- Mar 3, 2017
- 1,623
- 5,894
- 136
3500 seems to be the rumor.What then, will the expected GB6 score of Zen 5?
I think that was supposedly done on an ES that had clock regression (which is supposedly now addressed).It would be really crazy to see Zen 5 clock the same for 1T as Zen 4, with all the stuff they added/changed. But AMD did it with Zen 2 -> Zen 3 on the same node, so it would not be unheard of.3500 seems to be the rumor.
Guess i am just slightly annoyed by the state of modern "journalism".@Timmah!, money.udn.com is not a tech news site. (Well, neither is wccftech, arguably.)
(Try Homes & Gardens for more in-depth reporting on Zen 5, maybe...)
Nah no way, only 10-15% ipc increase with clock regression. All that for $999 launch date Q4, availability Q1 2025. Tame your expectation bro you're getting 10% single thread uplift at most. /sI think that was supposedly done on an ES that had clock regression (which is supposedly now addressed).It would be really crazy to see Zen 5 clock the same for 1T as Zen 4, with all the stuff they added/changed. But AMD did it with Zen 2 -> Zen 3 on the same node, so it would not be unheard of.
This isn't needed on RISC-V ISA's part only on the Architectural ISA.I don't know if there's coming clustered - split register file - risc-v design but it's sure that risc-v even considered to split registers in ISA level.
The software-facing ISA doesn't need to handle clustering. The pure reason for clustered cores is to handle decrease in figure in Dynamic EPI given extremely large monolithic cores. Clustered cores always win in the wider Energy per ILP(EPI) race. Large Monolithic cores have high energy consumption for their growth in ILP. While Large Clustered cores have low energy consumption given same IPC/ILP.ARM and x86 are adding instructions for every possible side-case. Risc-v is better but even there is million proposals to add very hardware-specific instructions. And ISA does not have to support clustering - compiler can cluster every ISA as it wants( only practical with enough hardware registers) and Risc-V so far have not added any clustering-specific instructions - only have kept path for clustering open for not implementing instructions that makes clustering difficult.
Clustered Micorarchitectures doesn't touch the ISA or programming models. It is either a CMP core or a SMT2 core to software.Nobody is going to change the entire programming model, which is what clustering would require.
What like 16 Zen5c cores? (Mostly joking btw)
Inb4 new RGT video where he now claims to have another "source" which backs up his claim of Strix Point - N4P and Strix-Halo - N3E.
You can blame WTFTech for proving that you can be relatively successful in the tech journalism biz by just vomiting mass quantities of low-level regurgitated articles. Look at everyone else now: Videocardz, Tom's Hardware, a bunch of other tech outlets do essentially the same thing. Honestly, I believe that if Twitter were to go under, WTFTech and a bunch of other outlets would lose a significant portion of their sources for content.Guess i am just slightly annoyed by the state of modern "journalism".
It's a great sorrow, that we do not have high standard tech journalism like old Anandtech, nowadays.You can blame WTFTech for proving that you can be relatively successful in the tech journalism biz by just vomiting mass quantities of low-level regurgitated articles. Look at everyone else now: Videocardz, Tom's Hardware, a bunch of other tech outlets do essentially the same thing.
It's symbiotic in a way, no? One is a platform for aggregating often-time-sensitive-juicy-info (arguably most of it belonging in a cesspool) meant to affect or sway population sentiments while the other disseminates it.Honestly, I believe that if Twitter were to go under, WTFTech and a bunch of other outlets would lose a significant portion of their sources for content.
Nobody is going to change the entire programming model, which is what clustering would require. Also, the kind of "weave independent execution paths together" that you propose really would not help as much as you think, because in practice execution time of those independent paths are going to be wildly variable, because of unpredictable memory latency. As in, as soon as you do anything more complicated than compute Pi, you should expect fully independent code paths to diverge by thousands of instructions, not tens. And at that point, one of them is going to just lag enough that everything else stalls on it.
Proverbially, no-one actually does that. In the entire industry a few weirdoes recompile code, and everyone else runs the same binaries they have for two decades. Only slightly exaggerating.
You obviously don't know/remember Daily Tech!It's a great sorrow, that we do not have high standard tech journalism like old Anandtech, nowadays.
In a way, yes, but if anything WTFTech is the leech when all they do is just repackage what's being written on Twitter into a format that nets them ad revenue. The worst part is that a basic news aggregator would do the same job perfectly fine, yet WTFTech in their efforts to regurgitate and not purely plagiarize often times misrepresents the leaker(s) by either adding info that was NOT said by the leaker(s) OR misinterpreting the info entirely. It's self-evident that the writers there have sub-par tech knowledge (at least not beyond what the average PC gamer would know) and it shows. If someone like Ian Cuttress were to interview them, they couldn't explain what they were writing about beyond just stating what marketing slides say. I mean, c'mon. I've seen one of their "deep dive" CPU architecture articles before, and all they did was just summarize bullet points off the architecture slide deck. If all they can do is reiterate marketing slides, what are they besides a corporate plug?It's symbiotic in a way, no? One is a platform for aggregating often-time-sensitive-juicy-info (arguably most of it belonging in a cesspool) meant to affect or sway population sentiments while the other disseminates it.
And people get rumors/gossip without loading the Twitter domain on their computers.
WIN WIN!
That's why I try not to go there. Videocardz and Tomshardware are vastly more bearable.If all they can do is reiterate marketing slides, what are they besides a corporate plug?
The WTFTech comments section is 100% definitive proof that the owners DGAF about journalistic integrity. What self-respecting news outlet would let a cesspool like that fester, let alone condone it. Since that toxic environment is what drives their regulars to keep visiting their website (read: bring in more ad revenue), they won't ever moderate or curb their comment section. It's always been about money to them, and is why I'll never take them seriously. Every single review of a product they've written about has always been positively reviewed; they simply will write positive reviews if given money or free products.That's why I try not to go there. Videocardz and Tomshardware are vastly more bearable.
WTFTechTV YT channel has one-third the number of subscribers as MLID's channel. So one-third as "respectable"WTFTech is the website equivalent of a youtube channel like MLID.
Dang, that's pretty solid. PHX gets about 5.1 GHz ST boost if I recall correctly for the 8840HS. If DT Zen 5 has a small or non-existent clock regression (i.e. 5.5 GHz) then it should hit 3700 points.3500 is for Strix I guess?
Phoenix is doing 2600 in GB6 ST.
3500 is a 900 point increase, aka about 35%, which fits with the IPC gain rumours.
According to adroc Zen5 DT will not have any NPU, so it'll show zero in task manager if that is to be believed. See below:NPU coming to task manager!
So what are your thoughts about what NPU will be included on the desktop variants of Zen5, and what performance can be expected from it?
None.
Granite Ridge is Raphael with a new CCD.
Where’d you hear that?
Very late Engineering Samples of AMD’s Zen5-Flagship, the Ryzen 9 8950X score 3500 points (ST) and 24k points (MT) in Geekbench 6.2.
Massive AMD Zen5 Leak Reveals 20% Single-Core Uplift, Sarlak 'Strix Halo' Achieves RX 6750 XT Levels of Performance at 95W-125W
RedGamingTech has dropped bombshells by revealing a large number of AMD Zen5 benchmarks, expressing serious confidence in team red for 2024.appuals.com
Of course, take that with a boulder of salt.
P.S. And just like that, I discover that there is a company called https://www.bouldersaltcompany.com/