I am a new PI looking to purchase a desktop computer for running QIIME2. I have only ever used QIIME2 on institutional servers so I am unfamiliar with the hardware requirements and am looking for recommendations. I would prefer a PC (and would install the Windows Subserver for Linux), but am open to other options. I hope to be able to use this computer for a long time with out needing to upgrade components. Thank you for any recommendations!
1 Like
Hi @ashore,
I’m a heavy QIIME 2 user, and just to give you a data point as a Mac owner, I have successfully run QIIME 2 on:
- Macbook Pro 2015 (8 GB of RAM, intel chip)
- Macbook Pro 2020 (8 GB of RAM, intel chip)
- Macbook Pro 2020 (16 GB of RAM, intel chip)
- Macbook Air 2020 (8 GB of RAM, M1 chip)
- Mac mini 2020 (8 GB of RAM, M1 chip)
I currently own 4 and 5, and run QIIME 2 everyday with sample sizes ranging up to 400. As a side note, I perform heavy tasks that need multiple threads on my company’s server (e.g. denoising reads with DADA 2 or taxonomic classification). Hope this helps!
5 Likes
Welcome to the forum, @ashore! Thanks for a great first question, and congratulations on your new PI-ship. Here are some of the main questions I’d consider:
- What scale of data are you usually working with? (e.g. one sequencing run, or meta-analyses with tens of thousands of samples?)
- Do you need a machine to handle your peak computing loads, or for most of your work, most of the time.
- Do you have still have access to a high-performance compute cluster, cloud compute resources, or similar?
- Will this workstation be your daily driver, or an additional tool for heavy analysis?
If you’re committed to doing everything locally, 32-64 gb ram is enough for most users to do most things. More and faster CPUs will often get you faster run times. IIRC, GPU parallelism is supported by a select few QIIME 2 commands, but isn’t common. (It may be useful for other computational work you’re doing, though.)
Also, consider the opportunity cost to tying up your desktop for hours with heavy computing loads. Running computationally intensive commands in a cluster environment will save you from locking up your machine, and will give you the ability to scale up as needed to more CPUs and memory than most fancy-pants desktops make available, without needing a super-powerful machine all the time. Your cluster probably won’t age as quickly as your workstation.
I tend to run my analyses (up to ~1600 samples) on good-enough laptops (8-16GB ram, core i7 processors, no dedicated GPU), because that’s what I’ve got. I use our cluster selectively to do the heavy lifting (denoising, many-sample beta-diversity calculations, tree-building, classifier training). In some ways, it’s less convenient than running everything on one machine, but it’s very nice to be able to queue a job remotely and move on with my work, too.
Just my two (or three?) cents,
Chris 
2 Likes
Thanks for the reply! I currently have access to a computing cluster my at previous institution, but I want to build something that will be able to run QIIME2 from start to finish (so including denoising/dada2), because my current institution does not have any computing clusters and I will eventually loss access to the server at my old institution. It would be a bonus if the system could also run Trinity or SOAPdenovo or SPADES for small scale transcriptome and genome assembly (but maybe that’s asking for too much).
Good to know that the 32-64 GB RAM range might work.
This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.