If they iterate upon Keller's design, maybe they have something decent to start with? AMD is in a unique position to have the first heterogeneous ISA CPU in the world. Boot with ARM or x86-64 OS or boot with one of them and use the other through a hypervisor. If they even use their Bobcat core as the x86-64 part of the heterogeneous ISA equation, it would still be a super exciting SoC. Why emulate when you can run Android applications at native speed? Or dual boot between Windows x86-64 and Windows on ARM. Would certainly open up interesting applications.Even with ARM, AMD becomes another generic ARM vendor. I doubt they can be competitive with even top 5 vendors.
It is a daring synthesis but Zen 5 telephone speculation gets views.Make a telephone speculation thread.
This is the Zen 5 speculation thread.
That's more likely, yes.What's next? Zen 5 in smart refrigerators?
Nvidia and Transmeta were close (through unconventional means) but neither shipped anything like that.If they iterate upon Keller's design, maybe they have something decent to start with? AMD is in a unique position to have the first heterogeneous ISA CPU in the world. Boot with ARM or x86-64 OS or boot with one of them and use the other through a hypervisor. If they even use their Bobcat core as the x86-64 part of the heterogeneous ISA equation, it would still be a super exciting SoC. Why emulate when you can run Android applications at native speed? Would certainly open up interesting applications.
Interesting OS details: When Linux uses 4K pages and turns off THP, Zen's TLB coalescing cannot be triggered because the memory mapping granularity is pages, and the TLB coverage is 4K*entry. However, under Windows, since the granularity is 64K, the TLB coverage of the equivalent 16K page can be observed normally. In theory, the hardware can be seamlessly expanded to a larger 64K in the future. This is also one of the few places where I have seen Windows perform better than Linux in recent years.
TLB coalescing refers to this feature, which is basically a 16K page that is transparent to software.
a) laptop volumes are lesserOnly a problem for penny pinchers who don't want to actually sell anything at volume
I don't think you understand the BOM for any of those.Qualcomm, MediaTek, Apple and even Intel are all making N3 designs shipping in cheaper devices than Strix Point
M4 in anything resembling a laptop starts at about $1699 or whatever the baseline 14" MBP config is.M4 is 160mm2 of N3E and is shipping in far cheaper devices than HX370
It's too expensive because you have to keep the costs to OEMs palatable, and no one lived or died by 15% perf 3-2 N3e offers.If N3E was too expensive for AMD you have to ask what's blowing out their transistor counts relative to the competition
It's $1400 at Amazon. Which is cheaper than any HX 370 laptop I can find on Amazon. The cheapest HX 370 mini PC starts at $1000 vs $600 for Mac Mini. I am not even convinced M4 is more expensive to make at all given the die size difference.M4 in anything resembling a laptop starts at about $1699 or whatever the baseline 14" MBP config is.
Vivobook S14 is 1.2kilobuck for a 24G/1TB config. Don't. This is embarrassing.Which is cheaper than any HX 370 laptop I can find on Amazon
Any actual volume NUCs with Strix from the likes of Asus aren't even out.The cheapest HX 370 mini PC starts at $1000 vs $600 for Mac Mini.
It is, and then there's a platform premium.I am not even convinced M4 is more expensive to make at all given the die size difference
Doesn't look like it. There's one (1) laptop HX 370 laptop that's available (where? not Amazon; nor Best Buy apparently) for less than the MBP 14 which I'm sure is better built (inb4 notch and jello screen).It is, and then there's a platform premium.
Next.
And they won't be, too costly.Any actual volume NUCs with Strix from the likes of Asus aren't even out.
So they made a product that is bigger than all the competition. And what is the plan for next year? A refresh with a 10% faster NPU again? How do they expect to gain laptop market share. The entire strategy seems to rely on Intel sucking forever (that might work out).It's too expensive because you have to keep the costs to OEMs palatable, and no one lived or died by 15% perf 3-2 N3e offers.
So true.Doesn't look like it.
Took me the whole of 15 seconds to find the exact SKU.There's one (1) laptop HX 370 laptop that's available (where? not Amazon; nor Best Buy apparently)
It's also plenty mean. Hope that helps.So they made a product that is bigger than all the competition
The plan is called Kraken1.And what is the plan for next year? A refresh with a 10% faster NPU again
Aggressively playing HPT1/2 and KRK1 in the channel. Commercial in particular.How do they expect to gain laptop market share
Well, only Intel understands PC as far as Windows vendors go. So yeah, gotta kick them and you win.The entire strategy seems to rely on Intel sucking forever (that might work out).
Amazon goes to greats length to hide the CPU name, it’s actually the HX 365Took me the whole of 15 seconds to find the exact SKU
That's the 365 version.Took me the whole of 15 seconds to find the exact SKU.
Not compared to the M4. It's more comparable to the most common SDXE SKUs which are smaller and also cheaper. Only thing giving HX 370 any market is the x64 ghetto and a Radeon GPU which 86% of people would not choose if Nvidia was an option.It's also plenty mean. Hope that helps.
Ouch.Amazon goes to greats length to hide the CPU name, it’s actually the HX 365
Ouch. But Apple also sells binned stuff aplenty.That's the 365 version.
Different things.No compared to the M4
SDXE has a joketastic GPU. Next.It's more comparable to the most common SDXE SKUs which are smaller and also cheaper
The ghetto that understands PC, yes.Only thing saving it is the x64 ghetto.
Even with the WoA handicap Qualcomm has more appealing laptops than AMD this year. Maybe that changes next year.The ghetto that understands PC, yes.
No one else besides Apple does.
They do not.Even with the WoA handicap Qualcomm has more appealing laptops than AMD this year
Nothing changes as far as Qualcomm goes.Maybe that changes next year.
It's better because it's an expensive part on an expensive platform made to maximize the expensiveness. All that is gone H2'25 with PTL.LNL has better still but that's because as you say Intel understands OEMs
This is the funniest joke I've ever seen.AMD seems to lack OEM muscle even compared to newcomer Qualcomm.
Sure, as long as you compare Strix Point laptops to SDXE laptops it'll be pretty close.Should we start counting the units shipped in % of 2024 PC TAM? gonna hurt.
It's not gonna be even remotely close.Sure, as long as you compare Strix Point laptops to SDXE laptops it'll be pretty close.
"Vertical integration" means they just get to pocket the margin delta that OEM takes. I.e. they get to stuff a more expensive SOC/panel/yaddayadda into a given laptop without hurting the final gm.Apple has vertical integration, so it’s unfair to compare AMD with them.
Why? AMD wants to compete in laptops. I see them as mainly interchangeable and so do many others. There are factors that AMD cannot control (like Windows 11 sucking) but that doesn't stop people from buying an Apple laptop instead of an AMD-powered laptop.Apple has vertical integration, so it’s unfair to compare AMD with them.
It means they have to keep the OEMs from starving by keeping SoC/platform costs at bay. Next.
That's actually true and I forgot. They really were a joke until they tried this year.Also QC isn't a newcomer, they've been doing it for 7 years. It's a case of utter ineptitude at volume building.
They haven't tried this year either.They really were a joke until they tried this year.
The plan used to be the 40% IPC dominance. I'm confused.The plan is called Kraken1.
I live in Romania, and here we have on the greatest e-tailer (link, if you wanna check) 50 models... So yeah.... different. The cheapest is under 1000 euro (1 euro=5 lei). Also, I wouldn't count base MacMini, that's just a hook from Apple to catch the plebs... while the CPU is great and ram is acceptable, fixed 256GB in this day and age is not acceptable. And if you go to a devoce that has 24GB and 1TB of memory, you are over twice the base price and higher than a Beelink SER9, which is a 32GB Ram and 1TB sdd. And looks exactly like a Macmini before the Macmini was this sizeIt's $1400 at Amazon. Which is cheaper than any HX 370 laptop I can find on Amazon. The cheapest HX 370 mini PC starts at $1000 vs $600 for Mac Mini. I am not even convinced M4 is more expensive to make at all given the die size difference.
I mean it’s not just that, it should be obvious to anyone that AMD is selling CPUs to vendors at a profit, who are then turning around and trying to ship a completed system for a profit. Apple cuts out the middleman there so they are automatically getting their CPUs for “cheaper”It means they have to keep the OEMs from starving by keeping SoC/platform costs at bay. Next.
The excuse for the less competitive performance is that it is cheaper to make. This may apply to Granite Ridge but not Strix Point. Its die is so much larger than the competition that it matches or exceeds the rumored 25% to 37% price increase between N4 and N3E.I think it’s fair game for outright performance discussions and architecture but I don’t want to listen to bickering about laptop prices