I think the protocol is the same but physical layout is bound to be different if they can drop clocks 10x and achieve the same BW. The only way I know of is to route more wires (they can bc new pacakging allows for denser links) but that requires more contacts.Conversely, in Strix Halo you simply have the same on-die fabric as in Granite Ridge and Turin, but routed to reasonably small vias/ contact pads for hybrid bonding.
I undestood from the article that now there is no serdes on this link, hence my expectation this would be somehow reflected in silicon, but yea it might be a droplet in otherwise busy chip.While it's ten times wider than what's going from SERDES to substrate, it's not terribly wide compared to what is already there on-chip. Actually, what's going from the Coherent Master over to the fanout contacts is likely at the same order of magnitude of clock and width as what's going from the Coherent Master to the SERDES.
Further, as @Joe NYC pointed out, this stuff may not show up on die shots at all because it happens in the "lowermost" metal layers.
i would buy it in a heartbeat if ever releasedA niche product which in its mere existence would consume two 9800X3Ds. And at best the reviews will say to skip it because it is indistinguishable in most tests and even worse in some others at higher cost. It isn't a halo part. It offers no performance over the mixed configuration unless you're running niche work like fluid simulation. It's just an expensive part. AMD has the benchmarks and said "I’m scared to death of it" because it's a very hard product to sell.
The easiest way to avoid reviews is to launch it as part of AM5 Epyc and/or toward the end of Zen 5's life span when no one cares.
Nvidia works very hard to actually deliver more performance at their top end prices.Nobody would buy any GPU at $4000 10 years ago just for gaming
Then they should do the next best thing. Unify the V-cache across the two CCDs and let all cores share it!Why cannibalize two 9800X3Ds when that's what people actually want?
Most likely not happening on AM5.Then they should do the next best thing. Unify the V-cache across the two CCDs and let all cores share it!
Yes. They even had a 5950X3D2. Never shipped. Guess why.is it easy for AMD to manufacture an ES prototype 9950X3D2? so it's certain they have it and tested it
It doesn't make sense but it doesn't take any of their attention to slap a new label on.They are too busy launching renamed AM4 cpus for some weird reason.
A smart company sells people products they need. A smarter company sells people products they want. I'm tired of your personal vendetta against the very concept and those of us who want it. Your words are changing no one's mind and are borderline insulting. We know exactly what it is, what it will do, and what it won't do.Nvidia works very hard to actually deliver more performance at their top end prices.
Wasting another cache chip for hwbot enthusiasts is not gonna be a financial success. Get out of your bubble. The 7800X3D is 6-7x more popular than the 7950X3D because it offers what people want (gaming performance). The 9950X3D2 would be even worse due to price and offering almost nothing measurable out of the box. Why cannibalize two 9800X3Ds when that's what people actually want?
Luckily the minds in charge don't need changing. But maybe they will ship it out when demand tapers off for the actually useful configurations.A smart company sells people products they need. A smarter company sells people products they want. I'm tired of your personal vendetta against the very concept and those of us who want it. Your words are changing no one's mind and are borderline insulting. We know exactly what it is, what it will do, and what it won't do.
It isn't +any%. Unless you overclock it. In which case an overclocked 9800X3D does the same. It is not a halo configuration unless they go out of their way to 1% bin X3D CCDs too. In which case gamers would again be better served by a cheaper, binned 9800XT3D.what matters is +5% over 9800X3D, the #1 position on benches has psychological impact, everyone will buy it just like 9800x3d
It isn't +any%. Unless you overclock it. In which case an overclocked 9800X3D does the same. It is not a halo configuration unless they go out of their way to 1% bin X3D CCDs too. In which case gamers would again be better served by a cheaper, binned 9800XT3D.
No, only in MT where it won't be any better than the 9950X. In a combination of ST/MT/gaming the 9950X3D matches it too, even exceeding in some benches where stock clock & latency regression hurts slightly. This is why it is not a halo part without special binning.but 7950x3d is definitely higher than 7800x3d in many cases. which means workload does spill to other cores, so there will be difference
And how big are "we"? I mean if you would be able to gather 10k people who agree with you, then maybe AMD would be able to do a short run series where you would pay 1k $ for the part. Or maybe 100k people is needed. But this is beside the point, thing is if they release 1k$ SKU that doesn't have a noticeable lead over 9950x in Cinebench and 9800x3d in games, then techtubers will start another round of "waste of sand" media campaign that will hurt the brand image in eyes of general public, even if "you" know better where the strengths of the part are, I am afraid the public will not agree with you.A smart company sells people products they need. A smarter company sells people products they want. I'm tired of your personal vendetta against the very concept and those of us who want it. Your words are changing no one's mind and are borderline insulting. We know exactly what it is, what it will do, and what it won't do.
That's a good point. The older players, a lot now, have gotten wealthier.Seriously 9950X3D2 would be a big success
There is a big chunk of adult millenials who work in IT and can easily dish out top money for top computer parts
Price increased massively the last years but people still buy.
Nobody would buy any GPU at $4000 10 years ago just for gaming
Also, for apps that need lot of cache for ALL cores, it would be great for DC and other applications.That's a good point. The older players, a lot now, have gotten wealthier.
for zen 6 they really need to make a super superchip dedicated 100% to real performance, max tdp max vcache everything maxed out, without wasting silicon for NPUs and AVX512 and whatnot
like Halo but 100% CPU
hold the indisputable crown even higher than the popular *800X3D models
It's an opportunity to kneecap NVL-S my friend.I would have thought ... that Zen 6 on N2 would be an opportunity to truly compete against ARM solutions (Apple's included) ...
Could you test at different limits in Blender 4.2?
Was never my CPU. Sorry.Could you test at different limits in Blender 4.2?