Macbook M1 pro/M2 pro or Dell (with Ubuntu/other linux distribution) for bioinformatics stuff?

Question

[l.gallucci]

Hi!

I'm in the phase of choosing a new pc for work stuffs for my PhD, I'm working basically on Amplicon, metagenomics, and various things related to microbial ecology stuffs. I'm undecided among a macOS system (with M1 pro or M2 pro) or a Dell with similiar computational power (we talk about around 16/32 GB of memory and general powerful characteristics). I need to specify to the department in my institute which one I could prefer. I will use it for analysis in python/bash/R and using programs like FastQc, Cutadapt, Trimmomatic, Kraken2, Barrnap, Porechop and similar. In addition to this we have powerful HPC server on which usually we perform more important computational work. I want to make me free from server work for base/medium weight work. My doubt is based on compatibility of M1/M2 processor (in particular the pro version) with software and tools. What you suggest?

[Chris Dean]

Most of the bioinformatic work will probably have to be offloaded to the HPC, especially for the shotgun metagenomic data. The statistical analysis and data visualization component can be done on any of the machines you described in your post. I personally use a 2020 M1 Macbook Air with 8GB of RAM and 256 GB SSD. The machine is currently going for 750 USD new. I am extremely happy with the purchase.

[Sasha]

Yup - this exactly.

Plus if you're using any tooling outside of the bioinformatics universe; usually those tools run very well on Macs.

[ATpoint]

I would not use a Mac these days since they're hilariously overpriced, not upgradable and need another layer of VMs to run something like Docker and Singularity which I use to manage software versions. Any Windows machine with WSL2 to have native Linux (I use standard Ubuntu 22) is a good choice. So in a nutshell, I'd go with the Dell.

[steve]

I am not sure that "overpriced" is a valid criticism anymore. Ever since the introduction of M1, Mac (and especially MacBook) has had better performance compared to x86 systems at the same price point. And that is before considering the power efficiency which has allowed for truly full-powered fanless systems, and even the MacBook Pro with a fan is pretty much inaudible, combined with a 20hr battery life and superb build quality. "Upgradeability" is a bit of a red herring as well, since a lot of the incredible performance comes from having the entire system on a single unified chip. I dont think its a valid criticism. Especially when most users never upgrade their "upgradeable" systems either for the life of the devices.

[l.gallucci]

Thank you all for the reply. I think that I will choose a MacBook pro, since it is well integrated into the institute's ecosystem and many here use it, having full support from IT with a series of software and any hardware problems/faults. As some of you also suggested is a matter of choice and preferences, I read carefully all the reply and is always fascinating the different way in which different people approach to this things, so really thank you. As the institute will provide me this machine, I think that I will buy a personal laptop (I think I will choose a Dell) to be more free with other things and also in relation to managing of personal data and research data with are not linked to my actual institute.

[Ram]

That's a good call. It's been a while since I had a different OS on my home machine vs work machine, so I'm no sure if this point is still valid, but context switching between Windows and macOS machines can make for a bit of frustration. Hitting the Win key because muscle memory says it's the Cmd key is super frustrating.

[Alex Reynolds]

I suggest you go with a Mac. You'll get to do all the same science, it'll look great, and you'll annoy the Linux and Windows ppl, as they deserve.

[Mensur Dlakic]

To look great and annoy others may be one of the main reasons people spend extra money in general.

[GenoMax]

ATpoint has provided reasons for what to choose in his comment. It is your choice and money at the end of the day.

My doubt is based on compatibility of M1/M2 processor (in particular the pro version) with software and tools.

At a technical level all software (in theory) should work on M chips subject to limits of unified memory architecture M chips use (i.e. you are permanently limited to RAM you choose to buy and can't add more later). Much of software should be installable via conda.

[steve]

surprised that no one has mentioned Rosetta 2, which lets x86 software run seamlessly on Mac's ARM based M1/M2 CPU's. You can docker run containers built on x86 Intel on your M1 Mac and it will just work as if it was native, very few issues. The same goes for a lot of other legacy software that doesnt yet have a native ARM build available

Answer 1

[Istvan Albert]

MacOS M1 is probably the best system you could purchase - from the build quality, consistency, reliability, user experience, screen quality, quietness, etc. - no other system I have ever owned comes even close.

All of our bioinformatics staff works on iMacs to work out pipelines and then log into HPC like systems to execute these on a larger scale.

WSL is an inferior choice because you have to constantly put up with the various limitations.

No question that WSL is a major step in bringing Linux to Windows, but it is still an emulation layer. Lots and lots of things can go bad as the permission systems, the UNIX flag system are not quite compatible, you can't run graphical UIs from Linux etc. Then you find yourself ending up with two different versions of R one in Windows one in Unix ... it is a constant battle of keeping two systems in sync

The only limitation that I have ever experienced is the issue that ATpoint also mentioned, that the singularity containerization does not work on Macs.

PS: I will also admit that I am a bit disappointed that the latest iMacs come with 24'' screens, whereas the previous generation had the 27" inch option. I'm so used to the 27" retina display that I will keep using the old-school Intel Macs as long as possible.

[steve]

the biggest issue I have experienced with WSL, which is pretty relevant here, was that of storage space. Last I checked, the default WSL installation volume is only 250GB, and resides on your OS install disk. Trying to move it and resize it seemed like such a headache. And trying to do fast high-throughput file IO onto another larger mounted volume ran into strange disk bandwidth issues related to the virtualization-type stuff going on. Overall it was fine as long as your data fit into the 250GB default volume on the OS disk but if you wanted something like a fast secondary NVMe scratch disk, it proved to be a big headace and had horrid performance. Maybe it could have been fixed with some configuration changes but at that point you might as well just ditch Windows and go back to real bare-metal Linux installation

[jared.andrews07]

I'm not sure how this is different than any other OS or setup. Yeah, if you have lots of data, you will need more storage. The WSL installation volume will expand automatically up to 1 TB, and can be expanded beyond that as you want. I have also mounted many different drives and have noticed no IO difference at all between Windows and WSL.

[steve]

if I am building up a local workstation, I typically do not need a huge volume for the base OS installation (e.g. Windows boot disk), and I definitely do not want tons of heavy IO happening on that disk in order to preserve SSD lifespans. At least over the past years, it was very common for me to build the base system with only a 256GB or 512GB boot disk meant to hold only the OS, and offload all data on larger secondary volumes. SSD prices have dropped a lot the past year+ so this might not be as big a deal on a new system.

I dug up these links to info on managing the WSL volume, for reference;

• https://learn.microsoft.com/en-us/windows/wsl/disk-space
• https://medium.com/@sky.tienyu/move-wsl-to-another-drive-e548a264953c (strange that I havent found first-party docs on this easily)

And I think this was the disk IO performance issue I was running into. It seems to be a niche issue that might only crop up under certain circumstances.

https://github.com/microsoft/WSL/issues/4197

[jared.andrews07]

Huh. I may not have run into that issue locally because most of my big data is located on network drives, where my guess is the network is often the bottleneck rather than my PC.

[GenoMax]

You could use your 27 iMac as a target display with a Mini or Studio and join the M* revolution.

[Mensur Dlakic]

PS: I will also admit that I am a bit disappointed that the latest iMacs come with 24'' screens, whereas the previous generation had the 27" inch option. I'm so used to the 27" retina display that I will keep using the old-school Intel Macs as long as possible.

Decidedly a first world problem. I don't want to sound judgemental, but it strikes me as less than a major issue whether the display is 24" or 27". Maybe that's what a bit disappointed means as well.

[jared.andrews07]

No question that WSL is a major step in bringing Linux to Windows, but it is still an emulation layer. Lots and lots of things can go bad as the permission systems, the UNIX flag system are not quite compatible, you can't run graphical UIs from Linux etc. Then you find yourself ending up with two different versions of R one in Windows one in Unix ... it is a constant battle of keeping two systems in sync

The bolded statements are patently false, just for the record. WSL2 runs a native linux kernel and most common GUI applications work fine. I also don't know of any Linux-specific GUI applications that are not available for Windows with any relation to bioinformatics/'omics. I don't know what permissions issues are being referred to here, but the previous Windows <-> Linux filesystem issues from WSL1 have been largely resolved for years now.

That isn't to sway you one way or another, just facts.

As for my actual opinion, I've used WSL since its release and had pretty much zero issues. I agree with others though that most big computing you do will likely be on an HPC or other shared computational resource. And clearly plenty of people use Macs with great success. Personally, I don't find them worth the price tag given the absurd upcharge they tack on just for decent memory.

[steve]

whereas the previous generation had the 27" inch option. I'm so used to the 27" retina display that I will keep using the old-school Intel Macs as long as possible.

some of my friends are very upset that the IT department is forcibly retiring these now due to no longer meeting OS security requirements or some such business. Very sad. I keep hearing "mac rumors" about an upcoming M1/M2/M3 27" iMac, who knows if it will ever happen, I guess the best thing right now to replace it is simply a Mac Mini or Mac Studio plus Apple Display of some type.

Answer 2

[steve]

In addition to this we have powerful HPC server on which usually we perform more important computational work.

This is the really important part of the situation.

The ideal configuration, from my experience, has been to use a macOS system to ssh in to your institution's HPC (or cloud). iTerm2 + CyberDuck + VS Code is pretty much all I have needed for many years. I have got some notes here with the typical extra software I ended up using in my macOS systems, among other things.

I think the question in the OP is itself a little misguided, because in real life you will pretty much never want to be using Windows for bioinformatics (genomics) work. However if you had a Windows PC you could just install Linux (such as Ubuntu) on it and have a much more appropriate system. But at this point, I would still have a hard time recommending it to anyone for their lab's machine, simply because you want your lab workstation to "just work" and not require a tech-head to know how to manage a local Linux workstation. Not that its hard, especially if you are gonna be using Linux on the HPC anyway, but I really think its important that if you are requisitioning something it needs to be as easy as possible and reliable to use. macOS is really what you would want here. Others mentioned the ability to use MS Office; I think this is a critical consideration as well. Also consider the ease of integrating the system into your institution's IT infrastructure. I run a lot of Linux systems at home for personal use, and despite the huge strides that Ubuntu has made in their out-of-the-box experience and driver support, there's still occasional hiccups that I would loathe to deal with in a lab setting.

use it for analysis in python/bash/R and

Python and bash are by far the easiest to work with in macOS over Windows. R (with R Studio) is surprisingly painless on Windows and the experience is about the same on both. Linux makes all three trivial.

I want to make me free from server work for base/medium weight work.

I do not think this is really a worthwhile expectation to have. Just about the only work that I have ever experienced that was feasible on the local system, instead of on the HPC, was downstream prototyping with local R and Rstudio, which would ultimately get back ported to the HPC for final execution. But no matter how beefy of a workstation you get, Mac or PC, you will simply never get the type of throughput that is available on the HPC, to the extent that it makes it not worth your time to bother trying to do much work outside of it. Also consider the headaches involved in keeping your analysis' software configurations in sync between two different systems. I think its far more worthwhile to do two things;

• petition for some sort of reserved or dedicated resources on the HPC (or other system such as a lab-specific compute server)

• get a laptop (MacBook) so you are not chained to your lab workbench

For software, conda has pretty much got you covered, especially if you stick with newer package versions (some old packages dont have M1 versions available for macOS), and you can pull and run pretty much all Docker containers flawlessly. Building containers on macOS ARM (M1, etc.) to run on x86 Intel HPC is a bit of a headache however. Singularity does not run natively on macOS or Windows, but that is not really an issue if you just stick with Docker (and convert to Singularity from within the HPC environment).

Ultimately my best suggestion is to get a MacBook Pro with the most memory and storage space you can budget (32GB / 1TB or 2TB would be good to shoot for), and utilize the HPC for most everything possible, and utilize network storage for your lab's data archives (dont store all your research data on the laptop). Also worth considering that if you want to be mobile, the 14" laptop models are significantly lighter and easier to take with you e.g. to a conference or to the breakroom to do work while you drink the professors' coffee in the faculty lounge.

[Mensur Dlakic]

Not sure if you realize it, but most pros you gave for Macs are either your own bias, or matters of convenience. Given your final suggestion - to use the HPC for heavy work and a laptop to manage connections - one could do all of these things with a PC laptop that would have 64 GB RAM and 4 TB disk space and still cost less. As to carrying around something that is 2 or 3.5 lbs, that is decidedly a matter of convenience. I am yet to see anyone who developed a chronic back pain or knee problems by carrying around a laptop, regardless of its exact weight.

[steve]

matters of convenience

yes that is exactly correct. Because in the lab you need things to "just work", and that is what macOS and Mac do best. Trying to run genomics on a PC laptop with 64GB and 4TB is a fool's errand. Offload that to the HPC or a cloud server. You will not get any significant amount of work done on the laptop and you will end up waiting around for hours and hours to do a single analysis step (out of a dozen+) on a single sample (out of a dozen+) while your laptop is burning up and fans spinning loud as a jet engine and you cannot do any other work the whole time. Nothing of value is going to get done this way on your laptop when you could have just ran it in a screen session on the HPC or submitted it as a e.g. SLURM job or similar, and then retained usage of your laptop to orchestrate more activities. There is simply no way to recommend in good faith that someone like a lab student should be bothering with such things on a laptop, and in a Windows or Linux desktop environment no less. To make such a suggestion is to simply set them up for infinite headaches when all they wanted was to get their Thesis project done. macOS surely strikes the best balance between having an OS and filesystem that is nearly seamlessly similar to the HPC environment and OS you'll be doing the real "work" on, while having a robust local software ecosystem itself and still having access to the standard Microsoft Office Suite to facilitate collaboration with your colleagues and integration with your department's IT, in addition to the fantastic build quality across the board. As far as local laptops and workstations go, of course you can get higher spec systems and you can get faster or cheaper systems, but they all come with significant trade-offs and restrictions especially if your end goal is to facilitate genomics in an academic setting.

[Mensur Dlakic]

Trying to run genomics on a PC laptop with 64GB and 4TB is a fool's errand. Offload that to the HPC or a cloud server.

I don't think you read my replies carefully. I said:

Given your final suggestion - to use the HPC for heavy work and a laptop to manage connections - one could do all of these things with a PC laptop that would have 64 GB RAM and 4 TB disk space and still cost less.

So I agree with you that everything should be moved to the HPC. I am suggesting to buy a PC with more memory and a larger disk only to connect to HPC where the work would be done. Other benefits of such a PC are to play games, be able to connect to non-Apple devices, and in general save some money compared to buying a Mac. There is no doubt that Macs are better computers than PCs if one doesn't care about money, gaming, general connectivity to other devices, and Apple hegemony in particular. Yet for those of us who care about one or more of those things, better computer is not necessarily a better value.

[Ram]

be able to connect to non-Apple devices

Can you elaborate? I use TeamViewer and it connects to and from everything.

I feel like both you and steve are coming from a bit of an outdated perspective on this. Yes, Apple had a monopoly on the OS yeas ago. Yes, Linux on Windows is a little more complicated than on macOS. But there is no problem connecting to most devices from a MacBook Pro, and WSL on Windows is nowhere near as much a pain as dual booting or using VMs used to be.

A Mac works best for me because I need a powerful machine with a much closer linux integration than WSL, and I need to run local apps that perform insanely fast, and most of all, I'm used to macOS. PC would be a lot less expensive but laptop money is not a concern at work for me. Also, using PC would give my IT a LOT more control over my machine and I would hate it.

To the OP, I'd recommend either a medium powered M series MacBook Air or a PC with higher power that one can get at a similar price range. Buy a machine with a slightly conservative estimate and focus on getting a much closer match for your next computer. Only you can factor in your personal comfort and how much that matters because in the end, both machines will be good enough at objectively doing what you need outside of a few select possible deal-breakers.

[Ram]

Cyberduck looks interesting but pricey. For my cross-storage-locations purposes, I prefer rclone. It's got an rsync-y feel to it and connects to almost any remote AND is FOSS, which is awesome.

[steve]

Cyberduck is free. But it will ask you to donate occasionally. Since I have used it daily for years I just paid the fee to help support them, their team does good work.

[Ram]

I visited the Mac App Store page for it and it was priced at $23.99, which is why I balked a little. I'll check out the free version, a GUI might be refreshing for a change.

[steve]

the link is here https://cyberduck.io/download/ :)

Answer 3

[Mensur Dlakic]

There is little doubt that Macs are better computers than PCs from an objective standpoint, yet I haven't had a Mac since 1999. It has to do with bang-for-buck reasons. For users who care about convenience, sleekness and style of their computers, and for those who don't care about Macs not working well with non-Mac computers and non-Apple-made devices, and especially if money is not a major consideration, Mac is your choice.

In a research setting, I would have a hard time justifying a Mac purchase because I have no idea whether it will work with devices I may want to couple them with a year or two from now. The number of free or freeware programs for PCs is staggering; haven't checked lately, but it used to be considerably smaller for Macs. That said, Macs are natively closer to Linux or one could even say equivalent to Linux, so for Bioinformatics applications some of the software issues may not be there for Macs. All things considered, I'd rather buy 1.5 or 2 good PCs than a Mac that may be better-looking and has a faster processor, but has a smaller disk, less memory and can be fixed only in 1-2 places in town.

To answer the specific question the OP posed, I don't think one should rely on a laptop to do this kind of analyses. I agree with Chris Dean that this should be farmed off to HPCs and dedicated lab computers. In this case I would still recommend a PC laptop as it will cost considerably less if we eliminate the esthetics and less relevant features for everyday use (to me) such as processor speed and retina displays.

Answer 4

[cfos4698]

FWIW, I work daily on a Linux workstation (Dell precision 'tower' running pop! os 22.04) and it's fantastic for bioinformatics work and much cheaper than buying a Mac. Our lab also runs the same OS on several Dell laptops successfully. In general we don't have many issues, but sometimes I do miss the sleek MS Office ecosystem for writing manuscripts, compatibility with colleagues etc.

Answer 5

[Zhenyu Zhang]

Our center has a policy of either macbook or dell ultrabook. Previously, less than 20% choose dell. Since M1 is out, no one choose dell anymore.

[steve]

I have seen the same trend elsewhere. I have also seen a fair number of "converts" turning in their Windows laptops back to IT and requesting MacBook's instead. I do not think I have ever seen the reverse. Pretty much everyone I have met has either just always used Mac, or once used Windows then switched to Mac and never went back. </anecdote>

Answer 6

[Almira]

I would recommend you to buy a MSI or Dell computer for bioinformatics work. Mac computers aren't compatible with some applications/softwares. Also MSI computers have high CPU and high GPU. For example, usually MSI computers are faster for doing MD simulations work than Mac computers. Also RAM capacity is important.

[Istvan Albert]

One problem with Windows laptops that I always ran into was that build quality varies immensely.

There are so many options they all compete on quality and price at the same time - but it is never clear which model is worth the extra money

When you have two identically specced systems, most people will buy the one that is $100 cheaper - because you can't tell whether the more expensive one is better built.

As an anecdote: students come to my office hours with their laptops - and it seems every student with a Windows laptop has to plug in after 5 minutes ... and they say the same thing ... my battery is crappy ... and sometimes they lug around a cooling pad.

An M1 MacBook Air is a technological marvel, thin and light, runs some 7 hours on a single charge, does not have a fan, does not get hot in your lap, does not make any sound. We should be in awe of that technology because after all, it helps make all laptops better if they want to compete.

PS. there was a time when I would have absolutely never gotten a Mac since playing games on my laptop was a form of entertainment that was necessary for my quality of life. I still play games, but I have a separate machine for that :-). I understand the sentiment of not wanting a Mac as well.

[Ram]

Mac computers aren't compatible with some applications/softwares

Any examples? This seems like a niche thing as the only piece of software I've had a problem with are Excel macros. And you know how I fixed that? By running Parallels on my Mac. You can run Windows on a VM in a Mac, but the other way around would be nowhere near as fast.

usually MSI computers are faster for doing MD simulations work than Mac computers

Are MSI computers any different than Dell latops? I've found that custom built PCs are slower than macOS machines - somehow, macOS machine components work together better. I think if you were to compare these MSI computers and macOS machines with matching configurations, any program would run faster on macOS than on Windows unless you're referring to old software built to be dependent on older Windows features.

[Almira]

I can give many examples but for instance autodocktools software and first of all, have you ever done MD simulation on Mac and MSI?

[Ram]

have you ever done MD simulation on Mac and MSI?

No I have not. I accept that there could be some Windows specific software, but most computationally intensive/distributed software will be written for linux as most HPCs use linux.

Also, I see AutoDock4 available for macOS here: https://autodock.scripps.edu/download-autodock4/

Am I missing something?

[Almira]

If you check it, that software is available only for some old MacOs versions not the current version Sonoma, Ventura or Big Sur. Currently, MacOs doesn't support 32bit softwares. Also if you check the processor speeds MSI has higher speed than Mac.

[Ram]

I see what you're saying. It feels like this is on the software products though - 32 bit software has been outdated for a while now, and these legacy products are yet to catch up. So I guess if one needs these software products, they must either use Windows or run a VM.

On the processor speed, trust me when I tell you that specs are not important when cohesion comes into play. You could have the best components and they would each perform at sub-optimal levels owing to small hiccups created by other components. This is why I use Apple machines - their hardware is extremely seamless. I could have an objectively slightly lower processor speed than on a Windows machine and still outperform Windows machines because the resources are managed so much better. Maybe Windows has improved and my experience is outdated but I've been on Apple machines for ~10 years now and I'm yet to regret a single device based on performance. I say performance because between 2016 and 2021, MacBook Pro machines were awful in terms of build quality and pro user feel.

[Almira]

I have been using MacBook Pro machines since 2011 and also used Windows machines. In my opinion, for compatibility with software and applications MSI is more useful. I can run both 32bit and 64bit softwares and the MSI machines' speed is fast not like HP laptops.

[Ram]

I might look into this MSI for personal machines, thank you. Are they seamless and performant in gaming?

Answer 7

[Jean-Karim Heriche]

Just to provide another perspective, I would suggest considering virtualization. For the last few years, my main computer has been a VM in a data centre to which I connect either from a Rasberry PI 4 or a Dell XPS 13 laptop (which I bought mostly for travelling and with Linux pre-installed). The advantage of virtualization is that depending on the setup, the resources can evolve with your needs.

[Ram]

The principal downside to this approach is the latency in forwarding GUI. Remote-connections might work well but again keyboard shortcuts don't translate well even with the best of remote software and VPN can grind things to a halt.

[Jean-Karim Heriche]

YMMV, so far it has worked well for me (shortcuts included) or I wouldn't have stuck with it for so long :)