- Mar 3, 2017
- 1,747
- 6,598
- 136
Will be good for those that primarily are interested in max MT perf on DT though.I presume that's no clock increase on desktop, but if the efficiency increases are there then might the large server chips might get a huge increase in guaranteed all core base clock?
That would explain the big uplifts rumours we got if they came from hyperscalers. Great for DC too I would imagine. Not that great for desktop though.
Now, this somehow doesn’t work with Nvidia. That guy always manages to still hire and retain excellent engineers. Even if some great ones leave, later on other great engineers join.Company's performance is tied to abilities of engineers working in it, and they come and go. Saying that the current issues are the same as those of 15 or more years ago is just silly, as it's the same company in name only.
Look I didnt compare the last 4 years. I compared. i compared Zen 4 to Zen 5 and M3 -> M4. Both companies gen on gen in the last year.If we compare Apple M4 to M1, the difference is nearly 8% on the same clock in Geekbench 5 (149.3% * 3.19 / 4.4 - 100% = 8.2%, link to results). If we use SPEC 2017, the IPC difference will be nearly the same (~11%), according to Geekerwan.
That's what we have for 4 years. Yes, increasing clocks from 3.2 to 4.5 GHz is also not an easy task by any means, but, as a result, a P-core in M4 consumes more than twice as much power as the one in M1.
In the case of AMD, we expect nearly 25-30% IPC improvement for Zen 3 -> Zen 5 with the increased clock speeds for the same period (2020 -> 2024).
To be fair, AMD did not hype Zen 5 performance to the moon. Leakers who believe they know more than anybody else about what future products would be like created this narrative.People who lived through the hype cycles of K9, Barcelona, Sandtiger, Bulldozer (and, of course, Tunnelborer), or K12 know this.
Not all CPU talents are famous or want to be famous.Someone clone Keller and GWIII.
I hope so, but rumors are all over the place, so I'm just going by what they said regarding Z4, which was that they had reviewed the alternatives and found that it didn't make sense to them then. Safest assumption would be that that still applies for Z5.You won't have that issue if they put 3D V-cache on both chiplets, as rumored.
It was not in 7 months. It was developed for several years as always.if true, Apple got an avg improvement of 7% in 7 months and achieved more than AMD did in 22 months. This makes M4 look very good.
This is embarrassing. Like I think it’s fair to say AMD lost talent. I mean it looks like it. The cadence slowed down as well after Zen 3.
I agree that Apple have slowed down too but so did AMD by the looks of it.It was not in 7 months. It was developed for several years as always.
What clearly happened is that M3 was pushed back a lot due to process delay, almost to the point where M4 was ready. There was originally going to be much shorter gap between the M1 and M3 architecture and thus larger gap between what is now M3 and M4, that's all.
That's pretty clear given how Apple went three years without a new performance core before that. There's roughly 4 years between M1 and M4, so it was clearly intended to be closer to ~2 years spacing between new architecture originally. And both of those new uarchs (M3, M4) show single-percent IPC uplift.
Those 18-24months gaps are pretty much what AMD does, and the IPC, well. Apple owes the lion's share of their success after M1 to clock speeds chasing, which may be that effect of starting from low level, but also having the tailwind of being the richest kid on the block with most reckless customers so they can afford the node upgrade goodies 1-2 years before the rest.
To be fair, AMD did not hype Zen 5 performance to the moon. Leakers who believe they know more than anybody else about what future products would be like created this narrative.
And YouTube clowns like RGT just ran with it seeing click opportunities.
"The best is yet to come" etc. are just optimistic statements which they don't repeat everywhere.
No need to stoop so low, and take an engineer's passion about his work out of context. He did not promise the moon.
It's a meme my dude.No need to stoop so low, and take an engineer's passion about his work out of context. He did not promise the moon.
But you can see someone doctored this picture and generated hype.
Idk....Pretty efficient at least it looks.
I thought rembrandt was extremely based. Good battery life, and pretty good perf/watt. I'm looking to upgrade from my 12900h asus m16 because charging between classes is now getting kind of a pain.... and I don't think ADL was all that competitive in battery life even when it released, IIRC, lol.Sooo....we went from +40% to +32% to +16% and now closer to +10%. Ooooff. Performance seems to not be exciting.
The only exciting aspect for consumers now may be power efficiency. But given that AMD didn't say anything concrete about that, maybe it's not actually much better either...
I am hoping to be wrong when strix releases, but I think I'm just on hopium at this point because I was looking forward to upgrade from Rembrandt.
To be fair, AMD did not hype Zen 5 performance to the moon. Leakers who believe they know more than anybody else about what future products would be like created this narrative.
And YouTube clowns like RGT just ran with it seeing click opportunities.
"The best is yet to come" etc. are just optimistic statements which they don't repeat everywhere.
Taking one step back, to jump two steps forward!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.
Is there a step back?Taking one step back, to jump two steps forward!
Ye, when reviews come out 🤣If we wait some days more we will get -20% /s
Saw that on reddit a couple days ago too IIRC.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.
According to Huang, who apparently has already started micro benching Zen 5, there is. But who knows, his methodology might be a bit scuffed...Is there a step back?
The link's broken for me.Saw that on reddit a couple days ago too IIRC.
Found it (at least the reddit comment), it was Keller
Honestly... kinda feel like its cap, I think AMD was still aiming for a higher IPC uplift with Zen 5.
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.According to Huang, who apparently has already started micro benching Zen 5, there is. But who knows, his methodology might be a bit scuffed...
Honestly... kinda feel like its cap, I think AMD was still aiming for a higher IPC uplift with Zen 5.
EDIT: removed confd. info.Saw that on reddit a couple days ago too IIRC.
Found it (at least the reddit comment), it was Keller
Honestly... kinda feel like its cap, I think AMD was still aiming for a higher IPC uplift with Zen 5.