- Mar 3, 2017
- 1,626
- 5,910
- 136
But which parts will reap a benefit from this? Just the "cloud-native" ones, i.e. the Bergamo successor?there's improvements in L3 cache / Ringbus.
No, all versions it seems. I know nothing about the details though.But which parts will reap a benefit from this? Just the "cloud-native" ones, i.e. the Bergamo successor?
Same general feelings here. This "Plebs have not even the slightest idea on how fast the thing really is" sounds like utter BS. Even if its the more than respectable 40 percent speed increase would turn out to be true, it just means that 95 percent of the time you would not see difference, and the rest of cases it would still be not good enough.I'm all for Zen 5 being great, but honestly I'm a bit confused about the recent twitter hypetrain (not to mention adroc).
It better be something like that, or like a 40% IPC increase, otherwise the hype-train just seems .... stupid. Zen 4 also ended up roughly 30% faster and definitely didn't make older stuff useless (e.g. 5800 X3D). Yeah, i know it was mostly clocks, but the average Joe doesn't care.
Same general feelings here. This "Plebs have not even the slightest idea on how fast the thing really is" sounds like utter BS. Even if its the more than respectable 40 percent speed increase would turn out to be true, it just means that 95 percent of the time you would not see difference, and the rest of cases it would still be not good enough.
Tell me how fast the thing is, when it finishes tasks, that takes Zen 4 whole hour, in mere seconds. If it cant, then there is nothing special about it, that we havent seen before.
I guess you're right in that Intel HAS to market some kind of advantage of ARL over RPL, otherwise what's the point?
Just sad that they could only muster a few % points improvement, but to be honest, I'm totally cool with AMD gaining more market share for a few generations only because I know that in the grand scheme of things, the Intel-AMD market share needle is still very far from 50-50, which is where I believe it needs to be for maximum consumer benefit from a competition point of view.
C'mon, how would you power CS bots without AI 😂But ARL will be the first gaming CPU with AI! That's been their push so far. Color me unimpressed.
Or maybe what i am saying, is that hyping Zen5 as the best thing since sliced bread, as if 30 percent more perf for another 700 bucks 2 years later is anything but bare minimum, is kind of daft.Wow, that has to be one of the stupidest takes I have ever seen. So you basically want Zen 5 to be 10 times faster or so, otherwise it's garbage? Ok...
Doesn’t that honor technically belong to AMD? Hawk point has “AI” and you could definitely consider that to be a gaming CPU. Zen 5 will also have it and launches before Arrow Lake. (unless something has changed — I haven’t had much time to stay on top of things)But ARL will be the first gaming CPU with AI! That's been their push so far. Color me unimpressed.
Or maybe what i am saying, is that hyping Zen5 as the best thing since sliced bread, as if 30 percent more perf for another 700 bucks 2 years later is anything but bare minimum, is kind of daft.
But sure, you do you, and keep twisting my words and pretend i said its garbage.
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.I'm all for Zen 5 being great, but honestly I'm a bit confused about the recent twitter hypetrain (not to mention adroc).
How exactly is it going to "instantly make everything older obsolete" or be "almost a Conroe level jump" ?
Let's look at what we know:
- AFAIK Most of the hypers tend to agree that the leaked architecture slides from AMD are correct. Yet seem to think the actual gains are bigger than the 15% mentioned (some say 20% some ~30-35%).
- Those changes essentially make it a bit wider and better executed Golden Cove. Now I know AMD has managed to extract way more out of a similar-width CPU than Intel, but nothing from that leaks is anywhere near a "Conroe level" upgrade:
- There are also no packaging, uncore or major I/O changes. The same I/O die probably limited to the same bandwidth /latency numbers as Zen 4.
- The FPU part looks beefed up, but that has very little relevance to the general user, outside of benchmarks.
- And there will be a ~200 Mhz frequency drop.
So what is that "secret Conroe sauce" that causes people to hype it endlessly (provided there is any)?
If I had to bet, I'd wager it's some sort of architectural change not yet listed that will disproportionately affect some workflows:
- A godly new branch predictor?
- A radically different way to built the ROB buffer? That would allow a way bigger OoO windows in some scenarios (there have been some papers on that).
It better be something like that, or like a 40% IPC increase, otherwise the hype-train just seems .... stupid. Zen 4 also ended up roughly 30% faster and definitely didn't make older stuff useless (e.g. 5800 X3D). Yeah, i know it was mostly clocks, but the average Joe doesn't care.
See, you're not that dumb.If I had to bet, I'd wager it's some sort of architectural change not yet listed that will disproportionately affect some workflows:
- A godly new branch predictor?
- A radically different way to built the ROB buffer? That would allow a way bigger OoO windows in some scenarios (there have been some papers on that).
Implying that someone is relatively dumb is dumb. ShameSee you're not that dumb.
Major front-end innovation + bigger core ???? good stuff.
Well he almost spelled why it's special, just in a weird, irony-poisoned terminally online skeptic™ way.Implying that someone is relatively dumb is dumb. Shame
Well he almost spelled why it's special, just in a weird, irony-poisoned terminally online skeptic™ way.
Whataboutism. Intel got away with 5 percent perf increases between generations once, so now even 10 percent by AMD is to be cherished and approached as godsend. How about no.If intel were doing 30% improvement every 2 years from Sandy onwards then Zen would not even exist. To call it bare minimum is laughable.
I don't recall such sustained performance progress since OG Athlon / P3 through to Sandy.
Oh no lol, everyone here is underselling it.thats doing AMD more disservice than it helps.
I'm honestly perfectly fine with that. Sandbagging AMD is the best AMD. It's whenever that fails and falls apart (*cough RDNA3 *cough) that I have serious issues with them.Oh no lol, everyone here is underselling it.
It's never supposed to.It's whenever that fails and falls apart (*cough RDNA3 *cough) that I have serious issues with them.
Oh come on now.It's never supposed to.
AMD says RDNA 3 is projected to provide more than a 50% uplift in performance per watt.
Experience up to 50% greater performance per watt with AMD RDNA™ 3
This isnt really called for, simply confirming he was on the right track would have been a nice way to respond.See, you're not that dumb.
Major front-end innovation + bigger core ???? good stuff.
I'd say that's a given, especially with halo products. And even then the development in the past years wasn't all that bad considering the general inflation in pretty much all markets.I expect the performance-per-cost-of-ownership gains to be notably lower than the absolute performance gains.
Yeah I know.Oh come on now.
No, it's appropriate, most terminally online skeptics™ have no idea what they're talking about.This isnt really called for, simply confirming he was on the right track would have been a nice way to respond.
Your a smart and (apparently) well informed guy, no need to talk down to the less informed who are most people!
Be better than that, it makes the world a better place.
It's still a decent perf/$ bump, particularly for anything 1t-sensitive.I for one belong to those for whom per-core or per-socket performance is of secondary interest at best, while performance/$ (purchase and operation) does interest me quite a bit.
So what will we get, and when?What we get: not that and later too