We have tried a whole bunch of benchmarks and the laptop was on par or faster than the older Threadripper workstation and my 5950X desktop. Most benchmarks were multithreaded, but there was some singlethreaded stuff as well. He uses the system to run simulations for work and that software also runs faster than the old workstation. I can’t run that on my system, so I wouldn’t know how it compares.
I don’t have the exact bench results as we didn’t write them all down, just ran and compared. But I do have a result screenshotted of 27.9 in Speedometer 3.0, which is pretty good I think.
As it’s the laptop from work he runs Windows on it, the new Windows Arm version which wasn’t even fully released at the time he got it. That version seems to be a big step up from the old Arm Windows which was used for budget Bing books. His model is the most high end one, 15" with 64GB of memory and 1TB SSD with a Qualcomm Snapdragon X Elite X1E-80-100 cpu. That one has some pretty good cooling inside.
I was sceptical at first as well, as I would have thought the performance wouldn’t be great and there would be compatibility issues. But he’s been using it for a while now and says everything works just fine. Replacing a big box workstation with a thin and light notebook and have it perform better is pretty wild. There would absolutely be faster systems available, for example a 9950X system or latest gen Threadripper workstation would be faster. But that would have been more expensive because those systems are more expensive to start with and he would then need a separate laptop as well. Having something in a thin laptop form factor and it be an upgrade in performance is pretty mind blowing.
I thought you said your friend used applications not benchmarks. You know something like x265 encoding, AI video upscaling, mining XMR (CPU based crypto) or complex single-thread dependent strategy games; I say this as someone involved in computers for hobbies and leisure, not in a professional manner.
No offence, but what you are saying does not sound convincing in the least.
A high level X Elite GB6 ST/MT score is around 2,800 ST and 14,300 MT.
A high level 5950X GB6 ST/MT score is around 2,400 ST and 14,300 MT.
And this is a short GB benchmark (i.e. not sustained for hours on end).
You’re saying if I give you a complex (lots of water, storms, seas) 2+ hour BD source to encode into x265, your friend’s X Elite laptops won’t start to throttle in ~5 min and it will complete it in the same time as your 5950X? You’ll have to provide proof.
But he’s been using it for a while now and says everything works just fine. Replacing a big box workstation with a thin and light notebook and have it perform better is pretty wild.
Some specifics would be interesting. What applications were being used on the big box workstation?
But I do have a result screenshotted of 27.9 in Speedometer 3.0
Not aware of Speedometer 3.0, this does seem like a very solid result, but what does it show? Do you have any context on it? This is the first time I’ve encountered this particular benchmarks, would be interesting to hear about what it means.
It’s a work machine, he uses it for work. He runs a custom simulation package from work, I can’t name the app without doxxing my friend. It scales well with CPU and memory and uses an optimal number of threads for the amount of cores (and even works well with stuff like multiple CPUs or different cores having access to different cache). For running most stuff at least 32GB of memory is required and for the stuff he does 64GB of memory is an absolute must. Simulations take between 20 mins and 8 hours depending on what he wants from it. The simulation tool does not use the GPU at all, so that’s a non-factor. The tool is x86 based, with an Arm version coming soon™, so there might even be performance improvements in the future. The simulations run faster in all scenarios as compared to his old workstation, even the long ones. Cooling is not an issue on this particular machine and due to the many core load boosting isn’t done anyways.
We ran Speedometer because many laptop reviews include that one and it’s very quick and easy to run. Specifically the 3.0 version because we had a source open with Apple M benchmarks that included that one. The result was somewhere around Apple M2, maybe a bit faster than an M2 but def slower than an M3.
You can try it yourself and read about it here: https://browserbench.org/Speedometer3.0/
It benchmarks regular use in web-apps, as a lot of apps these days are web-apps. So it gives an impression of every-day tasks in websites and web-apps.
I think the results you mention back up what I said? It regularly outperformed my 5950X desktop machine in benchmarks or was at least on par in other cases. My desktop is a big case with water-cooling and when benching the fans do make a bit of noise. That little notebook outperformed it and the fan was barely noticeable.
Like I said, I was sceptical, but that thing impressed me a lot. You can draw your own conclusions, that doesn’t really matter to me. If you think Arm laptops suck by definition, that’s fine, you do you. But don’t say you can’t use it for heavy applications, because at least for some cases that’s just not true. I think the GPU (especially the driver) is a weak point for these system, so anything that leans on them probably should use a systems with a separate GPU and not the builtin one. But this is also true for Intel and AMD, so not really any difference there.
You misunderstand, I don’t have any allegiance to Intel/AMD or ARM or Apple.
I am interested in factual reality though. This is not about Qualcomm versus AMD or whatever. It’s common sense. If you have a use case (e.g. x265 encode or god forbid an AV1 encode) that takes many hours, you laptop is going to suffer due to cooling issue. This is even true for an ARM laptop versus an ARM desktop.
I think the results you mention back up what I said?
So you’re saying that throttling after 5 min is not an issue with MT (or even ST) workloads?
I would be happy to be proven wrong (I am not kidding), but I would need solid proof from a 3rd party.
No offence, but random claims online is not how it works.
Mind you, I am not saying you or your friend are lying. There are likely other factors at play.
I didn’t think the thing would be good. When he got it in, we spent a day running benchmarks and fooling around with it. We compared it to his old workstation and my desktop system. It wasn’t a very controlled environment, we were just having fun and putting the thing through it’s paces.
I asked my friend yesterday how he liked the machine having worked with it for some time now and he’s really happy with it. It is across the board faster than his old machine and is wonderful to work with. He can setup complex simulations and take it with him to the office. This was always a bit of a pain point in the past, where he would run the simulations at home on his workstation, but then could only share the results. Sometimes they would rent server time to run the simulation on a cloud system, but that was a bit of a hassle and had costs. Now he just uplugs his notebook, puts it in the bag and off he goes. He also now doesn’t have 2 systems from work he needs to regularly log into and keep up to date. Sometimes he had a couple of months where he didn’t need the laptop and it would get fussy over missing updates etc. So for him at least it’s a big win and to me shows you can run some pretty heavy stuff on those machines.
Are there faster machines out there? Absolutely. Are there even better notebooks out there? For sure, Apple M3 is faster and M4 is even faster still. And with Apple the performance per watt is better as well. But running Windows on those is (for now at least) not something that’s suitable for work. The security department would certainly not approve of a highly modified version of Windows.
The whole point of this post was Arm chips might be huge in the future and I have to agree. These current gen Arm CPUs are impressive and the next gen will be even more so.
You also seem to indicate running benchmarks and running applications is somehow not the same thing? Sure not all benchmarks are realistic, but are more of an indication and relative performance thing, as to easily compare different systems. And not all applications stress the system the same way. But every benchmark I’ve seen says that notebook is on par and exceeds the performance of my 5950X desktop and to me that’s impressive. In the real world if we are using simple office applications or websites/web-apps, I doubt we would notice the difference in performance, both are equally fast and perhaps the latency of the internet connection is a bigger factor there. But something like Speedometer shows the real world browser performance of the laptop is better than on my desktop.
Did the engineers at Qualcomm spend a couple of weeks with a small team to optimise a custom Linux environment for Geekbench and put a boatload of cooling on the chip? Sure, I believe that. They want to show the CPU in the best possible conditions. Is the real world performance still very good? Yes, it is. And there are so many notebook reviews that back this up.
Are there also terrible notebooks with a CPU throttled all the way down and lacking enough cooling? Also yes. But the same can be said for x86 notebooks. Especially Intel notebooks of 12th and 13th gen, those ran hot and slow all the time.
If you are convinced all Arm notebooks suck, I’m not here to change your mind, I’m not here to provide any kind of proof. All I can tell you is I know of one real life case where I saw with my own eyes the thing was pretty damned good. If you don’t believe me, that’s just fine. It’s just a discussion on the internet, don’t take it too seriously.
It’s not like anyone can afford a new laptop in 2026, with the RAM prices being what they are. So it probably won’t be the year of the Arm CPU, no matter how good those chips actually are.
We have tried a whole bunch of benchmarks and the laptop was on par or faster than the older Threadripper workstation and my 5950X desktop. Most benchmarks were multithreaded, but there was some singlethreaded stuff as well. He uses the system to run simulations for work and that software also runs faster than the old workstation. I can’t run that on my system, so I wouldn’t know how it compares.
I don’t have the exact bench results as we didn’t write them all down, just ran and compared. But I do have a result screenshotted of 27.9 in Speedometer 3.0, which is pretty good I think.
As it’s the laptop from work he runs Windows on it, the new Windows Arm version which wasn’t even fully released at the time he got it. That version seems to be a big step up from the old Arm Windows which was used for budget Bing books. His model is the most high end one, 15" with 64GB of memory and 1TB SSD with a Qualcomm Snapdragon X Elite X1E-80-100 cpu. That one has some pretty good cooling inside.
I was sceptical at first as well, as I would have thought the performance wouldn’t be great and there would be compatibility issues. But he’s been using it for a while now and says everything works just fine. Replacing a big box workstation with a thin and light notebook and have it perform better is pretty wild. There would absolutely be faster systems available, for example a 9950X system or latest gen Threadripper workstation would be faster. But that would have been more expensive because those systems are more expensive to start with and he would then need a separate laptop as well. Having something in a thin laptop form factor and it be an upgrade in performance is pretty mind blowing.
I thought you said your friend used applications not benchmarks. You know something like x265 encoding, AI video upscaling, mining XMR (CPU based crypto) or complex single-thread dependent strategy games; I say this as someone involved in computers for hobbies and leisure, not in a professional manner.
No offence, but what you are saying does not sound convincing in the least.
A high level X Elite GB6 ST/MT score is around 2,800 ST and 14,300 MT.
A high level 5950X GB6 ST/MT score is around 2,400 ST and 14,300 MT.
And this is a short GB benchmark (i.e. not sustained for hours on end).
You’re saying if I give you a complex (lots of water, storms, seas) 2+ hour BD source to encode into x265, your friend’s X Elite laptops won’t start to throttle in ~5 min and it will complete it in the same time as your 5950X? You’ll have to provide proof.
Some specifics would be interesting. What applications were being used on the big box workstation?
Not aware of Speedometer 3.0, this does seem like a very solid result, but what does it show? Do you have any context on it? This is the first time I’ve encountered this particular benchmarks, would be interesting to hear about what it means.
It’s a work machine, he uses it for work. He runs a custom simulation package from work, I can’t name the app without doxxing my friend. It scales well with CPU and memory and uses an optimal number of threads for the amount of cores (and even works well with stuff like multiple CPUs or different cores having access to different cache). For running most stuff at least 32GB of memory is required and for the stuff he does 64GB of memory is an absolute must. Simulations take between 20 mins and 8 hours depending on what he wants from it. The simulation tool does not use the GPU at all, so that’s a non-factor. The tool is x86 based, with an Arm version coming soon™, so there might even be performance improvements in the future. The simulations run faster in all scenarios as compared to his old workstation, even the long ones. Cooling is not an issue on this particular machine and due to the many core load boosting isn’t done anyways.
We ran Speedometer because many laptop reviews include that one and it’s very quick and easy to run. Specifically the 3.0 version because we had a source open with Apple M benchmarks that included that one. The result was somewhere around Apple M2, maybe a bit faster than an M2 but def slower than an M3.
You can try it yourself and read about it here: https://browserbench.org/Speedometer3.0/ It benchmarks regular use in web-apps, as a lot of apps these days are web-apps. So it gives an impression of every-day tasks in websites and web-apps.
I think the results you mention back up what I said? It regularly outperformed my 5950X desktop machine in benchmarks or was at least on par in other cases. My desktop is a big case with water-cooling and when benching the fans do make a bit of noise. That little notebook outperformed it and the fan was barely noticeable.
Like I said, I was sceptical, but that thing impressed me a lot. You can draw your own conclusions, that doesn’t really matter to me. If you think Arm laptops suck by definition, that’s fine, you do you. But don’t say you can’t use it for heavy applications, because at least for some cases that’s just not true. I think the GPU (especially the driver) is a weak point for these system, so anything that leans on them probably should use a systems with a separate GPU and not the builtin one. But this is also true for Intel and AMD, so not really any difference there.
You misunderstand, I don’t have any allegiance to Intel/AMD or ARM or Apple.
I am interested in factual reality though. This is not about Qualcomm versus AMD or whatever. It’s common sense. If you have a use case (e.g. x265 encode or god forbid an AV1 encode) that takes many hours, you laptop is going to suffer due to cooling issue. This is even true for an ARM laptop versus an ARM desktop.
So you’re saying that throttling after 5 min is not an issue with MT (or even ST) workloads?
I would be happy to be proven wrong (I am not kidding), but I would need solid proof from a 3rd party.
No offence, but random claims online is not how it works.
Mind you, I am not saying you or your friend are lying. There are likely other factors at play.
Alright, I’ve got nothing for you then.
I didn’t think the thing would be good. When he got it in, we spent a day running benchmarks and fooling around with it. We compared it to his old workstation and my desktop system. It wasn’t a very controlled environment, we were just having fun and putting the thing through it’s paces.
I asked my friend yesterday how he liked the machine having worked with it for some time now and he’s really happy with it. It is across the board faster than his old machine and is wonderful to work with. He can setup complex simulations and take it with him to the office. This was always a bit of a pain point in the past, where he would run the simulations at home on his workstation, but then could only share the results. Sometimes they would rent server time to run the simulation on a cloud system, but that was a bit of a hassle and had costs. Now he just uplugs his notebook, puts it in the bag and off he goes. He also now doesn’t have 2 systems from work he needs to regularly log into and keep up to date. Sometimes he had a couple of months where he didn’t need the laptop and it would get fussy over missing updates etc. So for him at least it’s a big win and to me shows you can run some pretty heavy stuff on those machines.
Are there faster machines out there? Absolutely. Are there even better notebooks out there? For sure, Apple M3 is faster and M4 is even faster still. And with Apple the performance per watt is better as well. But running Windows on those is (for now at least) not something that’s suitable for work. The security department would certainly not approve of a highly modified version of Windows.
The whole point of this post was Arm chips might be huge in the future and I have to agree. These current gen Arm CPUs are impressive and the next gen will be even more so.
You also seem to indicate running benchmarks and running applications is somehow not the same thing? Sure not all benchmarks are realistic, but are more of an indication and relative performance thing, as to easily compare different systems. And not all applications stress the system the same way. But every benchmark I’ve seen says that notebook is on par and exceeds the performance of my 5950X desktop and to me that’s impressive. In the real world if we are using simple office applications or websites/web-apps, I doubt we would notice the difference in performance, both are equally fast and perhaps the latency of the internet connection is a bigger factor there. But something like Speedometer shows the real world browser performance of the laptop is better than on my desktop.
Did the engineers at Qualcomm spend a couple of weeks with a small team to optimise a custom Linux environment for Geekbench and put a boatload of cooling on the chip? Sure, I believe that. They want to show the CPU in the best possible conditions. Is the real world performance still very good? Yes, it is. And there are so many notebook reviews that back this up.
Are there also terrible notebooks with a CPU throttled all the way down and lacking enough cooling? Also yes. But the same can be said for x86 notebooks. Especially Intel notebooks of 12th and 13th gen, those ran hot and slow all the time.
If you are convinced all Arm notebooks suck, I’m not here to change your mind, I’m not here to provide any kind of proof. All I can tell you is I know of one real life case where I saw with my own eyes the thing was pretty damned good. If you don’t believe me, that’s just fine. It’s just a discussion on the internet, don’t take it too seriously.
It’s not like anyone can afford a new laptop in 2026, with the RAM prices being what they are. So it probably won’t be the year of the Arm CPU, no matter how good those chips actually are.