You DO? Cause lord knows that's not what I'm seeing.
Two weeks. Stop projecting.
I will not repeat myself.
You DO? Cause lord knows that's not what I'm seeing.I will not deny that I have a preference for ARM, but I strive to be factual and open-minded.
Two weeks. Stop projecting.Last time I checked, as per the leakers, Zen5 is said to bring a 40-50% ST performance improvement.
Geekbench 6
Hawk Point: 2700
M3 : 3150
M4 : 3900
Strix Point : 3700-4000
So Zen5 in Strix is going to match M4 ST performance at best, but it will most certainly use significantly more power to do so.
Attributing Apple's success to them being on an advanced node is a ridiculous presumption.
The node is just the foundation. It's not magic. The microarchitecture is where the magic is.
Case in point: Ryzen 7840HS vs Apple M2
I will not repeat myself.Both score about 2600 points in GB6 ST, but the Ryzen uses about 2x the power to do so. That 2x difference isn't coming from the node, and the irony is that Apple doesn't even have a node advantage here (N4 vs N5P). You can't use the newer uarch excuse, because both Zen4 and M2 are 2022 uarches.
Unless Zen5 is bringing like a 2x ST performance improvement or even better- 2x the performance-per-watt (both of which are unlikely), I don't see any issue for Apple. They'll be fine.
I see no issue in comparing Zen5 (Strix Point) and M4 P-core. Both are laptop chips.It is more accurate to compare the M4 P-core to Z5c as both are N3E and aim to offer a strong balance of performance, power and area.
Z5 classic is N4X and is built for pure performance.
The Zen5 hype train has reached truly dangerous velocities. I have been recieving some threatening messages for trying to dispute this (in another platform, that is).Two weeks. Stop projecting.
Good thing there is the Strix Z5c and the Turin Z5c to compare the node diff.I see no issue in comparing Zen5 (Strix Point) and M4 P-core. Both are laptop chips.
There’s nothing wrong with comparing both at least for science. After all both Zen cores and M4 P will end up on laptops.It is more accurate to compare the M4 P-core to Z5c as both are N3E and aim to offer a strong balance of performance, power and area.
Z5 classic is N4X and is built for pure performance.
Yes, I did remember that Strix has both.There’s nothing wrong with comparing both at least for science. After all both Zen cores and M4 P will end up on laptops.
Beyond laptops, hyperscalers use of Arm is another thing to consider. Desktop is another discussion (I wouldn't consider anything but AMD here), but all other markets are seeing a push for Arm. Except for Apple, Arm laptops might prove a complete failure, but I doubt we'll see hyperscalers drop Arm in the medium term.@Mahboi I don't disagree, in tech logically at the current rate (that is Zen 5 keeping all the promises) AMD with x86 should be the clear winner even with all the added competition.
But we are mainly talking about the mainstream laptop market, which is dominated by Intel up to now and AMD even with all its clearly superior tech had a rather hard time getting into. Microsoft (mostly through its Surface brand) is now heavily pushing ARM which will mix that up some more. Intel is bound to be the biggest loser from this, and while I don't expect ARM to endanger x86 it's likely to take a significant share.
And Intel/x86 rose through PC. Everyone in the workstation and server market laughed at them for a dozen of years. And I was one of them. I was proven very wrong, a good lesson in humility some lack.ARM rose through phones.
Not really comparable now is it? ARM isn't getting a massive 10 year higher income vs an x86 world that would be slowly losing its edge. Which was the case for workstations/servers back in the day. x86 just managed to cover everything well enough and had a bigger TAM and income. Today, ARM's market isn't crushingly higher than x86, so they don't have that kind of strong advantage.And Intel/x86 rose through PC. Everyone in the workstation and server market laughed at them for a dozen of years. And I was one of them. I was proven very wrong, a good lesson in humility some lack.
True but Intel is with Lunar Lake. It would be interesting to see M3/M4 vs Lunar.Yes, I did remember that Strix has both.
At the end of the day AMD isn't bothering with a tablet level chip yet, extremely low power operation is a future design goal.
Yep, same design goals, Lunar will lose but not by as much as people expect. At the end of the day to compete in power you need to xtor bloatmaxx like Apple does.True but Intel is with Lunar Lake. It would be interesting to see M3/M4 vs Lunar.
Do you happen to know how many CPU Arm sells in the embedded and hyperscaler markets? I don't.Not really comparable now is it? ARM isn't getting a massive 10 year higher income vs an x86 world that would be slowly losing its edge. Which was the case for workstations/servers back in the day. x86 just managed to cover everything well enough and had a bigger market. Today, ARM's market isn't crushingly higher than x86, so they don't have that kind of strong advantage.
Or Arm could have just disappeared. I don't make any bet that far away, we might be at a turning point... or notBut sure, in 10 years ARM could have broken through every tier of the x86 bunker. It's 10 years away though.
yep Nvidia does this too but let’s not go there. Don’t want to die yet😂Yep, same design goals, Lunar will lose but not by as much as people expect. At the end of the day to compete in power you need to xtor bloatmaxx like Apple does.
Huawei will be artificially limited to China, or at best Asia.Qualcomm vs Nvidia vs Mediatek vs AMD vs Samsung vs Huawei
No clue how many, but it's more a question of margin and volume isn't it?Do you happen to know how many CPU Arm sells in the embedded and hyperscaler markets? I don't.
Of course Arm (as a company) revenue is not high as they don't sell any chip contrary to Intel, AMD or NVIDIA. But they seem to have enough revenues to pump out cores that are good enough for many of their customers.
I think they'll be fine unless their niches get strongly munched by RISC-V, but really no clue, we'll have to see.Or Arm could have just disappeared. I don't make any bet that far away, we might be at a turning point... or not
Different markets for the most part with Mx being fixed to MacOS.Unless Zen5 is bringing like a 2x ST performance improvement or even better- 2x the performance-per-watt (both of which are unlikely), I don't see any issue for Apple. They'll be fine.
Let's wait for BW vs RDNA4 PPA comparisons.yep Nvidia does this too but let’s not go there. Don’t want to die yet😂
Yep,, x86 cores cannot show any weaknesses across the whole shebang of workloads or the other x86 vendor will punish you hard.Different markets for the most part with Mx being fixed to MacOS.
Especially when you account for server/datacenter too.
Yep,, x86 cores cannot show any weaknesses across the whole shebang of workloads or the other x86 vendor will punish you hard.
Apple Si just needs to do things on MacOS and Geekbench and they can call it a win basically.
Again Arm doesn't sell chips. They sell licenses and get royalties. You can't compare their business model and revenues with AMD, Intel or Arm customers.No clue how many, but it's more a question of margin and volume isn't it?
Phones have insane volume, they're way more sold than PCs nowadays (most people can internet enough with a phone), but it's a highly competitive market where margins for chips cannot be too high. We're miles from the top tiers where duopolies are common.
The best for hyperscalers is "cheaper for them". Where cheaper is a good enough chip which TCO is less than competition, and that satisfies their customers. Definitely not a question of x86 vs Arm as ISA; as I wrote above, they don't care about the now obsolete RISC hype.As for server/hyperscaler, honestly, I don't think it's that much. Every time someone mentions ARM server CPUs, it seems to be some IHFH (in house for house, take that semiconductor industry, I can make mad acronyms too), or it's designed for a fairly specific task. Or it's cheaper than x86 for that task. Or all three. I've yet to see ARM server presented as "taken because it was the best for us". It's always "taken because it's cheaper" or "takent because it's not x86".
Yeah, big question mark here in the medium term for R-V.I think they'll be fine unless their niches get strongly munched by RISC-V, but really no clue, we'll have to see.
Might I remind you that Apple take CPU performance very seriously, so much so that they switched vendors/architectures if they felt their platform lacked CPU performance in laptops and desktops for their designs and “prestige”.Yep,, x86 cores cannot show any weaknesses across the whole shebang of workloads or the other x86 vendor will punish you hard.
Apple Si just needs to do things on MacOS and Geekbench and they can call it a win basically.
Don't get semantical, you know I meant the ARM-based SoCs.Again Arm doesn't sell chips. They sell licenses and get royalties. You can't compare their business model and revenues with AMD, Intel or Arm customers.
As I said in another discussion in this thread, that is an irrelevant debate. If you boil down to "it's not a question of ISA", then the only question becomes in house talent. AMD has a ton there, so they still get my vote.The best for hyperscalers is "cheaper for them". Where cheaper is a good enough chip which TCO is less than competition, and that satisfies their customers. Definitely not a question of x86 vs Arm as ISA; as I wrote above, they don't care about the now obsolete RISC hype.
It'll be interesting to see where it grows or stumbles.Yeah, big question mark here in the medium term for R-V.
Don't get semantical, you know I meant the ARM-based SoCs.
As I said in another discussion in this thread, that is an irrelevant debate. If you boil down to "it's not a question of ISA", then the only question becomes in house talent. AMD has a ton there, so they still get my vote.
And up to now, "cheaper for them" has simply not been anything but x86 yet. And I find no reason to think that the ARM world, who comes from smaller, lesser chips, is going to bridge the gap and leap above the x86 vendors.
Well, I say vendors, I mean AMD. Intel is dying due to its own problems.
ARM most assuredly has been cheaper for hyperscalers. Hence Graviton, Cobalt, and Axion. Google said outright that they've been running a bunch of workloads on ARM servers for a while - do you really think they'd be doing that if it weren't cheaper?
Memes, all of them.
Graviton is the poster child for these chips: done in-house for the dozen or so services that AWS could either put on a Xeon/EPYC for full price, or design at home, buy at TSMC, and get 75% of an EPYC/Xeon's perf for 50% of the price. When you control the entire software/HW stack that the chip is meant to serve, it's just a smart call. None of those sell as high performance generic chips in the server market, except Ampere. The rest is all in-house-for-house, or a literal "we taped this awesome Grace CPU to your GPU and you can't buy the GPU without the CPU bruh".