You’re making it seem like they’re hiding that information under a footnote. The real text on the page, which is quite visible, is:
> Up to 23x faster than fastest Intel‑based MacBook Air
And right next to it:
> Up to 2x faster than MacBook Air (M1)
The footnotes are there to expand on the conditions of the measurements.
So not exactly misleading. On the contrary, it seems to me they’re quite clearly saying “if you have an Intel or M1 MacBook Air you have reason to upgrade. Otherwise, don’t”.
"Up to" is still doing a lot of work there. What kinds of workloads are we talking that get the big numbers, and what can we realistically expect on real workloads?
I'm reminded of 90s advertisements in which the new G3 processor was supposed to be so many times faster than the Pentium or even Pentium II. Their chosen benchmark: how long it takes to run a Photoshop plugin. On Mac OS pre-X, a Photoshop plugin got 100% of the CPU because there was no preemptive multitasking. Windows 9x versions of Photoshop had to share the CPU with whatever else was running.
> Testing conducted by Apple in January 2025 using preproduction 13-inch and 15-inch MacBook Air systems with Apple M4, 10-core CPU, 10-core GPU, and 32GB of RAM, as well as production 1.2GHz quad-core Intel Core i7-based MacBook Air systems with Intel Iris Plus Graphics and 16GB of RAM, all configured with 2TB SSD. Tested using Super Resolution with Pixelmator Pro 3.6.14 and a 4.4MB image. Performance tests are conducted using specific computer systems and reflect the approximate performance of MacBook Air.
As someone that migrated to the M1 Macbook Air from a Mid-2014 Macbook Pro... the Intel customers are still the ones they're trying to target, amusingly.
If they'd just give me onboard mobile connectivity, I'd upgrade to the next Air sooner, otherwise this thing will run until it dies... and maybe some day they'll start comparing performance against their original M1.
> the Intel customers are still the ones they're trying to target
Definitely. I have ZERO rational reasons to upgrade from my lowest-spec first-gen Air M1. I use it everyday and speed and battery life are still way more than I need.
Literally the only material difference between using my M1 Air and my work M1 Pro is the somewhat-better port selection on the Pro. Though even that doesn't have the single-most-useful port it could (aside from USB-C): a USB-A port.
few weeks back a professional ios dev looked at my m1 pro and ask why i had an air instead of pro. i might go air when i finally upgrade bc the new pros are giant compared to the m1
This mimics my experience. I bought the absolute bottom barrel M1 when they launched to replace my 2014 MBP, 8gb RAM and 128gb of space. The HD space is annoying, but otherwise this machine is untouchable. I do game dev work bouncing between the MBA and my gaming rig, which is Ryzen 7 2700, 64gb RAM and a 3070, and with certain benchmarks, the MBA still wins, silently, on battery for hours. Still blows my mind.
I see a lot of people requesting cellular modems in MacBooks, but the integration with iPhone hotspot connectivity is so good that I don’t really see the point of it for most people.
Battery consumption and antenna efficiency are two major pain points. iPhones suck battery like a horse drinks water when in hotspot mode, and the large surface area of even a small MacBook Air would allow for some pretty interesting antenna design.
And it isn't really ironed out to behave in Germany where on a train you have frequent losses of phone connectivity. Every time it loses signal, the hotspot drops out and disconnects.
This is a weak argument. Qualcomm is charging for iphone and ipad too. They could do it if they wanted.
The real reason is Apple wants you to buy an ipad for on the road. Laptops, according to them, are strictly for office/home usage where wifi is available.
The integration is fine, but it's not perfect. It kicks my wife off her iPhone Hotspot every time she closes the lid on her laptop. It also burns the battery on her iPhone, which is a concern in the exact situation you'd want cellular connection (places with no wifi often don't have outlets either).
Anecdotally, I've also seen her get issues when going from an area with bad connection to an area with good connection (iPhone will disconnect).
The experience with a non-iphone is also not seamless, though that's to be expected.
Point being that reliable and easy cellular access on a MacBook would be a pretty nice improvement. This is especially true given how much of what people do on computers relies on the internet these days.
The page up to 2x faster than M1, but it's not worth upgrading from for the average person, your laptop should last longer than 4 years hence why they market to Intel Mac users.
I think that was around the time when macbooks were "fast enough", especially since that was when SSDs became the default. I remember I got my first macbook around 2011/12 and at the time doing your own upgrade of memory and replacing the hard drive with an SSD was a pretty popular DIY upgrade (N=1).
> the Intel customers are still the ones they're trying to target, amusingly
yeah i just checked mine, it says MacBook Pro 16" 2019 and the cpu is an intel i7. i don't know what to say, it still meets all my requirements, i don't feel any need to upgrade.
My work laptop is a 2019 i7, my personal is a m3. There is a huge - very very noticeable difference. The thing that actually annoys me the worst with the intel though isn't the 'speed' per se but it's the shitty battery life and heat it generates (and the fan noise that causes).
If it works for you... but I had an Intel MacBook Pro (2019, 16", i9) for a work machine and the fans would sound like a jet engine. Installing NPM dependencies was particularly bad, since all the writes would make the corporate file scanning spyware go crazy. It ran like crap.
I have an M1 Pro (which is considered old now) and it's like night and day.
My 2014 got a little screwy around 2022 and eventually wifi stopped working entirely (I suspect battery swelling putting pressure on something) but if not for that I'd still be using it. Hell, I probably could have gotten it fixed, though I'd prefer to put that money toward another machine that'll last me 8+ years.
I'm on a 2020 [edit: I got it as part of comp for a contracting gig, is why the overlap in years with my 2014 MBP ownership, but didn't switch to using it for personal stuff until after that was over and my MPB wifi broke] M1 Air now, so close to or in year 6 for that. No issues yet and battery life still stellar, should get at least 2-3 more years.
(Folks who are like "LOL who even needs 18 hours of battery life?", which is a common sort of post on Apple laptop announcements: well for one thing it's extremely nice to be hunting for outlets even less often, and to maybe go on a whole light-laptop-use 3-day trip and not charge it the whole time and it's still alive at the end of it, or to have that battery as reserve for charging your phone, but also and perhaps most importantly, it means that a 30% degraded battery after several years of ownership still gets you 10+ hours of real-world use)
Not who you replied to but I’m on a Mid-2014 15 inch MBP retina, bought new and used nearly every day since and taken on dozens of trips.
I had the battery replaced, the tab key replaced, and the screen refinished (anti-glare coating removed) for about $240 a couple years ago and aside from the fact it can’t be updated beyond Big Sur 11.7.10 I have no issues.
Same, except my 2009 Mac Pro made for a better space heater, until I replaced it with an MBP M1 that doesn't have the decency to make noise to let me know it's working. Only downside of upgrading is that I had to get off of Mojave.
I have a Late 2013 MBP still going strong. Original battery, original charger, no repairs whatsoever, hours of battery life still. Wife stopped using it just two months ago when I upgraded her to my M1 Air.
> the Intel customers are still the ones they're trying to target, amusingly.
Yeah, particularly for the Air that makes complete sense, though. Consumer laptops tend to get replaced pretty slowly. I'll be upgrading from a _2016_ MBP (though not to the Air, given the lack of the 120hz screen; going to go for the Pro).
Yeah, I have a personal m2max. The only thing that might get me to upgrade to the m4 is just being able to hand this laptop down to my sister or my parents for whom it is severe overkill for but they will use it for like 10 more years.
Why would you need onboard mobile? It’s 2 clicks to trigger a mobile hotspot from your iPhone and there are very cheap LTE dongles on eBay. Not sure how much service would cost, most of us have reasonable download caps on our mobile plans. The dongles have better data plans than phones.
Why wouldn’t I want onboard cellular connection instead of having to be dependent on the more finicky and less reliable Hotspot connection, hurting my ability to use my phone freely, and burning both my laptop and my phone’s batteries at the same time.
Besides, having a cellular modem also allows you to tap into both WiFi and Cellular seamlessly like your phone does to make your overall connection much more reliable.
Convenience, security, and power-savings. I currently also use a Thinkpad X13s with onboard 5G and it's nice to not have to screw with it when you want connectivity.
On my Verizon plan (Unlimited Ultimate), I qualify for two 'connected devices' to be discounted. My Thinkpad is $10/mo extra on my account for unlimited LTE. I'm not a heavy data user by any means and this works out well for me.
Yeah, they trash idle/sleep battery life—or, at least, used to, back when I had access to lots of differently-configured iPads for my job—so you don’t want it on there unless you really need it.
1) Apple releases incremental upgrades! Why won't they make huge strides every year so I can upgrade!
2) People who upgrade every year are sheeps!
3) Apple support devices for longer than Android, that's nice! (yes, not Windows though).
4) God, why do their benchmarks compare devices that are 3-5y old?!
Apple is marketing to people who have devices that are old, because they are old.
"Hey, you noticed things are slow? Well, this thing is a lot faster" is pretty good marketing if it's true, nobody except the very wealthy are dropping thousands of euros/dollars on a new device for 10% performance gains, however if it's twenty-three times the performance of the Mac I currently own? Maybe it's enough to convince me or someone like my Mum to splurge on a new device.
Maybe my current Mac is not "good enough" anymore when 23x is the number on the box if I buy new.
It's fair to compare with devices that you expect actual people to actually upgrade from, there's a lot of Intel macbook airs in the field.
Since graduating from college in 1993, working in the graphic design industry full-time through 2019, I had two brand-new Macs (a PowerMac G3/800MHz, and a G5), the balance were hand-me-downs from other employees --- the G5 in particular was especially long-lasting, though ultimately it was supplemented by an Intel iMac.
Each year when Apple came out with new machines, we would make a game of putting together a dream machine --- ages ago, that could easily hit 6 figures, these days, well, a fully-configured Mac Studio is $14,099 and a Pro Display w/ stand and nano texture adds $6,998 or so.
> these days, well, a fully-configured Mac Studio is $14,099
Not surprising considering the CPU in the fastest "desktop" Mac before today was slower than an old Intel chips you can buy for ~$350 (e.g. the 14700k).
TBH, for non-tech folks that upgrade cycle has likely stretched a good bit beyond 3-5 years. 3-5 was the norm 10 years ago, but I’d wager needs-driven upgrades, opposed to marketing driven, are closer to 7-10 years outside of obvious niches.
Sample size one: My spouse is using either a 2013 MBA and wants to upgrade, mostly b/c the enshitification of web sites. Basic productivity was okay-ish for her work (document creation, pdfs, spreadsheets, etc), but even Gmail now suffers with more than a tab.
Edit: thinking more, I don’t know if I agree with myself here.
> Apple is marketing to people who have devices that are old, because they are old.
It still makes claims like that arbitrary and meaningless. What does "23x faster" even mean, it's not like there are that many people who are upgrading from an Intel MBA yet are also fulltime Cinebench/etc. testers.
> It's fair to compare
Well yes. It's reasonably fair (realistically its not like any of those people this is targeted at would feel a difference between 10x, 15x or 30x) and obviously smart.
The point is that benchmark is pretty useless and likely does not line up to what a user that is still running a intel air would expect the word "faster" even means.
When normal users are thinking "faster" they are really thinking about snappiness/responsiveness, not number crunching.
Those benchmarks seem to be more GPU based as well. e.g. something like Geekbench (not that it's necessarily that representative either) is just 2-3x faster.
Well yeah, I understand that this is based on some specific benchmark. Yet it's still some random arbitrary number effectively picked to mislead consumers.
Especially when for the M1 (2x faster) they decided to use an entirely different Photoshop benchmark YET they they still show it alongside the 23x for the Pixelmator one (presumably the M4 is NOT 2x faster than the M1 there..).
That's just objectively slimy (even if mostly harmless) marketing...
Also presumably Pixelmator's "Super Resolution" and Photoshop's "radial blur, content aware scale, diffuse, find edges" are also mostly GPU bound these days? Which again.. might not be the best indicator for "performance" for most consumers.
Edit: Looking at some more general benchmarks the the i7 (I7-1060NG7) from the last Intel MBA is "only" 4x (Geekbench MT), ~2.7x (Single-Core) or 2x (Cinebench single core) slower than the M4. Picking some highly specific "benchmark" that's several times higher than that is just dishonest.
Depends what you mean by 'faster' ... I wouldn't be surprised if the AGC was more responsive (faster response on the screen to user input) than a modern computer. Early computers were often quite snappy.
Considering that a modern Ryzen is 1375 times faster than a VAXstation 4000/60, and a VAXstation 4000/60 is around 1280 times faster, at least in clock, than an AGC, that would mean the M4 would need to be about 5.6 times faster than that modern Ryzen.
Hmmm... The M4 might be ten million times faster than the AGC, depending on the instructions per clock of the AGC and the VAXstation 4000/60 with which we're comparing it.
I've got an M1 Air and there's still no really compelling reason to upgrade. MagSafe and a nicer camera don't really justify it, especially when Continuity Camera is better than on the M1 or M4.
As I said in another comment, probably the benchmark is done just using some hardware instruction that didn't exist on those models and gets compiled to several instructions (possibly by a very very old compiler, while we're at it) vs something handwritten in assembly for the purpose of one specific benchmark.
Does this mean it's 23x faster for normal workloads? Nah.
Apple when they were pumping clang were also claiming that binaries produced with clang were much faster than those made with gcc. This was because they used a 15 years old version of gcc that didn't have any vector instructions (because they didn't exist at the time) and benchmarking using some code that was solely doing vector stuff.
Haha. Well, I guess it kind of makes sense in some way, Apple doesn’t want to say anything negative about any generation of “M” processor, maybe?
Up to 23x faster. Of course, the fastest Intel MacBook Air is pretty old. But 23X is pretty crazy, right? I wonder what they are comparing against. Int-8 matrix multiplications or something else that’s gotten acceleration lately, maybe?
That’s roughly the Air I have still. I hate using it (prior to recently adding the cooler shim mod, it would thermal throttle constantly) but between a Hackintosh and my work Mac I haven’t felt the need to upgrade. I think sometime in this M4/M5 gen is when I’ll pull the trigger and retire the Hackintosh to gaming rig only status.
I don't think it's silly to state. That message is probably for intel macbook air users who may be considering an upgrade.
(Anyway, I just ordered one for my wife, a soon-to-be-ex-intel-mac user. She'll probably be pretty happy about this, especially since she doesn't have an intel air as powerful as that one.)
People don't upgrade every year. I still have an Intel MacBook Pro (2020 I think?) that I don't plan on upgrading anytime soon because it still works great.
And the benchmark is probably jut using one hw instruction that didn't exist on that model and now exists, and is not representative of anything at all.
I love how even fair and justifiable critique of Apple needs to be hedged with the "Apple is great" prefix, such is the terror of the Apple downvote mafia on HN.
/typed from my Macbook Pro M4 — Love Apple — This is great!
The first thing I noticed in all of these announcements is that every main comparison is against M1. Why are they comparing with hardware 2-3 generations ago? I don't care whether my Intel i9 has 50x the performance of a Pentium processor from the 90s, it seems like a disingenuous attempt to make the numbers as high as possible.
Maybe, but I disengaged partway through (right after “I’m not going to bury the lede” and seeing there was a bunch more engagement filler immediately after, burying the lede). I will not read prose written like this.
If the article was compellingly written I guess I would have gotten to it. But the author is a bad writer and uses tricks reminiscent of “read this list, item #9 will shock you!”
No thanks.
I opened another article someone posted by the same author and now that I know they write like this, u couldn’t make it through the first paragraph. Absolute trash.
Complete hardware + software setup for running Deepseek-R1 locally. The actual model, no distillations, and Q8 quantization for full quality. Total cost, $6,000. All download and part links below:
Motherboard: Gigabyte MZ73-LM0 or MZ73-LM1. We want 2 EPYC sockets to get a massive 24 channels of DDR5 RAM to max out that memory size and bandwidth. https://t.co/GCYsoYaKvZ
CPU: 2x any AMD EPYC 9004 or 9005 CPU. LLM generation is bottlenecked by memory bandwidth, so you don't need a top-end one. Get the 9115 or even the 9015 if you really want to cut costs https://t.co/TkbfSFBioq
RAM: This is the big one. We are going to need 768GB (to fit the model) across 24 RAM channels (to get the bandwidth to run it fast enough). That means 24 x 32GB DDR5-RDIMM modules. Example kits: https://t.co/pJDnjxnfjghttps://t.co/ULXQen6TEc
Case: You can fit this in a standard tower case, but make sure it has screw mounts for a full server motherboard, which most consumer cases won't. The Enthoo Pro 2 Server will take this motherboard: https://t.co/m1KoTor49h
PSU: The power use of this system is surprisingly low! (<400W) However, you will need lots of CPU power cables for 2 EPYC CPUs. The Corsair HX1000i has enough, but you might be able to find a cheaper option: https://t.co/y6ug3LKd2k
Heatsink: This is a tricky bit. AMD EPYC is socket SP5, and most heatsinks for SP5 assume you have a 2U/4U server blade, which we don't for this build. You probably have to go to Ebay/Aliexpress for this. I can vouch for this one: https://t.co/51cUykOuWG
And if you find the fans that come with that heatsink noisy, replacing with 1 or 2 of these per heatsink instead will be efficient and whisper-quiet: https://t.co/CaEwtoxRZj
And finally, the SSD: Any 1TB or larger SSD that can fit R1 is fine. I recommend NVMe, just because you'll have to copy 700GB into RAM when you start the model, lol. No link here, if you got this far I assume you can find one yourself!
And that's your system! Put it all together and throw Linux on it. Also, an important tip: Go into the BIOS and set the number of NUMA groups to 0. This will ensure that every layer of the model is interleaved across all RAM chips, doubling our throughput. Don't forget!
Next, the model. Time to download 700 gigabytes of weights from @huggingface! Grab every file in the Q8_0 folder here: https://t.co/9ni1Miw73O
Believe it or not, you're almost done. There are more elegant ways to set it up, but for a quick demo, just do this. llama-cli -m ./DeepSeek-R1.Q8_0-00001-of-00015.gguf --temp 0.6 -no-cnv -c 16384 -p "<｜User｜>How many Rs are there in strawberry?<｜Assistant｜>"
If all goes well, you should witness a short load period followed by the stream of consciousness as a state-of-the-art local LLM begins to ponder your question:
And once it passes that test, just use llama-server to host the model and pass requests in from your other software. You now have frontier-level intelligence hosted entirely on your local machine, all open-source and free to use!
And if you got this far: Yes, there's no GPU in this build! If you want to host on GPU for faster generation speed, you can! You'll just lose a lot of quality from quantization, or if you want Q8 you'll need >700GB of GPU memory, which will probably cost $100k+
I'd assume that the existing llama.cpp ability to split layers out to the GPU still applies, so you could have some fraction in VRAM and speed up those layers.
The memory bandwidth might be an issue, and it would be a pretty small percentage of the model, but I'd guess the speedup would be apparent.
Maybe not worth the few thousand for the card + more power/cooling/space, of course.
The former is a reason why there actually is a lot of research into this already/couple drugs approved (cancer is usually very profitable), the latter is why all drugs, not just these, need extensive clinical trials and strong regulation. Because it's measles virus and herpes virus etc etc, and often genetically engineered, so the safety risks are ... obvious
