The focus is that the GFX IP is much much better than anything gfx10.something.
The focus is that the GFX IP is much much better than anything gfx10.something.Focus will be on more memory, faster RT and better CPU.
This does not really answer my question.Because L2 is physically not configurable.
PS6 won't have zen6 core
N2 isn't that pricey.A double node shrink onto the most expensive process on the planet
Nope.If true, this can only mean that Intel really does have the goods with 18A
Not even close.Then server will come down to who can do the most work with the bandwidth they can shove through their memory control fabric.
N2 isn't that pricey.
Nope.Sure seems like it.
That one has BSPDN so maybe.A16 is going to be even more so.
Oh no it's not about power at all.It does make sense if you have customers willing to pay for the power savings
Nope.
?You should have more skepticism from your so-called "Leaker" buddies.
By prior standards, it's just a double half-node shrink.A double node shrink onto the most expensive process on the planet?
No, this can mean all kinds of things.If true, this can only mean that Intel really does have the goods with 18A.
Name something pricier.N2 isn't that pricey.
18A.Name something pricier.
"Nope" means nope.Hard to formulate a rebuttal to "nope".
It's not "desktop". Doesn't exist anymore.Even if AMD goes with N3P for desktop Zen 6
Yeah, you know, TSMC is not Samsung or Intel, their stuff actually yields on ramp.Do we really expect the first GAA to clock higher than the most mature FinFET right out the gate (no pun intended)?
The density bump on N2 is close to nothing.I expect N2 to be more dense. I expect N2 to be less leaky, and more power efficient
N2 risk production started last July. Mass production will be starting in H2 2025, not risk. China Times had an article (referenced by wccftech) that seemed to be claiming (automated translation is always a big unknown with such things) that mass production of N2 will begin around the end of April. If so, the first chips would be delivered to customers in Q4. We'll have to see if TSMC makes any statements about recognizing any N2 revenue in this fiscal year in their next quarterly call. The article (or translation) could be wrong though and the first deliveries won't happen until H1 2026.
Apple moved M4 production forward by months to soak up excess production capacity until it was needed for iPhone SoCs, I expect they would do something similar to insure they are consuming all the N2 they've prepaid to get early access to. There is now the additional wild card of internal production of chips for their internal server needs, which could soak up a ton of capacity and isn't dependent on a calendar like iPhone chip production.
Why is this even a discussion point. These chips still support and accelerate AVX-512.There is no engineering effort to gut the FPU, AMD has already designed it to be configurable. You just pick whether you want 256-bit or 512-bit ALUs, and I don't see a world where the die area would be better spent on doubling the SIMD execution width, and not in the GPU.
And again, 256-bit execution width doesn't mean that the FPU is crippled. Zen4 and mobile Zen5 with half-width FPU are still really good! If the data path between L2 and L3 doubles in Zen6, I would bet that half-width Zen6 beats full-width Zen5 on a lot of real FPU-heavy loads.
Not without redesigning the chip.This does not really answer my question.
It is not configurable out of the box, could be very well true. But hey, AMD can design CPUs and Caches and whatnot. They had 512kB L2$ from Zen 1 to Zen 3. If Sony asks for 512kB L2$, AMD could design it. 100%.
AMD wouldn’t want Intel dead, and despite Intel’s current state, they still have significant marketshare.By prior standards, it's just a double half-node shrink.
Also, cost is relative.
What's 50% more cost per mm² vs. N4P if it lets you put 50% more cores and L3 into the same area, with higher clocks and IPC to boot?
No, this can mean all kinds of things.
It could mean AMD feeling they got Intel in the ropes and going in for the kill.
It could mean 18A is a dud and Intel going all-TSMC for the relevant parts, so AMD saw a need to do the same.
OR... it could mean AMD just looked at what's best for their planned products and chose N2 variants because N2 R&D went with less hiccups than N3B/E/P and therefore just hit a good time-to-market spot for AMD, with solid enough cost/benefit ratio.
18A is very real.It takes time for things to be designed, developed, and manufactured, but chips are coming soon. Possibly from other vendors before Intel even.I meant something real.
The whole gimmick of N2 is that density bumps are incremental
It is very real, but it's also the most complex process node on the market.I meant something real.
No lmao, the cost bumps are also incremental.That's what makes it so expensive.
Turin-D CCDDo we even have any AMD chip on N3E? They dont seem to want to be early adopters for cost/yield issues?
As mentioned, Turin Dense Zen 5c is on N3E.Do we even have any AMD chip on N3E? They dont seem to want to be early adopters for cost/yield issues?
Oh noes my friend they're climbing onto N2X.This is why it seems strange to me that AMD would jump over N3E, N3P, N3X and climb onto N2P.
N5X doesn't exist in TSMC's official roadmaps as far as I know, but a perf-oriented "N5HPC" was publicly mentioned years ago, so I guess it's basically the same thing?Oh noes my friend they're climbing onto N2X.
From N4X (and N5X before that) but I digress.
Not kill kill, but "beating Intel across the board for good and taking even more marketshare and margins" kill.AMD wouldn’t want Intel dead, and despite Intel’s current state, they still have significant marketshare.