14 ms·

Intel and Nvidia just dunked on Apple’s M1 Max. Should you believe the hype?

33 points by carlycue 5y ago

n7pdx 5y ago
Looks like Intel's "AC performance" marketing worked on PC World's executive editor.
olliej 5y ago
I’m not sure comparing benchmarks of yet to be released hardware is super useful, especially given intels recent track record. I’m more likely to believe nvidia’s shipping estimates there. I’ll given them both props for not pulling that ludicrous bullshit graph that apple had in the original M1 announcement. An interesting feature of SPEC that strongly disadvantages Intel is the a number of SPEC tests use ‘long double’ which every platform except Intel is always 64bit double running on modern execution units. For Intel (and AMD) the ABI for 32bit, and 64bit on linux and macOS is to use 80 bit floats. That necessarily means x87. So no SIMD, a vastly slower execution engine, and higher precision. There is no case where x87 performance comes close to the primary floating point pipelines.
6sp 5y ago
The M1 Max is pretty awesome and I can’t recall anything from Intel as of late that amazed me all that much. Intel made plenty of incremental improvements but nothing necessarily leaps ahead of their previous generations imo. The M1 and M1 Max were a refreshing change to years of underwhelming releases at crappy price points. For what I paid for my Intel-based MacPro my MacBook Pro M1 Max runs circles around it at a much lower price point. If they really beat the M1 Max my question is at what cost in dollars and battery life.
ksec 5y ago
This certainly fit into certain type of perspective that is common on the internet and HN. Absolute performance regardless of energy usage. Another problem is people blindly trusted Apple's number on GPU. And while Apple's benchmarks in terms of CPU and Battery life tends to be fairly accurate if not often under stating it. Their GPU numbers and graphs are a marketing spin to say the least.
robthebrew 5y ago
Surely it is all about the Watts? OK, I think the M1 iMac is a lame duck, but for laptop use, the M1 is just more useful. Let's see what the M2 (or whatever) delivers in a Mac Pro (which is what pros use anyway).
phyalow 5y ago
As the article alludes, most of the benchmarks for the M1 presented in the article are meaningless as the software is either running virtulised under rossetta (AND) OR not utilising the GPU/Neural Cores via the Metal API (Cuda equivalent). Expect things to even out over the next couple of months as software is patched and recompiled to run under an ARM architecture.
- sagarm 5y ago
  The benchmarks are meaningful for people who need to use that particular software. They also won't extrapolate well to software that is better optimized for ARM.
kahrl 5y ago
No. You shouldn't. This is clickbait. All their benchmarks are meaningless, as they're comparing native x86 using GPGPU to virtualized Rosetta using no GPU acceleration.
- usbqk 5y ago
  People want to run x86 code.
  - zepto 5y ago
    No they don’t. People want to run fast code. Rosetta is just a stopgap.
  - Uehreka 5y ago
    Not me. I have assembly-olfactory synesthesia and x86 code smells like the bad parts of gasoline. Kidding aside, unless you’re looking to play games, most apps have either ported to Apple Silicon, or have a Beta of their port that is releasing soon. I think the M1 Pro/Max release was the kick in the pants that companies/projects like Adobe, Unity, Unreal and Blender needed in order to go “OK yeah, professionals use these, time to get porting”. (I know some of these betas preceded the release of these laptops, but for a few months before these laptops came out it was clear they were going to be nuts)
  - kahrl 5y ago
    That's fine. We absolutely need benchmarks comparing native x86 performance against virtualized performance thru Rosetta 2. A lot of code isn't getting re-compiled for arm anytime soon. All I'm saying is this article is clickbait because it frames these comparisons in the dumbest way possible as if to say 1 processor is better than the other, with no explanation that these are not 1-to-1 comparisons. It is a complete lie by omission.
daniel-thompson 5y ago
Original announcement: https://www.nvidia.com/en-us/geforce/news/geforce-rtx-laptops-3080-ti-3070-ti/ https://www.nvidia.com/en-us/geforce/news/geforce-rtx-laptop... As far as I can tell, the benchmarks are done with software packages^ that support a GPU rendering path for NVIDIA chips via CUDA / OptiX, but not yet with M1 via Metal (or are even CPU-emulated on M1 via Rosetta). So it's "fair" in the sense that this is what the market currently offers for users, but it's not an apples-to-apples comparison on the hardware. ^ e.g. Blender 3.1, currently in alpha, adds a Metal GPU backend to Cycles, which in early reports gives about a 3x - 4x improvement in render times vs the CPU path on M1 machines (https://devtalk.blender.org/t/cycles-apple-metal-device-feedback/21868/8 https://devtalk.blender.org/t/cycles-apple-metal-device-feed...).
- bilalnpe 5y ago
  >but it's not an apples-to-apples comparison No, it's an apples-to-nvidias comparison
  - olliej 5y ago
    Slow clap :D
horsawlarway 5y ago
While I'm not a fan of Apple in most regards, I think their approach to mobile hardware is much more coherent than NVidia and Intel right now. Generally speaking - The CPU just isn't the performance bottleneck on most machines that people are intending to carry with them during the day. The performance bottleneck is the battery and the temperature. I think it's incredibly telling that from this chart (produced by intel, designed to promote intel products) my takeaway is intel has missed the mark entirely. The relative performance of their chips just doesn't matter when you realize that the M1 chips are SO MUCH BETTER at power usage. The ranges for power draw barely even overlap - the M1 is getting you ~100% of the performance, using half the power of the lowest power intel chip. That's the win. Full stop. It means the machine with that CPU is going to run cooler for longer. It means that cpu isn't going to be thermally throttled so you can actually use all of it (unlike that intel). It means you still have a running machine 15 hours later, instead of having a dead brick after just 5. Ethically - I refuse to buy any general purpose machine Apple makes, but I'm sad Intel is so far off the mark here.
- lukevp 5y ago
  What are your ethical issues with Apple? Why mention them in the first place without further explication? Your other points rang true for me but this seems a little tangential.
  - pcdoodle 5y ago
    For me it's the hostility towards developers and users. I can't even lower the dang contrast on my iOS or MacOS devices to make them a little less dopamine inducing. Sure you can switch to B&W but then I lose usability (Is that link clicked?). The massive cut they take from creators is another pain point. These are just off the top of my head, I'm sure others can chime in.
    - zepto 5y ago
      The dopamine inducing element is addictive apps like Facebook, not the color levels of the display. You can simply delete the apps that have that effect.
darthrupert 5y ago
Well, of all the reasons this might be the biggest one to buy Apple. Must feel amazing when your competition does your marketing for you.
- andreime 5y ago
  have you seen the charts apple uses in presentations? they deserve any mockery going their way
vmception 5y ago
Does it really matter? People that spend more than $1,000 on a laptop buy Apple. Edit: Some outliers really want me to write “overwhelmingly most” or find that old market study? Maybe. But I can also spend 4 points so fire away.
- avgDev 5y ago
  False. I like windows and have been using XPS laptops for a long time.
- LegitShady 5y ago
  do you really think this is true? because it isnt.
- acdha 5y ago
  This is a pointless distraction from the article. Professional users pay more for their laptops, no matter what platform, and given that the kind of laptop I need to do my job has gone down in price by at least a factor of two that price matters less than whether it does the job well.
- kahrl 5y ago
  I had no idea my Lenovo Thinkpad X1 was actually made by Apple! TIL!
notriskfree 5y ago
Interestingly it is the base M1 that is still the most interesting device, lower cost, lower heat, good single core performance, capable of most tasks, and reasonable graphics for a lower power device. Other fast laptops are all about the heat and the fan noise (my tiny work laptop is the loudest device in my room) A number of times I have got off the train with a laptop almost on fire in my bag, because windows failed to notice the laptop lid was closed. Ill be getting the M1 laptop next time I need one.
Shadonototra 5y ago
i have an M1 macbook, Intel/NVidia can lie all they want they can't hide facts M1 > to what ever they produce TODAY and tomorrow, it beats the hell out of them and let's not talk about how macOS puts to shame windows 11 thanks journalists for spreading their corporate propaganda, i wouldn't mind if it was to improve everyone's life, but spreading crappy bloatware.. nope this article is on par with the verge "stadia bad but xcloud good" let me laugh for a moment.. lol
qeternity 5y ago
I switched to Apple when they dropped PPC for Intel. I've had just about every x86 Macbook they built, and have built a handful of high-spec hackintosh desktops. When the M1 came out, I ran out and grabbed the cheapest one I could (8gb M1 Air) to play around with. After about 30 minutes, I decided to transfer my daily driver to this little passively cooled notebook. The next day I bought the M1 Mini to replace my overclocked hackintosh. My high-end x86 gear just couldn't compete with the low-end M1s. I don't care what Intel and Nvidia say. I've used all their products. The M1 is black magic.
- faeriechangling 5y ago
  One thing I’ve noticed about a ton of benchmarks is a focus on throughput instead of latency. Apple has had some of the lowest latency computers around and the M1 has only improved the situation with QoS features on the CPU and such.

33 points by carlycue 5y ago

Looks like Intel's "AC performance" marketing worked on PC World's executive editor.

I’m not sure comparing benchmarks of yet to be released hardware is super useful, especially given intels recent track record. I’m more likely to believe nvidia’s shipping estimates there. I’ll given them both props for not pulling that ludicrous bullshit graph that apple had in the original M1 announcement. An interesting feature of SPEC that strongly disadvantages Intel is the a number of SPEC tests use ‘long double’ which every platform except Intel is always 64bit double running on modern execution units. For Intel (and AMD) the ABI for 32bit, and 64bit on linux and macOS is to use 80 bit floats. That necessarily means x87. So no SIMD, a vastly slower execution engine, and higher precision. There is no case where x87 performance comes close to the primary floating point pipelines.

6sp 5y ago

The M1 Max is pretty awesome and I can’t recall anything from Intel as of late that amazed me all that much. Intel made plenty of incremental improvements but nothing necessarily leaps ahead of their previous generations imo. The M1 and M1 Max were a refreshing change to years of underwhelming releases at crappy price points. For what I paid for my Intel-based MacPro my MacBook Pro M1 Max runs circles around it at a much lower price point. If they really beat the M1 Max my question is at what cost in dollars and battery life.

ksec 5y ago

This certainly fit into certain type of perspective that is common on the internet and HN. Absolute performance regardless of energy usage. Another problem is people blindly trusted Apple's number on GPU. And while Apple's benchmarks in terms of CPU and Battery life tends to be fairly accurate if not often under stating it. Their GPU numbers and graphs are a marketing spin to say the least.

robthebrew 5y ago

Surely it is all about the Watts? OK, I think the M1 iMac is a lame duck, but for laptop use, the M1 is just more useful. Let's see what the M2 (or whatever) delivers in a Mac Pro (which is what pros use anyway).

phyalow 5y ago

As the article alludes, most of the benchmarks for the M1 presented in the article are meaningless as the software is either running virtulised under rossetta (AND) OR not utilising the GPU/Neural Cores via the Metal API (Cuda equivalent). Expect things to even out over the next couple of months as software is patched and recompiled to run under an ARM architecture.

sagarm 5y ago
The benchmarks are meaningful for people who need to use that particular software. They also won't extrapolate well to software that is better optimized for ARM.

kahrl 5y ago

No. You shouldn't. This is clickbait. All their benchmarks are meaningless, as they're comparing native x86 using GPGPU to virtualized Rosetta using no GPU acceleration.

usbqk 5y ago
People want to run x86 code.
- zepto 5y ago
  No they don’t. People want to run fast code. Rosetta is just a stopgap.
- Uehreka 5y ago
  Not me. I have assembly-olfactory synesthesia and x86 code smells like the bad parts of gasoline. Kidding aside, unless you’re looking to play games, most apps have either ported to Apple Silicon, or have a Beta of their port that is releasing soon. I think the M1 Pro/Max release was the kick in the pants that companies/projects like Adobe, Unity, Unreal and Blender needed in order to go “OK yeah, professionals use these, time to get porting”. (I know some of these betas preceded the release of these laptops, but for a few months before these laptops came out it was clear they were going to be nuts)
- kahrl 5y ago
  That's fine. We absolutely need benchmarks comparing native x86 performance against virtualized performance thru Rosetta 2. A lot of code isn't getting re-compiled for arm anytime soon. All I'm saying is this article is clickbait because it frames these comparisons in the dumbest way possible as if to say 1 processor is better than the other, with no explanation that these are not 1-to-1 comparisons. It is a complete lie by omission.

daniel-thompson 5y ago

Original announcement: https://www.nvidia.com/en-us/geforce/news/geforce-rtx-laptops-3080-ti-3070-ti/ https://www.nvidia.com/en-us/geforce/news/geforce-rtx-laptop... As far as I can tell, the benchmarks are done with software packages^ that support a GPU rendering path for NVIDIA chips via CUDA / OptiX, but not yet with M1 via Metal (or are even CPU-emulated on M1 via Rosetta). So it's "fair" in the sense that this is what the market currently offers for users, but it's not an apples-to-apples comparison on the hardware. ^ e.g. Blender 3.1, currently in alpha, adds a Metal GPU backend to Cycles, which in early reports gives about a 3x - 4x improvement in render times vs the CPU path on M1 machines (https://devtalk.blender.org/t/cycles-apple-metal-device-feedback/21868/8 https://devtalk.blender.org/t/cycles-apple-metal-device-feed...).

bilalnpe 5y ago
>but it's not an apples-to-apples comparison No, it's an apples-to-nvidias comparison
- olliej 5y ago
  Slow clap :D

horsawlarway 5y ago

While I'm not a fan of Apple in most regards, I think their approach to mobile hardware is much more coherent than NVidia and Intel right now. Generally speaking - The CPU just isn't the performance bottleneck on most machines that people are intending to carry with them during the day. The performance bottleneck is the battery and the temperature. I think it's incredibly telling that from this chart (produced by intel, designed to promote intel products) my takeaway is intel has missed the mark entirely. The relative performance of their chips just doesn't matter when you realize that the M1 chips are SO MUCH BETTER at power usage. The ranges for power draw barely even overlap - the M1 is getting you ~100% of the performance, using half the power of the lowest power intel chip. That's the win. Full stop. It means the machine with that CPU is going to run cooler for longer. It means that cpu isn't going to be thermally throttled so you can actually use all of it (unlike that intel). It means you still have a running machine 15 hours later, instead of having a dead brick after just 5. Ethically - I refuse to buy any general purpose machine Apple makes, but I'm sad Intel is so far off the mark here.

lukevp 5y ago
What are your ethical issues with Apple? Why mention them in the first place without further explication? Your other points rang true for me but this seems a little tangential.
- pcdoodle 5y ago
  For me it's the hostility towards developers and users. I can't even lower the dang contrast on my iOS or MacOS devices to make them a little less dopamine inducing. Sure you can switch to B&W but then I lose usability (Is that link clicked?). The massive cut they take from creators is another pain point. These are just off the top of my head, I'm sure others can chime in.
  - zepto 5y ago
    The dopamine inducing element is addictive apps like Facebook, not the color levels of the display. You can simply delete the apps that have that effect.

darthrupert 5y ago

Well, of all the reasons this might be the biggest one to buy Apple. Must feel amazing when your competition does your marketing for you.

andreime 5y ago
have you seen the charts apple uses in presentations? they deserve any mockery going their way

vmception 5y ago

Does it really matter? People that spend more than $1,000 on a laptop buy Apple. Edit: Some outliers really want me to write “overwhelmingly most” or find that old market study? Maybe. But I can also spend 4 points so fire away.

avgDev 5y ago
False. I like windows and have been using XPS laptops for a long time.
LegitShady 5y ago
do you really think this is true? because it isnt.
acdha 5y ago
This is a pointless distraction from the article. Professional users pay more for their laptops, no matter what platform, and given that the kind of laptop I need to do my job has gone down in price by at least a factor of two that price matters less than whether it does the job well.
kahrl 5y ago
I had no idea my Lenovo Thinkpad X1 was actually made by Apple! TIL!

notriskfree 5y ago

Interestingly it is the base M1 that is still the most interesting device, lower cost, lower heat, good single core performance, capable of most tasks, and reasonable graphics for a lower power device. Other fast laptops are all about the heat and the fan noise (my tiny work laptop is the loudest device in my room) A number of times I have got off the train with a laptop almost on fire in my bag, because windows failed to notice the laptop lid was closed. Ill be getting the M1 laptop next time I need one.

Shadonototra 5y ago

i have an M1 macbook, Intel/NVidia can lie all they want they can't hide facts M1 > to what ever they produce TODAY and tomorrow, it beats the hell out of them and let's not talk about how macOS puts to shame windows 11 thanks journalists for spreading their corporate propaganda, i wouldn't mind if it was to improve everyone's life, but spreading crappy bloatware.. nope this article is on par with the verge "stadia bad but xcloud good" let me laugh for a moment.. lol

qeternity 5y ago

I switched to Apple when they dropped PPC for Intel. I've had just about every x86 Macbook they built, and have built a handful of high-spec hackintosh desktops. When the M1 came out, I ran out and grabbed the cheapest one I could (8gb M1 Air) to play around with. After about 30 minutes, I decided to transfer my daily driver to this little passively cooled notebook. The next day I bought the M1 Mini to replace my overclocked hackintosh. My high-end x86 gear just couldn't compete with the low-end M1s. I don't care what Intel and Nvidia say. I've used all their products. The M1 is black magic.

faeriechangling 5y ago
One thing I’ve noticed about a ton of benchmarks is a focus on throughput instead of latency. Apple has had some of the lowest latency computers around and the M1 has only improved the situation with QoS features on the CPU and such.