- Mar 3, 2017
- 1,747
- 6,598
- 136
Successfully memed Zen5% into existence.I see, now we are in the Zen±5 zone
Oh that's my bad, I think I just failed copy pasting....The link's broken for me.
What? Sorry I'm very sleep deprived lolIt is either 16% now or 20%+ in 8 months, at some point time to market has more importance than a few % more IPC.
What?No Geddagod, it was not a wish(can't say why).
🤣Successfully memed Zen5% into existence.
Oh no I was talking specific structure sizes. Apparently some stuff in the core for Zen 5 is significantly scaled back? (Presumably in the execution engine or some structures past the rename stage?) I'm assuming they expected the new massive front end to compensate. Who knows.I mean across the board. It is not uncommon for a new architecture to show some little regression on some tests while being noticeably faster in the average. And If I understood correctly, he measured an average of slightly less than the advertised +16% iso clock in his set of application. So calling a +13-15% a "step back" is a kind of stretch, maybe disappointing if we think about the increase in the core area and time to market. It will depend on how many low hanging fruits there are, and if Zen6 will take advantage of those when using a denser process.
This seems very vague, probably when creating the front end not all the components have been duplicated, so in general there is an increase of the total resources available but less than it would have been expected in a comparison with Zen4.Oh no I was talking specific structure sizes. Apparently some stuff in the core for Zen 5 is significantly scaled back? (Presumably in the execution engine or some structures past the rename stage?) I'm assuming they expected the new massive front end to compensate. Who knows.
What? Sorry I'm very sleep deprived lol
EDIT: erased confd. info.What?
What he's saying makes a lot of sense (who would've thought that about Jim Keller of all people).
This sounds like fabricated hopium.The planned Zen5 was really 30% faster than Zen4. But AMD suffered a lot of turbulences and decided to delay the "Scary fast Zen" to some future Zen iteration. Can't say more.
Let's hope it's not the comeback of Bulldozer.I guess we'll only see the full impact of Zen 5 uarch changes only in the future generations.
This was never launching in Q3 2024. Strix always had 3.5 in the roadmap.RDNA4
I remember watching interview with either Keller or Mike Clark few years back, in which one of them stated that new architecture does not mean major jump in performance, it means reworking things to not regress too much, but open new possibilities and options of improving no longer possible on older, fully developed architecture.
Already back then it was indicated that normal IPC jump is what they aim for in general.
If I find that quote I will link it here.
Z6 probably design complete by now, should be in implementation.I guess we'll only see the full impact of Zen 5 uarch changes only in the future generations.
There is a neat Zen 5 table by instlatx64:Leaked slides already showed different FP width variants, why not?
View attachment 101542
Doesn't look like that. Long development times with so so results.That's 10 years like dead.
People say:Same here, it s better to release a new CPU at 16% better IPC now than next year at 20%, financiarly it makes much more sense.
Well, it doesn't look that good since the MLID's roadmap slide painted Zen 6 as a very light Zen 5 optimization with that 10% IPC figure. Note this figure is lower than Zen 4.DisEnchantment said:Z6 probably design complete by now, should be in implementation.
But it is going to get a clock bump moving to the new process N3E/N3P, so it is not going to look as bad as Z5.
It’s amazing how much of this Apple and AMD get this right.Mike also made this presentation, which stresses on multiple aspects of performance beyond IPC.
They are selling their A0 silicon, wich say that they were pressed to release the chip as soon as possible, as for up to 6 months with no competition they can make a lot of money out of DT Zen 5.People say:
* AMD nowadays got a lot of money from EPYC sales & gamerz buying X3D like 100x compared to Intel
* Intel not being a threat with the poor broken Lion Cove
* ARM is laughable
Besides, Zen 5 got a scope cut since it still reuses that Zen 2 IF architecture...
What's the rush then?
Well, it doesn't look that good since the MLID's roadmap slide painted Zen 6 as a very light Zen 5 optimization with that 10% IPC figure. Note this figure is lower than Zen 4.
Lol this is so not trueThe planned Zen5 was really 30% faster than Zen4. But AMD suffered a lot of turbulences and decided to delay the "Scary fast Zen" to some future Zen iteration. Can't say more.
Some healthy banter is always welcome, it's just pieces of sand after allLet's hope it's not the comeback of Bulldozer.
Hey it's Friday, a bit of trolling is acceptable
Zen 5 did indeed have problems. Notice the 2 months gap between release date and announcement.They are selling their A0 silicon, wich say that they were pressed to release the chip as soon as possible, as for up to 6 months with no competition they can make a lot of money out of DT Zen 5.
Beside they are suley more concentrated in their mobile offerings, that s where there s actual competition and bigger volumes, so better to get rid of the DT parts and concentrate on the few mobile chips that are still under developpement like KRK and Strix Halo.
Zen 5 did indeed have problems. Notice the 2 months gap between release date and announcement.
I think the problem is that they wanted to scale too fast.They may have held back Granite Ridge because No AI and/or the gaming perf isn't better than X3D.
No is true yes.Lol this is so not true