Can Intel finally beat ARM on performance per Watt?
hackaday.com
hackaday.com
https://www.youtube.com/watch?v=A2B7oI0FYqo
and post:
https://www.jeffgeerling.com/blog/2026/excited-for-intel-eff...
He's basing this on his own "super computing benchmark" measuring only FP64 performance. Meanwhile, any normal person using these $600 laptops will never run FP64 supercomputing tasks on the CPU.
He also only showed perf/watt in this one benchmark but skipped the fact that Apple Silicon is significantly faster.
Obviously, all this efficiency is on the low end—the laptop only achieves a little over 120 Gflops, while Apple's faster M3 and M4 Macs are pushing through up to 1.2 Teraflops.
... if it was within orders of magnitude of being true, perhaps.
Further, I'm basing my opinion on a number of benchmarks and power measurements besides just HPL. I just favor that because it's one of the few that push a CPU hard, fills system memory, and can't easily be gamed/made irrelevant with new architecture.
The main point of my blog post is to say this is the first Intel chip that's even punched in the same weight class as any Apple Silicon, wrt efficiency. It quite surprised me.
"Intel just matched Apple Silicon. Seriously."
If your headline was being honest, it'd say "Apple A18 Pro is still 3-4x more efficient and significantly faster than Intel's low end chip in common tasks".
If this isn't an undisclosed sponsored review, then it's just misleading clickbait for Youtube views. It's clearly working for you.
You can do better.
For technical readers, I write my blog posts. I don't have to (I lose money hosting that thing and all it's mostly just free training data for Anthropic, OpenAI, et all), but I enjoy writing, so I do it anyway. If you don't like the YT game, please subscribe to the blog's RSS feed. I much prefer text over video.
You can make an even more outlandish title and get even more views. Doesn't make it right.
Here's a few of them:
- Humane AI Pin Review: The Worst Product I've Ever Reviewed
- I Tested the World's Craziest Smartphone
- Do NOT Buy These Fake Tech Products!
- This Is Not a Normal Laptop...
- Unboxing the $10,000 Mystery Tech Box
- Don't Make This PC Building Mistake
- This Desk Accessory Changes Everything
- I Didn't Expect This Phone to Be This Good
- I Bought a Mystery Tech Box from Amazon
- Testing the Craziest Cheap Tech on Temu
Jeff Geerling's videos have fine titles.
Declaring that Jeff can't be a serious reviewer because he uses clickbait is kinda silly. Have you seen the state of journalism? Somehow the tabloids are frequently less clickbaity.
That’s not “the same amount of people were shown the video and 2x the people clicked in”. The algo is always skewing.
Just because lots of people are doing something that sucks does not make it OK.
It's not kicking dogs. Or, if it is kicking dogs, society has been doing it for at least as long as publishable media has been a thing, probably longer with the likes of bards.
This is ridiculous
A good hook does not have to be clickbait.
In my experience, when a person struggles to come up with a good hook, they write clickbait.
I don’t think it’s wrong to use clickbait, but come on, own up to it. You consciously chose to “let [YouTube] pick it.”
If you don't like it, either install DeArrow or subscribe to my blog, where I tone it down for the more technical audience. (Or block/ignore my content, no skin off my back!)
Even your blog post does not tell the truth nor does it have highlights of benchmarks that represent how people will actually use a $600 laptop.
The people who are buying these laptops are likely not very technical and it's your job to guide them. They wouldn't know that an FP64 test on the $600 laptop is near useless - yet you base the entire conclusion on it.
My problem is that both the clickbait title and the conclusion are straight up wrong and misleading.
Even clickbait titles on Youtube will often have the correct conclusion. This is not the case here.
You have an opportunity to be continue to be taken seriously like you have been so far.
Surely there must be some way to avoid mis-leading clickbait titles.
Separately, asking your fellow community-members to not bother, and that too with an unaffected shrug, just kills the conversation we would want to have with you.
Then... people doing scam or misinformation can also say that
- ignore it.
No. We like your content.
While you may get more clicks etc at the moment - eventually you will lose subscribers.
You could have made the title at least like
- Intel tries hard to win against apple silicon - Almost there but not yet...
I think it’s pretty normal that you are incentivized by YT. Too bad the internet favors screaming and yelling. Pretty clear line you draw between writing and video.
An absence of open expression of disapproval is not the same as approval - people may just be avoiding dogpiling a member of our community.
I would add that people have a right to express disagreement or disappointment in a respectful manner in accordance with the rules of this moderated private space, and we have not had moderation/interventions suggesting otherwise. The criticisms are reasonable, to my mind at least, and probably ought to be taken as a massive hint that something is wrong. We are less than a day in, and it's not too late to fix things.
I am trying not to be too critical, as I have enjoyed previous content from this channel and I wish the poster well for the future.
“Did Jeff […] get paid to post misleading headlines”
“I can't believe this crap is getting upvoted. Hopefully this guy loses all his credibility going forward. This kind of post is seriously misleading and smells like like an undisclosed sponsor review.”
Ad hominem, cursing, opinion as fact with a barrage of replies.
That could have gone three steps up in politeness without losing any of the criticism.
All of us have watched enough Jeff Geerling videos to know that he's not a secret spy for Intel or the Manchurian Candidate for Apple haters. He just wants to try out new technology with the curiosity that most people have, and apparently that's enough to outrage some people. Accusing him of avoiding disclosure is against HN's guidelines and generally an absentminded accusation to level. Methinks somebody doth protest too much...
I think you’re being mid leading here Jeff. You almost make it sound like YT actually came up with the title.
For anyone that is not a YouTuber, Jeff wrote three different titles himself, then YouTube ran an A/B test and just kept the one that resulted in the most watch time.
Jeff consciously chose to write a click bait title, he could equally have chosen not to.
My understanding is the Wildcat lake 304 is that replacement, and it's been "released" for a while. It can be configured down to 10W vs 6W for the N100. Both draw about 7W from the wall when idle. The 304 is about 80% faster for the same power draw, has a better GPU in includes a 15 TOPS NPU.
Despite the 304 being released, it's not easily available yet.
I went through the whole thread of this conversion, and as the old adage goes, don't feed the trolls.
> this was not sponsored in any way
Aren't you contradicting yourself here? For what purpose was the laptop loaned to you, and by whom? I'd consider a review loan a kind of sponsorship, even if you weren't paid to come to a specific conclusion.
You should _want_ that to happen, otherwise the only people who can do reviews will be the few who can afford to spend thousands for hardware they don't want or need.
So I'm not sure what your original post about smaller cores is all about considering that the A18 Pro CPU is almost certainly smaller in actual die size.
Maybe you can articulate what you're arguing from the start?
If you put 4X of the same cores in the same system then the CPU uses 60W, the WiFi etc. still use 10W, so the whole system uses 70W. It scores 4000 points and the performance per watt is 57 points per watt. You got four times the multi-thread performance and performance per watt went up.
So where is the reviewer worth their salt you're talking about?
Many top quality CPU reviewers also do this. Chinese reviewer Geekerwan is probably the best at it. In the past, written reviews like Anandtech will always do it. Today, anyone with video editing software thinks they can review CPU performance.
Ever since Youtube, there have been plenty of low quality reviewers that mislead people. I didn’t know who this Jeff guy was but now he’s on my list as one. There are still some who do proper reviews.
That's not actually isolating CPU power. To begin with, the CPU itself has idle power draw. Measuring that way would favor a CPU that has worse power management because it would be excluding the draw the CPU itself fails to shed at idle.
It also gets it wrong the other way. The SSD etc. is using however much at idle but that isn't the amount it uses when running the benchmark. Many of them don't have very fine grained power management and will draw ~100mW at idle but consume close to full power under even moderate load.
The most accurate is reading from the wall. Thats what Notebookcheck does.
For all intents and purposes, it is good enough to get a read on CPU efficiency. Furthermore, if you are comparing equivalent systems, it is good enough to compare the two.
And what results from this when you compare system with a 15W CPU to one with a 60W CPU on performance per watt when both systems have an additional 10W draw from non-CPU components?
All we care about here is performance for the CPU, performance per watt, and performance per die area.
If it's utter Apple shill bullcrap like Geekbench, caught 100x rewriting results in favor of Apple (for what rea$on it'$ not hard to figure out), it's perfectly OK because it supports what you believe in.
So agreed, this is extremely misleading.
That Intel chip is competing against the A series in the Neo. To get to M you need to spend at least 50% more (and there are considerably better Intel chip in laptops at that price range as well)
Also benchmark performance hardly matters for people buying laptops like this. It just has to be fast enough and not lag when running a handful of apps and have decent battery life (which presumably is the case here)
Also benchmark performance hardly matters for people buying laptops like this. It just has to be fast enough and not lag when running a handful of apps and have decent battery life (which presumably is the case here)
The A18 Pro is 50% faster in ST and maintains the ST while being fanless. The Intel chip will throttle almost immediately. The difference in efficiency is 3-4x.
> no headphone jack
No need for wireless and battery, just not the plug you were expecting.
Selling wired headphones to wireless-iPhone users for wired-MacBooks could (arguably, maybe) create Apple sponsored interoperability dissonance.
MacBook charging mats and wireless Thunderbolt coming in early 2018... /h
Framework laptops have another decent strategy, where you've got on-motherboard ports that have relatively few usage cycles, and are protected from most forces because the easily-replaceable custom-dongle-squares are held in place quite securely. They take the damage, protecting the motherboard. And as long as the motherboard ports are still good, you can even forego the modular thing and just plug in directly if needed.
Can you explain what do you mean by that? It is quite rare to see broken trrs jack on a laptop, at least I have never experienced one, nor on walkman or anything getting heavy plug/unplug use.
I thought it is an issue in reducing the cost to manufacture the case or something.
Doesn't this make including the jack even less justifiable?
So yeah, they break, but they are (or used to be?) quite cheap to fix (and this was one of the super thin 13" xps)
I had a slim XPS13 and IIRC the headphone jack was on it's own little board. That makes it quick/easy/cheap to repair if you know what you're doing / have the service manual / are comfortable taking it apart.
Not all things are designed that way; sometimes the jack is right on the main board and most people that buy dell laptops are not going to DIY through ali express; they're going to use up phone support time, require a shipping label, technician time ... etc.
You take care of your hardware, even during "heavy use," many people absolutely do not - and are downright abusive.
Many of these parts exhibit partial or intermittent failure modes where it's more expensive and difficult to determine warranty eligibility than to just fix the problem. Peanut butter in the jack, loose solder joints, whatever.
Furthermore, the warranty determination isn't free, either: in addition to the repair itself, some (likely multiple) employee's time was spent determining whether or not an issue existed at all, whether or not the issue was covered, whether it could be fixed, and facilitating the whole transaction.
Repairs also don't result in 100% yield. Depending on your company policy you might be eating the whole unit because of a mishap or just plain stubborn components/physical realities.
Repair fees don't cover all of the organizational moving parts required to actually carry out repairs.
Nothing for the desktop in the last couple of decades is going to struggle with that.
> I insist on a wired headset
usb audio generally works great, and because it's powered, you often get better audio quality (with microphone arrays for anc)
and even neat features like system volume integration with buttons and ANC
i have a yealink uh48 but you can get almost-as-good functionality for a lot less (or pay more for even better!)
Does it have to be wired via headphones jack?
Of all the parts on electronics I have owned i would put headphone jack just behind power supply for least reliable and that it was set to override digital settings when something was connected felt like a loss of control.
I am very forgetful, wireless has been a godsend by comparison and I don’t get hurt with cable snags. I’d pay extra to remove it before shipping if it were an option.
I bought a pair of AirPods and have never looked back. If you have a pair of wired studio headphones, there are a million USB-C adapters you can buy to have this functionality. Or migrate eventually to a set of USB-C native headphones. This is such a non-issue.
> i would put headphone jack just behind power supply for least reliable
I've never broken a headphone jack, or known many people who have, honestly. Your experience sounds pretty unique.
I suspect way more people have had issues with Bluetooth connectivity, though thankfully, that's improved over time.
Cheap wired headphones are the default choice in the classroom, in my experience.
Bluetooth headphones that need charging aren’t really what you want.
It’s easy enough to get cheap USB headphone adapter, but still, I t would nice if it was built in.
Apple knows what they are doing sometimes it seems! :-)
1) a few cents of components
2) the opportunity cost of selling you a high margin dongle
3) minus the risk of a lost sale if you buy a competitor
Did it need to be that thin? Heck no. But it is, and yeah there is absolutely no room for one because of that decision.
The Apple Neo was still 2x faster for graphics and around 1.4x faster for single core cpu. And remember the Apple Neo is running an iPhone CPU that is much slower than the M series.
It is strange that this chip is so crazy efficient though. I would love to know the details of exactly why it is? It seems like some new tech or process?
I do want 1.5x battery life for free out of my next laptop. Who wouldn’t?
The exact design of the cores, how they’re scheduled, the way background stuff is handled on MacOS. All of it. The chips are made on a fantastic process too, but that alone isn’t it. For example they’ve been putting the ability to drop refresh rate really low when nothing changes to more and more of their machines.
It could be that the Intel chip is close, IDK.
But add in other motherboard components, the screen, the OS, the software and system libraries, all are optimized together to hit those numbers. And any of those alone or in combination could be part of the differences.
The Intel chip holds its own for multicore in less bursty workloads, where the fans help.
They do in audio processing. ;)
* Web rendering and almost all web apps (see the JS number type)
* Most things in a game not done on the GPU
* Audio processing
* Most GUI processing/rendering
* Every miscellaneous/scalar calculation these days
Believe it or not, on a CPU this adds up to a lot more than the work you get doing AI processing.
You have a point about games, but the CPU is seldom the limiting factor.
Very few people are handling raw audio. For the rest, codec hardware beats everything else by a massive margin.
Being good at floats matters, but even float-heavy code generally has way more integers than floats making integer performance king.
Is this actually a thing? Macbook Pro goes up to 100 Wh and almost nobody goes above that because it's the highest capacity battery you don't need special permission to have in a carry on at the airport.
> But add in other motherboard components, the screen, the OS, the software and system libraries, all are optimized together to hit those numbers.
I always find this argument weird because there is so little actual "together" having anything to do with it. Are TSMC's current process nodes efficient? Yes. Can you save power by dropping the refresh rate? Yes. Do the teams doing these things need to coordinate with each other somehow rather than each doing their part independently? Not really.
We really need to get over generalization like "phone CPU". Times have changed. Phone or not really isn't the focus.
Per core the A series (cough phone) isn't that much slower than equivalent M series. It's literally the same family just with some extra features, more power, etc.
It’s good Intel can match anything while staying cool and efficient. That’s a great step. But I’d like to see them match the M5.
The processor is this one: Intel Core 5 processor 320 https://www.intel.com/content/www/us/en/products/sku/246018/... in the Wildcat Lake series.
Jokes aside, I don't think the engineering incentive case can be overstated. Apple has been working to squeeze every milliwatt of energy efficiency out of their chips since they started making their own iphone chips in 2010. Their chip is a win if it lets the user have phone battery longer.
Intel and AMD have never been on a power budget. Their chips are a win if they can crank the benchmark performance numbers just a little higher, electricity cost be damned.
This is what you get when you spend 15 years optimizing for energy efficiency per $ instead of raw power per $: a workstation cpu that doesn't even need a fan.
Its only strange if you believe Apples lies.
The truth is, architecture doesn't matter. Single core performance, Apple was always on par with the best chips. Compute is compute, under the hood moving data costs energy, an you can't get away from that.
The reason for "efficiency" of Apple silicon it is because they spent a long time basically optimizing their chips for their specific hardware while running specific programs. Its like building an ASIC to specifically watch youtube vides. When it came to things that are not pipelinable (i.e a lot of branching), the efficieny/battery life statistics flew out the window.
A couple of years ago I got tired of dealing with OpenWRT and various Tomato based firmware or hunting down compatibility between ARM and MIPS based routers. I finally took the plunge and bought myself a mini pc with an Intel N100.
The thing has 4 cores (no SMT), a 6W TDP and boosts to 3.4 GHz and can support 16GB of ram. The mini pc I bought cost me $200, includes the processor as well 4x Intel 10Gbps Ethernet ports. At 6W, it needs no fan, and the whole chassis is metal and specifically designed to act as a heatsink, which means 0 noise and 0 moving parts.
I ran a power virus on it for a few hours to test the thermals, and on a warm day in Melbourne, after hours of having all 4 cores running full throttle, CPU temperature never exceeded 90 C and the chassis barely hit 55 C.
The best thing about it? Full upstream support for any Linux distro and FreeBSD. No more wondering if this particular Broadcom chipset needs a weird binary blob, no more being hostage to an OEM and having to run old forked kernels.
I installed OPNSense on it, VESA mounted it behind our TV, and it hasn’t been seen or heard since. Not only has it been the most reliable router I’ve ever had at home, it’s providing various other services such as DNS server, apt cache, WireGuard server and client etc for my homelab.
I’ve been waiting for ARM to get here since ARMv8 was announced in 2011, but it just hasn’t happened.
Long live x86.
It's also super easy to spec out your own devices in this regime on x86. There are mini itx boards with n100/n305 with pcie slots so you can add your own cards, and the n100 has a surprisingly number of available lanes for its small size.
There just seems to be no viable "just works" solution in the below 10 or 20W range apart from raspberry pi, and Intel n100/n305.
Instead of Raspi I've got a N100 uATX mainboard with N100 (120 EUR). Fortunately it does work with normal 32GB DIMM (55 EUR, those were the days!). But I'm still missing a pico-PSU, which can be powered directly from 220V (for reasonable costs).
There were a series of other benchmarks too, but the Hackaday article only highlights HPL as that was the benchmark that changed my perspective on Intel a bit. This was the first X86 chip to trade blows in efficiency (and idle power draw) with Apple's M3/M4 (I don't have any M5 to test with right now).
I know some have mentioned the $699/$599 price could also change a bit in the US come November. The internal assembly and parts seem like they'd cost Dell a little more than Apple's cost building the Neo, so maybe in the US they're selling them closer to at cost, and subsidizing it with the Windows exclusivity and price elsewhere? Not sure, but it stinks the price only applies here. It's a good deal at $699. Harder to justify at $1,000+.
At least for Europe, and more specifically France, laptop use in classrooms in high school, is still not a thing. So that would explain why they don't do it everywhere the same and the huge price difference.
MacBooks simply have the best value proposition. Compared to the competition at similar price point they are much faster, have much better build quality, better displays, better battery life, less fan noise, etc. Also much higher resale value.
Macs are great if you want a Mac. If you want a flexible workstation, a gaming PC, a retrocomputing build, an inference server, a programming environment, or a headless VPS, macOS doesn't even attempt to cover your use-case. It's the reason why the Mac install base is still vying for pole position after 40 years of "you're holding it wrong" to their users.
My only reservation wrt the core of it is that it tests the cpus on a matrix operations task, and power efficiency in this case refers specifically to doing matrix operations. This is fine if that's the kind of task one wants to optimise for, but it does not necessarily translate to general energy efficiency for a broader spectrum of tasks that most people may do most of the time.
And thus, if one wants to optimise for matrix operations, wouldn't it make sense to use apple's accelerate framework instead of blis/openblas (I think you used blis?)? I compiled hpl to use accelerate [0] on my m1 max (macbook), and ran it with settings almost identical to the original m1 max settings [1]. I get 422 Gflops vs 264 Gflops in m1 max mac studio (docker) [2]. My average draw (using a wall power meter, but not interfaced with computer to keep a log, just me looking at it) is around 55-57W (probably less because it peaked at 57W but the last half was getting around 52W). With the most conservative estimate of 57W, this gave a 7.4 Gflops/W which is second place, just below the m4 mac mini in the benchmark table. The m1 max in the original did 4 Gflops/W (at 66W).
So tbh I do not really trust the conclusion of the benchmark page, except if I miss sth. It is quite nice, organised work anyway.
[0] https://github.com/tycho/hpl/blob/master/setup/Make.MacOSX_A...
[1] I just fixed the p:q ratio, which was prob not really meant to be 1:10 there, I assume, also judging from the settings in the other tests. I put p=2,q=4. Also I was unable to put more than 8 processes in total.
[2] The original m1 max benchmark: https://github.com/geerlingguy/top500-benchmark/issues/4
- Performance
- Performance/watt
- Die area
These are the 3 things determine a competitive chip. You can increase perf/watt but decrease performance. You can increase performance by giving the chip a lot of cache and transistors but it's more costly to manufacture.
The problem is that Intel chips are losing in all 3 areas simultaneously against Apple chips. They are slower, less efficient, and the CPU cores are bigger physically.
If you don't want to run Windows, is there even a choice? My issue now with the non-Apple computers if you want to run Linux it's really hard to predict how awful Chromium/Firefox will be. My work Dell Laptop is only a 3 year old CPU with integrated GPU (and unused Nvidia GPU) but Chromium and Firefox are both slow as shit just loading pages like Slack and Outlook. Even resizing the browser windows seems 4-5x times slower than my M3 Macbook Air.
Maybe this is strictly a Linux issue but it hardly matters when the Windows ecosystem is pretty stagnant, full of junk-ware, and software development is much, much better on Linux/MacOS.
I tried using the flatpak Chromium in case it was some NixOS build optimization oddities but didn't see much of a difference.
Intel keeps trying to say their integrated graphics are finally decently fast, but that's only a select few special models. Their high-volume chips continue to have integrated GPUs that are barely fast enough for web browsing.
Yep, and the linked article is also kind of interesting.
Shouldn't we be concerned that the company instead of aiming for the leap is producing legacy hardware destined for landfill?
Wake me up when Intel produces something groundbreaking.
Like, we're comparing a new Dell to a Mac released only 4 months ago. How did you feel about the Mac Neo then, and now? Is it also legacy hardware destined for the landfill?
I think the bigger news is Intel's made a chip that isn't just pushing clocks to hit some performance number, they've actually tuned clocks for efficiency... which also leads to much better battery life.
And the fan on the system in the original article[1] also allows the Intel system to run substantially faster for longer tasks (where the Neo thermally saturates and has to stay throttled).
(The iGPU is another matter... sadly still a bit lacking.)
I'm hoping Intel makes a follow-up to the N100-series chips with just four of their newer low power E-cores.
[1] https://www.jeffgeerling.com/blog/2026/excited-for-intel-eff...
Efficiency and performance is another.
We've long known that pushing clock speeds results in diminishing efficiency, so it's about time we see Intel stop pushing the clocks (and power) so hard that the chip melts down and the battery drains quickly.
If they've managed to fix their process then they don't have to do that anymore.
It’s great that they are making progress. Competition is good. They certainly need good news right now. I have exactly one intel cpu in my house in a NUC I got off eBay and it’s about to be replaced by my M4 mini. My other x86 are Zen (threadripper, 5800x3d, 5950s).
Do some tests pre and post upgrade. It’s ridiculous how efficient the mini is. My power usage is about 1/12 what it was with a Nuc.
Also, I think the battery life most people are practically interested in is more like running (close to) idle and less like the number of Watts spent when running maxed out, and I wouldn’t be surprised if Arm chips still win on that metric…
It’s good Intel got to this point, but you can’t declare them tied with Apple. They qualified for the real race, which us great, but they're not going to win with that chip.
Moreover, from tfa:
> This Dell also beats out both M4 and M3 Mac Studios, only failing to the M4 Mac Mini at 7.57 Gflops/W.
It’s good that this performs good and can go low power, but the big question to me is whether they can scale it up without it becoming a power sucking furnace.
There’s no other comment somewhere in here, where someone is pointing out that that was in FP64, which is not something most people use. Especially people who are going to buy a computer like this.
I suspect for me personally, integer use matters far, far, far more than anything else.
It's used in audio processing. But large DAW sessions need something beefier than both tested devices (Macbook Neo, XPS).
But as you said, if your workflow really needs FP64 you’re probably not reaching for a MacBook Neo or this Intel thing.
We still get these comments. Now about AI. China can only distill and are still years behind. Well well...
I can envision a day when Apple properly supports Linux. I cannot envision a day when their OS doesn't suck.
Apple leads in IPC, perf/area in CPU, and raw performance. We already have TSMC N7, N5, N4 AMD vs Apple CPUs to compare.
My guess is that even the designs of the transistors and perhaps logic gates are IP blocks owned by TSMC and not whoever wants to get their chip fabricated.
The processor is a Intel Core 5 320 Wildcat lake, with 4 "low power efficient cores" and 2 non-hyperthreaded Performance-cores.
It's variant of Panther Lake which has some models with Arc GPUs: https://en.wikipedia.org/wiki/Panther_Lake_(microprocessor)
https://www.intel.com/content/www/us/en/products/sku/246018/...