I never realized how much heat, vibration, air blowers negatively affected me before. A framework laptop with some type of ARM or ARM-like cpu could do a lot with the space savings on cooling.
I just recently stumbled across these two otherwise identical laptops.
https://psref.lenovo.com/syspool/Sys/PDF/Lenovo/Lenovo_Slim_...
https://psref.lenovo.com/syspool/Sys/PDF/Lenovo/Lenovo_Slim_...
The AMD version is rated for 20% greater battery life just from the CPU difference.
In some instances Alder Lake CPUs slightly edge out Apple's CPUs in performance, but with a terrible battery life (~4 hours of real world battery life vs >10 hours on Apple laptops) and noisy fans.
AMD on the other hand seems to have focused on getting a good balance this year. 6800U models have significant performance improvements over the last gen while improving the battery life as well.
I am thinking of buying a new laptop this year, but I am waiting for the AMD models. For anyone interested, these are the models I am waiting for:
https://psref.lenovo.com/Product/ThinkBook/ThinkBook_13s_G4_...
https://psref.lenovo.com/Product/ThinkPad/ThinkPad_X13_Gen_3...
Both have much better battery life than their Intel counterparts:
ThinkBook 13s Gen 4 (AMD): 14 hours
ThinkBook 13s Gen 4 (Intel): 10.9 hours
ThinkPaf X13 Gen 3 (AMD): 15.68 hours
ThinkPad X13 Gen 3 (Intel): 11.8 hours
These numbers are usually lower in real life, but they are comparable with each other.
There are data sheets and then there is real life. People were raving about AMD laptops before. I got a ThinkPad T14 AMD with a 4750U. The stated psref battery life is MobileMark 2014: 14 hr, MobileMark 2018: 11.5 hr. In practice that was more like 7 hours on Windows (with a mild workload) and 3 hours on Linux. The CPU performance was merely ok and I could get the fans blowing quite quickly.
Life with the laptop was quite miserable. There were many other paper cuts and despite being Linux-certified, I used Windows with WSL most of the time since basic things like suspend would at least work relatively reliably.
I sold the T14 after 7 months (thank deity that there was a chip shortage, so I could recoup quite a lot of the 1400 Euro I paid for the T14), got a MacBook Air M1 and didn't look back. It's a world of difference.
The machine that I'm currently using comes with a Ryzen 6800H CPU and LPDDR5-6400 RAMs, made by Lenovo. On Linux, the builtin keyboard don't work because of IRQ problems (See [1], a fix is also available at [2]), and it constantly spams out "mce: [Hardware Error]: Machine check events logged" messages.
[1]: https://bbs.archlinux.org/viewtopic.php?id=277260
[2]: https://lore.kernel.org/all/20220712020058.90374-1-gch981213...
If you read the code in [2], the patch basically disables IRQ override for every new Ryzen machines (`boot_cpu_has(X86_FEATURE_ZEN)`). Based on that, I assume every new Ryzen CPU has the same problem(???)
Edit: wait... blaming it on the CPU might be unjust, it's more like a kernel adoption problem.
Luckily, compiling Linux kernel with 16 threads is not really too big of a struggle, you can just apply the patch manually every time the kernel updates :-) :-| :-(
The exact problem is that an IRQ/ACPI workaround is not needed anymore on modern Zen platforms, and it (the workaround) now breaks compatibility. It's already in linux master, and will be released for v6.0
Commit message of the fix:
commit 9946e39fe8d0a5da9eb947d8e40a7ef204ba016e
Author: Chuanhong Guo <gch981213@gmail.com>
ACPI: resource: skip IRQ override on AMD Zen platforms
IRQ override isn't needed on modern AMD Zen systems.
There's an active low keyboard IRQ on AMD Ryzen 6000 and it will stay
this way on newer platforms. This IRQ override breaks keyboards for
almost all Ryzen 6000 laptops currently on the market.
Skip this IRQ override for all AMD Zen platforms because this IRQ
override is supposed to be a workaround for buggy ACPI DSDT and we can't
have a long list of all future AMD CPUs/Laptops in the kernel code.
If a device with buggy ACPI DSDT shows up, a separated list containing
just them should be created.But it's not CPU, it's Lenovo. Lenovo has terrible firmware team.
You can read more about this in my blog
https://127001.me/post/ten-years-of-thinkpadding/#_the_ugly_...
The reason I'm asking, according to a Lenovo talk from DebConf22 [0], the guy said it takes some time for patches to reach upstream, but with Fedora and Ubuntu they have some connections to shortcut some patches before they reach upstream.
Sad news is, according to my test, the latest kernel Fedora 36 offered, which is kernel-5.18.16-200.fc36.x86_64, does not come with the keyboard (IRQ override) fix.
Another thing is the mce errors, or more specifically, errors similar to:
[ 332.990256] mce: [Hardware Error]: Machine check events logged
[ 332.990267] [Hardware Error]: Corrected error, no action required.
[ 332.990276] [Hardware Error]: CPU:0 (19:44:1) MC15_STATUS[Over|CE|MiscV|AddrV|-|-|SyndV|CECC|-|-|-]: 0xdc204000000c011b
[ 332.990299] [Hardware Error]: Error Addr: 0x00000000efcf8bc0
[ 332.990304] [Hardware Error]: IPID: 0x0000009600050f00, Syndrome: 0x000001ff0a240701
[ 332.990312] [Hardware Error]: Unified Memory Controller Ext. Error Code: 12
[ 332.990316] [Hardware Error]: cache level: L3/GEN, tx: GEN, mem-tx: RD
Of course, those were just the most annoying two. There are many smaller problems such as: 1) Screen won't turn off after timeout if an external display is plugged in via HDMI, 2) Linux S2 (Suspend-to-RAM) never wakes up, 3) Builtin mic don't work, 4) Fingerprint scanner crashes when you put your finger on.
I guess it takes time for those engineers in Lenovo to address those problems, and Lenovo is not a Linux-friendly company (they are more like a Linux-meh company).
Framework on the other hand, cares about Linux more. Sadly they don't operate in my country :(
I did have plenty of freezes on my old 2500u machine though. Was very disappointed. Thankfully, those days are long gone
Plenty of people have new Ryzen laptops that work just fine.
- Kernel logs show that the GPU driver crashes from time to time. When it happens the screen freezes for a few seconds but recovers.
- HVEC hardware decoder gives green artifacts on VLC using VA-API.
- It seems not to support advanced power management, on battery the CPU will happily boost to 4.7 GHz which in my opinion makes little sense. AMD has been steadily pushing work related to power management so I expect it to improve. As it stands total power consumption on battery averages 5.1 W.
The other tool you could check out (and use in conjunction or instead of ryzenadj) is a tool called auto-cpufreq: https://github.com/AdnanHodzic/auto-cpufreq - it lets you switch governor and turbo/clock speeds automatically base on battery if you're just looking for a simpler way to set that.
5.19 has a bunch of amdgpu fixes that might fix some things https://lists.freedesktop.org/archives/dri-devel/2022-April/... but sadly there are often also sometimes regressions/weird bugs: https://gitlab.freedesktop.org/drm/amd/-/issues?page=90&sort...
Of course, the next question is the potentially asymmetric RAM on the ThinkPads with one channel soldered and one SO-DIMM slot, e.g. the 48 GB version. Will it use dual-channel until ~32 GB and then single-channel? How will the GPU perform, e.g. will it use only one slot in such a configuration. So many questions...
This to me sounds like the experience people were praising the M1 chips for, and tbh Intel actually delivered for once. I say this as someone who had been solidly in the AMD camp for a long time. Thanks to competition, Intel now has to be good. Weird!
Also Samsung laptops don’t last in my experience. They make short term tradeoffs at the expense of long term reliability.
I haven't had trouble finding an outlet for years, even at airports, planes, coffeeshops, etc.
Not to mention your laptop not getting hot enough to boil eggs or sounding like a 747 when you open too ,any Chrome tabs.
Maybe that's a skewed perspective though. I have yet to try the Pro/Max chips seriously (or the mobile 12th gen chips, incidentally), but I don't really find myself pining for them either. If my laptop is too limiting, I hop on Tailscale and dev on my desktop.
Meanwhile my MacBook from work gets risking-burns-hot from just screensharing on a Zoom call.
I'm willing to put up with all the downsides of this laptop and the OS, just because of the battery life.
I get around 6 hours from 100% with a light Webpack dev workflow + Docker for Mac running in the background (arm64 postgres and redis containers).
Is that normal? powermetrics shows package power bouncing around 500mW - 10 watts when compiling, which would suggest I should be getting a lot more battery life. Is there a utility to reliably check power consumption of the whole system? Stats.app seems to think my system idles around 5 watts and goes to 40 watts when compiling which is much higher than the package power and my screen is only 1/3 brightness.
Not really. My M1, used for email, browsing, some xterms, some emacs, 3 or 4 video meetings and the Goland IDE, only gets about 5 hours.
When new it was a little longer than that.
My compute[1] is done locally only. My personal laptop is an HP Victus, it gets around 7 hours because I don't compute locally, which makes the M1 look okay in comparison.
I always wonder about these people who claim the M1 can last a full day without needing to be charged - they can't be developing on the thing, can they? If they're not doing computations on the thing, then the battery life is quite similar to all other non-gaming laptops in that price range.
[1] If by 'compute' you mean compiling code, running it, debugging it, etc
ETA: Comparable laptops reviewed by that site get 8hr runtime from a 71Wh battery and the P CPU, or 13 hours from a 60Wh battery and the U CPU.
It's like a constant drilling sound that gets into your brain and makes thoughts flow like through the mud in the full sunlight.
For that reason, despite that I don't like Apple and I am not a fan of macOS, I decided to buy M1 MBP.
This is a lifechanging experience, that laptop. You can even put it on low power mode and won't lose much perfomance and it is absolutely silent. Even if you run a prolonged heavy task and fans kick in, you will barely hear them (ha hope it won't get worse over time and that there will be a way to disconnect them if that is going to be a problem).
Oh and no coil whine. Even if my XPS 15 miraculously don't have fans on, you can be sure their noise is inconveniently replaced by distinct coil whine that is just as annoying as fans.
I am probably going to buy another macbook, this time the Air one just for leisure and learning.
[Edit] My understanding is, that a lot of power efficiency and performance comes from the large cache - not only ARM - which is more efficient than main memory and faster - also see AMD X3D. (per core AFAIK) Ryzen 7 L1/L2: 32/512, M1 L1/L2 192/3072
Samsung's internal processor division seems to be sucking. Google is trying it half-heartedly like they do with nearly everything. Nvidia and AMD haven't really tried to compete for the desktop ARM market, and Intel certainly hasn't. Qualcomm has rolled their mobile chips for laptops, but they remain mobile chips in a larger form factor. So Apple is in a class of its own for desktop-grade ARM CPUs.
Chromebooks are mostly low price, and may lack the software to truly push the chips often (without dropping into dev mode).
Windows on ARM seems to be a near failure. Very little native software and x86(-64) emulation is far slower than Rosetta. May not matter, they seem to be low cost machines anyway.
Linux just isn’t a big enough consumer market to drive it.
Server seems to be the big front for ARM, but that’s a very different chip. So I’m under the impression ARM laptops (outside Apple) are just phone chips, which don’t compete on the same level.
I'm very happy with my SP X running insider Windows 11 and WSL2. You couldn't pry it from my dead fingers for travel use right now.
Not saying they won't turn out to be, but they are still young considering the space they are in.
Similar holds for any other hw maker that is only starting out!
I've bought hardware before though from companies I'd never heard of, and either never received it or when I did it was much poorer quality than advertised or was so long after the order that I'd nearly forgotten about it. Would hate for that to happen with xcalibyte
That’s not true in any way?
And what does “Everything is in one chip” even mean? Because the memory certainly isn’t, it’s soldered on the package but it’s not part of the package, it just doesn’t take additional room on the main board. And there are a bunch of other chips on the mainboard
Finally, it’s pretty much just following mobile / phone chip SoC design, so any other manufacturer could do the same, if they wanted to create a giant and expensive SOC.
And I want to be really clear on the “giant and expensive” part: the M1 family is the sort of scale you usually see on giant workstation or server chips, the M1 Pro has 33.7 billion transistors, that’s more than a Threadripper 5995WX, a $6500 64 cores CPU. The M1 is just short of the 5950X’s transistor count (16 billions to 19.2), the M2 is above (~20).
I guess he's mainly thinking of GPU, which isn't unique. But there's not that many SoCs with that amount of power in one SoC. So it's close to competing with alternatives with discrete GPUs, which does increase power consumption.
I believe it has integrated Flash controller too, which is very unique for a laptop/desktop chip, no?
> Because the memory certainly isn’t, it’s soldered on the package but it’s not part of the package, it just doesn’t take additional room on the main board. And there are a bunch of other chips on the mainboard
It's on the package so it can be as close to the SoC as possible. That decreases the capacitance of the traces, which decreases power consumption.
It's not stacked on top of the SoC, which might have been even better (but harder to cool), but it's close.
> Finally, it’s pretty much just following mobile / phone chip SoC design, so any other manufacturer could do the same, if they wanted to create a giant and expensive SOC.
Uh, yeah, anyone could copy the M1 for a laptop/desktop product. But they haven't exactly done that yet have they? That's kind of the point?
> the M1 family is the sort of scale you usually see on giant workstation or server chips
Yeah, which again, almost never packs the kind of functionality Apple does into the M1.
With that many transistors, on such an advanced process, you're going to have a lot of leakage currents, so Apple must have put an impressive amount of work into power management.
I mean, it's not just about being an SoC with lots of things packed within the chip or extremely close to the cheap (memory). No. It's apparent that they've focused on across the entire design process. Even the choice of ARM factors into that (fewer transistors needed for instruction decoding). But I wouldn't say the original comment is completely wrong.
This isn't true AFAICT.
Cezanne SOC topology: https://cdn.mos.cms.futurecdn.net/ix6FrojFD7anKypadF2vKM.jpg
Rembrandt SOC topology: https://cdn.mos.cms.futurecdn.net/8wnuFCokmeHjESuSbzYWMJ.jpg
Even on the desktop processors, there are onboard lanes dedicated to the NVMe itself, and the entire processor (including NVMe) is capable of booting without a supporting chipset at all - the "X300 chipset" is actually not a chipset at all, it is just using the SOC itself without an external chipset, and you do not lose NVMe functionality.
Not sure if you are using some really weird meaning of "NVMe controller" that doesn't match what anyone else means?
(unless it doesn't - eMMC doesn't normally have a flash controller and you just do direct writes to the flash cells... as do some IOFusion PCIe cards that are just flash directly attached to PCIe with no controller. Sometimes that functionality is just software-based. Flash cards (eg microSD) usually also do not have a flash controller directly either.)
Anyway, it's true though that Apple does push the flash controller functionality into the SOC while AMD does not, Apple implements their SSD in a non-standard fashion, that's why it's not compatible with off-the-shelf drives. The flash is just flash on a board, I don't even think it's NVMe compatible at all either.
So if you want to be maximally pedantic... neither does Apple implement onboard NVMe controllers, just flash controllers ;)
The above also applies to eMMC and UFS storage as found in devices like smartphones.
pretty sure that's false as well... can an X300 system boot a NVMe drive? I'd assume yes. It does that with a UEFI which lives... where? Oh right, on the chip.
Most NVMe SSDs don't provide legacy option roms anymore either.
No, UEFI is firmware stored on a flash chip on the motherboard, usually connected to the CPU through the LPC bus. CPUs do not have any nonvolatile storage onboard (unless you count eFuses).