- Mar 3, 2017
- 1,687
- 6,243
- 136
I don't see a downside to that. Avoids the dual CCD shenanigans.Only give desktop one 16C CCD?
That's server and it's 32c.What about the rumors that Zen 6 CCD is larger than 8C?
Desktop isn't a thing with Zen6, it's just a luggable/mobile workstation platform transplant.How will they stick to 16C max on desktop with larger CCD's?
remember zen6 stops being server IOD/ CCD for DIY consumer market.What about the rumors that Zen 6 CCD is larger than 8C? How will they stick to 16C max on desktop with larger CCD's? Only give desktop one 16C CCD?
Nah.So i would expect monolithic for CPU + IOD + display controllers with maybe rest of GPU being a separate chiplet,
Zen5c means smaller and area compact cores at better perf/watt, with more cores per chip for non-peasant/gamer markets. Will also be possible with regular Zen5/6 cores.Bigger and meaner core for poverty markets.
As already proven, it does not make any sense to use 170W TDP with only 16C or below. Very little perf gained for the last few 100 Mhz while consuming an insane amount of additional TDP. 170W is intended for >16C while within AM5 lifetime.They intend to use it for fatter cores.
It's not really better perf/W, it's just cheaper.Zen5c means smaller and area compact cores at better perf/watt
Why are you saying things while having no clue about Granite Ridge or any other future AMD products?As already proven, it does not make any sense to use 170W TDP with only 16C or below.
Why are you saying this without profiling a living GNR-B0 silicon?Very little perf gained for the last few 100 Mhz while consuming an insane amount of additional TDP.
They won't. I don't think having a ton of MT performance really sells gaymer processors.Don't think AMD will go over 16C unless threatened by Intel's 40 core Beast Lake.
If you have a wider core, don't you also have more transistors switching at any one time, so more power used? Why are you locked on frequency?Zen5c means smaller and area compact cores at better perf/watt, with more cores per chip for non-peasant/gamer markets. Will also be possible with regular Zen5/6 cores.
As already proven, it does not make any sense to use 170W TDP with only 16C or below. Very little perf gained for the last few 100 Mhz while consuming an insane amount of additional TDP. 170W is intended for >16C while within AM5 lifetime.
Then Zen5c is poorly designed. AMD should look at what e.g. Apple / Qualcomm is doing with their efficiency cores.It's not really better perf/W, it's just cheaper.
It's obvious. Huge amount of TDP consumption on max 16C for the last few 100 MHz:es for nearly no perf gain is insane. AM5 is screaming for more cores to make better use of the 170W TDP.Why are you saying things while having no clue about Granite Ridge or any other future AMD products?
If you've got official profiling statistics from GNR-B0 silicon, please share.Why are you saying this without profiling a living GNR-B0 silicon?
How much more perf will the last few 100 Mhz:es gain on Zen5, while consuming how much more TDP? Do you have numbers?If you have a wider core, don't you also have more transistors switching at any one time, so more power used? Why are you locked on frequency?
No, it's designed for a very specific reason: cost.Then Zen5c is poorly designed
Qualcomm has no LITTLE.Apple / Qualcomm is doing with their efficiency cores.
It needs fatter cores and that is exactly what you're getting.It's obvious. Huge amount of TDP consumption on max 16C for the last few 100 MHz:es for nearly no perf gain is insane. AM5 is screaming for more cores to make better use of the 170W TDP.
I do, but you have to guess stuff on your own.If you've got official profiling statistics from GNR-B0 silicon, please share.
Even 16C was Threadripper area back in 2018. Until it was not.They're not peasants at all.
That's Threadripper area.
If you're too peasant for it then, well, too bad!
It still is, TR starts there.Even 16C was Threadripper area back in 2018
You are not getting more cores.Until it was not.
What I meant is that it's then poorly designed for perf/watt. Compared to e.g. efficiency cores by Apple / ARM.No, it's designed for a very specific reason: cost.
ARM has. Apple has.Qualcomm has no LITTLE.
The 170W is needed and intended for more cores on later Zen generations.It needs fatter cores and that is exactly what you're getting.
Yeah, right. Funny guy.I do, but you have to guess stuff on your own.
There is no official info pointing in either direction.You are not getting more cores.
Please, just give up and accept what matters in client: single-thread performance.
TR may start there, but its not exclusive to it, when it once was. Thats the point.It still is, TR starts there.
You are not getting more cores.
Please, just give up and accept what matters in client: single-thread performance.
They have a gigantic 1t advantage and care not even one bit about atom spam.I'm expecting AMD to follow, unless they want to be left behind in the dust.
The point is you're not getting more cores, they're not free lunch.Thats the point.
because he's too poor for threadripper and wants cinememe pumping in DT mainstream230W does allow 32 cores but why are you insistanr they'd do that when they're making the cores bigger and on a process that doesn't have much higher logic density?
Does not help in MT perf if AMD's 16C will be stomped on by Intel's 40C.They have a gigantic 1t advantage and care not even one bit about atom spam.
Indeed. If he did have profiling statistics from GNR-B0 silicon Intel would very much like to see them.If you've got official profiling statistics from GNR-B0 silicon, please share.
I do, but you have to guess stuff on your own.
Yeah, right. Funny guy.
MT isn't relevant in client to this degree.Does not help in MT perf if AMD's 16C will be stomped on by Intel's 40C.
Oh they do know them.Indeed. If he did have profiling statistics from GNR-B0 silicon Intel would very much like to see them.
Oh they do know them.
They also know Turin numbers.
It will be, if intel´s marketing do their thing As if people knew they dont really need 40C or that most of them are "atoms". As if they knew what Atom is.MT isn't relevant in client to this degree.
Plus, I mean, Atoms. not very useful outside of cinememe.
Oh they do know them.
They also know Turin numbers.
B0 step has been with us for a while, it just PRQd early January.But numbers from actual GNR B0 silicon as of today?
They can't do anything because 1t delta.It will be, if intel´s marketing do their thing
Actual Cinema4D is GPU-accelerated.Sure, its kinda overused, but its not like its completely artificial - its derived of Cinema 4D after all, is it not? Actual app people use for work.