Please help me make a decision

General Topics
1

Dear all,

I would like to ask your opinion on which laptop I should buy. I am currently in Taiwan, and I am deciding between laptop brands that have good Linux support. I will sometimes need HPC to run and simulate some models. I am considering whether framework 16 with GPU is a good choice. This will be my first experience of handling a laptop with individual GPU, so I don’t really know how to utilize that computer capacity for now. I guess there’s a way to utilize it in FORTRAN or Julia. Also, the competitor in my mind is Tuxedo computer and Dell computer.

One of the biggest concern I have comes from this video. It seems to me that there are some criticism on the building quality, battery performance, and keyboard:

  1. Seems like the overall building quality is not comparable to Dell and Macbook? Of course that flimsy feeling may come from the repairability of it, but it scares me when some videos show that if you open the laptop screen on one side, the whole screen bends like crazy. Let me know if the newer version resolved this issue, or this is still the case.
  2. Battery performance seems to be not optimal? What I saw on reddit says that it can only last for maximum 8 hours for extremely light usage, 4-5 hours for mid usage, and 1.5 hours for HPC. Is this still the case? Also, that video says that the 100W adaptor cannot keep up with the electricity need for GPU, and it will lead to the usage of battery even plugged in.
  3. It seems to me that the flexing keyboard issue is mostly resolved with the new kit launched one year ago. How was the keyboard typing experience compared to, say thinkpad?
  4. Last question seems to be obvious since I am asking in framework community, but how would Dell and Tuxedo computer compare to framework in your opinion?

Thank you so much for your reading, I appreciate your help into my decision making!

Best regards,

Hui-Jun Chen

2 Likes
2

  1. I don’t know the answer to it, but the work around is pretty simple. Open it at the center.

  2. I have not tested this. I am plugged into the mains most of the time.

  3. I have never had a problem with the keyboard. I probably got my FW16 after the flex fix.

  4. I don’t have a Dell or Tuxedo to compare.

But, other things to consider:

  1. You probably need to do work around for the sleep waking when you did not wish it to. (i.e. waking in a backpack). I have a script so that it only wakes with the power button. It works fine for me.

  2. Not all 6 USB ports are the same. Some can output more power than others.

See here:

Compatibility: Some USB devices (e.g. phones, SSD enclosures) are not compatible with the USB-C connections, but work fine on USB-A connections. Some don’t even work on USB-A. I.e. SSD enclosures that when the SSD is used a lot, it disconnects.

  1. My FW16 has no dGPU, just the GPU that comes on the APU (combines a GPU and CPU in one chip) Using the APU, you can run LLM models that take 96GB RAM. (If you install enough RAM) Whereas dGPU is limited to its onboard RAM. So, depending on your use case, you might not need the dGPU.

So, it really depends on what you need.

If you value the repair-ability, and can make do with a few niggles that might eventually get fixed, choose the FW16, but if you want a Laptop that just works and if there are any niggles with it, they will never get fixed, choose another brand.

For example, there are some Battery / Charge problems with the FW13/FW16 and I was able to fix the Embedded Controller (EC) source code myself so it works now exactly how I wish it to. Other manufacturers normally have only binary only EC code, so you cannot fix them.

One last thing.

There is a no-quibbles return policy for the first 30 days. So maybe just give it a try, and return in before 30 days if you don’t like it.

1 Like
3

just in addition, the framework 16 can be purchased alongside a 180W adaptor, which is enough to power the laptop and the extra gpu module (unless you are really pushing the system to its limits). a 100W power adaptor is all you need for just the laptop with an expansion bay shell instead of the dGPU (dGPU is an addon so the stock laptop can be powered entirely off 100W)

battery life is fine for me personally, laptops dont generally have good battery life to begin with.

also as the previous commenter said, some of the usb-c ports can output more than others, but the top 4 ports are the only ones that take an input, and they can take up to 240W (which incidentally is enough to power the laptop + dGPU at theoretical max power draw, its just that framework doesnt have any 240W power adapters because they did not exist when the laptop was first released, though it does support the specification for them)

4

Thank you so much for your reply!

As far as I can see right now, there is no GPU available for either DIY or prebuild right now.

If there is the case for the foreseeable future, may I ask your opinion on the choice between FW13 over FW16? It seems to me that the new FW13 CPU is more powerful than FW16.

5

Edit: in reply to the OP. Appologies to shadowblade8192.

All laptop brands can have issues when linux and HPC are concerned. Spend as much time as you can to define and what you need and then research the options based on your requirements and budget.

I have not directly worked in HPC, but have prior experience with HPC model “definition” and (experimental) validation. I have used linux since the mid 90’s. I am very happy with my framework 13 (for personal use) and would purchase it again… but consider carefully your choice especially if you have time to do so.

HPC is a very broad category and no one laptop (or any device) will perform well for all applications. Do you know if you will need to use intel’s MKL library (optimized blas and lapack) with a cpu? How about nvidia’s cuda library for use with a gpu? If your new to HPC (heading off to undergraduate university in the sciences), and must make a choice with limited time, budget, and knowledge, an intel cpu and nvidia gpu may be a safer choice for HPC applications.

Any well reviewed current gaming laptop (intel and nvidia based) in your budget range should suffice for undergraduate university. 32+ GB ram (more is better) and as much gpu memory as you can afford without purchasing a high end laptop 4090 or 5090 (due to thermal management and/or power issues - this advice comes from reviews I’ve read, use your own judgement). Cooling and not running into power limits would be key criteria for me.

For me, this rules out the framework 16.

As an example of the issues with a laptop, I run basic “cpu HPC” benchmarks in a self built optimized python environment using numpy (also self built and optimized including compiling either against intel mkl or openblas). When I compare my fw 13 results against an i9-10900k desktop (also running ubuntu), the desktop easily outperforms the much newer framework intel processor. Moreover, the desktop does better on sustained calculations (think hours at full load) due to better cooling options.

Note AMD has optimized blas/lapack libraries for its cpu’s but I’ve read these may be suboptimal for some applications. I’ve also read you can run MKL on an AMD cpu, but it likely will take more time to get it to work, intel likely wont provide support, and some problems may not run as fast on an AMD cpu (if it runs at all) as they might on an intel cpu using MKL.

The point is any laptop is likely a suboptimal choice for “HPC” except prototyping and model development. Even a top of the line consumer desktop with the latest gpu is likely a suboptimal choice for commercial work or graduate level university research HPC. That said, you can get pretty far with consumer hardware even on a 5-10k usd budget.

The problems with framework you read in the forums are real and you likely will run into some of them. Do you care more about the “clock on the wall” time to do your HPC work (as compared to the time for your model to run on a laptop)? If so, how much time troubleshooting your laptop are you willing to spend?

Lastly, I deliberately choose a lower spec intel core ultra based framework 13 due to the thermal, battery, and linux (mostly with amd cpu’s) issues I read about in this forum with either the higher spec intel or amd cpu’s. I also want to use intel’s MKL. The lower cost is a nice plus. I would use a commercial desktop or low end workstation (10-50k usd) for gpu work depending on the need and budget.

I am currently not aware of any other laptop on the market with frameworks combination for ease of repair, ability to upgrade, flexibility with ports, and linux support. I think these are the reasons to purchase a framework but only if you are willing to tolerate frameworks growing pains.

1 Like
6

Hi,

Regarding the inability to order the dgpu with a new laptop. FW are a bit odd with the way their website ordering works.

If it is out of stock, they don’t let you order it. You just have to wait until it comes back into stock before you can order it. I expect FW are loosing a lot of orders because of this.

I expect you, like me, would have liked to at least order the item, even if you had to wait for it to be restocked. It is far less effort to purchase that way.

7

Eventually I bought FW13 for its portability. Hopefully it will work well with Arch Linux!

8

How is that not optimal? If you have a laptop that can use 100W and a 100Wh battery, you can kill the battery in an hour with an HPC use case. FW16+dGPU can use much more than that and has a smaller battery, so 1.5hrs is actually very good in this case :sweat_smile:

That’s physics and it’s the same no matter what laptop you have. I can kill a 16 inch MBP battery in 1.5 hours too…

Laptop batteries only last all day if you are not using them. Hardware accelerated video playback, light browsing or text editing (dependent on the browser and editor), low screen brightness, and you can get through the day with a FW16

But if you are using GPU compute on any laptop, bring a charger

9

OK now I made a decision to buy framework 13 AMD AI 300, and I also bought Adata ram. Now with both ram inserted I cannot even boot. Only when the ram is inserted to the left slot it can boot.

What a shame. Will probably ask for a refund.

10

Hi,

Which RAM chips (exact part number) did you install? Are they from the list above?

There have been some cases of the RAM delivered from 3rd parties not matching what was ordered. So it is worth checking.

There is also a “memory training” step, that can take 15mins in some cases. So, just waiting might help.

11

Yes I am aware of that step and have waited at least 30 mins.

This is the ram: ADATA

AD5S560032G-S

32GB DDR5-5600 SO-DIMM

OK - Limited Validation

I don’t know what limited validation means.

12

That RAM should work OK.

Raise a support ticket via the FW web site.

1 Like