600k with 6p/8e is above 12900K/7950x, with 7950x3d only beating it by 10% stock vs stock on 6000mhz ram. with 13600k core/mem oc they'll be the same cause 7950x3d is at its limits there, locked for oc and memory can't take more than ~6000 without taking a performance penalty. No wonder r8000 is coming. x3d is both fast and efficient, but apparently it can't compete with intel on cost, can't have +500eur products competing with regular i5s in the long run. I'll take the 300eur big+little if performance is same, no matter what I like or don't like about hybrid designs. Seems like amd thinks likewise, value comes first. P+E will make it into their the mainstream line, P w. v-cache will be rarity cpus for twice-three times the price. The cheapest x3d will launch around 2.5x of the cheapest r7000, if that mainstream 6/12 ryzen had e-cores too, it'd benefit a lot more people than those who are willing to spend +500eur on a gaming cpu alone.
Alessio1989
keep the shit big.little in intel home, thank you. nobody cares to spare 1w in idle on a desktop system. Intel already messed out AVX-512,we don't need the same on AMD.
alanm
[youtube=LcQUUmi3rWI]
Undying
For gaming fastest core wins. We know that and it will always be the case. Even If ryzen 5 8600x have 6/12 it will still beat any previous gen R7/R9's just like 7600x did unless ofc huge v-cache plays a role.
Steve from GN did "what if" 7600x3d existed which is just 7900x3d with one ccd disabled and still just as fast.
[spoiler][youtube=fCTPvUhYFv4][/spoiler]
