ChunkServe dedicated server review: decent hardware, unreliable upstream (Aug–Nov 2026)

milkcartonuk

Provider: ChunkServe (AS214481)
Product: Dedicated server QDC-NWS3, €78.99EUR/month, Amsterdam, NL
Used for: Proxmox host for a gaming community (web, game admin panels, PostgreSQL, TeamSpeak)
Period: August -> November 2026, then migrated away

Specs

CPU	2× Xeon Gold 6132 (28c/56t)
RAM	128 GB (after correction, see below)
Disk	2× 960 GB Micron enterprise SSD
Port	10 Gbps
IPs	/29 from a leased range (IPXO)

What went wrong

• Delivered with 80 GB instead of the 128 GB ordered, in mismatched DIMMs (16 GB in one socket, 64 GB in the other). Fixed in a maintenance window about a week later.
• 14 Sep: short uplink blip.
• 7 Oct: repeated upstream outages. The provider gateway was unreachable 11:46–12:00 UTC and on and off 17:17–18:16 UTC; the longest single drop was 13 minutes. Host, NIC and internal network were clean the whole time (0 errors, internal probes at 0.4 ms / 0% loss). Traffic died at the first hop. Was eventually resolved, with no credit or true explanation.
• 2 Nov: the server rebooted itself at 12:34 UTC with zero explanation.
• Overall: slow or non-existent response times, I've had a ticket open for weeks with no reply.

What was fine

• The hardware itself: stable, no disk or ECC issues, after the original RAM issue.
• Day-to-day network quality outside the outages was good: 0.03–0.4% weekly packet loss in most weeks (0.84% in the week of the 7 Oct outage), jitter 0.08–0.25 ms, and ~10 ms to our game servers (Prometheus/blackbox, one probe every 15 s).
• Throughput was OK when the network was up: ~750–800 Mbit/s sustained outbound to our new host during migration.

Verdict: The hardware is fine for the money, but game servers and voice need a stable upstream, and we couldn't rely on it. 4/10.

Might suit batch or non-latency-sensitive workloads.