It's a leveraged core, not a new one. There's no reason for it to bring such a jump.
It's a leveraged core, not a new one. There's no reason for it to bring such a jump.Could Zen6 bring greater than 20% IPC increase?
It's notable that although every Zen core has brought "double digit" IPC gains, it's never been >20%.
I expect a better IPC. I have the feeling that with the new uarch, AMD has left some low hanging fruits to pick (due to time constraints, process, or strategy). Don't get me wrong, I think the Zen5 IPC improvement looks good on integer workloads, but with a new uarch there's always room for improvement in particular when a new process is invited to the party.
Is that enough to wait for Zen6? I'm unsure. But I'm eagerly waiting for desktop Zen5 benchmarks and deep technical reviews before making a decision.
Mike Clark: Yeah, as we think of Zen 5 we needed a new foundation for more compute to drive future workloads that continue to stay on this cadence of double digit IPC per generation. So you know we have been at the original Zen was 4-wide [dispatch and] 6 ALU’s and we had done a lot of innovation to really you know leverage all those resources [in] Zen, Zen 2, Zen 3, Zen 4. But we really we’re not to be able to keep that up, so we really needed to reset that foundation of a wider unit, more ALUs, more multiplies, more branch units, and then be able to leverage that like we did with the originals then to provide innovation going forward.
If I knew I could not tell you 😀 I think another 10% for integer might be good.Could Zen6 bring greater than 20% IPC increase?
It's notable that although every Zen core has brought "double digit" IPC gains, it's never been >20%.
Thanks for the quote!Mike Clark somewhat alluded to that in his interview with C&C. I'm hopeful for it anyway. I'd really like to see AMD be willing to give up some frequency to improve IPC and hopefully get a more efficient core out of it. The GHz race, to be honest, has been exciting and it's pretty remarkable how much the designers have been able to due to get near 6 GHz operation, but I would appreciate even more a "smarter" core that still improves performance but runs at more efficient frequencies.
It is a good target. The problem is the Zen cadence is nearly half the speed of the competition. Both ARM and Apple try to have yearly product updates.If I knew I could not tell you 😀 I think another 10% for integer might be good.
I know there was a discussion about the frequency/IPC tradeoff that x86 manufacturers are making and how they are chasing the highest clocks for the sake of marketing and I'm frankly not buying that theory, in AMD's case at the very least.Thanks for the quote!
Regarding the rest of your post, I fully agree: I'd prefer more IPC and less frequency, I really like the Apple and Arm approach. And if I consider getting an AMD it's because they are more efficient than Intel... and have AVX-512.
My speculation is that the optimal clock vs IPC for ARM and x86 is just different.Now answering the question of why the ARM and x86 vendors have arrived at such different conclusions regarding this trade-off requires someone who actually knows what he's talking about, I have no idea personally. Any guesses as to what might be going on here?
This is the answer really, imho. The zen compact cores are much more area-efficient than apple cores.The only reason I could think of that could make them wary of switching to lower freq, higher IPC design if it's indeed more performant is severe area penalty maybe? So severe that it would not work on server?
Did they? Except for Apple, which only sells consumer hardware, aren't all other arm vendors at similar PPC* and area levels compared to Zen (C) cores?why the ARM and x86 vendors have arrived at such different conclusions regarding this trade-off
First 16% is way over 10% plus. Second, until the actual benchmarks come out, that 16% is from AMD. The real number could be more, as it has been before.Why are we expecting big gains for Zen 6 when the same roadmap that halted 32% dreams for Zen 5 also showed 10%-plus?
Why don't you look back at the leaked slide, re-read my post and come back to me?First 16% is way over 10% plus. Second, until the actual benchmarks come out, that 16% is from AMD. The real number could be more, as it has been before.
We aren't.Why are we expecting big gains for Zen 6 when the same roadmap that halted 32% dreams for Zen 5 also showed 10%-plus?
What leaked slide ? How about a link. Nowhere on this page that I see.Why don't you look back at the leaked slide, re-read my post and come back to me?
There are benchmarks though, we know everything there is to know about Zen 5 mobile since it has officially launched.First 16% is way over 10% plus. Second, until the actual benchmarks come out, that 16% is from AMD. The real number could be more, as it has been before.
Flametail is expecting 20%, and Nothingness is talking about greater gains, and it isn't the only time where it was alluded to saying a big departure in architecture gives more room for future uarchs.Don’t think anyone’s expecting large gains with leveraged core. Zen 4-like jump at best (so 12-13%), but with much smaller clock gains, 100-200 MHz tops.
Yea, 10% on Int and 25% or so on FP.There are benchmarks though, we know everything there is to know about Zen 5 mobile since it has officially launched.
Which makes the 16% IPC score dubious if Zen 5 desktop doesn’t get close to that 16% increase on int performance.Yea, 10% on Int and 25% or so on FP.
Zen 5's goals were as follows IMO.Flametail is expecting 20%, and Nothingness is talking about greater gains, and it isn't the only time where it was alluded to saying a big departure in architecture gives more room for future uarchs.
If a big improvement is coming, it's not Zen 6.
Yea, 10% on Int and 25% or so on FP.
Flametail is asking whether it could, rather than straight up expecting.Flametail is expecting 20%, and Nothingness is talking about greater gains, and it isn't the only time where it was alluded to saying a big departure in architecture gives more room for future uarchs.
If I knew I could not tell you 😀 I think another 10% for integer might be good.
Flametail is asking whether it could, rather than straight up expecting.
And Nothingness is on record saying that another 10% would be a good result which is in line with the roadmap.
Exactly. If ARM sticks to their healthy IPC CAGR, they'll surpass AMD's absolute performance in a few years.It is a good target. The problem is the Zen cadence is nearly half the speed of the competition. Both ARM and Apple try to have yearly product updates.
Zen's cadence as of late is for fighting refresh-happy Intel and even then it may not be enough.
That's for another thread but I suspect they'll be better off with a focus on clock rate for a bit - at least it worked for Apple.Exactly. If ARM sticks to their healthy IPC CAGR, they'll surpass AMD's absolute performance in a few years.