CUDA vs ROCm vs CPU: Running AI on Whatever GPU You've Got
CUDA vs ROCm for AI on Linux: NVIDIA's easy path, AMD's emotional journey, and why CPU inference isn't dead yet. Real Docker setups included.
All the articles with the tag "performance".
CUDA vs ROCm for AI on Linux: NVIDIA's easy path, AMD's emotional journey, and why CPU inference isn't dead yet. Real Docker setups included.
DDoS mitigation for self-hosters: Nginx rate limiting, Fail2ban, Cloudflare free tier, CrowdSec, and iptables tricks that actually work.
GNU parallel runs tasks in parallel across CPU cores. It's faster than xargs and easier than writing a job queue. Here's when and how to use it.
Squeeze every MB/s from WireGuard: MTU sizing, GSO/GRO CPU offloading, AllowedIPs routing, PersistentKeepalive tradeoffs, and iperf3 benchmarks included.
Plex defaults to transcoding files your TV could already play. Fix direct play, enable hardware transcoding, kill the relay tax, and know when to quit for Jellyfin.
Learn how to set Docker resource limits for memory, CPU, swap, and PIDs. Practical guide with real-world sizing examples, OOM killer behavior, and cgroups explained.
Your Linux box takes forever to boot? systemd-analyze shows exactly which services drag it down. Read blame, critical-chain, plot, and fix the slow ones.
Master SSH ControlMaster, ControlPath, and ControlPersist. Reuse connections for lightning-fast SSH, SCP, and rsync operations.
Why Linux kills your app with no warning. Understand the OOM killer, read dmesg logs, and shield critical processes with oom_score_adj and swap.
Compare zram (compressed RAM) and traditional swap files. Learn when each works, setup, and real-world performance trade-offs for memory-constrained systems.
Docker layer caching explains most slow builds. Why copying files too early kills the cache, the layer ordering that fixes it, and a multi-stage benchmark.
Manage Linux process priority with nice (CPU) and ionice (I/O). Learn priority levels, renice running processes, and avoid tanking server performance.