Hacker News Reader

Hacker News

Comment by ksec | Hacker News Reader

Full thread

ksec

Here [1] is an actual test and benchmarks. Showing M1 is faster than a Laptop with Nvidia Discrete GPU.

Quite amazing.

[1] https://petapixel.com/2021/03/10/real-world-test-photoshop-f...

View on HN

They write that any GPU workload scores similar to the intel internal graphic[1].

For the total scores, they're intentionally using an old beta of the benchmark as it includes an older CPU workload that the M1 happens to excel at.

Seems a little like they're cherry picking results. Interesting, but a little dishonest.

----

[1]: Quoting the article:

> Unsurprisingly, the M1 Mac mini loses to the competition in raw GPU performance, more-or-less matching the onboard graphics of the quad-core Core i7 that’s in the 13-inch MacBook Pro (full review here).

Fact check: even if the M1's gpu performance matches that of an i7 as opposed to a discrete gpu it:

1) Is an entry level chip matching a high-end's performance.

2) it does so in around a 5th of 6th of the thermal workload.

Thermal load *

1) macs are not entry level, entry level is a 350$ laptop mounting an i3

2) it doesn't matter unless you work with the laptop sitting on the bare skin of your legs

EDIT:

sorry!

forgot the basic rules: don't dismiss bitcoins, don't engage with Apple users.

My fault.

The thermal load matters for battery consumption.

my 400$, 6 years old Chinese entry level laptop does 8 hours on battery...

My raspberry pi does 9 hours.

Your post is a poor fit for the collective mindhive here and it makes me uneasy. I better downvote it.

Keep calm and everything will be fine.

For the historical HN crowd, a MacBook Air less than $2000 is definitely entry level.

My high end Dell XPS was $2600, and the M1 Air blows it away for half the price.

> For the historical HN crowd, a MacBook Air less than $2000 is definitely entry level.

so basically 99% of the laptops are entry level...

Anyway the full optional Macbook Air on the Italian Apple website costs € 2.349,00

Which just has a bigger storage, but the performance is the same as in the cheapest one. There’s no point taking the most expensive option and try to invalidate the results because of it.

And yes, for development sub $1000 laptops are entry level.

16GB of RAM are the bear minimum in 2021

By simply upgrading the RAM from 8 to 16GB the MacBook Air cost goes up to € 1.389,00

Which is a lot to me for only 256GB of SSD storage

>My high end Dell XPS was $2600, and the M1 Air blows it away for half the price.

Isn't "blowing it away" a little hyperbole? An XPS with 1185G7 has around ~10% worse single core and ~25% worse multi-core CPU performance than the M1. An improvement sure, but not the leap some people make it out to be.

I don't notice a 10% worse single core performance, but what I do notice is going OOM because the Air is limited to 16GB RAM (base model 8GB is personally unusable in 2021). Or being limited to 1 external display due to the design of the M1.

Apple silicon needs to support at least 2 external displays before I can consider one and I'm honestly surprised they currently do not. The next generation of Apple chips might be useful, particularly if linux kernel support improves too. A 2023 Macbook Pro has a lot of potential.

The XPS 13 is also limited to 16 GB of RAM, and it's actually more expensive than a M1 Macbook Air, so I'm not sure why you're using that as a comparison. Even if you bump the Air to 16gb of RAM and 512GB SSD it's still $50 cheaper than the comparable XPS 13 which is $1499, which btw has a worse 1920x1200 screen vs the 2560x1600 one on the Air.

I'm not sure what source you're using for performance. But at least with Geekbench 5 results it's not as close as you're saying. The M1 Macbook Air scores 1699 in single core, and 7362 in multi-core[1]. The Intel Core i7-1185G7 scores 1446 in single core and 4924 in multi-core[2]. That's about 18% better single-core and 50% better multi-core performance, for me that's a pretty big difference.

[1] https://browser.geekbench.com/mac-benchmarks [2] https://browser.geekbench.com/processor-benchmarks

>The XPS 13 is also limited to 16 GB of RAM

This isn't true, I have an XPS 13 (9310) with 32 GB of RAM. You can also get the XPS 13 with a 4k screen, although I do not understand why you would. 1920x1200 is already much higher DPI than I care for, 2560x1600 and 4k are just a battery drain.

I'm going off my own benchmarks using Geekbench 5, where the M1 scored 1700 single core and 7500 multi core, and the 1185G7 scored 1550/6000 multi core. (There are people with better results than this on Geekbench as well.)

I wasn't making an argument about the price by the way, I do understand the Air is cheaper, but in my opinion it is cheaper because it is a lesser quality device with key functionality missing. Of course, for some people the Macbook Air is the best value because they don't need or care about these things. Which is totally OK, but having used both devices both the Air and the MBP are non-starters for me.

As I said, I hope future revisions can improve things and be more competitive.

Yes, it isn't only about pure performance numbers. Better performance with much lower heat, much better battery life. "Blows it away." Every laptop seems to have a drawback or two however.

A 15 W mobile chip would be considered entry level for most laptops.

A MacBook Air is 100% entry level.

MacBook Air costs 3x the price of a Dell Inspiron. The Mac laptops seem to be great (I've never owned one, have ordered one though), but they're not entry level. It might be Apple's entry level, but it's not the markets entry level.

For someone who is using Photoshop as a professional, $999 really is not far off from entry-level. People easily spend $2500+ on computers for professional work that requires high performance. Sure you can use a $350 Dell Inspiron for Photoshop work, but it's not going to be very practical. Performance, battery life, and even the actual screen will be pretty poor at that price point.

though, Photoshop professional work is not entry level.

My uncle is a woodworker, his tools are not entry level.

He doesn't buy the best of the best the market can offer, far from it, but his tools are not entry level.

they can be considered entry level only if you compare them with the state of the art in his field of work.

Entry level of the high end segment maybe.

I think they might be referring to the power used by the chip, which is indeed comparable to entry-level chips.

The thought is that if it runs this fast right now, imagine what it’ll be like when you put a big heat sink and fan on it and let it use 100w or something, on a Mac Pro. And more - this is a first generation product. Imagine what the second generation will be like.

I think people are just excited for the possibilities. We’ll see how it all turns out, but it really does seem promising.

> I think they might be referring to the power used by the chip, which is indeed comparable to entry-level chips.

Is it?

the Core i3 530 consumes 73 Watts

the AMD Ryzen 9 5980HS consumes 35 Watts

The i3-530 is a 10-year old desktop CPU, so of course it uses more power. The M1 can go up to 39w, which is comparable to the Ryzen you mentioned. I didn’t know either, and looked it up before commenting to make sure I didn’t make any wildly incorrect claims.

The M1 is an incredible achievement, but let’s try to remain honest about the numbers.

https://support.apple.com/en-gb/HT201897

that's absolutely true

The i3 desktop CPU is an entry level CPU because it can still handle entry level workloads, consuming a lot more energy than the state of the art because providing high end performances while consuming less energy than the competitors is not worth the entry level segment money.

Entry level means (IMO) good enough for as little money as possible in a non pro field (as in: your salary do not depend on the performances of the device)

M1 Macs are not in that segment.

> The M1 is an incredible achievement, but let’s try to remain honest about the numbers.

100% agree

Actually, I agree with the sibling commenter that power consumption is not a good predictor of whether a chip is entry level.

Estimate of TDP for the M1 silicon are about 20-24W [0].

The i7-1185G7E can be offered at anywhere within 12-28W TDP [1].

This 12-28W figure is exactly the same for the i3-1115G4E [2].

Yes, there's some trickery in Intel's TDP numbers, but based on Anandtech's results for that i7 [3], it looks like lower TDP means that you can't burst for as long.

[0] https://www.anandtech.com/show/16252/mac-mini-apple-m1-teste...

[1] https://www.intel.com/content/www/us/en/products/sku/208076/...

[2] https://www.intel.com/content/www/us/en/products/sku/208079/...

[3] https://www.anandtech.com/show/16084/intel-tiger-lake-review...

Power usage is not linear with perf. Doubling power wouldn't double clockspeeds or something.

They're comparing entry level Macs to high end Macs, or processors equivalent to them. That's the context.

> 1) macs are not entry level, entry level is a 350$ laptop mounting an i3

It depends on how you see it:

The M1 is the entry level chip for the whole upcoming M-series, this is a logical FACT.

It's hard to estimate how much an M1 costs but I would say it's closer to an i5 than to an i3 simply because the price tags can afford it.

Apple could make a profitable $350 plastic laptop with some kind of watered down m1 chip, they just don't want to.

> 2) it doesn't matter unless you work with the laptop sitting on the bare skin of your legs

It's in the name.... "LAP" "TOP", regardless this is not about temperate, this is about efficiency. The 18 hours battery life works both on and off your laps ;).

The m1 is basically the a14, which is rumored to go into the $350 base ipad real soon now. That’s their profitable entry level laptop.

They're similar, but it is not "basically the a14". This is like saying that an Intel Core i9 is basically the same as a Core i5. Sure it has a lot of similarities, but they have very different thermal and performance characteristics.

https://www.extremetech.com/computing/318715-comparison-of-a...

The i7 (and Intel) has always had mediocre graphics performance, mostly just for 2D, office/productivity and media encoding/decoding.

What'll be interesting is how the M1 will compare upcoming 4th gen Ryzen laptop parts with RDNA2.

Those chips will probably not be out until later this year and will take another couple months to show up in laptops. By then we will see the midrange versions of the M1. In Macbook pros.

I agree wholeheartedly, but a small correction: those chips are available now if you're very very lucky. For example, the latest ROG Zephyrus is available from my neighborhood Best Buy. The M1 has been readily available for months, and we'll probably see Apple's new pro chips by the time AMD's supply chain issues are sorted.

My bet is Ryzen will be slightly competitive this time around (at maybe double the power usage), but Apple will increase performance by 10-20% per year and eventually AMD will start lagging behind.

Not only is your post an example of the Moving goalposts fallacy [0].

Is also an apples to orange comparison.

The M1 was being compared to an i7, and that was I was responding to.

It will also not "be interesting", since if we part from "M1 does not beat current discrete GPUs" as per grandparent comment, it will obviously not beat an upcoming one, whatever the wattage will be.

[0] https://en.wikipedia.org/wiki/Moving_the_goalposts

4th generation Ryzen parts are expected to replace their Vega-based integrated GPU with an RDNA2 integrated GPU. That comment was not about discrete GPUs.

can it handle svga high resolution ?

Beating the integrated graphics in an i7 is nothing, especially on the older chips. Intel integrated has never been about performance. It really isn't saying much when comparing to the 10th gen chips. They use the old Iris graphics. Apple didn't adopt the 11th gen which are actually contemporary to the M1 and use an entirely new integrated gpu.

It is an entry level chip. And it’s a fantastic one.

But a comparison is a comparison even if lopsided, and when people are pushing this as “As fast as high end Intel”, it’s going to be squared against high end intel setups with discrete GPUs.

The M1 is absolutely fantastic. But it does absolutely fall short in some (arguably small) ways. Pointing those out isn’t unfair.

I don't understand what is behind the "filter" benchmarks, and it smells a bit fishy. Image filtering are (should be) highly GPU intensive tasks, so what does it mean to score low on "GPU" but somehow high on "filtering".

Results are a little strange. Theoretically speaking, the M1 is a 2.6 TFlop GPU at FP32. The Macbook Pro they are benching it with is just 1 TFlop. to get the exact same result on a GPU with 2.6X the FP perfromance seems....unusual. I would be interested to see if their benchmark is a limiting factor here as realword is dramatically different.

I think this is more a testament to the quality of Adobe's GPU pipeline (or lack thereof). M1's GPU is surprisingly capable for a 10W part, and more importantly, it supports low-latency CPU/GPU communication and opt-in memory page sharing. The later is especially interesting for creative software, as CPU and GPU can work on the same data structure, eliminating the need for copies or GPU buffer management. Knowing Adobe however, it's probably most likely that they are still using OpenCL, so the lack of performance is not surprising.

P.S. Or maybe the GPU test is simply memory bandwidth limited...

The workload on Photoshop should be a mix of CPU with GPU acceleration. And possibly NPU with M1.

So the reference to GPU was even with a discrete GPU doing whatever acceleration Adobe has done for Nvidia, the M1 is still faster. And doing so with a Power Usage that is far less than its competitor. Of course if we only focus on "GPU" category than the results as listed is quite clear.

Which leads me to believe the NPU and CPU may have played a big part of it.

None of those laptops has a modern Ryzen CPU in them though? For a more fair comparison?

Also your post is a bit misleading: GPU performance is comparable to Intels, not nVidias which is explicitly called out.

Ryzen isn’t popular in the high end laptop class.

Any guesses on why? I'm trying to get a new dev laptop with a Ryzen 4xxx or 5xxx but none of the _big_ enterprise vendors (dell, lenovo) have it yet.

Lenovo is much better than Dell and has announced a lot of Ryzen 5xxx laptops. Not sure if the Thinkbooks/Thinkpads are available yet but they are coming.

AMD just caught up (to Intel) in 2020 in laptop efficiency, and the OEMs are starting to catch up in offerings in 2021. However, availability is a mess in general lately. Might be 2022 before really high-end laptops offer Ryzen regularly.

The M1 is honestly very good value and the overall best mobile chip right now which is rare for Apple. Unless you need a PC. Goes toe to toe with my big desktop machine on dev workloads.

Pretty much all major laptop vendors have Ryzen 4xxx models. Some like Dell only put them in their lower tier models.

https://www.amd.com/en/shop/us/Laptops should get you started on where to look.

5xxx models are just appearing, look up the Zephyrus G15.

> M1 is entry

> Only compare high-end laptops

Wait...

Depending on what you're doing, compare Best (most powerful), Value, Efficiency, etc...

M1 wins in Efficiency. Ryzen wins on high-end power. Value varies by laptop but currently Ryzen wins on any sub-$1200 laptop comparison for value.

Thing is, Ryzen Zen3 is supposed to be pretty close to M1 on efficiency as well - this is why I want to see numbers comparing them because they're closer in capability than some obsolete Intel chip.

I suppose it depends largely on whether you consider a laptop with a 1080p laptop and a sluggish SSD a good value. Seems to me often the “Value” you get is simply going cheap on components. Mediocre display, barely useable trackpad, slow SSD, fan noise, thermal throttling, etc.

The current MacBook Air is a phenomenal value with almost no compromises.

The M1 chip macbook pro I bought recently is only able to support a single external monitor natively, whereas non-M1 macbooks can support more.

That's a massive compromise, and I'll be returning this version

1080p is perfect for my eyes (and games). My $1030 laptop has a mid-high end NVMe 512GB SSD. Display is 144Hz, 330-nit, 100% sRGB, muffled fan noise only when playing high-end games (RTX 2060 6GB), zero heat issues or thermal throttling (max ~75°C gaming). So I guess it's just another great value with almost no compromises.

(The Air and mine have different compromises. High-end gaming, wider selection of games, bigger screen vs minimal/no heat output, amazing battery, better portability. But I do think some of your assumptions about what is possible in the value segment aren't current.)

MacBook Air isn't marketed as a "Gaming" laptop, nor would anyone buy it as such. Claiming it's compromised because it's not great for gaming is like claiming the Xbox is compromised because it's not good at video editing.

My original comment was about each comparison fitting. The M1 is an expensive entry-level thin and light, but you get excellent performance and battery life. So it does that job well. But to say it has "no compromises" would say it could work for everyone. But it isn't a well-rounded laptop because of the screen size, OS, lack of GPU power, etc. Those things don't hurt you if you're happy with MacOS, the screen size, don't care about many games, etc. so it's not a "compromise" for your use case.

Not sure what that has to do with marketing - I mean you can buy based on marketing, but the original discussion was about comparing entry-level laptops and high-end laptops to the Macbook Air, and excluding Ryzen because they aren't in high-end laptops.

> But it isn't a well-rounded laptop because of the screen size, OS, lack of GPU power, etc.

You are off in the sticks. By your definition, nothing is "Well rounded". Every product is designed for a purpose. Value is about how well a product suits its that purpose. When I buy a hammer, I don't complain because it doesn't work as a screwdriver.

Your continued assumption here is that every computer needs to be suitable for gaming and that assumption is nonsense.

> original discussion was about comparing entry-level laptops and high-end laptops to the Macbook Air, and excluding Ryzen because they aren't in high-end laptops.

Your original comment which I replied to was talking about how Ryzen wins on value. Which is only the case if you are willing to accept lower end components, less battery life, and fan noise.

> Which is only the case if you are willing to accept lower end components, less battery life, and fan noise.

But this is only the case if you compare $1200 versions of the Macbook Air to $400-500 laptops...

My previous comment already discussed the existence of $1000-1200 laptops with Ryzen that don't have "lower end components" or "fan noise" (outside of gaming, which we can't compare on because we're excluding that as a possible compromise.)

Sure we're getting a bit into terminology. Long battery life is a feature, just like a full selection of games - it is a compromise to have less of either. It's not a critical compromise depending on what you buy the machine for.

Critical compromises (again mostly in the eye of the buyer) could arguable be bad quality or unusable screens, like those with poor color representation that prevent professional work.

So do this just mean that the legacy CPU architectures we've been working with for decades are just really outdated?

Is this less about M1 being amazing and more about Intel being inefficient, in terms of utilization of power/speed/features?

Because I have to assume the M1 chip can ignore a host of legacy compatibility/performance tradeoffs that must exist in x86 architecture.

If this is the result, hopefully it means a more generalized leap forward (not just for Apple hardware) at some point in the coming decade..

Reply on news.ycombinator.com