- Mar 3, 2017
- 1,682
- 6,197
- 136
are we seriously posting random ebeggars nowAnd here it is. Leak about Zen 5 IPC increase.
Sometimes he gets genuine stuff but makes all the wrong conclusions.But thiiiis time he must be right? No l34k3r is wrong *checks notes* 8 times in a row or something.
I won’t a believe a thing he says. After RDNA 3 , stuff him.
And here it is. Leak about Zen 5 IPC increase.
Everyone who shilled that was wrong.After RDNA 3 , stuff him.
He says IPC increase is 30% and calls it absurdly high, and is instead expecting 15-20%. I'd call 30% low, especially when considering the clock regression.
And here it is. Leak about Zen 5 IPC increase.
Did you believed 128 MB Infinity Cache in RDNA2, that he leaked, then?I won’t a believe a thing he says. After RDNA 3 , stuff him.
This is the statement: "And then, we also see Turin, our Zen 5 product coming in the second half of the year."
30% IPC is too good to be true, despite clock regression?Paul is dancing around the topic of IPC, because the number, as Adroc written many times already is too ridiculous to be believable.
you're shadowboxing wrong things again.30% IPC is too good to be true, despite clock regression?
If it's instead 15-20% with clock regression, we're looking at only a net 10-15% performance increase over ~2 years. That's approaching Intel sandbagging territory, when they were giving us 7% per generation. If Adroc is correct AMD will also bump the price to $999. Who'll be paying that when you can a 7950X3D which is not much slower, for about half the price already now?
So what's the right direction to think in? And why?you're shadowboxing wrong things again.
not thinking in the right direction.
You're thinking how it could be worse.So what's the right direction to think in?
Well they weren't cramming much more stuff, that's the issue.Intel were getting much smaller IPC gains from and with node shrinks were far more impactful though, allowing to cram much more logic with every shrink.
Who could've thought that hardware clickbait rumor mill websites would just run with incomplete quotes without fact checking... oh wait.I think this is the statement from AMD which just being misunderstanded:
Well they weren't cramming much more stuff, that's the issue.
Compare the size of Sandy 4c and Skylake 4c.
Focus on Apple.
Apple delivered large IPC gains 2013-2019.
Cyclone and Twister were in particular abhorrently mean for their era.
No?Apple started at a far lower level of IPC
everyone does 4 cycle L1's now brother.It is also easier to make IPC gains at lower clock rates like Apple's - wider designs have a better ability to result in IPC gains since caches are "closer" (in terms of clock cycles) than they are in higher clocked designs.
They're clocking higher because that's all they can do now.While Apple isn't clocking as high as Intel/AMD, at 4 GHz these days they are a lot closer to them than they were in 2019 when they were in the mid 2s.
I'm think it could be what a lot of people and some leaks think it'll be.You're thinking how it could be worse.
That's not the way to go.
ugh.And Zen5 DT looks bad.
yea.Expensive
No? No.poor performance increase
Don't need that.no core count increase
Do you really need a bigger iGP on a desktop/luggable part of all things?no iGPU improvement,
We'll see if it's for real.and slowing generational release cadence
AMD people are damn good at what they do, and they have quite a few comp roadmaps to bury.So why should we think any other way?
But there is no official info on any clock speed regressions either, yet you decided to run with it to come at the conclusion you wanted. Why is that?I'm think it could be what a lot of people and some leaks think it'll be.
And Zen5 DT looks bad. Expensive, poor performance increase, no core count increase, no iGPU improvement, and slowing generational release cadence. Back to Intel-style sandbagging of the past.
Since there's no official info that's the best we've got. So why should we think any other way?
=yesugh.
=badyea.
=yes (or you think 10-15% over 2+ years is good?)No? No.
=you don't need that, others doDon't need that.
=yes (AMD is lagging behind Intel iGPU now, despite AMD claiming to be an (i)GPU company)Do you really need a bigger iGP on a desktop/luggable part of all things?
=You got any official evidence of the opposite? You made dead certain claims about April 2024 release, all the way up until ~3 months before release. What do your 100% certain sources tell you now w.r.t. release date?We'll see if it's for real.
=Ok, so for Zen5 DT we should be seeing ~30% performance increase (via IPC and/or clocks), 32 cores, new IOD with better iGPU, lower price, etc? Any official info to back that up?AMD people are damn good at what they do, and they have quite a few comp roadmaps to bury.
Neoverse V in particular has to die a horrible death and for that you need gigantic IPC bumps.
I think @Fjodor2001 missed Kepler's recent excellent summary as to why we should error on the positive side of IPC gains....But there is no official info on any clock speed regressions either, yet you decided to run with it to come at the conclusion you wanted. Why is that?
It's not that complicated. Zen was designed to be a very balanced uarch, as it replaced both the "HPC" Bulldozer line of cores and "Low power" Cat cores. So they were very careful with any power and area increases, as going ham (like Intel) could hurt markets like Cloud and low-end Notebooks.
With Zen5 they now have the Dense/Compact cores that can serve markets that need max core density, low cost and low power cores. So the Zen5 design team was allowed to make a more performance focused core. As a result, the performance increase is bigger than usual.
IOW, Zen1 to Zen4 were "Medium" cores. Zen5 is AMD's first "Big" core in well over a decade.