Snapdragon based Framework laptop to address fan noise problem

Now that we have seen that ARM processors can be successful in laptops, namely the Macbooks and perhaps the Thinkpad X13s and to a lesser extent the fanless Pinebook Pro, doesn’t it seem reasonable to consider using a laptop-grade ARM such as the Snapdragon 8cx Gen 3 in a Framework laptop? (This is in the Thinkpad X13s.) The main thing that is holding me back from buying a Framework is the fan noise, which is due to the 12th gen Intel CPUs, although truth be told the 6000-series AMD Ryzens have the same problem. It seems reasonable then to switch to an ARM as these naturally run cooler than CISC processors.

3 Likes

Use case: No heavy work ~ Disabled tubo ~ set may frequencies etc. 6 hours watching video many times. I have an i7 1165 and no fan noise, except when updating the bios as control is disabled at that point.

But yes maybe an ARM one distant day but for now put it out of reach on not your arm but your mind :slight_smile:

If you use Linux definitely try https://github.com/TamtamHero/fw-fanctrl

1 Like

I suffer from the noise, too. Sitting in the library and running a short (45seconds) meshing in cfMesh/OpenFOAM is a guarantee for being hated :wink:

But I hesitate to install this tool as I am afraid of melting my CPU.
Are there safety measures for preventing a damage due to overheating?

I’ve actually started under clocking my Framework. Taken it down to max 2.4GHz.

Still doing what I want it to do.

Does anyone know, was fan noise a problem with the 1135G7 based Framework? I’ve read a few reviews of other laptops (e.g. Acer Swift 3) where they say that CPU ran quiet.

I’ve personally tested three two non-FW laptops that had the 1165G7, 1235U and 1240P and each of them was anything but quiet. They were as loud as the Core i9 Macbook Pro.

As of late, the only CPU I’ve encountered that was quiet while in Windows performance mode and running Geekbench was the Ryzen 5 5650U (Lenovo X13 Gen 2).

If the latest generation CPUs from Intel and AMD are too hot, something has to be done. They’ve engineered themselves into a corner, the way that Intel did with the Pentium 4.

Wouldn’t something like the upcoming AMD Mendocino be a more elegant solution? For those that don’t know, it’s four Zen2 cores paired with two RDNA2 compute units on a tiny 6nm die.

This would also make for a great budget Framework offering… at least in the future since I think it might only be compatible with DDR5. Though, at whatever point Framework switches to DDR5, Mendocino will probably have been succeeded by like an even tinier 4nm Zen4 based-equivalent or something (or heck, maybe even a Zen4c design with “only” 8 cores instead of the usual 16 that Zen4c uses per CCD).

I really don’t think AMD’s are as, the more you reduce the power, the greater performance-per-watt AMD gains compared to Intel.

Also, with the sub-ideal situation of chipset fans on the launch versions of AMD X570 motherboards, it became clear to me that one of the industry’s dirty secrets is that a small & noisy fan is cheaper than a similarly-sized, properly-finned, passive heatsink, and it’s not until you get to ridiculously low power consumption such as those seen on phones and tablets that the minimal amount of heatsink material can be cheaper than a fan (heck, even on Apple, the heatsink they use on their fanless Macbooks is absolutely dinky).

Fun fact - I can run a Ryzen 4800U (eight Zen2 cores + eight Vega compute units @ 7nm) completely passively on a OneXPlayer (it’s a handheld gaming PC like the Steam Deck) even under load if I set Linux’s CPU governor to “powersave” which seems to cap it to something like 1.4GHz at 0.75v, and I can even get away with the “conservative” governor setting if I don’t keep the CPU under sustained load for too long.

Honestly I get the feeling that no matter how many people in the forums proclaim “I want this, I want that”, Framework seems to be inclined to change very little about the laptop, and they seem to be wedded to Intel.

But if I were them, I’d at least offer a 9W Intel processor like the i7-1250U for the people who don’t want a computer that sounds like a jet engine.

That said, yes the Linux settings can help. You can also use cpufreq-set to change the frequency ranges per core.

1 Like

As I stated in the AMD CPU thread, it could be simply due to wanting to continue to use DDR4 due to the upgrade path it allows from 11th gen, and the ability to upgrade and replace individual parts is kind of the Framework Laptop’s entire point.

So I’ll say the same thing here that I said in that thread: I would not at all be surprised if AMD variants don’t appear until Intel no longer supports DDR4.

Virtually every AMD laptop I’ve seen (and Snapdragon for that matter) has used LPDDR4/5, which is never upgradeable because it’s soldered. So it is plausible they’re rejecting AMD because it won’t be upgradeable and as you assert, upgradeability is their selling point. However that wouldn’t stop them from just providing a large enough one-size-fits-all amount of LPDDR5 like 32GB to satisfy 99% of AMD buyers.

Some more rationales for avoiding AMD might be

  • they don’t want the hassle of another set of unsold motherboards when it turns out some AMDs aren’t popular.
  • their motherboard supplier is more comfortable producing and testing Intel than AMD.
  • their firmware supplier isn’t writing EFI / BIOS software for AMD.

In addition, they may have done their research and learned about how AMD Ryzen 6000 CPUs have been exhibiting some problems. I’ve seen hardware errors myself involving the L2 cache. Others (Thinkpad buyers) have complained about a range of issues with AMD 6000 based Thinkpads, as you can read about here:
https://www.reddit.com/r/thinkpad/comments/wfk5c4/thinkpad_p16s_amd_problems/

I know at least the HP Dev One and the HP ENVY x360 uses SODIMM DDR4, so such things are definitely supported.

What I meant by upgrade path is that Ryzen 6000 does not support SODIMM DDR4 at all, so retaining that DDR4 upgrade path for 11th gen Intel Framework owners was quite possibly an important key point considering that Intel 12th gen supports both DDR4 and DDR5 and, sure enough, Framework went with DDR4. So whether Ryzen 6000 even supports SODIMM DDR5 rather than just exclusively LPDDR5 is kind of a moot point.

Thing is, Ryzen 6000 is the first AMD CPU with USB4 support in any capacity, so it’s possible that Ryzen 5000 and older are just non-starters, while Ryzen 6000’s incompatibility with SODIMM DDR4 makes it a non-starter as well.

Nevertheless, I’m definitely aware of things like the USB4 troubles on Ryzen 6000, not to mention the Pluton shenanigans where Microsoft mandates that Pluton-enabled devices sold with Windows 11 pre-installed must have 3rd party certificates disabled and, if you want to install Linux, you must have 3rd party certificates enabled - and Framework is obviously very Linux-friendly.

(more info on the 3rd party certificate thing: https://www.phoronix.com/review/rembrandt-linux-boot)

1 Like

Ah thanks for that detail. I wasn’t aware of it. I guess I’ll be avoiding Pluton-afflicted laptops from here on out.

That might be hard to do though. I observe that all new Thinkpads have Pluton. Not sure about the Linux-specific ones. Also the new Acer Swift Edge does as well.

Update: It seems Pluton is included in the SoC, so all 12th gen will have it, including Framework’s laptops.

I would say fw-fanctrl is pretty safe because even with the laziest config your fan speed reaches 100% when temperature is 85 degrees. It just allow you to run at a reasonable temperature without much fan noises. When you really need your fan to speed up you can always adjust the config. I would also underclock my cpu when using battery to further reduce fan noise and battery life using tlp.

Time has come.

Is this the moment to re-live this request?

Out of curiosity, what OS would you guys be running if there were an ARM-based chipset available?

Microsoft’s offering is in a relatively poor state as it is, their x86 emulation is not up to par. The Linux side might be better but there are still plenty of projects that don’t have native ARM support built yet, or it’s not quite baked fully.

Changing a processor architecture isn’t as simple as popping out an Intel/AMD chip and dropping in a Qualcomm one. Apple made it look easy mostly because they control both the hardware and the software in one package. And also they make their money off making things look easy.

1 Like