CPU Wars 2022 - Apple Silicon vs. Intel/AMD

[maflynn]

This has been on my mind for a while. How does Apple follow up its success of the M1 or does it have too? I think their main bread and butter for Macs is the laptop sector, and this falls right into Apple's strengths with its ARM processors. The issue is they could be viewed as falling behind if they roll out an M2 pro/max/ultra that can't compete with the latest offerings from intel and AMD.

Performance wise we're seeing significant gains from Intel and AMD. I could spend hours and hours going through each benchmark, but Cinebench r23 is one that is generally accepted. The Intel 13th gen numbers fall into the unverified category - so take them with a grain of salt. Also these numbers are not written stone. I can run Cinebench 10 times and get slightly different results, so keep that in mind.

 

[Xiao_Xi]

Cinebench R23 is not a good cross-platform benchmark. It is better to use Geekbench 5.

Cinebench uses hand-optimized x86 SIMD code. Instead of actually rewriting that for Arm, Cinebench's Arm port relies on a library which autotranslates every x86 SIMD instruction to an equivalent sequence of NEON SIMD instructions.

https://forums.macrumors.com/threads/cpu-only-apple-m1-2-max-ultra-tsmc-5nm-vs-amd-zen-4-tsmc-5nm-technical-analysis.2354712/post-31559101

 

[leman]

To be honest, I will be surprised if Apple will face string pressure on the mobile sector this hardware iteration. Some leaks of the upcoming Intel CPUs place them very similar to the M1 Max. And while the final processors might end up faster, I don’t expect dramatic changes. Zen4 might end up slightly faster than that based on what we’ve seen, but still unlikely to outmatch even the most conservative projection for M2 Max.

For this year, I almost expect the M2 series to be rather bland incremental upgrades without too many new features or capabilities. Big upgrades will probably come next year. But I very much doubt that any x86 maker will manage to catch up with Apple in the low-power performance any time soon.

 

[TSE]

Apple's strategy is to maximize efficiency. Power/watt has always been a metric that they view as most important.

Intel's strategy is to maximize pure performance. Benchmarks has always been a metric that they view as most important.

AMD's strategy in the last five years or so has been somewhere in-between the two.



Personally, I'll never buy an Intel device ever again unless they really start focusing on power consumption.

 

[maflynn]

Personally, I'll never buy an Intel device ever again unless they really start focusing on power consumption.

That's how I feel about Nvidia.

I think Intel and AMD's paths are converging. I'm really surprised (and a bit disappointed) at how hot Ryzen 9 runs. Its been reported that Raptor lake chips will not run hotter even though they'll be running faster, so intel made some progress.

I'm not doubting Apple, and competition is good for the consumer, but its clear that both Intel and AMD are pushing hard at making their CPUs faster, regardless of power consumption or heat.

For Apple, the M series really shines in the laptop

 

[pshufd]

That's how I feel about Nvidia.

I think Intel and AMD's paths are converging. I'm really surprised (and a bit disappointed) at how hot Ryzen 9 runs. Its been reported that Raptor lake chips will not run hotter even though they'll be running faster, so intel made some progress.

I'm not doubting Apple, and competition is good for the consumer, but its clear that both Intel and AMD are pushing hard at making their CPUs faster, regardless of power consumption or heat.

For Apple, the M series really shines in the laptop

The vast majority do not require a 7700 or 7750 or a 12900 or 13900. Most people are fine with an M1 or M2 class CPU. That you can get great performance from your CPU may get people to look at your product line in the first place but most will buy something far less capable. So Apple doesn't need to be the absolute fastest and they won't be because they actually care about efficiency.

I think that we need to reassess when the mobile parts come out from Intel and AMD as they may make decisions that are more efficient than their desktop parts.

I don't really care that Apple's Ultra isn't as fast as Intel or AMD's stuff. I'm never going to own an Ultra.

 

[TSE]

That's how I feel about Nvidia.

I think Intel and AMD's paths are converging. I'm really surprised (and a bit disappointed) at how hot Ryzen 9 runs. Its been reported that Raptor lake chips will not run hotter even though they'll be running faster, so intel made some progress.

I'm not doubting Apple, and competition is good for the consumer, but its clear that both Intel and AMD are pushing hard at making their CPUs faster, regardless of power consumption or heat.

For Apple, the M series really shines in the laptop

I agree with everything you just said. I really like AMD's approach - sometimes they lean more towards the performance like they seem to be doing with the 7000 series, other times they are maximizing the efficiency. They are doing a solid job.

The last Intel laptop I had, the XPS 15 9570 with it's Intel 8750H CPU, benchmarked great but throttled immensely after about 10 minutes. This wasn't the typical expected laptop throttling. I'm talking slowing down to a crawl. And this was a widespread issue with that generation of XPS 15, so in real world use it really killed any performance advantage that the H series had at the time. I had to undervolt and reapply thermal paste on that darn thing thing to get somewhat ok performance, but then Intel released a BIOS update along with Dell a couple years later that permanently disabled all undervolting. That was my final straw with Intel, and Dell for that matter.

For desktops, like you mentioned, efficiency doesn't quite have the real world productivity increase like it does on laptops, but I still like what Apple's doing with their desktops.

 

[ArkSingularity]

In the benchmark space, competition is going to be heating up. Intel just released Raptor Lake and it has turbo clocks up to 5.8ghz, and the overclockers haven't even gotten started yet. Intel has terrible IPC compared to Apple processors, but they have clocked them so high that it doesn't really matter, they're winning on the benchmarks.

The issue is that you cannot take one of these processors and throw them into a laptop. Apple learned that lesson when they tried to put an i9 in a MacBook Pro and it couldn't even maintain its base clock without throttling. That was before Intel started going crazy on clock speeds, the problem is likely even worse now.

Intel will shine on the desktop with this strategy, but they've sort of forfeited any sort of mobile dominance by relying on ludicrous clocks to compete. This was part of what convinced Apple to migrate away from Intel to begin with, so I don't think they will have much trouble remaining competitive.

 

[Kazgarth]

I think Intel and AMD's paths are converging. I'm really surprised (and a bit disappointed) at how hot Ryzen 9 runs. Its been reported that Raptor lake chips will not run hotter even though they'll be running faster, so intel made some progress.

I'm not doubting Apple, and competition is good for the consumer, but its clear that both Intel and AMD are pushing hard at making their CPUs faster, regardless of power consumption or heat.

AMD is being forced to choose that path.

Intel, due to lack of any governmental guidelines or consumer awareness, has been doing nothing but feeding more wattage into their ancient 10nm process node. Instead of improving performance with newer more power efficient node and architectural improvements. Their latest 12900KS uses 300W at full load.

AMD's Zen 4 is insanely power efficient if they didn't have to push it to the limit due to the lazy competitor.
For example Ryzen 9 7950X achieves 80% of its peak performance at almost 1/3 of the power (65W vs 170W).

 

[pshufd]

AMD is being forced to choose that path.

Intel, due to lack of any governmental guidelines or consumer awareness, has been doing nothing but feeding more wattage into their ancient 10nm process node. Instead of improving performance with newer more power efficient node and architectural improvements. Their latest 12900KS uses 300W at full load.

AMD's Zen 4 is insanely power efficient if they didn't have to push it to the limit due to the lazy competitor.
For example Ryzen 9 7950X achieves 80% of its peak performance at almost 1/3rd of the power (65W vs 170W).

View attachment 2083774

California has done something about it at the OEM level but I don't think that they are going to do this at the enthusiast level. I'd love to see Intel and AMD sales breakdowns by processor. I think that the 6 and 8 core models way outsell the 10, 12, 16 core models. The 6 core models may even do better with newer generations because the cores are stronger.

In gaming reviews, I've seen recommendations for 6-core 5xxx Zen 3 models over Zen 4 models, especially if you can reuse motherboard and RAM.

 

[ArkSingularity]

AMD is being forced to choose that path.

Intel, due to lack of any governmental guidelines or consumer awareness, has been doing nothing but feeding more wattage into their ancient 10nm process node. Instead of improving performance with newer more power efficient node and architectural improvements. Their latest 12900KS uses 300W at full load.

AMD's Zen 4 is insanely power efficient if they didn't have to push it to the limit due to the lazy competitor.
For example Ryzen 9 7950X achieves 80% of its peak performance at almost 1/3 of the power (65W vs 170W).

View attachment 2083774

The advantage of AMD's approach is that they will probably be better poised to take advantage of next-generation fabs when they are ready. Intel fumbled badly when 10nm/Intel 7 was first released, the old 14nm fab could reach 20-25% higher clock speeds and it basically negated the advantages of ICE lake's higher IPC. They obviously figured out how to tweak 10nm and have gotten higher clock speeds since, but that took years.

Intel might be making the same mistake again if "Intel 4" isn't able to reach similarly high clock speeds right off the bat. They've set the bar ridiculously high for their fabs by relying on these kinds of clock speeds.

 

[mi7chy]

The real war is to be on the latest TSMC node but CPU is no longer the primary processor and just a slave for GPU compute that does most of the heavy lifting for modern workloads like Stable Diffusion, Blender, etc. More interested in Nvidia RTX4000 vs AMD RX7000 war.

 

[leman]

I think Intel and AMD's paths are converging. I'm really surprised (and a bit disappointed) at how hot Ryzen 9 runs. Its been reported that Raptor lake chips will not run hotter even though they'll be running faster, so intel made some progress.

That's the unfortunate consequence of the market reality. The users made it very clear that they prefer bigger numbers over saner numbers and that manipulative marketing and reporting benchmarks at maximal (not maintainable in reality) boost clocks works. Intel was probably the company who pioneered this strategy (as they were stuck with the same architecture and node for years and had to get creative with their marketing) and Nvidia is now copying it. AMD is forced to up the power consumption of their CPUs as well to stay competitive and alas they are also cheating (for example their TDP does not reflect the real power consumption of the CPU — you need to look at the "max socket power" instead).

Apple is probably the only company who can resist going down this path, simply because of their size and influence. They can do their own thing, and having the most power-efficient tech in the world doesn't harm either. But as already discussed in the other thread, the power draw inflation has reach such ridiculous figures on the desktop that Apple will also have to review their TDPs if they want to compete at the high-end.

 

[pshufd]

That's the unfortunate consequence of the market reality. The users made it very clear that they prefer bigger numbers over saner numbers and that manipulative marketing and reporting benchmarks at maximal (not maintainable in reality) boost clocks works. Intel was probably the company who pioneered this strategy (as they were stuck with the same architecture and node for years and had to get creative with their marketing) and Nvidia is now copying it. AMD is forced to up the power consumption of their CPUs as well to stay competitive and alas they are also cheating (for example their TDP does not reflect the real power consumption of the CPU — you need to look at the "max socket power" instead).

Apple is probably the only company who can resist going down this path, simply because of their size and influence. They can do their own thing, and having the most power-efficient tech in the world doesn't harm either. But as already discussed in the other thread, the power draw inflation has reach such ridiculous figures on the desktop that Apple will also have to review their TDPs if they want to compete at the high-end.

One hopeful thing that I see to offset this marketing is the review sites. The ones that I follow also talk about value and efficiency and they tell you that you'd probably be better off with a 6-core if you want a CPU for gaming. They suggest the higher core counts if you are doing production. And they talk about the total cost of upgrading, not just the CPU price. The PC review sites have huge followings too.

 

[russell_314]

For the most part only geeky or nerdy people care about benchmarks. No one I personally know that’s not some kind of gamer nerd is able to tell you what the benchmark scores are for their computer.

What people look at is how fast it does things and how long does the battery last. That is for more professional people. Unfortunately, so many regular consumers just go by how it looks and maybe what their friends have. Oh you have a Lenovo, I think I will get one of those.

If Apple can keep this up, it’s not going to matter what Intel can do with more power because they won’t be able to achieve the battery life, and if you unplug the laptop it slows it down.

 

[ArkSingularity]

If Apple can keep this up, it’s not going to matter what Intel can do with more power because they won’t be able to achieve the battery life, and if you unplug the laptop it slows it down.

This is probably one of the biggest selling points for Apple Silicon for me. Of course they did a good job preventing Intel Macs from throttling, but most other Intel-based laptops (especially the higher end ones) operate on significantly lower TDPs on battery power.

 

[maflynn]

Intel will shine on the desktop with this strategy,

I'm not so sure, I'm an Intel fan, but its hard to pick a 12th Gen processor over AMD or Apple. My desktop PC rocks a 11th Gen i7 11700k, and I worked hard at keeping the temps down in the low 40s on idle and 60s while active. I really don't want a processor that is sitting at 80c while idle, and both Intel and AMD now have processors that seem to live in the 90c neighborhood.

Jayztwocents (an Enthusiast youtuber) was saying its the new normal, I really hope not.

For the most part only geeky or nerdy people care about benchmarks. No one I personally know that’s not some kind of gamer nerd is able to tell you what the benchmark scores are for their computer.

Yes and no, while most consumers never heard of cinebench or care geekbench scores, reading reviews they'll be seeing what machine will be the fastest. its no mistake that Apple has seen a huge uptick in sales with its shift to ARM. That wasn't because people hated Intel, it was because the MBP was marketed as being fast with fantast battery life. So performance is a metric that will continue to drive marketing and sales.

 

[russell_314]

This is probably one of the biggest selling points for Apple Silicon for me. Of course they did a good job preventing Intel Macs from throttling, but most other Intel-based laptops (especially the higher end ones) operate on significantly lower TDPs on battery power.

What sold it for me is responsiveness like when I click something how fast it opens and battery life. I can go somewhere for four hours straight doing work and don’t even have to think about plugging it in. I don’t have to remember to bring my charger.

 

[russell_314]

I'm not so sure, I'm an Intel fan, but its hard to pick a 12th Gen processor over AMD or Apple.

I think it depends on what you’re doing. If you’re doing productivity work, then it’s going to matter if the software is compatible with what you need. If you’re gaming there’s not even a question here. This is why I have an Intel desktop PC. I would like to see Apple get into that market, but I don’t think they will.

 

[maflynn]

Apple is probably the only company who can resist going down this path, simply because of their size and influence.

If they start falling behind in performance (I'm not saying they will) then they will be hard pressed to avoid that. I'm sure we all remember the days of the G3 and G4, where powerpc processors started off faster then Intel, but Intel overtook and surpassed them. Suddenly Macs were viewed as much slower (and more expensive) then PCs. That certainly didn't help their sales. There were other issues with PPC platform, but performance and keeping up with Intel was one problem

 

[ArkSingularity]

I'm not so sure, I'm an Intel fan, but its hard to pick a 12th Gen processor over AMD or Apple. My desktop PC rocks a 11th Gen i7 11700k, and I worked hard at keeping the temps down in the low 40s on idle and 60s while active. I really don't want a processor that is sitting at 80c while idle, and both Intel and AMD now have processors that seem to live in the 90c neighborhood.

Jayztwocents (an Enthusiast youtuber) was saying its the new normal, I really hope not.

Apple has a habit of running the Intel Macs close to about 100C before kicking the fans at full blast. Thermal Junction (the temp at which the processor is supposed to fry itself) is typically around 105-110C, and Apple gets dangerously close. It seems to work out for them somehow, I haven't heard of very many processor failures on them. Some of the M2 Macs run even hotter at 108C before throttling.

I was always a bit surprised that Apple took this approach though. The processors physically use more power when they get this hot (the internal resistance changes and they literally consume more power to perform the same workloads when operating at these kinds of temps). I assume Apple knows what they are doing, but in my limited understanding, it still doesn't make a whole lot of sense from an efficiency standpoint.

 

[leman]

If they start falling behind in performance (I'm not saying they will) then they will be hard pressed to avoid that. I'm sure we all remember the days of the G3 and G4, where powerpc processors started off faster then Intel, but Intel overtook and surpassed them. Suddenly Macs were viewed as much slower (and more expensive) then PCs. That certainly didn't help their sales. There were other issues with PPC platform, but performance and keeping up with Intel was one problem

They might fall behind in desktop performance, but at the same time they will be the only company delivering fast ultracompact desktops. As to mobile.. we already have the rather silly situation where Apple's passively cooled laptop is within 5-10% single core performance compared to AMD's latest and greatest. So I don't really see Apple falling behind here any time soon. As I wrote before, I am not 100% certain whether the upcoming prosumer Mac laptops will be much faster than the x86 offerings, but they won't be any slower while consuming significantly less power. Basically, same situation as we saw with M1.

 

[russell_314]

I always wonder how this works with TSMC making chips for Apple and competitors like AMD. It just seems like some of that engineering would leak over into their competition’s chips.

 

[leman]

Apple has a habit of running the Intel Macs close to about 100C before kicking the fans at full blast. Thermal Junction (the temp at which the processor is supposed to fry itself) is typically around 105-110C, and Apple gets dangerously close. It seems to work out for them somehow, I haven't heard of very many processor failures on them. Some of the M2 Macs run even hotter at 108C before throttling.

The CPU is far from frying itself at thermal junction. Tjunction is not the "oh noes, alarm!" temperature, it's the maximal temperature at which operation is considered perfectly safe. It's just that experimental research shows that 110-130C is the point where the semiconductor degradation really starts to accelerate. There is a reason why Tjuction for most modern devices is 105C or close to it. We should really forget this "high temperature is bad" notion, it's not that it is necessarily wrong, but it's entirely useless for all the practical purposes. Yes, semiconductors will degrade quicker at higher temperatures, but it doesn't matter unless you intend to keep your CPU for multiple decades.

I was always a bit surprised that Apple took this approach though. The processors physically use more power when they get this hot (the internal resistance changes and they literally consume more power to perform the same workloads when operating at these kinds of temps). I assume Apple knows what they are doing, but in my limited understanding, it still doesn't make a whole lot of sense from an efficiency standpoint.

From my similarly limited understanding it's the approach that makes the most sense. Operating the CPU at its maximal safe temperature means that you are running at the maximal safe power and thus are getting the best possible performance. If you try to keep the CPU under Tjunction it just means that you either need to limit the power (and thus performance) or engineer a much larger cooling solution, which again means that you are probably limiting your performance (with the larger cooling solution you could give the CPU more juice and let it run at Tjuctnion). The way to Apple does it is quite beautiful really. Their systems are precision-engineered to hit Tjuction exactly at the target TDP, which shows how meticulous the different parts are fitted together. There are of course other benefits — quieter operation on average (delay the fan ramp up and let the heatsink absorb the activity bursts), marginally lower consumption of cooling fans, more compact cooling system overall etc.

And finally, let's not forget the often overlooked difference between heat and temperature. Intel CPUs dissipate a lot of power, which makes them difficult to cool. Apple Silicon dissipates much less power (on average 2-3 times less to hit the same performance levels) which makes it easier to cool. Running the M2 at 100C is not the same as running an Intel i9 at 100C — the former will still only dissipate around 15-20W. Which means less heat transferred to components that are actually sensitive to higher temperatures, like the battery.

 

[leman]

I always wonder how this works with TSMC making chips for Apple and competitors like AMD. It just seems like some of that engineering would leak over into their competition’s chips.

It surely does, but there is engineering and there is engineering. Apple's outrageous performance/watt is not just the product of them using more advanced processing nodes, it's also the design of the CPU itself.