That 20+ months cadence has not prevented such weird launches as AM5 or Zen 5.
AMD used to be nimble in Zen 1-3 times.
That 20+ months cadence has not prevented such weird launches as AM5 or Zen 5.I think they are choosing to go slowly and carefully rather than rush things through and failing to catch some devastating bug (like the Arrow Lake regression in games).
Wrong. Zen 1 and Zen 2 gap is 28 month. Again I have to point this out: the only positive exception is Zen 3.AMD used to be nimble in Zen 1-3 times.
Wrong. Zen 1 and Zen 2 gap is 28 month. Again I have to point this out: the only positive exception is Zen 3.
I doubt Zen 6 will be 2027. It'd be effective corporate suicide. The cadence must accelerate, ARM isn't slowing down. But 22 months as is average is the bare minimum to survive.
Why do people think AMD won't release a Zen 5 refresh in the intervening period?
Ya know, the XT series aka 9950XT etc.
Another possibility would be the opposite of that: a Zen5 Refresh that uses the same CCD as Zen5, but with a new IOD (that will also be used for Zen6).And, if this is one, it'll likely be no different. At best, they could release an N3e version of the CCD, assuming such a thing was already under development, but missed the production deadline. What would that bring though? Maybe 200Mhz peak ST boost, perhaps 300Mhz+ for all core MT clocks under load. It won't solve the iod link restrictions. I doubt it brings even a consistent 5%.
Maybe there's a notable hardware eratta that an XT can solve that gives a bump.
But it isn't going to be a big gain.
Zen5 needs some work on the clustered decode in non-throughput intensive situations.Maybe there's a notable hardware eratta that an XT can solve that gives a bump.
Zen 1 got a healthy upgrade with Zen 1+. The 14nm->12nm frequency gain with many bugs got fixed resulted in a ~10% perf gain. Those embarrassing XT cash grabs can't be compared to that.Wrong. Zen 1 and Zen 2 gap is 28 month. Again I have to point this out: the only positive exception is Zen 3.
I doubt Zen 6 will be 2027. It'd be effective corporate suicide. The cadence must accelerate, ARM isn't slowing down. But 22 months as is average is the bare minimum to survive.
It's a shrink of Zen 1 with improved clock rates that took 13 months to follow Zen.Zen 1 got a healthy upgrade with Zen 1+.
IMO they are considering the budget and manpower increase compared to the Zen 1-3 days.It's a shrink of Zen 1 with improved clock rates that took 13 months to follow Zen.
AMD already launched the N3E shrink of Zen 5C only months after the N4P version and you're saying they're less nimble?
They finally managed to get the APU to launch in time with their leading GPU arch instead of Vega. The budget is working.IMO they are considering the budget and manpower increase compared to the Zen 1-3 days.
I would be surprised if AMD didn't release a 3nm revision of zen 5 for consumersAnother possibility would be the opposite of that: a Zen5 Refresh that uses the same CCD as Zen5, but with a new IOD (that will also be used for Zen6).
It's a shrink of Zen 1 with improved clock rates that took 13 months to follow Zen.
AMD already launched the N3E shrink of Zen 5C only months after the N4P version and you're saying they're less nimble?
The problem is cadence. Zen 5 is a Q1 2024 product that launched in Q3. Zen 6 is a competitive product in Q3 2025 but it launches in... 2027?
Why wait? They don't have a problem with the same core on multiple nodes. Launch on N3 and later for N2.What if launch in 2027 is because they are waiting for TSMC 2nm?
It was more than that, they achieved around a 10% memory latency reduction IIRC, and there was a cache latency bug that got solved.
Sure, the current budget manifests itself - the range of server and mobile products has grown considerably. But the core IP roadmap not really.They finally managed to get the APU to launch in time with their leading GPU arch instead of Vega. The budget is working.
But the problem is they do not yet have enough teams and simultaneously in-development architectures like ARM. Previous products slipping should not impact the next but for AMD it still does. And so Zen 6 will only be competitive in the server space. And against Intel only if Intel keeps messing up.
Where did you get that info?with a 10+% IPC gain
It is "a challenge" though. Someone even called it the "biggest" one.They don't have a problem with the same core on multiple nodes.
As always, source is the official 2023 roadmap. That gain makes sense since Zen 6 is the evolution project of revolutionary Zen 5.Where did you get that info?
There is a process bump from 4nm to 3nm/2nm
That's not AMD's fault but the messianic cult. The only blame AMD has is that time to market needs to be quick. They need to iterate like ARM. And ARM's budget is smaller so there are some questions why AMD won't hire as many teams.it was not the x86 messiah
Going 2 ALUs, 16k L1d & 2-way L1i, horribly slow L2, and doing FMA in 6 cycles?What questionable features did bulldozer have?
With the current cost of nodes, I think if Z6 is 2027 it is partly because its 3nm, not 2nm.What if launch in 2027 is because they are waiting for TSMC 2nm? I mean, Zen 5 was supposed to be a revolution some years ago and probably part of that was being on 3nm with the added transistor budget. Maybe AMD doesn't want to do the same mistake?