Hey everyone! My professor has tasked me with building a PC for our proteomics lab, and I have a budget of about $4000, but I could go up to $5000 if necessary. I'm located in Honolulu, Hawai'i, and our primary operating system will likely be Linux.
The PC needs to efficiently handle several tasks including:
- Running batches for molecular docking, aiming for about 10,000 jobs at once, so I need solid multi-core performance.
- Processing protein mass spectrometry data.
- Analyzing cryo-EM data.
- Simulating hydrology and wind models.
- I need some short-term storage for large datasets, as cryo-EM projects can generate a lot of data daily.
I'm looking for a build that prioritizes a strong multi-core CPU, a good amount of RAM, a powerful GPU (mainly for cryo-EM and visualization), and fast NVMe SSD storage with room for future upgrades. It's worth noting that most of our work is more CPU-bound rather than GPU-bound. However, I'll definitely need a capable GPU for the cryo-EM part.
Thanks for your help!
4 Answers
Have you considered whether your university has access to a computing cluster? It might save a lot of headaches to use that instead. Building a solid system nowadays can be quite pricy, especially when GPUs and RAM prices can fluctuate dramatically.
If you do decide to build, try to figure out exactly what software you'll be using, as some can be highly CPU-intensive while others may not utilize GPUs as much. That information can really guide your component choices!
Honestly, for that budget, you could be hitting some serious limitations with consumer-grade parts! SSDs and RAM prices are insane right now, and for the tasks you describe, you might end up needing a significant investment beyond just 4-5k. Just keep all that in mind!
Good opinion! I’m aware of the costs, and I’m hoping to find a balance for our needs. Thanks for the heads-up!
You’re on the right track focusing on a non-gaming setup! For your CPU, consider the AMD Ryzen 9 9950X; it's got 16 cores which will be great for running multiple jobs in parallel. Pair it with 128GB of RAM to handle the heavy data loads, and a solid 4TB PCIe 5.0 NVMe SSD for fast storage access.
For the GPU, the RTX 5080 16GB should work well, especially since you'll need it for cryo-EM tasks. Don't worry too much about aesthetics; just go for good airflow and reliability with a well-ventilated case.
You can check out a build list I made if you want to dive deeper into the components!
Thanks for the suggestions! I’ll definitely check out that CPU and GPU and the build list you mentioned.
Have you thought about whether a consumer build can actually handle your workload? I manage a CFD simulation setup that runs on a school's cluster with about 128 cores and 500GB of RAM, which is more powerful than most consumer setups. If your tasks are similar, it might be worth looking into local or cloud-based HPC options instead, especially since prices for memory and storage can get out of hand. Just a thought!
That's an interesting perspective! I think the molecular docking and cryo-EM might require different setups though, and we really need something we can upgrade in the future. Our previous supercomputer option isn't available now.
Unfortunately, the supercomputer nearby isn’t available anymore, which is why we’re looking to build. I’ll keep your advice in mind about checking the software requirements.