Hacker News Reader

Comment by 1_player | Hacker News Reader

Note: 30% faster than a thermally challenged i9 on a MacBook Pro, not a desktop one. Given the comments on similar threads, I feel this needs to be mentioned.

View on HN

To go another layer deeper in analysis though, it is still >20% faster on the thermally throttled M1 MacBook Air. That's a laptop without a fan, and it's still faster than an i9 with a fan.

This right here. I was so skeptical of getting an Air. And yes, during heavy compile sessions with make -j8, it can hang after a half hour or so. But (a) you can make -j7 instead, and (b) it’s impressive how long it lasts without hitting that point.

I’ve been thinking of doing the cooling mod too, where you pop open the back and add a thermal pad to the cpu. It increases conductivity with the back of the case, letting you roast your legs while you work, aka cool the cpu. :)

Do any of the laptop cooler systems with fans help the M1 Air thermal issues? I used one on an older 2011 MBP, and it definitely helped that laptop. It might have just been a placebo of getting the laptop off of a flat surface to allow air to circulate around it, but the fans could only help in that.

> And yes, during heavy compile sessions with make -j8, it can hang after a half hour or so

You bought a computer that crashes reliably when used "heavily"? I mean, if you heard that same statement from an individual PC user in past decades, you'd roll your eyes about the ridiculous overclockers and go back to work on your slightly slower but actually reliable development machine.

But this isn't a too-l33t-for-her-own-good overclocker. It's Apple Computer! Jobs is clearly still directing the reality distortion effect from the grave.

I assume he means step down in clock due to thermal pressure… not literally hang (stop working)

Well, depends on your definition of "literally hang." It does indeed literally hang (i.e. system becomes unusable, activity monitor impossible to open) for minutes on end if you're not careful; a fact I was somewhat dismayed to discover.

But it's rare enough that I just don't care much. The performance in other areas is too good to gripe about "if I run make -j8 for an hour, it'll hang for a few minutes."

It's running out of memory and that's when the memory pressure becomes notable enough to affect the terminal app. The same would happen on any other system - in fact it'll become unusable more quickly on Intel.

That seems unlikely. In fact kernel builds don't really produce much memory pressure, no more than a 1-200 MB per parallel make process in general. (The final link takes more, but that doesn't really parallelize anyway).

I don't know about kernel builds, but memory pressure is the usual limiting factor when building for macOS/iOS builds when you step up parallelization like this.

Another possibility is system bugs that can be exposed when it thermal limits because it will stop running background QoS tasks, and things waiting on them will hang. It's pretty hard to stress it to that level.

I have done both these things by building compilers on M1. The memory pressure wasn't always from the host compiler but rather tools like dsymutil that may not be perfectly optimized themselves.

Limiting factor for being able to use the machine that is, not for compile time.

Apple seem to have actively chosen to skimp on memory capacity, so this seems very much like something that's on them.

I would say it's unreasonable to expect 'make -j8' to perform well on a low-end laptop (M1 is a low end CPU) and that the system should be picking the -j number for you.

Geez. It ran 11% slower when it had to throttle a bit. That is not crashing. The only reality distortion occurring is a misreading of the article and overstatement of the effects of the reduced cooling in the Air.

Statement upthread, and confirmed in another comment, is about a system that really does "hang" when presented with heavy load. I'm responding to the commenter I quoted, not the linked article.

My misunderstanding. Thanks for the clarification.

there should be a term like “reality distortion field” that describes whatever it is that gets people to come out of the woodwork and nit pick acceptable performance compromises in Apple products

My 32 core threadripper hangs reliably if I run make -j64 and it starts to page out. Buckling under load isn't a phenomenon reserved for new hardware.

Just for your information, thermal pad means more (on youtube test +10 degrees celsius) hot chassis and more hot battery. Just be aware:) Good luck

To be fair, a fan takes away from potential heatsink space and eliminates contact from heatsink to shell. A fan based cooling system can actually be worse than a fanless setup... especially if they're the Macbooks with constrained intakes. (virtually all of them)

https://forums.macrumors.com/threads/macbook-pro-17-my-cooli...

https://hackaday.com/2017/12/14/cncd-macbook-breathes-easy/

https://www.ifixit.com/News/6882/why-i-drilled-holes-in-my-m...

I will never understand why Intel stuck with the i3, i5, i7, and later i9 branding across so many generations.

I’ve lost track of how many times I’ve heard people wonder why their 10-year old computer is slow. “But I have an i7”

Thank you for saying this.

Honestly - it's not even the i3, i5, i7, i9 thing. It's the fact that two i5s, etc; can be ludicrously different in terms of performance from one another because of the sub-generations within the named generations.

Yes - it's ridiculous that I could buy an i7 ten years ago, buy an i7 today, and yet - of course - they are absolutely nothing close to each other in terms of performance.

IIRC the Pentium line did not make this mistake. (Though the Celeron line could be very confusing, if I recall correctly.)

To play devil’s advocate, I can buy a Corvette today that is nothing like the one from ten years ago too.

In fact, lots of things are like this.

How much does the top speed differ?

More important than top speed, 0-60 went from 4.2 to 2.9 seconds in that time. So the analogy stands

https://media.gm.com/media/us/en/gm/news.detail.html/content...

About 10 years ago. 7:19.63 (ZR1 / i9)

https://www.automobilemag.com/news/2020-chevrolet-corvette-s...

About a year ago 7:29.90 (base model / i3)

Better measure than top speed anyway

Corvettes like all cars are identified by their model year. Hence there is no confusion that a 2021 Corvette is "better" than a 2011 Corvette.

Are the '21 models better than an '11? I don't think anyone would say they'd rather have a '21 than a '69.

I do love collecting my old classics, but every time I'm the one driving my wife's Civic, I think, huh, everything about modern cars is better in every single way.

Including pervasive tracking of driving habits and who you associate with, sold to the highest bidder. :-(

In terms of speed, definitely - the '69 Corvette's zero-to-60 mph time was about 7 seconds, the '11 Corvette took about 4 seconds, and the '21 Corvette takes under three seconds. Sustained performance, maximum speed, etc. has also improved.

They would if they were looking for top speed.

Depends on if I have a garage and a team to keep it running good, if I'm taking it for a nice drive or just showing it off in a museum.... The long, swooping '69 is iconic, but there are lots of things to prefer about a modern model.

Almost everyone knows that the model year is part of the, idk, "minimal tuple" for identifying vehicles, though, and you can count on it always appearing in e.g. advertisements.

In CPU land, the architecture codename or process node might be part of such a "minimal tuple" but these are frequently omitted by OEMs and retailers in their advertising materials.

The point is that people think the numeral in the brand is something like a version number in which larger numerals are better. I.e., an i7 is always better than an i5 when in fact a new i5 might exceed the performance of a dated i7 for some particular metric.

Or frankly one new i5 might be powerful than a new i7 if one's a desktop or H series and the other is a Gx or U series.

2009 C6 Corvette ZR1: 0-60 in 3.3 seconds 2019 C7 Corvette ZR1: 0-60 in 3.2 seconds

I have always hated 0-60 because it depends greatly on transmission shift points, driver ability, rear end ratio, and most of all traction, which OEM street tires have little of.

Let’s go to a full quarter mile, which show a bigger difference, but again not a huge one.

2009: 11.2 seconds/130.5 mph

2019: 10.6-seconds/134 mph

This would be a better comparison of Ford Mustang GT’s of the same year. Massive improvement.

It's a crappy metric. What about:

Time to stop from 60 mph?

Effective turning radius at 60 mph?

Interior noise in decibels at 60 mph?

Maximum shock absorption at 60 mph?

Nürburgring times.

Or 0-100-0.

>> Nürburgring times.

7:06 vs 7:24 for the old car. Although in 'segmented' times it had went something like 6:54

This is valid since Intel's naming scheme is influenced by BMW.

I can't wait for Intel to release an i7 variant as the i8 Gran Coupé.

Instead they could have just called them i2017, i2018... going with year of manufacturing. That way it is useful to make some sense out of performance with an understanding iN is always better than i(N-1)

That gives you the opposite problem, where someone gets a brand new dual core and is confused by it being slower.

Best is to give them a number that approximately maps to performance.

The "pro" version might be an i8, while the budget version is i3. In a few years time, the pro version will be up to i12 while the budget version is now i8.

You have model numbers for when someone needs to look up some super specific detail.

Right. The issue I have with this is that we have i7 models that are from 5 and 10 years ago with vastly different performance due to their generation. If it was more iterative it would make more sense.

Year of manufacturing says nothing; you can have two different gens manufactured in the same year, one for lower price tier and the other for the higher one. Just like Apple still produces older iPhones, same thing.

Instead, you have designations like "Intel Core i7 8565U Whiskey Lake" or "Intel Core i7 10510U Comet Lake". The first one is 8th generation (=8xxx), the second one is 10th generation (10xxx, but the 14nm one, not the 10nm "Ice Lake"), and most OEMs do put these into their marketing materials and they are on their respective web shops (these two specifically were copied from two different Thinkpads X1 Carbon models).

Then there's the Acer way of naming products like "XR382CQK bmijqphuzx" so that each one is unique. I like Intel's clearly defined 9 > 7 > 5 > 3. However, I do wish that Intel made the model generation part of their marketing material so that retailers and OEMs would be forced to include that information in their product marketing too. Intel i7.2019, for example.

> Then there's the Acer way of naming products like "XR382CQK bmijqphuzx"

This take is deceitful. Acer typically uses a very simple scheme to define their products, which goes something like "Acer <product line> <model>".

The scheme "XR382CQK bmijqphuzx" is more a kin to a product ID, which goes way beyond make and model and specifies exactly which hardware combination is shipped in a particular product.

Complaining that Acer names its products like "XR382CQK bmijqphuzx" makes as much sense as complaining that Apple names its laptops like MVFM2xx/A, which Apple uses for the same effect.

The model number is placed on the beginning of the title. Or you argue that we should call it as """ 37.5” UltraWide QHD (3840 x 1600) AMD FreeSync Monitor (USB 3.1 Type-C Port, Display Port, HDMI Port) """ ?

https://www.amazon.com/Acer-XR382CQK-bmijqphuzx-UltraWide-Fr...

WTF, XR382CQK bmijqphuzx is really the model of an Acer monitor?

This is not at all how monitor marketing works. Every company gives them insane nonsense names. There is no other identifier. for XR382CQK bmijqphuzx there is no product line or model other than "monitor". That's all you get.

> IIRC the Pentium line did not make this mistake. (Though the Celeron line could be very confusing, if I recall correctly.)

Current Pentium and Celeron chips can either be mobile variants of the Core line or... Atoms.

I have a Celeron N4100 in my cheap netbook. It's listed as "Gemini Lake" in Intel ARK, which is a rebranding of its original lineage... Goldmont. (Goldmont Plus)

https://en.wikipedia.org/wiki/List_of_Intel_Celeron_micropro...

If you were to look at other Gemini Lake cores... You'll see that "Pentium Silver" also falls into that group.

https://ark.intel.com/content/www/us/en/ark/products/codenam...

TLDR: Intel has made it nearly impossible for anyone to know if they're getting Core based or Atom based mobile chips unless significant research is done beforehand.

This is not to say avoid Atoms at all costs. The N4100 I have is reasonably on par with the Core 2 Q6600 quad, which was quite the beast of a chip back in its day.

The current Atoms are nothing like the in-order-execution only monstrosities they originally launched as... but I still find Intel's branding a bit more than confusing. If incredibly deceitful.

Except this is an article about the Apple MacBook Pro 16" which came out aprox. 1 year ago (edit, 1 and a half years ago).

To be fair, it's probably not the CPU that's making their computer feel slow. Most likely, the lack of SSD and small amount of RAM.

No joke. I still daily use a 2008 MacBook Pro(with a Core2Duo CPU), with an SSD and 8GB of ram and that machine is perfectly fine for what I need it for - browsing the web, replying to emails, listening to Spotify, watching YouTube/Netlix. And people have bought a 2019 iMac with a normal 1TB 5400rpm HDD(!!!!!) And complain that it's slow. Yeah, of course it is, but it has nothing to do with the CPU in there.

Yeap. It helps that Intel has missed the ball in the last 5 years or so. And tech savvy users are used to watching CPU/RAM usage, which helps a lot too.

My wife had a “my computer is slow” issue the other day. Fans blowing like crazy, extremely hot, battery draining fast. Turns out, lsof was stuck on some corrupt preferences file. Killed the process and file and all was fine. Regular people lack the tools to diagnose the problem and just deal with it, restart or buy a new one. We really should make diagnosing easier.

Replace Geek Squad with a very small (AI/ML) script.

No, seriously. Once the root cause is found, humanize the description of the problem and ask the user what they want to do about it.

Have you seen the software the Apple Genius Bar uses to troubleshoot problems? It’s all scripted out with lots of automation.

Actually, no, I never use them.

I bet they would to reduce labor costs.

Desktop Reliability Engineering ;-)

I gave a responsible, sane but poor elderly homeless guy I was acquainted with a MacBook Pro (13-inch, Mid 2012), chargers, and canoeing dry bag. The MBP was originally a base 4 GiB and 500 HDD that, at some point, I upgraded to 16 and 480 SSD (OWC when it was $1.5k) + 500 HDD - optical.

OSes, platform toolchains, and apps need to prioritize UX (UI response time and time-to-first-window) to lessen the perception and the reality of waiting on apps doing unimportant activities rather than appearing and beginning interaction.

https://youtu.be/H7PJ1oeEyGg

https://web.archive.org/web/20110703091817/http://buyafuckin...

The memory hierarchy of a modern CPU with spinning rust might as well be the Parker Solar Probe waiting on a sea anemone.

I am remiss how someone these days can buy a laptop with spinning rust and 4 GiB and then complain it's "slow." Maybe they should've bought a real laptop for real work that's repairable and upgradable?

>repairable and upgradable

Unfortunately, the best ones aren't anymore.

Marketing "the best" (most expensive / newest) aren't the best, for most purposes.

I bought a T480 with dual batteries that does run for 10 hours. It has a 2 TiB SSD and 32 GiB. Works fine for me. Water-resistant, repairable, awesome keyboard, and MIL-SPEC drop rated too. Optional magnesium display top assembly cover.

Those aren't the best ones then.

But it's a 9.21 Jigawatt, 11 GHz, i21 processor!?

And it's cycle- and power-efficiency figures are?

Only the frequency wars ended, the marketing and features wars continue the prohibition on computer architecture enlightenment unabated.

Exactly. They should have named them Air, Pro and iMac. That way you clearly can see which is fastest and which year they came out.

It seem pretty common for people who own Apple products to say "2013 macbook air" and that is easy to find on the device or in the software. Apple makes it pretty clear newer is better. Intel has a bunch of lakes and numbers and an easier way to tell would be nice all around. Though Apple has less skus to differentiate making it much easier for them.

It's pretty common for people who own Intel cpus to say 5th gen or 9th gen or to easily find on the device or in the software the name of cpu like i5-4670k or i9-9900K

I think GP is talking about a branding problem though. Non tech savvy people may not realise that the i7 in that new computer at the store is different from the i7 written on the front of their PC.

For years CPUs had numbers that went up: 286, 386, 486, Pentium.

After that, it was Mhz and Ghz that people used to rate a CPU.

But none of those things are as relevant as they used to be.

So what number goes up? i3, i5, i7, i9. That's what Intel is telling us. That's the number they want us to see.

Even on Mac OS, when you go to "About This Mac" -> "System Report", you only see the iBranding and not the generation (I see "6-Core Intel Core i7" on mine for example).

from about this mac:

> MacBook Pro (Retina, 15-inch, Mid 2015)

> Processor 2.8 GHz Quad-Core Intel Core i7

prettly clear there

But Mac date year doesn’t always correspond with the intel processor year/generation.

So which one in the following is fastest: i7-960, i5-2400K, or i3-1115G4? And where does Pentium Gold 7505 fit in?

Except for the Pentium, I can tell the age from the number. The i3 is probably the fastest, but since stuff uses less power these days, there's a small chance the old i5 is faster. I'd look up benchmarks if about to make a purchase. Is the Pentium equivalent in age to 7th gen core i, or is their numbering different?

Now consider that you are an average person, not so tech-savvy, don't really follow CPU industry news.

How would you differentiate?

These aren't realistic choices unless you are buying a used PC. But even less savy consumers understand generations, if you present a fast 1st gen model versus a not the fastest 11th gen model I think the majority will go with the newer model.

So which of these is faster Macbook Air or Macbook Air?

Those are all fast(to the extent matters to their marketing)

Intel would be a lot more motivated to show you which products were newer if their newer products were actually better.

I've seen the generation mentioned first in almost all places. For example https://www.dell.com/en-uk/shop/laptops/xps-15-laptop/spd/xp... lists "9th generation Intel core i5". How come people pay attention to the other number more?

They pay more attention to the other number because that's how the English language works. "9th generation" is a descriptive prefix to the actual CPU model name.

I don't get why not a single brand other than Apple gets product names remotely right. Every time a relative asks me for a laptop recommendation and shows me a list of models I have absolutely no clue what to tell them. All the model numbers look like they came from a password generator and the only discernible difference without hours of looking at spec sheets is the price.

PlayStation. The name is brilliant and they’ve incremented the index each release since the first about 25 years ago.

(I know it’s a different product category. It’s relevant because it’s name and number is perfect)

Most cryptic product number for consumer product I've ever know is Acer's monitors like XF270HBbmiiprx.

So that OEMs can keep selling their 5 year old chip designs at the same margins while their manufacturing costs plummet.

Combined with generation it says plenty. It's unfortunate when the public doesn't understand. I'm not sure what should be done.

It is worth noting that if you think GPU model numbers are fine, CPUs are actually very similar. There's the 2600k, 3770k, etc. (sandy bridge and ivy bridge CPUs respectively) The first number goes up each generation and the rest is how powerful it is within the generation. Similar to Nvidia having a GTX 980 and later a GTX 1080.

I get your point and it's only 5 years but my i5-7300HQ and i7-3520M were pretty comparable due to the number of cores, despite what pure benchmark numbers would say (and I'd actually prefer that i7)

It’s a problem of intels own making - marketing vastly different capabilities under the same brand in order to segment the market.

or Apple doing a bad job with the previous generation

Or both

The 2019 macbook ironically had better heat dissipation than the previous generations, but it's still pretty bad.

We can blame Apple for using chips that are too intense for their laptops, and we can blame Intel for making garbage chips that can't really perform in real world cases while spending a decade not bothering to innovate. Apple at least is moving away from Intel as a result of all of this, and I'm really impressed with how well the M1 transition has been going.

Ehh. I take the view that Apple has been intentionally sandbagging their laptops for a while to facilitate an ARM transition.

Not to say that M1 isn't amazing, but I think Apple has been preparing for this for a while and needed to make sure it would succeed even if their ARM CPUs weren't quite as groundbreaking as they turned out to be.

Are there any benchmarks you can point to that have a similarly spec'd laptop (ideally similar size & weight too) that would show that Apple is sandbagging?

Possibility 1: Apple was making do with what Intel gave them, because their profit margins didn't care and they were busy naval gazing into their post-Jobs soul

Possibility 2: Apple had a master plan to intentionally torpedo performance in order to make their future first-party chips appear more competitive

What Intel supplied was the bigger problem, but Apple was definitely not trying to make the chips perform well. They were hitting thermal limits constantly, and more directly toward "sandbagging" the recent macbook airs have a CPU heat sink that isn't connected to anything and has no direct airflow. They could easily have fit a small array of fins that the fan blows over, but chose not to.

I've seen the video that makes that claim but it's just plain wrong, if you look at an Intel Air the whole main board complete with heat sinks lives in a duct formed by the rear section of the chassis.

Some air goes over it but it's not enough with the tiny surface area. A proper block of fins would have done so much more while still being small and light.

If it’s that easy, do you have an example of another vendor doing it? It seems like Intel would love to highlight, say, Lenovo or HP to say Apple was cooling it wrong.

Almost every laptop in the world connects a block of fins to the CPU with a heat pipe.

I went to ifixit and clicked the very first laptop teardown I saw, it has one. https://guide-images.cdn.ifixit.com/igi/gnBU5dVYrIhInIXV.med...

Or, let's see, razer blade stealth, that's a 3 pound laptop. https://guide-images.cdn.ifixit.com/igi/nHPaankElOWQndsv.ful...

What's a light 15 inch laptop, XPS 15 7590? https://i0.wp.com/laptopmedia.com/wp-content/uploads/2020/01...

Acer swift 5 SF515, "lightest 15 inch laptop on the market" https://i1.wp.com/laptopmedia.com/wp-content/uploads/2019/01...

LG Gram https://guide-images.cdn.ifixit.com/igi/fFlFJB3PKThJs3TA.hug...

Apple is actually just really terrible at thermal design, simple as that. It's surprising, because you'd kinda expect them to at least be slightly competent (like they are in other areas), but they're just not.

They've gotten around it with M1 by building an entirely new chip that's power efficient enough that it practically doesn't need good thermal design to perform near it's peak. It's an impressive technical accomplishment, but it's also hilarious that Apple had to go this far to cool a chip properly.

Yes, we knew that design was common - that’s why that PC fan made the video picking it as a point of criticism – but that doesn’t tell us whether it’s as big a deal as being claimed. That would require some benchmarks running for long enough to see substantially greater thermal throttling, higher CPU core temperatures, etc.

Apple could have had good thermals with what Intel gave them, they just didn't because they seem to value "as thin and light as possible" as opposed to "pretty thin and light, and with sane thermals". Even then they seem to make straight up poor decisions around cooling a lot even when doing the right thing wouldn't affect size/weight. Do they just not have engineers on staff who are good at this or something?

It's absolutely possible to do high end Intel in a compact laptop with relatively good thermals - look at thin Windows gaming laptops for plenty of examples of this (and these have way beefier GPUs than macbooks too).

It's worth noting the engineers actually seemed to get management to allow the 2019 model to be _thicker_ with more thermal mass than the earlier models. But in a sense, Intel's chips have never been amazing for laptops.

The power draw drains batteries, and the heat is... very annoying at best. I have used a couple ThinkPads and a Dell XPS over the past 5 years too, and all of them had the same issues where they'd constantly be pushing out very hot air, and the battery life was never more than 4-6 hours.

Compared to other amazing Intel laptops of similar form factor ? All Intel laptops are insanely loud and generate tons of heat for any reasonable performance level. They are just generation(s) behind in process, plus they start from an architecture designed for servers and desktops and cut down, Apple went the other way so it's reasonable they will do better on thermals and power consumption.

For _five years_?

Or longer, really; while everyone, of course, loves the 2015 MBP, they're mostly thinking of the integrated GPU one; the discrete GPU version was pretty thermally challenged. Arguably Apple's real problem with the post-2015 ones was that Intel stopped chips with fast integrated GPUs, so Apple put discrete GPUs in all SKUs.

While I don't think it's a likely theory, 5 years from "we should do it", to design, to testing, to preparing fabrication at scale, to design of the system around it, etc doesn't sound unreasonable. I would be really surprised if Apple decided the Arm migration after 2015.

> I take the view that Apple has been intentionally sandbagging their laptops for a while to facilitate an ARM transition

That's just not a defensible position to take.

Intel's inability to execute a node transition has led to a situation where for years their only way to increase performance has come at the cost of major increases to power and heat.

>Whilst in the past 5 years Intel has managed to increase their best single-thread performance by about 28%, Apple has managed to improve their designs by 198%, or 2.98x (let’s call it 3x) the performance of the Apple A9 of late 2015.

https://www.anandtech.com/show/16226/apple-silicon-m1-a14-de...

I bought a 2019 Core-i9, top of the line, and plugging a non 16:9 resolution screen into it out of clamshell mode causes the GPU to consume 18W, and basically cause the system to unsustainably heat-up until it starts throttling itself.

https://discussions.apple.com/thread/250943808

It's clear they have hit a limit of what they could really do with Intel's top of the line processors from a thermal perspective, with the form factor they want to deliver.

Now I have sort of an expensive paper-weight sitting next to my M1.

> GPU to consume 18W

That was a bug on Apple’s part, not a feature. Case in point my Kaby Lake Windows laptop package power draw is less than 2W when plugged into 1440p.

So Macbook Air that came out in 2008 is as fast as the one that came out in 2020?

I blame Intel. For years i9 wasn’t just a “beefed up i7”, it was a high end line of CPUs with 10+ cores costing $1000+, meant for professional workstations that didn’t need Xeon’s ECC RAM support. Then suddenly i9 started to also mean “a beefed up i7”.

Maybe they did this in the hope that the i9 HEDT “brand halo” would help them sell more of these consumer CPUs. But if that’s the case, they don’t get to complain when someone posts that an M1 beat an i9 and people accidentally interpret that as “the M1 competes against Intel’s HEDT line”.

Its also a 9th gen i9, a two and a half year old chip.

You know what's interesting, China and Russia have been struggling for years to get something on the level of intel Westmere. And here comes Apple out of the blue with a proprietary arch and hardware emulator; cinebench showing it to be around a x5650 xeon (Westmere). Easy.

M1X and M2X in the making, too?!

Apple have at least 13 years experience.

Not to dismiss the hard work the engineers at apple put in but China and Russia hasn't poached as many engineers over the years as Apple has.

That's correct, China just poaches the tech when it lands on their soil without paying a thing. At least Apple pays those they've poached a salary.

They also don't use TSMC's sub 10nm nodes.

The whole point behind the effort of China and Russia is to have a CPU under their control, starting with the design but also including the manufacturing.

And TSMC uses technology for their nodes which is not under their control at all.

The amount of expertise China already has and will accumulate in the next decades should not be underestimated.

> And TSMC uses technology for their nodes which is not under their control at all.

Note it's also not under TSMC's control - a lot of fab technology comes from other suppliers like ASML in Europe.

Definitely not out of the blue. This has been a long and steady march

Not only do Apple have decades of experience both as themselves and PA Semi, they can also probably outspend efforts that this countries could do politically (Russia yes, China probably not, but you get the idea) especially when weighted against their ease of acquiring information.

Also Intrinsity.

> And here comes Apple out of the blue with a proprietary arch and hardware emulator...

Apple is designing processors and GPUs at least since iPad2's tri-core GPU. They're neither coming out of the blue, nor newbies in this game.

It's not out of the blue and not even surprising.

Look at this from this POV:

- Apple started on custom ARM many years ago.

- Apple isn't really smaller then Amd, and Amd also rewrote and restructured their architecture some years ago.

- Apple hired many greatly skilled people with experience (e.g. which worked before with Intel)

- Apples uses state of the art TSMC production methods, Intel doesn't and the "slow" custom chips from China and Russia don't use that either as they want to have chips controlled by them produced with methods controlled by them. (TSMC production methods are based on tech not controlled by Taiwan).

- Apples had a well controlled "clean" use-case, where they bit by bit added more support. This includes that they could drop hardware 32 bit support and don't have any extensions they don't need for their Apple products, this can make thinks a lot easier. On the other hand x86 has a lot of old "stuff" still needing support and use cases are much less clear cut (wrt. thinks like how many PCI lanes need to be supported, how much RAM, etc.). This btw. is not limited to their CPUs but also (especially!) their GPUs and GPU drivers.

So while Apple did a grate job it isn't really that surprising.

> On the other hand x86 has a lot of old "stuff" still needing support and use cases are much less clear cut (wrt. thinks like how many PCI lanes need to be supported, how much RAM, etc.).

Good point overall - the performance of Rosetta 2 helped facilitate much of this and - at least from what I've read - does seem to be surprising to folks in the space. So that also helped them out.

> ... (TSMC production methods are based on tech not controlled by Taiwan).

Do you mean "TSMC production methods are based on tech controlled by Taiwan", or "TSMC production methods are based on tech not controlled by China/Russia"?

Update: after reading the thread I think I understand that you meant TSMC production relies on tech/equipment from ASML, which is not controlled by Taiwan.

>China and Russia have been struggling for years

Sometimes it's hard to figure out all of the things left out of the plans that were stolen. Some engineer saw something not working, looked at the plans, and then noticed where the plans were wrong. The change gets implemented, but the plans don't get updated. Anyone receiving the plans will not have those changes. Becareful of those plans that fall of the back of trucks.

The team behind the M1 has been doing high performance chips since they did a PowerPC chip in _2005_ and has been doing custom low-power apple arm CPUs for almost 14 years.

Being thermally challenged is part of the design, huh...

This was a famously bad CPU/cooling design, actually. LOTS of people complained about it at the time. You can place blame on either party according to your personal affiliation, but similar Coffee Lake chips from other vendors with more robust (and, yes, louder) cooling designs were running rings around this particular MacBook.

After having to use i9 MBP for 1,5 years, my conspiracy theory is that Apple created those purely to emphasize how their own processors are better.

Also, many MB(|P)s get thermally-challenged and frequency-challenged because their sensors go tits-up.

Why? It’s all the more apples-to-apples as a comparison because the form factor remains the same and thermal limitations are similar between two system.

Why would you want to compare a desktop class i9 with a 10 watt M1 chip?

Not sure why you are being downvoted.

Because there are a lot of issues with i9s in those form factors that leads to less perf than even an i7 from the same generation.

There was a Linus Tech Tips the other day about how even current gen Intel laptops can see better perf on an i7 than an i9. It looks like the i9s simply don't make sense in this thermal form factor and are essentially just overpriced chips for people who want to pay the most to have the biggest number.

The title is implying M1 is always better than every intel, given that I9 is the best intel (consumer) chip.

It's been known for years that apple has been limiting the intel chips by providing insufficient cooling. I don't overly care about how fast an M1 chips in a macbook is compared to an intel chip in a macbook. I want to know how fast an M1 is compared to a desktop I9 (given mac minis have M1 chips now), or compared to a properly cooled laptop latest-gen I9.

All this experiment shows is that insufficiently cooled processors perform worse than sufficiently cooled ones. It's a classic example of cherrypicking data. Admittedly, my solution would be different to the article's author. Instead of using a badly cooled laptop to compile stuff, I'd setup a build server running linux.

If it's only a matter of cooling why does the Intel also produce 4x the power consumption (not highlighted in the article, but still...).

If I wanted absolute best compile times I'd go build a threadripper workstation, but I need mobility and can't always rely on an Internet connection.

Well that's why the M1 is much better than the I9, not strictly because it's a faster chip, but because it's a more efficient chip. Efficiency = less cooling required.

I think part of it is because I don't care about mac computers. I'm interested in how the M1 - a consumer ARM chip - stacks up against the best x86_64 chip, not how it performs in one particular laptop.

Oh come on, comparing it to the "best" intel mac laptop makes sense when you only care about that form factor/specific model, which is the case for a big part of many industries (Software included)

Not to a desktop CPU no, but I'd rather see comparisons to laptops that don't have terrible thermal design. And to AMD as well as Intel - they're doing significantly better than Intel in the high power efficiency space afaik.

It just seems like a cherry picked fight - there's more to the laptop market watt/performance wise than 2019 macbooks.

Reply on news.ycombinator.com