4 ms·

Ask HN: RPi Cluster or a Dual Xeon? (How to compare Arm to Intel?)

2 points by jason_slack 8y ago

I have a text processing project (and I asked a few questions about this here already). It's currently Bash, GNU Parallel and other command-line tools. It's getting moved to a C++ queue type design. I'm trying to decide on an 8 node Raspberry Pi 3B+ cluster (about $500) or buying a used Dual Xeon (12 core) desktop (about $800, HP Z820). The RPi cluster will help ensure my code scales well. However, I don't quite know how speeds compare. 8 RPi's x 4 cores each = 32 cores 2 x Xeon (12 cores) = 24 cores. The RPI's are 1.4GHz 64-bit quad-core ARMv8 CPU, 1 GB RAM. The dual Xeon is 2.5ghz each, 64gb RAM. How can I accurately think about the cores and speeds of each to decide what to do? This project will get larger within the next year and there will be more files to parse, manipulate, etc. Edit: typos Edit 2: typos Edit 3: clarify RAM Z820 comes with Edit 4: formatting

snazz 8y ago
I say the HP workstation. Not only will the work required to set up a bunch of individual RPis to talk to each other effectively eclipse the work required to configure a single machine, but the overhead from truly parallel, separate computers is huge compared to a single machine where the individual cores share CPU cache and memory. Disk performance, which I imagine to be a possible issue with your project, is going to be orders of magnitude better on a real SSD than slow, unreliable microSD cards. And 64GB RAM is 8 times as much as the RAM on all the RPis combined. Considering that there’s more duplication of code in memory required on each Raspberry Pi, you’re going to have much less than 8GB total to work with. Per watt, the ARM cores will beat the Intel ones, but per dollar you’re getting a much better deal on the desktop. The Raspberry Pi setup will cost quite a bit more than $500 once you include fast microSD cards, power supplies, cooling, an Ethernet switch, and cabling to hook everything together. Get the serious workstation-class machine, or consider buying a couple of used rackmount servers on eBay for cheaper than $800 (minus the astronomical power costs; you can heat your house with server hardware).
- jason_slack 8y ago
  Thank you for the thought our reply. I was actually thinking how to take the heat from machines and moving it through my house. At least maybe heating my laundry room so the water pipes stop freezing up :-)
detaro 8y ago
The workstation wins that easily. Faster, more RAM, not as IO-limited, ...

2 points by jason_slack 8y ago

snazz 8y ago

I say the HP workstation. Not only will the work required to set up a bunch of individual RPis to talk to each other effectively eclipse the work required to configure a single machine, but the overhead from truly parallel, separate computers is huge compared to a single machine where the individual cores share CPU cache and memory. Disk performance, which I imagine to be a possible issue with your project, is going to be orders of magnitude better on a real SSD than slow, unreliable microSD cards. And 64GB RAM is 8 times as much as the RAM on all the RPis combined. Considering that there’s more duplication of code in memory required on each Raspberry Pi, you’re going to have much less than 8GB total to work with. Per watt, the ARM cores will beat the Intel ones, but per dollar you’re getting a much better deal on the desktop. The Raspberry Pi setup will cost quite a bit more than $500 once you include fast microSD cards, power supplies, cooling, an Ethernet switch, and cabling to hook everything together. Get the serious workstation-class machine, or consider buying a couple of used rackmount servers on eBay for cheaper than $800 (minus the astronomical power costs; you can heat your house with server hardware).

jason_slack 8y ago
Thank you for the thought our reply. I was actually thinking how to take the heat from machines and moving it through my house. At least maybe heating my laundry room so the water pipes stop freezing up :-)

detaro 8y ago

The workstation wins that easily. Faster, more RAM, not as IO-limited, ...