most reviewers are hardware elitists who think every single person needs at least a core i7 (or ultra 7), 16gb ram, and 1tb storage. X Plus does 99% of what X Elite can do, but more efficient.
@@ASEM-1123Microsoft also sends out the highest spec models to reviewers. Alex bought the Plus model with his own money which is a rarity for YT reviewers.
@@AZisk Hi, I have a question. The surface with x plus comes with a 39w charger. What is the maximum wattage for charging via usb-c? Thanks in advance for your answer
Thank you for your continuous efforts in testing these units. I have many customers waiting me to port my software to ARM and you already highlighted many potential problems that would take ages to figure out.
You are making some incredible videos. Even when you’re talking about things that are slightly over my head, you explain in a way that makes perfect sense. I feel like I’m learning something every time i watch.
When using sorting algorithms as benchmarks, it would be prudent to lock in the seed of the vector that holds the out-of-order numbers. That way the algorithm performs an identically hard problem in all cases in terms of iterations needed to complete the operation, rather than relying on the wall-time required as the sole indicator of performance. P.S.: There's a running joke in the optimization community that BOGOSort is the best sorting algorithm, you just have to "high-roll" a good enough vector that is practically already sorted, highlighting the need for exactly reproducible problems.
As someone who uses remote desktop to connect to my PC in the office to work, this kind of laptop is perfect for that. Can't wait for even more options and maybe some price cuts
Running a program without optimizations turned on is really odd, even for debugging if you need performance you'd often at least use -O1 it might also explain why prism was somehow faster then running natively, the x64 code is just not that great and prism might be able to do some of the optimizations the compiler would have done if you turned them on
@@seeibe it's not about spending time, it's about passing the "-O2" flag when you compile. I guess compilation might be slightly slower, but not normally noticeable
you know what when i saw your content i was like "this is what i was looking for" but then when i saw your sub count on channel... it broke me. fr this king of content deserve at least 5 million subs.
I think Thunderbolt compatibility testing would be best performed with Thunderbolt-only docks/accessories (the venerable CalDigit TS3 Plus, for instance). If the dock is also designed to work fine over standard, non-Thunderbolt USB-C, we don't get a good idea of how well the laptop plays with Thunderbolt. The dock you linked can fall back to a standard USB connection (probably 10 Gbps) if Thunderbolt bandwidths are not available on the port. If you are testing with that OWC dock, it would be best to connect a Thunderbolt accessory to it and verify if it's achieving Thunderbolt levels of bandwidth.
Don't run very short tests like 5 seconds. CPU core might not fully wake up after working on very low frequency (idle state), and might not get to turbo boost frequencies yet.
This was a very cool one Alex. Thanks for making and share this content. As a true NET developer (XAML moving to Web, and increasingly to Linux for containers, and JVM for PDF generation) this was very insightful, practical and to the point. Thanks and cheers from Florida today.
Sadly the manufacturers are completely cheaping out on RAM... The CPUs might have enough power to compile bigger projects in a reasonable amount of time, but I'm missing the 128 GB RAM option to test it out...
128 GB RAM? Where the fuck do you work at bro, CERN, NASA? I think some of them have the option for 64 GB, but I find even 32 GB way too much. I have 16 on my fuckin DESKTOP which is used for everything (gaming, development, etc).
@@MadafakinRiowhy are you cussing so much 😅 Not many people need that much ram. People that run Huge LLM’s or people editing huge video projects or even music producers who use lots of sample instrument libraries. Not much people run 128 of Ram but it’s good that there’s an option to.
@@MadafakinRio 128GB is a lot for a laptop, but 16GB is mediocre. 32GB minimum for professional work. I am constantly maxing out my 16GB work laptop and having to close programs. I'm a software engineer. Need to get them to ship me out a 32GB model.
@@MadafakinRio Lets just say that I am involved in the development of medical imaging solutions. 128 GB is enough for compiling most of our projects. Testing needs much more, between 1 and 2 TB, depending on the input data.
These things are not bad machines and will only get better. Rocky start, but a short time later things will iron out and they will become critically hard to compete against.
Alex at 6:50 I would like to point out bad working conditions for schwarzenegger 2.0. It is trashed to the corner, really suffering and being ignored. Not cool man!!!! not cool.
I would like to see a long term testing scores for these computers after the daily system updates that are always require for Windows. Maybe monthly to see how performance degrades over time.
Hey sir, when you're benchmarking Intel Chip, please try it with Lenovo/Asus as Surface really doesn't have good Intel models! Lenovo 14th Gen laptops are quite quiet! Great comparisons tho!
Perhaps next time a storage intensive test, taking into account the price tag of having a machine with a storage of 2TB or 4 TB (which let apple look like a discounter offer?) The interesting time will come if snapdragon equivalents for The Pro and Max apple Silicon appear, currently there is only competition on MBA level
really appreciate these reviews. I love my surface pro 7 but its battery life is terrible. It still does everything I need it to for now but may end up getting one of these beautiful surface laptop 7 soon..
Maybe MSVC for ARM64 is less refined / optimized than for x64? Or maybe like @squadb3 said Prism may optimize unoptimized Debug code, because it's ultimately a JIT compiler.
You are mentioning that the multicore sort is faster on the X-elite, having 10 or 12 cores, compared to the Macbook Air's only 8 cores. But in reality, the Air only has 4, as in 4 performance cores compared to 10 or 12 performance cores on the X-elite. I wish reviewers would point out this detail.
moreover, E-cores are mainly only used for background tasks, and MacOS only schedules threads onto the E-cores when P-cores are fully saturated, and even then, it won't fully use the E-cores as they are reserved for the OS' maintenance processes. So Apple mentioning the total number or cores is misleading.
The Surface 6 with the Core Ultra 7 165H honestly sounds like a great Business laptop. None of the ARM crap, but with real performance compared to the wimpy i7-1355U we had previously
I may have missed it in the video - but in Windows, you must set to "best performance" power scheme. Otherwise, power management features will dramatically limit your performance. Interestingly, on the Qualcomm CPUs (similarly to M-arch on Macs) - if you stay on Best Performance while on battery - performance doesn't suffer. On Intel you can expect ~7% drop, on AMD a lot more. Using "Best Performance" while on battery will significantly hurt battery life though.
I'm loving my 32gb ARM Surface laptop. The only issue is Cisco VPN client, probably because they are investors in Intel. I also love me intel laptop as well so I'm good 🙂
Alex, all your videos comparing ARM devices are getting me hyped, although I still think they need to be cheaper for people to buy them. The great battery life and low fan noise in everyday use isn't a decisive factor because people have simply gotten used to it and see it as "meh." I think if Microsoft changes the screen on the "x plus" chip versions to one that isn't so reflective and isn't touchscreen (as a developer, I don't know why anyone would want a touchscreen on a laptop; they should get a tablet instead), it would lower the price and make it a better product. Otherwise, I don't see many people choosing these over the MBA M3, and soon they will be even cheaper when the M4s come out.
How about taking us though a SSD upgrade process? Using a Mac for so long I cant even remember all the steps e.g. download a ISO, make USB bootable, actually replacing the SSD, is it really easy? Install windows, does all the drivers work out of the box?
i have the huawei matebook 14 with intel core ultra 5 125h. it comes with a huawei pc manager. it lasts for 18hours in power saving mode. the app throttles the cpu but browsing and streaming experience is still great. but i doubt i can use that setting when i actually do real work. but at least it's good to have an option like that when you don't really need the cpus full power.
13:29 you probably compiled for x86. The maximum stack size is way smaller on x86 compared to x64, so you would need to compiler for x64 (and for timing also use -O3)
Having tried compiling different software with -O3 vs -O2 it seems it rarely produces much difference in runtime but increases compile time significantly.
@@lambdaprog O3 allows experimental optimizations that may change the behavior even of perfectly legal C code without undefined behavior, and in most cases will not be significantly faster than O2. In fact, it may be slower than O2.
that's it I'm going to buy one of this, especially when android studio support the ARM Version, i thought it wouldn't be too long, since they already have the ARM Version for Mac
I don't think they push this one any further, some of them are already discounted the Laptops just a few weeks after the release.. so they still had too many what nobody wanted..
Hi Alex, first of all i wanna say thanks to you because of your super helpful videos. And secondly i wanna ask you something, personally i 2, 3 times reinstall windows in a year, so can you make video about it on arm chips. Is it still easy to reinstall windows on arm chips like x64, or should we expect more problems? I hope you will read this!
To prevent C++ people to yell at you wrongly setting the flags, just use Rust and dont forget set itu to release build (I assume the windows would use native windows compiler and mac would probably use clang, thus this yup this is not apple to apple benchmark)
I bought the 15in Surface Laptop 7, really baffles me though that the battery is only 66Wh, while the 13.8in has a 54Wh battery. You’re telling me with an extra 19.2 square inches of space (based of chassis dimensions) they could only fit another 12Wh worth of battery? Considering the cost and how it’s supposed to be their top end device, they really should’ve crammed a 100Wh battery in like Apple does for the largest MacBook Pros. Would’ve given a 50% boost in battery life and helped with long term degradation.
12:39 You need to quick sort an array with same array for both builds. quick sort could just finished with fewer iteration on x64 build since you shuffled array on runtime based on source code available on 11:14
I have the base model Surface Laptop 7 with the Plus CPU (because that's pretty much the only sensible choice :P), and I was fully expecting games to be atrocious, based on the reviews I've seen. But actually everything I tried ran, and most games even ran well enough to properly play. I've tried Subnautica, System Shock remake, Scrap Mechanic, Prey, Skyrim SE, Grounded, Portal 1 and Portal 2, etc. Oddly enough Skyrim gave me the most trouble, because if I didn't use native resolution it started up with 3/4 of the rendered image being off-screen.
VMware Hoziron also totally messes up the system when you try to run the newest version. The laptop literally locks up after a few minutes, you can barely even shut it down. There are some older versions which seem to run okay, but still it's not reassuring if that's how you connect to your corporate work VM and you'd like to work from a Surface Laptop 7.
Hmm. I've never heard that quote from Tim Cook. So, the way Apple copies some features from Android is the sincerest form of flattery. 😂 I am eager to see some tests and reviews of the X1 ThinkPad Elite since ThinkPads have been my daily driver for quite a while.
Josh reviewed this and said the battery life on the surface pro 11 is not good. Battery life is the main reason I’m considering one. Now I’m on the fence.
I'm a Visual Studio C# developer and host my SAAS app on Azure. I can't decide between Surface Pro 10 for business with 64GB RAM (expensive!) or the maxed out Surface Pro 11. I would sure love to speed up my compile times! Currently takes ~1 minute on my Surface Pro 7 with 16GB. I don't really care about battery life -- I'm always plugged in. Or the price. The only thing I really care about compile times. Any thoughts, Alex? Great video.
It's simple. For apple get m3 base or pro, unless you need capabilities of max or ultra. Or wait for coming m4. For snapdragon, it's only matter for Xelite. You don't want to pay for that price? Don't go low ball and get XPlus, not worth it. And Microsoft "translation" layer is not refined like Apple, so stick to intel ones for production usage. And on one use their laptop without charging regularly or always plugged in. But there must be "content", isn't it?
You really need to turn on O3 for any software benchmarks you compile yourself. Production software will never release without optimizations enabled. I would guess this is why you saw Prism running better than native, assuming Prism does optimizations for ARM during the translation phase which your native binary missed out on. Not certain though. A 5 second test also isn't very helpful as it's just not enough time for the operating system's process scheduler to kick in and set your high throughput task as needing priority. At least 30 seconds would be better. You should also be more specific when making battery claims. It's been shown that while the Snapdragon laptops can last longer at low loads like looping a Netflix video which uses the chip's low-power dedicated decode hardware, they fail to maintain that power efficiency as loads increase, unlike M3 which has much better power efficiency as workload scales up. It's also been shown that some Snapdragon laptop models have different power budgets with some significantly throttling the CPU after a short period of high usage, which makes it look like the battery life is excellent when in reality the performance was nuked, which you might not notice in a hands-off automated test. Just Josh did some excellent performance analysis in this regard. I think for the average person the battery life will be good either way, but it's not really accurate to suggest they can outlast a Macbook with normal day-to-day use. I'm also curious about the memory bandwidth results, I wonder if running inside WSL is creating a bottleneck by running in a VM. Would be nice to use a native benchmark, might show better results. Though to be fair, most developers are probably running AI stuff inside WSL, so maybe it's a more useful benchmark even if its not the best it can do. Overall, really good video, and I appreciate taking the time to benchmark tasks aimed towards programmers. Not a ton of people out there other than Josh doing that.
Any chance you can tune the power settings ? Intel chips are known that can be fed with a lot of watts. And, unfortunately, many laptop manufacturers actually do that, then fiddle with their thumbs when people complain about battery life. I'm also curious how the difference is when the laptops are tuned to perform the same. Or simply... yeah, that should be much easier, a battery (pun intended) of tests at different power levels. Complete with the performance results. Maybe Intel can last just as long if it's set up at minimum values, like running at 800 MHz on the CPU. But it might be 3 times as slow. Or maybe instead of being 10% faster, it can be just as fast at half of the power draw it had initially and actually compete in battery life. We need MOAR tests! Thing is that the battery life can vary A LOT, really, A HUGE LOT, between one laptop and another. The motherboard design of the laptop matters. How many ports you have matters. What type of RAM you have matters. And lastly, like I said above, how the CPU is configured also matters. I don't remember the exact case, but, you can have 2 laptops with the same chip and one to be less efficient, less performant, at higher temperature all AT THE SAME TIME, just because one is properly designed and configured and one not.
I bought the 15 inch Surface Laptop (elite model), and also a Galaxy Book4 Pro 360 (155H model)... so torn on which one to keep. I think the Surface is more powerful, and I think the build quality is slightly better... but the Galaxy's 3k amoled display, ability to fold into a tablet, and pen input is so good... ugh. What do I do Alex? lol. Will I actually notice much of a performance boost with the X Elite over the 155H?
Love all your geeky tests. These are the things I do and appreciate there are others like me. I am specifically interested in the rendering compatibilities of Qualcomm chips. I use Resolve for video editing. I realize that Resolve ARM version is in beta but just how well does it work? Does hardware video acceleration work when exporting a video in AV1? How about Handbrake (which has an ARM version I believe)? Does it support hardware video decode/encode?
Was there any mention about mAh? battery sizes? as when I googled,.. MBA 13" M3 = 4380 mAh, Surface 7 Snapdragon X Elite = 5806 mAh... Surface 7 Snapdragon X Plus = 4755 mAh With that, I assume, both should last longer than the macbook, if not, that's bad. But also means that the Elite power efficiency is really bad if it looses against the Plus.
Where did you get those ? Also, unlike the smartphones, where you know the voltage (as 3.7 - 4.2, so roughly 4 volts), the battery capacity in ... everything else, is usually measured in Wh. Your numbers tell me nothing, since I don't know the voltage or if at least is the same between the devices. If it wasn't for the voltage, you could say that many smartphone have a bigger battery (at 5000 mAh) than M3 and Surface 7 Snapdragon X Plus, while it's beyond obvious that that's not the case here.
check out my video on the battery test where I go through that: How long can they last? | ULTIMATE BATTERY TEST ua-cam.com/video/u1XJAOf_W5w/v-deo.html
when using random stuff, you either store the random vector and reuse it for each test or you count the number of operations and compare time per operation ... your test is not so accurate :( ... but congratulations on doing a great job with this channel!
I'm going to tell you guys something that'll hopefully clear your mind about the surface laptops 7. For the 13.8 inch, get the x plus its been proven that its better and more preferred than the x elite. However, if you will get the 15.6 inch, get the x elite since its literally the best and you're kind forced, lol, since there aren't any x plus for the 15.6 inch. All of this is proven thanks to the big man "Just Josh". Additionally, Alex is a top tier man. Go check out the video where he compared the intels, with the snapdragon and even with the m chips all by himself, What a beast man.
Great review. My big question is how is this handling 1/2 monitors, one running a 1080p/4K video(on Chrome) and one running Intellij or a game(no fancy titles)? As these laptops have no dedicated graphics cards, from personal experience with other ultrabooks, I wonder if this has any performance issues(video stuttering, drops in framerate, etc).
The question is which powerprofile did you use? I notices on my Surface Laptop 7 with the Snapdragon X Elite that the recommended powerplan in the windows settings disables the dual core boost and therefore limits the clockspeed to 3400 Mhz. Would love to know if the X Elite performs similar to the x plus in battery life with the recommended power plan. That would mean you have best battery life that way while having the option to use dual core boost.
As much as I love the Surface laptop design, i know I'll have to get a framework for proper Linux support. I hope the surface don't have any BS proprietary stuff that needs custom kernel patches.
The huge problem with ARM Windows is the lack of support. If your workflow requires Google Drive - forget it. Google does not plan to offer Win ARM verison of its Drive app, nor it can be emulated 😞 That was the reason I had to return my Surface Laptop 7. Who knows what other things may not work. Apple has come a long way with ARM. Microsoft it still at the very beginning.
Quick question… which Surface model (by size) are you testing? The 13.8 or the 15? Apologies if this is a dumb question. I just can’t tell from the video. And do you have a preference and why?
judging from battery life it gives while having usable performance, X plus really is underrepresented in YT review video
Definitely.
most reviewers are hardware elitists who think every single person needs at least a core i7 (or ultra 7), 16gb ram, and 1tb storage.
X Plus does 99% of what X Elite can do, but more efficient.
@@ASEM-1123Microsoft also sends out the highest spec models to reviewers. Alex bought the Plus model with his own money which is a rarity for YT reviewers.
@@ASEM-1123 You do need 16gb of RAM, though, and electron apps will ensure that.
@@ASEM-1123 Lmfao? 16gb and 1tb storage is very much necessary. Well maybe really light users won't need 1tb, but 16gb def necessary.
This is the best channel for developers. You are doing a fantastic job.
@@AZisk Would like to see the same video with the new AMD Ryzen IA 370
@@jlparcerisa me too. That chip looks slick from the paid reviews I saw yesterday :)
@@AZisk Hi, I have a question. The surface with x plus comes with a 39w charger. What is the maximum wattage for charging via usb-c? Thanks in advance for your answer
Thanks for the thunderbolt display test. Noone else seems to test this on these new snapdragon X laptops...
@@pcbona agreed, it was very informative!
Thank you for your continuous efforts in testing these units. I have many customers waiting me to port my software to ARM and you already highlighted many potential problems that would take ages to figure out.
Glad to help
What idiot has bought these laptops and hasn't returned them?
You are making some incredible videos. Even when you’re talking about things that are slightly over my head, you explain in a way that makes perfect sense. I feel like I’m learning something every time i watch.
Freaking most detailed tech review. I'm thinking Alex talks in his sleep about core, gpu, python, etc. 😅
When using sorting algorithms as benchmarks, it would be prudent to lock in the seed of the vector that holds the out-of-order numbers. That way the algorithm performs an identically hard problem in all cases in terms of iterations needed to complete the operation, rather than relying on the wall-time required as the sole indicator of performance.
P.S.: There's a running joke in the optimization community that BOGOSort is the best sorting algorithm, you just have to "high-roll" a good enough vector that is practically already sorted, highlighting the need for exactly reproducible problems.
As someone who uses remote desktop to connect to my PC in the office to work, this kind of laptop is perfect for that. Can't wait for even more options and maybe some price cuts
A Mac user since forever should not be watching this. Yet here I am. Thanks Alex for publishing videos that are both informative and entertaining!
wht the heck man. ur reviewing these like a pro level
Running a program without optimizations turned on is really odd, even for debugging if you need performance you'd often at least use -O1
it might also explain why prism was somehow faster then running natively, the x64 code is just not that great and prism might be able to do some of the optimizations the compiler would have done if you turned them on
Same reason why using a compiled language can be slower than a jit language ifd you don't spend time on optimization
@@seeibe it's not about spending time, it's about passing the "-O2" flag when you compile. I guess compilation might be slightly slower, but not normally noticeable
Meanwhile me compiling software with -Os -flto=thin 😂
@@TheoParis lol, i use -o2 on linux
3:21 ok so that's dope
Okay, those amputated fingers are the most awesome test I've ever seen :D
you know what when i saw your content i was like "this is what i was looking for" but then when i saw your sub count on channel... it broke me.
fr this king of content deserve at least 5 million subs.
I think Thunderbolt compatibility testing would be best performed with Thunderbolt-only docks/accessories (the venerable CalDigit TS3 Plus, for instance). If the dock is also designed to work fine over standard, non-Thunderbolt USB-C, we don't get a good idea of how well the laptop plays with Thunderbolt. The dock you linked can fall back to a standard USB connection (probably 10 Gbps) if Thunderbolt bandwidths are not available on the port. If you are testing with that OWC dock, it would be best to connect a Thunderbolt accessory to it and verify if it's achieving Thunderbolt levels of bandwidth.
yes. good point, i could have gone one more step and plugged in the monitors into the dock for all the machines
Best Channel for Developers !!! Someone is reviewing laptops, considering our use cases as well :)
Bro I could watch these Snapdragon videos ALL DAY! In fact I kinda do lol
If I had the money, I'd grab the Surface Pro 11 in a heartbeat!
Don't run very short tests like 5 seconds. CPU core might not fully wake up after working on very low frequency (idle state), and might not get to turbo boost frequencies yet.
Just because someone designed their chips poorly doesn’t mean we have to test around that. Zen 4 mobile boosts to max stored in 2ms on wall power, and
This was a very cool one Alex. Thanks for making and share this content.
As a true NET developer (XAML moving to Web, and increasingly to Linux for containers, and JVM for PDF generation) this was very insightful, practical and to the point. Thanks and cheers from Florida today.
which one suggest for .net developer bro
Great videos, chap. Loving your X Elite coverage.
Sadly the manufacturers are completely cheaping out on RAM... The CPUs might have enough power to compile bigger projects in a reasonable amount of time, but I'm missing the 128 GB RAM option to test it out...
128 GB RAM? Where the fuck do you work at bro, CERN, NASA?
I think some of them have the option for 64 GB, but I find even 32 GB way too much. I have 16 on my fuckin DESKTOP which is used for everything (gaming, development, etc).
@@MadafakinRiowhy are you cussing so much 😅
Not many people need that much ram.
People that run Huge LLM’s or people editing huge video projects or even music producers who use lots of sample instrument libraries.
Not much people run 128 of Ram but it’s good that there’s an option to.
@@MadafakinRio 128GB is a lot for a laptop, but 16GB is mediocre. 32GB minimum for professional work. I am constantly maxing out my 16GB work laptop and having to close programs. I'm a software engineer. Need to get them to ship me out a 32GB model.
@@MadafakinRio Lets just say that I am involved in the development of medical imaging solutions. 128 GB is enough for compiling most of our projects. Testing needs much more, between 1 and 2 TB, depending on the input data.
@@dirkg.3163little buddy has never loaded a dataset before no wonder why he thinks more than 32gb is unfathomable
Moar great content from you sir!
Moar to come. I hope :)
8:46 Copying into the memory is not only about the memory controller and ram chips, but also single thread performace
These things are not bad machines and will only get better. Rocky start, but a short time later things will iron out and they will become critically hard to compete against.
That’s right, this is the best channel for developers. Let’s hope Microsoft improves and focuses on the quality control of their laptops.
Thank you for testing Thunderbolt 4 compatibility (which I figure is just using USB-4 standards over Thunderbolt, but at least it works).
Alex at 6:50 I would like to point out bad working conditions for schwarzenegger 2.0. It is trashed to the corner, really suffering and being ignored. Not cool man!!!! not cool.
He’s lucky I took him out of retirement for this one.
@@AZisk 🤣
I would like to see a long term testing scores for these computers after the daily system updates that are always require for Windows. Maybe monthly to see how performance degrades over time.
Hey sir, when you're benchmarking Intel Chip, please try it with Lenovo/Asus as Surface really doesn't have good Intel models! Lenovo 14th Gen laptops are quite quiet!
Great comparisons tho!
Perhaps next time a storage intensive test, taking into account the price tag of having a machine with a storage of 2TB or 4 TB (which let apple look like a discounter offer?)
The interesting time will come if snapdragon equivalents for The Pro and Max apple Silicon appear, currently there is only competition on MBA level
really appreciate these reviews. I love my surface pro 7 but its battery life is terrible. It still does everything I need it to for now but may end up getting one of these beautiful surface laptop 7 soon..
Maybe MSVC for ARM64 is less refined / optimized than for x64?
Or maybe like @squadb3 said Prism may optimize unoptimized Debug code, because it's ultimately a JIT compiler.
You are mentioning that the multicore sort is faster on the X-elite, having 10 or 12 cores, compared to the Macbook Air's only 8 cores. But in reality, the Air only has 4, as in 4 performance cores compared to 10 or 12 performance cores on the X-elite. I wish reviewers would point out this detail.
E-cores and P-cores in the M-chips are structurally different, not just slightly lower clocked cores
moreover, E-cores are mainly only used for background tasks, and MacOS only schedules threads onto the E-cores when P-cores are fully saturated, and even then, it won't fully use the E-cores as they are reserved for the OS' maintenance processes. So Apple mentioning the total number or cores is misleading.
@@MrLocsei That's not anyone s problem
This is true. thanks for pointing it out
Disregarding price point these should be compared to a MacBook Pro with similar P core count M chips.
It would be awesome to hear a final verdict on xelite laptops.
BTW loved the videos, no one else is going to the depth to test xelite system.
The Surface 6 with the Core Ultra 7 165H honestly sounds like a great Business laptop. None of the ARM crap, but with real performance compared to the wimpy i7-1355U we had previously
It would be great to see comparisons between snapdragon elite laptops. The surface is not the only x elite.
I may have missed it in the video - but in Windows, you must set to "best performance" power scheme. Otherwise, power management features will dramatically limit your performance. Interestingly, on the Qualcomm CPUs (similarly to M-arch on Macs) - if you stay on Best Performance while on battery - performance doesn't suffer. On Intel you can expect ~7% drop, on AMD a lot more. Using "Best Performance" while on battery will significantly hurt battery life though.
I'm loving my 32gb ARM Surface laptop. The only issue is Cisco VPN client, probably because they are investors in Intel. I also love me intel laptop as well so I'm good 🙂
Alex, all your videos comparing ARM devices are getting me hyped, although I still think they need to be cheaper for people to buy them. The great battery life and low fan noise in everyday use isn't a decisive factor because people have simply gotten used to it and see it as "meh."
I think if Microsoft changes the screen on the "x plus" chip versions to one that isn't so reflective and isn't touchscreen (as a developer, I don't know why anyone would want a touchscreen on a laptop; they should get a tablet instead), it would lower the price and make it a better product. Otherwise, I don't see many people choosing these over the MBA M3, and soon they will be even cheaper when the M4s come out.
New engineering standard - surface temp not to go over 45 C - so systems choke to cool. Just implemented last year's end - a new thing to check.
How about taking us though a SSD upgrade process? Using a Mac for so long I cant even remember all the steps e.g. download a ISO, make USB bootable, actually replacing the SSD, is it really easy? Install windows, does all the drivers work out of the box?
i have the huawei matebook 14 with intel core ultra 5 125h. it comes with a huawei pc manager. it lasts for 18hours in power saving mode. the app throttles the cpu but browsing and streaming experience is still great. but i doubt i can use that setting when i actually do real work. but at least it's good to have an option like that when you don't really need the cpus full power.
Your videos are incredible! Keep it up!
13:29 you probably compiled for x86. The maximum stack size is way smaller on x86 compared to x64, so you would need to compiler for x64 (and for timing also use -O3)
-O3 for the compilers, please. Always.
Not always.
@@ranjithmkumar In general I fire the guys whose code doesn't behave well with -O3.
Having tried compiling different software with -O3 vs -O2 it seems it rarely produces much difference in runtime but increases compile time significantly.
@@lambdaprog O3 allows experimental optimizations that may change the behavior even of perfectly legal C code without undefined behavior, and in most cases will not be significantly faster than O2. In fact, it may be slower than O2.
@@millfi I agree. O2 code size optimisation is better in the embedded environment.
that's it I'm going to buy one of this, especially when android studio support the ARM Version, i thought it wouldn't be too long, since they already have the ARM Version for Mac
Thanks for the insight on SIMD and AVX2.
Please Alex, take a look at mini PCs from developer's perspective
I'm waiting on a Chromebook with the Snapdragon X Plus to come out. Hopefully with a big battery.
I don't think they push this one any further, some of them are already discounted the Laptops just a few weeks after the release.. so they still had too many what nobody wanted..
Hi Alex, first of all i wanna say thanks to you because of your super helpful videos. And secondly i wanna ask you something, personally i 2, 3 times reinstall windows in a year, so can you make video about it on arm chips. Is it still easy to reinstall windows on arm chips like x64, or should we expect more problems? I hope you will read this!
Tell us about UEFI and BIOS on arm
To prevent C++ people to yell at you wrongly setting the flags, just use Rust and dont forget set itu to release build
(I assume the windows would use native windows compiler and mac would probably use clang, thus this yup this is not apple to apple benchmark)
I bought the 15in Surface Laptop 7, really baffles me though that the battery is only 66Wh, while the 13.8in has a 54Wh battery. You’re telling me with an extra 19.2 square inches of space (based of chassis dimensions) they could only fit another 12Wh worth of battery? Considering the cost and how it’s supposed to be their top end device, they really should’ve crammed a 100Wh battery in like Apple does for the largest MacBook Pros. Would’ve given a 50% boost in battery life and helped with long term degradation.
Will you be making setting up linux for development?
support isn't complete yet in the kernel but qualcomm is working on it
Just made a video about this last week.
i like this guy...
12:39 You need to quick sort an array with same array for both builds.
quick sort could just finished with fewer iteration on x64 build since you shuffled array on runtime based on source code available on 11:14
I have the base model Surface Laptop 7 with the Plus CPU (because that's pretty much the only sensible choice :P), and I was fully expecting games to be atrocious, based on the reviews I've seen. But actually everything I tried ran, and most games even ran well enough to properly play. I've tried Subnautica, System Shock remake, Scrap Mechanic, Prey, Skyrim SE, Grounded, Portal 1 and Portal 2, etc. Oddly enough Skyrim gave me the most trouble, because if I didn't use native resolution it started up with 3/4 of the rendered image being off-screen.
That's a surprise. What about CS:Go 2? Any good? That's a common casual play for me.
VMware Hoziron also totally messes up the system when you try to run the newest version. The laptop literally locks up after a few minutes, you can barely even shut it down. There are some older versions which seem to run okay, but still it's not reassuring if that's how you connect to your corporate work VM and you'd like to work from a Surface Laptop 7.
Hmm. I've never heard that quote from Tim Cook. So, the way Apple copies some features from Android is the sincerest form of flattery. 😂 I am eager to see some tests and reviews of the X1 ThinkPad Elite since ThinkPads have been my daily driver for quite a while.
÷I really apreciate these videos and the effort put in them. I like the surface laptop 7 with the SD plus.
Tbh I learned a lot about x86 and arm thanks ❤🔥
I can't wait to see native linux support for x elite laptops.
Josh reviewed this and said the battery life on the surface pro 11 is not good. Battery life is the main reason I’m considering one. Now I’m on the fence.
i didn’t review the surface pro
@@AZisk my bad, I responded to the wrong video.
The Problem with Android Studio could be the emulator/simulator, which could be solved by running the code on a real device debugging over USB.
I think that it is this close we are finally at the turning point for windows on ARM.. I wonder if we will get desktop processors anytime soon
I'm a Visual Studio C# developer and host my SAAS app on Azure. I can't decide between Surface Pro 10 for business with 64GB RAM (expensive!) or the maxed out Surface Pro 11. I would sure love to speed up my compile times! Currently takes ~1 minute on my Surface Pro 7 with 16GB. I don't really care about battery life -- I'm always plugged in. Or the price. The only thing I really care about compile times. Any thoughts, Alex? Great video.
It's simple. For apple get m3 base or pro, unless you need capabilities of max or ultra. Or wait for coming m4. For snapdragon, it's only matter for Xelite. You don't want to pay for that price? Don't go low ball and get XPlus, not worth it. And Microsoft "translation" layer is not refined like Apple, so stick to intel ones for production usage.
And on one use their laptop without charging regularly or always plugged in. But there must be "content", isn't it?
You really need to turn on O3 for any software benchmarks you compile yourself. Production software will never release without optimizations enabled. I would guess this is why you saw Prism running better than native, assuming Prism does optimizations for ARM during the translation phase which your native binary missed out on. Not certain though.
A 5 second test also isn't very helpful as it's just not enough time for the operating system's process scheduler to kick in and set your high throughput task as needing priority. At least 30 seconds would be better.
You should also be more specific when making battery claims. It's been shown that while the Snapdragon laptops can last longer at low loads like looping a Netflix video which uses the chip's low-power dedicated decode hardware, they fail to maintain that power efficiency as loads increase, unlike M3 which has much better power efficiency as workload scales up. It's also been shown that some Snapdragon laptop models have different power budgets with some significantly throttling the CPU after a short period of high usage, which makes it look like the battery life is excellent when in reality the performance was nuked, which you might not notice in a hands-off automated test. Just Josh did some excellent performance analysis in this regard. I think for the average person the battery life will be good either way, but it's not really accurate to suggest they can outlast a Macbook with normal day-to-day use.
I'm also curious about the memory bandwidth results, I wonder if running inside WSL is creating a bottleneck by running in a VM. Would be nice to use a native benchmark, might show better results. Though to be fair, most developers are probably running AI stuff inside WSL, so maybe it's a more useful benchmark even if its not the best it can do.
Overall, really good video, and I appreciate taking the time to benchmark tasks aimed towards programmers. Not a ton of people out there other than Josh doing that.
need one more with the Intel Ultra 200 series!
Any chance you can tune the power settings ? Intel chips are known that can be fed with a lot of watts. And, unfortunately, many laptop manufacturers actually do that, then fiddle with their thumbs when people complain about battery life.
I'm also curious how the difference is when the laptops are tuned to perform the same. Or simply... yeah, that should be much easier, a battery (pun intended) of tests at different power levels. Complete with the performance results. Maybe Intel can last just as long if it's set up at minimum values, like running at 800 MHz on the CPU. But it might be 3 times as slow. Or maybe instead of being 10% faster, it can be just as fast at half of the power draw it had initially and actually compete in battery life. We need MOAR tests!
Thing is that the battery life can vary A LOT, really, A HUGE LOT, between one laptop and another. The motherboard design of the laptop matters. How many ports you have matters. What type of RAM you have matters. And lastly, like I said above, how the CPU is configured also matters. I don't remember the exact case, but, you can have 2 laptops with the same chip and one to be less efficient, less performant, at higher temperature all AT THE SAME TIME, just because one is properly designed and configured and one not.
Hey, have you considered doing streams on twitch? Just chating about your perspectives in any topic would be entertaining
why Twitch and not here?
I bought the 15 inch Surface Laptop (elite model), and also a Galaxy Book4 Pro 360 (155H model)... so torn on which one to keep. I think the Surface is more powerful, and I think the build quality is slightly better... but the Galaxy's 3k amoled display, ability to fold into a tablet, and pen input is so good... ugh. What do I do Alex? lol. Will I actually notice much of a performance boost with the X Elite over the 155H?
Love all your geeky tests. These are the things I do and appreciate there are others like me. I am specifically interested in the rendering compatibilities of Qualcomm chips. I use Resolve for video editing. I realize that Resolve ARM version is in beta but just how well does it work? Does hardware video acceleration work when exporting a video in AV1? How about Handbrake (which has an ARM version I believe)? Does it support hardware video decode/encode?
Was there any mention about mAh? battery sizes? as when I googled,.. MBA 13" M3 = 4380 mAh, Surface 7 Snapdragon X Elite = 5806 mAh... Surface 7 Snapdragon X Plus = 4755 mAh
With that, I assume, both should last longer than the macbook, if not, that's bad. But also means that the Elite power efficiency is really bad if it looses against the Plus.
Where did you get those ? Also, unlike the smartphones, where you know the voltage (as 3.7 - 4.2, so roughly 4 volts), the battery capacity in ... everything else, is usually measured in Wh. Your numbers tell me nothing, since I don't know the voltage or if at least is the same between the devices. If it wasn't for the voltage, you could say that many smartphone have a bigger battery (at 5000 mAh) than M3 and Surface 7 Snapdragon X Plus, while it's beyond obvious that that's not the case here.
check out my video on the battery test where I go through that: How long can they last? | ULTIMATE BATTERY TEST
ua-cam.com/video/u1XJAOf_W5w/v-deo.html
He makes great videos and useful
when using random stuff, you either store the random vector and reuse it for each test or you count the number of operations and compare time per operation ... your test is not so accurate :( ... but congratulations on doing a great job with this channel!
I'm going to tell you guys something that'll hopefully clear your mind about the surface laptops 7. For the 13.8 inch, get the x plus its been proven that its better and more preferred than the x elite. However, if you will get the 15.6 inch, get the x elite since its literally the best and you're kind forced, lol, since there aren't any x plus for the 15.6 inch. All of this is proven thanks to the big man "Just Josh". Additionally, Alex is a top tier man. Go check out the video where he compared the intels, with the snapdragon and even with the m chips all by himself, What a beast man.
Nice one, could you do a review with the lenovo yoga 7x with the x elite chip ?
Beautiful videos. Thumbs up
such a dope video
Great review. My big question is how is this handling 1/2 monitors, one running a 1080p/4K video(on Chrome) and one running Intellij or a game(no fancy titles)? As these laptops have no dedicated graphics cards, from personal experience with other ultrabooks, I wonder if this has any performance issues(video stuttering, drops in framerate, etc).
Snapdragon X Elite ❤️❤️❤️🔋
Pls do a test on Unity 3D Development ( I know unity had ARM version ) but I'm curios how it will perform.
which one do you recommend for programming and devops staffs. No concern about camera, display, speakers, battery life and physical design.
The question is which powerprofile did you use? I notices on my Surface Laptop 7 with the Snapdragon X Elite that the recommended powerplan in the windows settings disables the dual core boost and therefore limits the clockspeed to 3400 Mhz. Would love to know if the X Elite performs similar to the x plus in battery life with the recommended power plan. That would mean you have best battery life that way while having the option to use dual core boost.
On MacOS you should have added -O3 to the compile command
Thank you!
Will you be testing strix point and lunar lake laptops in the future?
Very curious about Lunar Lake. Seems promising and it doesn't have any of the incompatibility issues.
@@shintarouemiya it’ll be interesting. Strix point won’t be on the 3nm node but I’m expecting a performance jump for zen 5 and new igpu.
Do you think the plus will be good for web development or it will come with compromises?
No compromises for web developement. It's rly good with web apps
As much as I love the Surface laptop design, i know I'll have to get a framework for proper Linux support. I hope the surface don't have any BS proprietary stuff that needs custom kernel patches.
it is an IO limitation of the non-pro/max chips not apple limiting it
Can you make more videos on local AI use cases? spscially for auto complition.
Thank you for you amazing content
thank you for great Vid
Is it possible to also make Linux tests on these SnapX machines?
I'm curious how's the battery and performance there, as most devs utilize Linux envs.
The huge problem with ARM Windows is the lack of support. If your workflow requires Google Drive - forget it. Google does not plan to offer Win ARM verison of its Drive app, nor it can be emulated 😞 That was the reason I had to return my Surface Laptop 7. Who knows what other things may not work. Apple has come a long way with ARM. Microsoft it still at the very beginning.
Quick question… which Surface model (by size) are you testing? The 13.8 or the 15? Apologies if this is a dumb question. I just can’t tell from the video. And do you have a preference and why?
@@jrpool88 xplus, xelite:13.8
Intel:13.5
He did mention it in the video.
@@dr.orange6509 thank you. My ADHD brain missed it. 😀. FYI, same macuinebi purchased and Inreally like it (fwiw).
Would you recommend this for a computer science student?
A little sad you're not testing AMD based laptops. They are generally better than Intel ones in price/performance/battery.
"Off cause the Maceunuch is the more potent"! But it cannot run your favorite little
12:35 let me guess, you were benchmarking on a debug executable with all optimisations disabled? C++ compilers disable optimisation by default.
14:42, yes, as expected
Will you test surface pro with new snapdragon?