Last updated: 01 Mar 2026
Location: Bucharest, Romania
Host platform: AMD Ryzen 5700G nodes with NVMe storage
Network port: up to 1 Gbps
This page is the benchmark hub for TinyServers NVMe VPS plans in Romania. The goal is simple: publish reproducible performance data, explain what those numbers mean in practice, and map each workload to the right VPS size.
Benchmark methodology
- Fresh VPS deployment before each test cycle.
- No custom kernel tuning unless explicitly stated.
- Each network and disk test run multiple times, median reported.
- Tools used: speedtest, fio, and standard Linux monitoring commands.
- Tests executed at comparable times to reduce peak-hour variance.
How to reproduce locally
sudo apt update && sudo apt install -y speedtest-cli fio
speedtest
fio --name=randread --ioengine=libaio --rw=randread --bs=4k --size=1G --numjobs=1 --iodepth=32 --direct=1
If you run your own tests, use at least three runs and compare medians, not single spikes.
Recent network snapshots
| Month | Download | Upload | Latency | Plan |
|---|---|---|---|---|
| Sep 2025 | 384 Mbps | 450 Mbps | 2.1 ms | TSM1NVME10 |
| Aug 2025 | 443 Mbps | 379 Mbps | 8.4 ms | TSM1NVME10 |
Network throughput changes by destination, peering, and time of day. Use these values as realistic references, not fixed guarantees.
How to interpret benchmark numbers
CPU
For web applications, CPU matters most during concurrent dynamic requests, compression, and background processing. If response times degrade under moderate concurrency, move to a plan with more vCPU headroom.
Disk I/O (NVMe)
Fast disk I/O improves package updates, database operations, cache warmup, and deployment time. For CMS and API workloads, strong NVMe performance is often the difference between smooth and inconsistent behavior under spikes.
Network
Low latency to local users improves page responsiveness and API round-trip times. Throughput is more visible during file transfer, media delivery, and backup windows.
Workload mapping by plan size
- 1 GB RAM: VPN, static sites, lightweight API endpoints, low-traffic tools.
- 2 GB RAM: WordPress with plugins, small production APIs, app + database combinations.
- 4 GB RAM: higher concurrency, more workers, larger caches, multiple always-on services.
For most new production projects, 2 GB is the safest starting point because it avoids early memory bottlenecks while keeping cost low.
Detailed benchmark articles
- TSM1NVME10 benchmark: Ryzen 5700G VPS performance test
- TSM1NVME10 speed test: 443 Mbps down / 379 Mbps up
- 2 GB NVMe VPS: best value for small production apps
- Cheap 1 GB VPS use cases and upgrade path
Choose your VPS plan
- 1 GB RAM NVMe VPS (TSM1NVME10)
- 2 GB RAM NVMe VPS (TSM2NVME20)
- 4 GB RAM NVMe VPS (TSV2NVME40)
- 4 GB RAM NVMe VPS (TSM4NVME40)
- View all NVMe VPS plans
Benchmark update policy
- New measurements are added after major infrastructure or routing changes.
- When a metric changes significantly, methodology notes are updated on this page.
- Older data points remain visible for historical comparison.
Quick FAQ
Are benchmarks enough to choose a VPS?
They are a strong starting point, but real usage profile matters more than a single synthetic score.
Why can my speedtest differ from published results?
Route selection, test endpoint, and time of day can change throughput and latency.
What is the safest default for new production projects?
Most teams should start with 2 GB NVMe and scale up when monitoring shows sustained pressure.
How benchmarks map to real workloads
| Workload type | Main bottleneck | Metric to watch first |
|---|---|---|
| CMS websites | Memory and database I/O | RAM pressure + disk latency |
| API backends | CPU concurrency | Response time under parallel load |
| Backup and transfer jobs | Network throughput | Upload/download consistency |
Benchmark caveats and good practice
- Single speedtest results can be misleading; use medians and multiple endpoints.
- Do not compare different providers without normalizing region and test method.
- Measure during your own peak business window, not only off-peak hours.
- Validate synthetic benchmarks with application-level metrics.
Decision matrix for fast plan selection
- Choose 1 GB when budget is strict and workload is simple.
- Choose 2 GB for production WordPress, APIs, and mixed app stacks.
- Choose 4 GB when you need higher concurrency and wider safety margins.
Need migration help while scaling? Follow the zero-downtime migration guide: How to migrate to TinyServers NVMe VPS.
