Question x86 and ARM architectures comparison thread.

[DavidC1]

well I would’ve said Nvidia N1X but Nvidia isn’t capable enough to ship on time.

So yeah you are right, other than Apple there is no competent GPU on ARM.

Gamers are an extremely finicky bunch and AMD/Nvidia managed to keep up with them. It's an extraordinarily hard barrier to penetrate, because it's the decades of hard work, often hand-tuned to get to that point.

 

[Joe NYC]

I wish they would, because the x86 world is just sad.

If you identify one area, as AMD did, in MT performance, mostly by adding more cores, an find an audience willing to pay big bucks for it in hyperscalers, it is only natural that AMD pushed against this wall and milking it to the max.

But now, that approach has run its course, we may have reached the plateau on the server core counts, a new approach is needed.

And my impression is that AMD is already pivoting to optimize the cores for top ST performance and efficiency.

Zen 6 may be mostly using blunt tools of jumping to highest performing node and increasing fmax.

Zen 7 probably doing full reset of the core, growing the core size for greater ST performance instead of growing the number of cores. And now we see also adding accelerators.

On the server side, moving L3 off die to allow for bigger cores, on client, going up to 100mm2 for 16 fore from ~65mm2 8 core, after 4 process node jumps.

Using some projections, the same 8 core Zen 5 chip on A14 node would be 31 mm2, 16 core ~60 mm2 and the die size is projected to be ~100 mm2. L2 and ACE are, by themselves, not going to eat 40% of die size. The cores are likely to grow.

Using AMD internal projections (from MLID leaks), AMD had only modest IPC increase projection for Zen 6, but a lot bigger jumps for Zen 7.

But it seems the expectation by the general public that ARM ISA has insurmountable advantage of x86 is slowly and steadily becoming true, even though if everything was ISO with all teams executing, I don't think the differences are anywhere that close.

Exactly, the lead shrinks considerably once several variables that go into the final performance are normalized.

And the leaks show that AMD has full intention to reach level that is on par with Apple by Zen 7.

But we will get an early preview with Zen 6 being on the same node and having LP cores. Those two components alone will make the gap to no longer be viewed as mythical and insurmountable.

But Zen 6 will be only the preview and Zen 7 will be the real thing.

How many iterations of Zen did it take the execution machine Lisa / Papermaser to finally beat Intel for good in datacenter? You can argue 3 to 5.

How many iterations will it take them to do the same to Apple in client? I think 2 to 3.

 

[poke01]

is Medusa halo going to reach 6.5GHz?

I doubt it. Considering AMD limits the current halo by 600MHz vs the 9950X.

That’s the problem with x86 mobile design since the core isn’t wide they use clocks to reach peak ST performance rather than making the core wider.

 

[mikegg]

When I posted in Zen 7 thread that it seemed to me that AMD seems determined to take performance crown from Apple, it was an acknowledgement that Apple has the performance crown in many tasks / benches.

You followed that with "AMD can't because of" a number of exaggerated claims, such as having to make up 720% efficiency gap (that only exists in your head).

The real distance, the real lead of M series is nowhere near that. Trying to break down the components of the real lead does not equal "excuses". It is exploration of:
- contribution from process node
- contribution from instruction sets
- contribution from accelerators
- contribution from the core itself
- contribution from server vs. client orientation of the core

I didn’t make it up. It’s what Cinebench 2024 shows for M4 Pro and Strix Halo via power from wall measurement. Interpret the data as you will.

 

[AMDK11]

I didn’t make it up. It’s what Cinebench 2024 shows for M4 Pro and Strix Halo via power from wall measurement. Interpret the data as you will.

Provide a link to the CB 24 test where M4 has a 720% higher score in ST.

 

[511]

The only way to compare power correctly is plugging equipment to the CPU Power rail

 

[johnsonwax]

Provide a link to the CB 24 test where M4 has a 720% higher score in ST.

He didn't say 720% faster, he said more efficient.

 

[NTMBK]

Sidenote: Bulldozer actually aged well and closed the gap to sandy bridge. Too bad it lasted for 10 years without improving.

It really, really didn't. It's a bad architecture with a bunch of glaring flaws.

 

[Doug S]

And the leaks show that AMD has full intention to reach level that is on par with Apple by Zen 7.

How many iterations will it take them to do the same to Apple in client? I think 2 to 3.

And Apple is just going to stand still? Apple passed by Intel & AMD without them even noticing or caring, because Apple doesn't compete with them (especially true for AMD who never benefited when Apple used x86 chips for the Mac)

So why should AMD care about Apple now? Whether they beat Apple or are behind Apple whether in raw ST or in power/performance, is that going to make any difference to AMD's sales? I would say no, because as I've said many times before people decide between PC and Mac FIRST, then they decide on the configuration/performance/etc. they're willing to pay for. They are not cross shopping Zen 5 and M5, and will not be cross shopping Zen 7 and M7/8.

 

[mvprod123]

It would be naive to assume that Apple will stand still. Johny Srouji has repeatedly stated in interviews that Apple will continue to focus on industry-leading ST performance.

 

[poke01]

So why should AMD care about Apple now? Whether they beat Apple or are behind Apple whether in raw ST or in power/performance, is that going to make any difference to AMD's sales? I would say no, because as I've said many times before people decide between PC and Mac FIRST, then they decide on the configuration/performance/etc. they're willing to pay for. They are not cross shopping Zen 5 and M5, and will not be cross shopping Zen 7 and M7/8.

oh amd cares about apple in the high end laptop space.

 

[jdubs03]

And the leaks show that AMD has full intention to reach level that is on par with Apple by Zen 7.

If you’re talking about single core/thread, this is incredibly unlikely. As of right now, their current best offering is two-three generations behind in raw performance let alone power draw (another generation at least). And by the time Zen 7 is in actual products, the M7 will be released or nearing release (assuming late 2027-early 2028 timeframe). I would be astonished if they could recover effectively five-seven generations to level set on performance, and power draw.

 

[adroc_thurston]

If you’re talking about single core/thread, this is incredibly unlikely. As of right now, their current best offering is two-three generations behind in raw performance let alone power draw (another generation at least). And by the time Zen 7 is in actual products, the M7 will be released or nearing release (assuming late 2027-early 2028 timeframe). I would be astonished if they could recover effectively five-seven generations to level set on performance, and power draw.

What's a 'generation'

 

[LightningDust]

It really, really didn't. It's a bad architecture with a bunch of glaring flaws.

I actually think that, conceptually, there's a lot to like about Dozer.

Doesn't change that the shipping silicon was mostly very bad, and post-hoc attempts to rehabilitate it are insane.

 

[poke01]

I’ll go and say that Zen6 may match M6 in Cinebench 2024 ST.

Zen4 desktop matched M2 Pro/Max in ST on that benchmark. This was on the same node.

 

[adroc_thurston]

I’ll go and say that Zen6 may match M6 in Cinebench 2024 ST

Why would it?
Cinememe24 hates AMD cores for idk reasons.

Zen4 desktop matched M2 Pro/Max in ST on that benchmark. This was on the same node.

Power delta was bigger.

 

[AMDK11]

is Medusa halo going to reach 6.5GHz?

I doubt it. Considering AMD limits the current halo by 600MHz vs the 9950X.

That’s the problem with x86 mobile design since the core isn’t wide they use clocks to reach peak ST performance rather than making the core wider.

Apple M5 P core
16x execution ports
8xALU, 4xAGU(up to 3xLoad/up to 2xStore), 2xFP + 2xFP/Load/Store

AMD Zen4
14x execution ports
4xALU, 1xBranch, 3xAGU(up to 3xLoad/up to 1xStore), 4xFP + 1xLoad/Store +1xLoad

AMD Zen5
16x execution ports
6xALU, 4xAGU(up to 4xLoad/up to 2xStore), 4xFP + 2xFP/Load/Store

RaptorCove
12x execution ports
3xALU/FP, 2xALU, 5xAGU(3xLoad + 2xStore) + 2xStoreData

LionCove-CougarCove
18x execution ports
6xALU, 6xAGU(3xLoad + 3xStore) + 2xStoreData, 4xFP

AMD tries to keep the clock speed high in Zen because a lot of code, including games, averages 1.5-2 IPC.

For this reason, AMD focuses on a powerful BPU, low L1, L2, and L3 cache latencies, and high throughput. Unfortunately, the chiplet design offsets some of these benefits due to latency, which AMD must compensate for with the expensive X3D chip.

 

[Joe NYC]

Provide a link to the CB 24 test where M4 has a 720% higher score in ST.

He said (power) efficiency. That AMD would have to increase power efficiency by 720% from a (bogus) base line.

 

[Doug S]

oh amd cares about apple in the high end laptop space.
View attachment 134095

Eh, mentioning them in the occasional slide (so long as they compare their best laptop chip to Apple's midrange) doesn't indicate they care. Their marketers know that Apple's M series Macs receive a lot of positive press, so when they see an opportunity to declare a win they will take it. But they know people aren't sitting around looking at Macbook Pro and a Ryzen AI laptop trying to decide which to buy. Most of them will have "runs Windows" as their first requirement, which tosses Apple out of contention even if it tripled AMD's rendering performance.

 

[poke01]

Why would it?
Cinememe24 hates AMD cores for idk reasons.

Power delta was bigger.

I still expect excellent ST from AMD, if its clocking in at 6.5GHz ish.

Especially in SPEC and Geekbench 5/6. Apple will be at best reach 5GHz for M6. Unless Apple launches a clean sheet design which is very unlikely I expect AMD to have a lead.

 

[Joe NYC]

And Apple is just going to stand still? Apple passed by Intel & AMD without them even noticing or caring, because Apple doesn't compete with them (especially true for AMD who never benefited when Apple used x86 chips for the Mac)

Apple will not be standing still but there are more low hanging fruits for AMD from growing the core size.

So why should AMD care about Apple now? Whether they beat Apple or are behind Apple whether in raw ST or in power/performance, is that going to make any difference to AMD's sales? I would say no, because as I've said many times before people decide between PC and Mac FIRST, then they decide on the configuration/performance/etc. they're willing to pay for. They are not cross shopping Zen 5 and M5, and will not be cross shopping Zen 7 and M7/8.

It's not a direct competition, and when AMD market share was very low, any gains to AMD came at expense of Intel. But if AMD does get to 40% of the market, the competition will become more direct.

 

[Joe NYC]

Eh, mentioning them in the occasional slide (so long as they compare their best laptop chip to Apple's midrange) doesn't indicate they care. Their marketers know that Apple's M series Macs receive a lot of positive press, so when they see an opportunity to declare a win they will take it. But they know people aren't sitting around looking at Macbook Pro and a Ryzen AI laptop trying to decide which to buy. Most of them will have "runs Windows" as their first requirement, which tosses Apple out of contention even if it tripled AMD's rendering performance.

If you look at leaked AMD roadmap for Zen 6, there are several premium SKUs.

That's another "tell" that AMD is going to be aiming for same market currently served by MacBook Pro.

 

[adroc_thurston]

I still expect excellent ST from AMD, if its clocking in at 6.5GHz ish.

Especially in SPEC and Geekbench 5/6. Apple will be at best reach 5GHz for M6. Unless Apple launches a clean sheet design which is very unlikely I expect AMD to have a lead.

Oh yeah I'm just saying cinememe24 hates AMD cores and Zen6 won't change that.

If you look at leaked AMD roadmap for Zen 6, there are several premium SKUs

Those are premium in a way table scraps are premium.
(they're dGPUs with a CCD or two taped to the side).

 

[Joe NYC]

Those are premium in a way table scraps are premium.
(they're dGPUs with a CCD or two taped to the side).

The comparison was with MacBook Pro, which does not have a high end dGPU that Windows laptops do.

BTW, another reason why I think AMD aims to compete in Windows parallel of MacBook Pro.

The high ASPs of AMD desktop CPUs seem to make Lisa and Jean's eyes light up. I bet they want the same from the laptop market.

 

[johnsonwax]

Eh, mentioning them in the occasional slide (so long as they compare their best laptop chip to Apple's midrange) doesn't indicate they care. Their marketers know that Apple's M series Macs receive a lot of positive press, so when they see an opportunity to declare a win they will take it. But they know people aren't sitting around looking at Macbook Pro and a Ryzen AI laptop trying to decide which to buy. Most of them will have "runs Windows" as their first requirement, which tosses Apple out of contention even if it tripled AMD's rendering performance.

Will note that not competing effectively against Apple is why Microsoft created the Surface line, and why they've now got Qualcomm making chips. I keep noting that Microsoft had a choice - they've contracted with AMD to make Xbox silicon but for their PC silicon they not only went to Qualcomm, they've committed massive resources to commit to ARM Windows parity. Why? Why on earth would they do that if AMD could just tweak the formula and deliver an equivalent x86 product? Either AMD can't do that, or AMD won't do that on the terms that Microsoft wants - cost, etc. But what matters is that Microsoft doesn't have that deal, and does have one with Qualcomm. And AMD may not care about MBPs but I'm pretty sure they do care about X2 Elite and X2 Elite Extreme laptops carrying Microsoft logo on them, and that product only exists because the MBP does because Intel/AMD/Dell/HP couldn't get their sh!t together to make a competing product. That Microsoft logo is something that AMD takes seriously because they can't compete against that.