Apple just has a big budget to buy out TSMC process nodes a generation early, their designs and architectures aren’t actually faster or more power efficient than AMD’s x86 cpus.
Am I blind? I don’t see any information in there to draw any conclusions about power efficiency. The little information that I do see actually seems to imply the apple silicon chip would be more efficient. Help me out please?
24 threads at 2.00 GHz vs. 8 threads at 0.66 GHz with a 40% difference in TDP. The AMD chip may draw more power, but has much higher performance. Simplifying things, it can perform 9x the operations as the Apple silicon for only 1.4x the power draw.
That… is very naive and inaccurate approach. You can’t use frequency and core counts to guesstimate performance even when the chips in question are closely related. They’re utterly useless when it’s two very different chips that don’t even use the same instruction set. But anyway, there are benchmarks in that page and they clearly show that the amd chip is clearly not performing 9x the operations. It is obviously more powerful, though not nearly by that much.
I desperately want something to start competing with apple silicon, believe me, but knowing just how good the apple silicon chips are from first hand experience, forgive me if I am a little bit sceptical about a little writeup that only deals in benchmark results and official specs. I want to read about how it performs in real life scenarios because I also know from experience that benchmark results and official specs alone don’t always give an accurate picture of how the thing performs in real life.
That’s exactly how you guesstimate CPU performance. It obviously won’t be accurate to real life use cases, but you don’t necessarily need benchmarks to get a ballpark comparison of raw performance. The standard comparison is FLOPS, floating point operations per second. Yes different architectures have different instruction sets, but they’re all relatively similar especially for basic arithmetic. It breaks down with more complex computations, but there’s only so many ways to add two numbers together.
Both chips are 20w class cpus, but the AMD cpu is much faster.
Apple CPUs don’t report wattage, so it’s a bit tricky to measure actual power consumption, but I can’t imagine the AMD cpu uses 50% more power under load.
The Apple CPU might score some wins for idle power consumption though, considering the optimizations in MacOS, and the focus on power consumption across the whole system design.
You can definitely get fairly accurate power draw readings from these chips in macOS, even with Apple’s own debugging tools. If anything, it’s harder (or at least more confusing) to get accurate readings for AMD chips (TDP != power draw).
Also, the TDP the manufacturer states in the spec sheet pretty much doesn’t mean anything these days. These chips will be allowed to draw different amounts of power for different durations under different conditions. This is especially true for the AMD parts, as they run in a lot of different laptops with different power and cooling capabilities. But even for Apple’s M chips there are different configurations: a MacBook Air only has passive cooling while the same chip in a MacBook Pro can have active cooling, which will impact maximum allowed (sustained) power draw and with that, performance.
You also link to CPU Monkey, a website I wouldn’t use for anything but very rough estimates, because their seemingly random collection of benchmarks are likely just taken/stolen from somewhere else (I doubt they benchmarked every single CPU they list themselves) and it’s unclear with what power limits and thermal constraints these benchmarks were run.
Even with all the data, it’s still hard to make a 100 % accurate comparison. For example, the efficiency curves of these CPUs is likely quite a bit different. The M3 might achieve its highest performance/watt at 12 watts, while the Ryzen’s best performance/watt might be at 15 watts (these numbers are just an example). So, do you compare at 12 or 15 watts then?
And yes, there absolutely can be situations where the AMD CPU draws 50% or even 100% (or more) more power under load, and depending on the configuration of the chip in a specific system, the opposite can be the case as well. This in itself doesn’t tell you much about potential power efficiency though.
EDIT: Also, comparing the Ryzen 9 part with 12 cores to the smallest M2 doesn’t make any sense. You’d much more likely compare it to the M2 Max which has 12 cores as well (and again, trying to match the TDP in the spec sheet doesn’t make any sense, as especially for AMD, TDP isn’t even close to actual power draw under load - PPT is at least a somewhat better number here).
I also get that you’re trying to match the process node as closely as possible and TSMC N4 is “just” an improed variant of TSMC N5P, but it still differs. Also, the M2 was released two years earlier than AMD’s AI 300 series, so you ignore two years of architecture improvements which happen regardless of the process node, just look at the (supposed) performance and efficiency improvements from desktop Zen 4 to Zen 5 on the same.
Maybe the new AMD chips are better in many ways even compared to more recent Apple chips, but the comparison you are trying to make is so deeply flawed on so many levels that it’s completely useless and it doesn’t prove anything whatsoever.
The difference between N5P and N4 isn’t significant compared to architectural differences, and the fact that Apple’s architecture is inferior is exactly my point. If the AMD 370 and the Apple M3 are neck and neck, despite Apple being an entire process node ahead (5nm vs 3nm), that shows that Apple’s architecture is inferior.
I don’t think it’s a fair comparison to compare the 27w 370 to the ~50w M2 Pro.
It’s true that power efficiency is such a hard metric to compare, especially on laptops and across different operating systems, but that’s the point I’m making with the rough figures we have available.
Iirc the die area for Apple’s chips are also a lot larger and that’s expensive. It’s a lot easier for them to tank that cost because they are building them for themselves rather than selling them to vendors who manufacture products like AMD.
Where you see vertical integration, I see unnecessary and customer antagonistic siloing of function. Do you have any idea how impossible it is to send an apple user money from a non apple device?
What does unnecessary and customer antagonistic siloing of function have to do with Apple’s vertical integration of manufacturing process? One doesn’t prevent the existence of the other within the same company.
I believe both M2 and Zen 5 use 4nm. 4nm is just a slightly improved 5nm, though. It’s the same process node, not an entirely new process node like 3nm.
Don’t trust any silicon manufacturer’s marketing department. Let the processing and battery life benchmarks and real world tests do the talking.
AMD’s CPUs are faster and more power efficient on the same process node. (i.e. 5nm vs 5nm)
https://www.cpu-monkey.com/en/compare_cpu-amd_ryzen_ai_9_hx_370-vs-apple_m2
Apple just has a big budget to buy out TSMC process nodes a generation early, their designs and architectures aren’t actually faster or more power efficient than AMD’s x86 cpus.
https://www.macrumors.com/2023/02/22/apple-secures-tsmc-3nm-chips/
Am I blind? I don’t see any information in there to draw any conclusions about power efficiency. The little information that I do see actually seems to imply the apple silicon chip would be more efficient. Help me out please?
24 threads at 2.00 GHz vs. 8 threads at 0.66 GHz with a 40% difference in TDP. The AMD chip may draw more power, but has much higher performance. Simplifying things, it can perform 9x the operations as the Apple silicon for only 1.4x the power draw.
That… is very naive and inaccurate approach. You can’t use frequency and core counts to guesstimate performance even when the chips in question are closely related. They’re utterly useless when it’s two very different chips that don’t even use the same instruction set. But anyway, there are benchmarks in that page and they clearly show that the amd chip is clearly not performing 9x the operations. It is obviously more powerful, though not nearly by that much.
I desperately want something to start competing with apple silicon, believe me, but knowing just how good the apple silicon chips are from first hand experience, forgive me if I am a little bit sceptical about a little writeup that only deals in benchmark results and official specs. I want to read about how it performs in real life scenarios because I also know from experience that benchmark results and official specs alone don’t always give an accurate picture of how the thing performs in real life.
That’s exactly how you guesstimate CPU performance. It obviously won’t be accurate to real life use cases, but you don’t necessarily need benchmarks to get a ballpark comparison of raw performance. The standard comparison is FLOPS, floating point operations per second. Yes different architectures have different instruction sets, but they’re all relatively similar especially for basic arithmetic. It breaks down with more complex computations, but there’s only so many ways to add two numbers together.
Both chips are 20w class cpus, but the AMD cpu is much faster.
Apple CPUs don’t report wattage, so it’s a bit tricky to measure actual power consumption, but I can’t imagine the AMD cpu uses 50% more power under load.
The Apple CPU might score some wins for idle power consumption though, considering the optimizations in MacOS, and the focus on power consumption across the whole system design.
You can definitely get fairly accurate power draw readings from these chips in macOS, even with Apple’s own debugging tools. If anything, it’s harder (or at least more confusing) to get accurate readings for AMD chips (TDP != power draw).
Also, the TDP the manufacturer states in the spec sheet pretty much doesn’t mean anything these days. These chips will be allowed to draw different amounts of power for different durations under different conditions. This is especially true for the AMD parts, as they run in a lot of different laptops with different power and cooling capabilities. But even for Apple’s M chips there are different configurations: a MacBook Air only has passive cooling while the same chip in a MacBook Pro can have active cooling, which will impact maximum allowed (sustained) power draw and with that, performance.
You also link to CPU Monkey, a website I wouldn’t use for anything but very rough estimates, because their seemingly random collection of benchmarks are likely just taken/stolen from somewhere else (I doubt they benchmarked every single CPU they list themselves) and it’s unclear with what power limits and thermal constraints these benchmarks were run.
Even with all the data, it’s still hard to make a 100 % accurate comparison. For example, the efficiency curves of these CPUs is likely quite a bit different. The M3 might achieve its highest performance/watt at 12 watts, while the Ryzen’s best performance/watt might be at 15 watts (these numbers are just an example). So, do you compare at 12 or 15 watts then?
And yes, there absolutely can be situations where the AMD CPU draws 50% or even 100% (or more) more power under load, and depending on the configuration of the chip in a specific system, the opposite can be the case as well. This in itself doesn’t tell you much about potential power efficiency though.
EDIT: Also, comparing the Ryzen 9 part with 12 cores to the smallest M2 doesn’t make any sense. You’d much more likely compare it to the M2 Max which has 12 cores as well (and again, trying to match the TDP in the spec sheet doesn’t make any sense, as especially for AMD, TDP isn’t even close to actual power draw under load - PPT is at least a somewhat better number here).
I also get that you’re trying to match the process node as closely as possible and TSMC N4 is “just” an improed variant of TSMC N5P, but it still differs. Also, the M2 was released two years earlier than AMD’s AI 300 series, so you ignore two years of architecture improvements which happen regardless of the process node, just look at the (supposed) performance and efficiency improvements from desktop Zen 4 to Zen 5 on the same.
Maybe the new AMD chips are better in many ways even compared to more recent Apple chips, but the comparison you are trying to make is so deeply flawed on so many levels that it’s completely useless and it doesn’t prove anything whatsoever.
I’ve taken those facts into account.
I haven’t seen any benchmarks that include power usage for Apple CPUs.
AMD cpus are not like Intel CPUs, they don’t use more than the TDP under load. https://www.phoronix.com/review/amd-ryzen-ai-9-hx-370/3
The difference between N5P and N4 isn’t significant compared to architectural differences, and the fact that Apple’s architecture is inferior is exactly my point. If the AMD 370 and the Apple M3 are neck and neck, despite Apple being an entire process node ahead (5nm vs 3nm), that shows that Apple’s architecture is inferior.
I don’t think it’s a fair comparison to compare the 27w 370 to the ~50w M2 Pro.
It’s true that power efficiency is such a hard metric to compare, especially on laptops and across different operating systems, but that’s the point I’m making with the rough figures we have available.
Iirc the die area for Apple’s chips are also a lot larger and that’s expensive. It’s a lot easier for them to tank that cost because they are building them for themselves rather than selling them to vendors who manufacture products like AMD.
Yeah, Apple’s vertical integration and volume is enviable.
Where you see vertical integration, I see unnecessary and customer antagonistic siloing of function. Do you have any idea how impossible it is to send an apple user money from a non apple device?
What does unnecessary and customer antagonistic siloing of function have to do with Apple’s vertical integration of manufacturing process? One doesn’t prevent the existence of the other within the same company.
Minor nit, but I believe that AMD CPU is 4nm
I believe both M2 and Zen 5 use 4nm. 4nm is just a slightly improved 5nm, though. It’s the same process node, not an entirely new process node like 3nm.
Everything I see says the M2 family is 5. Vanilla, pro, max, and ultra.
The nm process for each CPU is listed in technical details on cpu-monkey
Rumors before the M2 release said that it used 4nm.
https://www.macrumors.com/2022/03/10/m2-macs-with-tsmc-4nm-process/
Apple says they use “second generation 5nm technology”
https://www.apple.com/newsroom/2022/06/apple-unveils-m2-with-breakthrough-performance-and-capabilities/
TSMC’s website says they have 6 different 5nm nodes: N5, N5P, N5A N4, N4P, N4X
https://www.tsmc.com/english/dedicatedFoundry/technology/logic/l_5nm
So the M2 likely uses N5P, N4, or N4P. N4 and N4P are usually called 4nm in marketing material.
There’s probably a leaker out there with more knowledge.
Not trying to start a debate, just saying the specs in the link were different than what was mentioned.
My point is that the M series transition went very well.
Sure, I was just explaining it because the whole 5/4 thing is confusing.