Measured firsthand via SSH on our Nuremberg (NUE · AS197540) production node under Debian 12 (Kernel 6.1, BBR+fq, Multi-bit ECC). Host IDs and final IP octets are redacted.
DC: Nuremberg, Bavaria, Germany (NUE) Multi-bit ECC RAM (SMBIOS Verified) Debian GNU/Linux 12 (bookworm) · Kernel 6.1.0-53-amd64
RPKI: VALID · BBR + fq (MTU 1500)
Geekbench 6 (1C / 2C)
1021 / 1816
sysbench 2T: 1037.8 ev/s
NVMe 1M Mixed / SEQ
6.87 GB/s
SEQ1M Read: 4,364 MB/s
4K IOPS & Q1T1 Write
57.0k IOPS
Q1T1 Write: 115.87 µs
DDR5 ECC RAM Read
74.1 GiB/s
4K Rand: 5,122 MiB/s
Dual-Stack iperf3
2.72 Gbps
IPv6: 2.71 Gbps · FRA 3.58ms
Redis 7 / SQLite WAL
915.8k GET/s
SQLite WAL: 239.7k rows/s
IP RBL & RPKI Status
17/17 CLEAN
Native DE · Proxy: false
60s Stress & Steal
0% Throttle
Steal: 1.26%–1.31% (60s)
Key Engineering Findings Across All 12 Test Suites
2.5 Gbps Dual-Stack Confirmed: Hit 2.72 Gbps IPv4 & 2.71 Gbps IPv6 in iperf3 (Amsterdam/London) and 2.50 Gbps transatlantic to NYC, with 3.58 ms RTT to Frankfurt DE-CIX. Enterprise NVMe RAID (CrystalDiskMark + YABS): 4,364 MB/s SEQ1M Q8T1 Read, 6.87 GB/s 1M mixed R/W, 57.0k 4K mixed IOPS, and 115.87 µs Q1T1 4K write latency. Production Database & Cache Speed: Redis 7 reached 101,420 GET/s (P=1, p50 0.22 ms) and 915,751 GET/s (P=16 pipelined); SQLite3 WAL hit 239,690 inserts/s & 33,074 autocommit fsync tx/s. Zero 60s Throttling + 74.1 GiB/s DDR5 ECC RAM: Continuous 60s 100% 2-core load stayed flat at 993–1,026 events/s with just 1.26%–1.31% hypervisor steal (QEMU EPYC-Rome 2.0 GHz live-migration profile). 100% Native DE IP, 17/17 RBLs Clean & Full Streaming/AI Unlock: Zero proxy flags, RPKI VALID, zero Google CAPTCHAs, full Netflix DE non-original catalog, ChatGPT/Claude/Gemini/TikTok DE unlocked, plus 1-click self-service Port 25/465/587 unlock in SCP → Firewall. 1A · YABS fio Mixed 50/50 R+W (/dev/vda4)
Profile Read Write Combined / Latency 4k 116.70 MB/s 117.01 MB/s 233.72 MB/s (57.0k) 64k 957.46 MB/s 962.50 MB/s 1.91 GB/s (29.2k) 512k 3.12 GB/s 3.28 GB/s 6.41 GB/s (12.2k) 1m 3.32 GB/s 3.54 GB/s 6.87 GB/s (6.5k)
Classic dd (1 GiB conv=fdatasync): 498 MB/s
1B · CrystalDiskMark-Style Direct NVMe Profile (libaio, direct=1)
Profile Read Write Combined / Latency SEQ1M Q8T1 4,364 MB/s (4,161) 4,211 MB/s (4,015) 1.92 ms / 1.99 ms SEQ1M Q1T1 2,058 MB/s (1,962) 1,696 MB/s (1,617) 507.8 µs / 615.7 µs RND4K Q32T1 231 MB/s (56.5k) 152 MB/s (37.1k) 565.9 µs / 862.2 µs RND4K Q1T1 33.9 MB/s (8,272) 35.0 MB/s (8,546) 119.7 µs / 115.9 µs
2 · Real Application, Crypto & 60-Second Zero-Throttling Stress Curve
Redis 7 (P=1 / P=16)
101.4k / 915.8k GET/s
SQLite3 WAL (Insert / QPS)
239.7k/s · 154.8k QPS
OpenSSL 2T (AES128 / 256)
5.39 GB/s / 4.97 GB/s
TLS Sign (ECDSA / Ed25519)
32,504/s · 16,330/s
00s–10s
1026.36 ev/s
steal 1.31%
10s–20s
1012.75 ev/s
steal 1.31%
20s–30s
995.2 ev/s
steal 1.31%
30s–40s
1000.03 ev/s
steal 1.26%
40s–50s
993.17 ev/s
steal 1.31%
50s–60s
997.36 ev/s
steal 1.31%
3 · iperf3 Dual-Stack Global Throughput (IPv4 & IPv6)
Endpoint (Link) IPv4 Send / Recv IPv6 Send / Recv RTT Amsterdam, NL (100G) 2.72 Gbps / 2.33 Gbps 2.69 Gbps / 2.31 Gbps 16.3 ms London, UK (10G) 2.67 Gbps / 1.47 Gbps 2.71 Gbps / 2.21 Gbps 28.2 ms New York, US-East (10G) 2.50 Gbps / 1.79 Gbps 2.48 Gbps / 1.62 Gbps 97.1 ms Tashkent, UZ (10G) 2.54 Gbps / 1.82 Gbps 2.52 Gbps / busy 78.0 ms Singapore, SG (10G) 2.40 Gbps / 1.70 Gbps 2.38 Gbps / 1.52 Gbps 154 ms Los Angeles, US-West (10G) 1.03 Gbps / 499 Mbps 1.01 Gbps / 896 Mbps 170 ms
4 · Dual-Stack Streaming, AI Platforms, Google CAPTCHA & 17-RBL Reputation Matrix
Platform / Check IPv4 Result IPv6 Result Google Search (No CAPTCHA) ✔ Clean 200 (126 ms) ✔ Clean 200 (121 ms) Netflix (Full DE Catalog) ✔ Full DE (842 ms) ✔ Full DE (801 ms) OpenAI / ChatGPT ✔ DE FRA (64 ms) ✔ DE HAM (88 ms) Anthropic Claude ✔ DE Supported (65 ms) ✔ DE Supported (98 ms) Google Gemini ✔ DE Available (471 ms) ✔ DE Available (240 ms) YouTube Premium ✔ DE Region (831 ms) ✔ DE Region (664 ms) TikTok ✔ DE Region (471 ms) IPv4-Only Endpoint Disney+ & Spotify ✔ DE Available (108 ms) ✔ DE Available (128 ms) ZDFmediathek (German TV) ✔ Native DE (81 ms) ✔ Native DE (81 ms) Steam Store ✔ EUR (€) DE ✔ EUR (€) DE GitHub / PyPI / Crates.io ✔ 58–61 ms TTFB ✔ Dual-Stack Reddit Unauth JSON ⚠ OAuth Required (403) ⚠ OAuth Required
17/17 RBLs Clean: Spamhaus ZEN · SpamCop · Barracuda BRBL · SORBS · UCEPROTECT L1 · NiX Spam (DE) · Blocklist.de (DE) · PSBL · Mailspike + 8 more
5 · Global IX & Dev Registry Latency (from Nuremberg NUE)
Falkenstein, DE (Hetzner)
2.75 ms
0% loss
Frankfurt DE-CIX (Cloudflare FRA)
3.58 ms
0% loss
Vienna, Austria (VIX)
9.6 ms
0% loss
Amsterdam, Netherlands
16.3 ms
0% loss
GitHub / PyPI / Crates CDN
58.2 ms
0% loss
Ashburn, VA (US-East)
103.57 ms
0% loss
Singapore (APAC)
159.32 ms
0% loss
6 · China 3-City / 8-Node Latency & Evening-Peak Return MTR AS197540 (netcup) → AS47147 (Anexia) → AS1299 (Arelion Tier-1 Frankfurt)
Beijing Unicom (North)
147.9–191.7 ms
0%
Beijing Telecom (North)
152.7–185 ms
0%
Beijing Mobile (North)
155.1–155.2 ms
QoS
Guangzhou Unicom (South)
163.9–169.4 ms
0%
Shanghai Telecom (East)
188.4–188.7 ms
16.7%
Shanghai Mobile (East)
195.2–196.4 ms
16.7%
Guangzhou Telecom (South)
214.3–214.6 ms
16.7%
Guangzhou Mobile (South)
218–218 ms
0%
Carrier & Region Verified Return AS Path Avg RTT / Final Loss Routing Verdict China Telecom (163 Backbone)
Beijing / Shanghai / Guangzhou
AS197540 → AS47147 → AS1299 (FFM) → AS4134 → AS4812 187.75 ms
/ 16.7%
Direct Europe-to-China 163 path (152.7 ms min Beijing, 188.7 ms Shanghai); pair public web apps with Cloudflare CDN during evening peak. China Unicom (CU169 AS4837)
Beijing / Guangzhou
AS197540 → AS47147 → AS1299 (FFM) → AS4837 → AS4808 169.4 ms
/ 0%
Fastest raw latency to China: 147.9 ms min to Beijing & 169.4 ms avg (0% loss) to Guangzhou Unicom via Frankfurt AS4837 without US detour. China Mobile (CMI AS58453)
Beijing / Shanghai / Guangzhou
AS197540 → AS47147 → AS1299 (FFM) → AS58453 (CMI) → AS9808 218 ms
/ 0%
Rock-solid evening stability: 4.0 ms handoff into CMI Frankfurt; 155.1 ms Beijing, 196.4 ms Shanghai, 218.0 ms Guangzhou (0.0 ms jitter, 0% loss).
SMBIOS, Kernel Mitigations & Self-Service SCP Mail Firewall Status
✔ SMBIOS: Multi-bit ECC · Meltdown/MDS/L1TF Not Affected
✔ Kernel Syscall: 4.28M/s · Pipe IPC: 1,774.7 MiB/s
✔ gzip -6: 180.7 MiB/s · xz Decompress: 440.5 MiB/s
✔ SMTP 25 / 465 / 587: 1-Click Self-Service Unlock in SCP → Firewall (Delete "netcup Mail block" → Save)
$ View Sanitized Raw Multi-Suite Benchmark Summary Log ▼ # ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #
# Yet-Another-Bench-Script #
# v2026-10-31 #
# https://github.com/masonr/yet-another-bench-script #
# ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #
Tue Nov 17 01:25:04 PM CET 2026
Basic System Information:
---------------------------------
Uptime : 29 days, 22 hours, 6 minutes
Processor : AMD EPYC-Rome Processor (SMBIOS: netcup VPS 500 G12.5 · Multi-bit ECC)
CPU cores : 2 @ 1996.250 MHz (AES-NI ✔ · AVX2 ✔ · VM-x ❌)
RAM / Disk : 3.8 GiB / 62.9 GiB (Debian 12 · Kernel 6.1.0-53-amd64 · KVM · BBR+fq)
IPv4/IPv6 : ✔ Online (152.53.xxx.0/22 · RPKI VALID) / ✔ Online (2a03:4000:xxxx::/64)
Location : Nuremberg, Bavaria (BY), Germany · AS197540 netcup GmbH
fio Disk Speed Tests (Mixed R/W 50/50) (/dev/vda4):
---------------------------------
4k : Read 116.70 MB/s (28.4k) | Write 117.01 MB/s (28.5k) | Total 233.72 MB/s (57.0k IOPS)
64k : Read 957.46 MB/s (14.6k) | Write 962.50 MB/s (14.6k) | Total 1.91 GB/s (29.2k IOPS)
512k : Read 3.12 GB/s (5.9k) | Write 3.28 GB/s (6.2k) | Total 6.41 GB/s (12.2k IOPS)
1m : Read 3.32 GB/s (3.1k) | Write 3.54 GB/s (3.3k) | Total 6.87 GB/s (6.5k IOPS)
CrystalDiskMark-Style fio Direct Profile (libaio, direct=1):
---------------------------------
SEQ1M Q8T1 : Read 4,364 MB/s | Write 4,211 MB/s
SEQ1M Q1T1 : Read 2,058 MB/s | Write 1,696 MB/s
RND4K Q32T1: Read 231 MB/s (56.5k IOPS) | Write 152 MB/s (37.1k IOPS)
RND4K Q1T1 : Read 33.9 MB/s (119.67 µs) | Write 35.0 MB/s (115.87 µs, p99=235 µs)
iperf3 Dual-Stack Network Speeds:
---------------------------------
Amsterdam, NL (100G): IPv4 2.72G Send / 2.33G Recv (16.3ms) | IPv6 2.69G / 2.31G
London, UK (10G) : IPv4 2.67G Send / 1.47G Recv (28.2ms) | IPv6 2.71G / 2.21G (21.0ms)
New York, US (10G) : IPv4 2.50G Send / 1.79G Recv (97.1ms) | IPv6 2.48G / 1.62G
Singapore, SG (10G) : IPv4 2.40G Send / 1.70G Recv (154ms) | IPv6 2.38G / 1.52G
CPU, Memory, Crypto & Application Suite:
---------------------------------
Geekbench 6 : 1021 Single-Core | 1816 Multi-Core (https://browser.geekbench.com/v6/cpu/19485064)
60s Stress : 1026 -> 1013 -> 995 -> 1000 -> 993 -> 997 ev/s (0% throttling, 1.26%-1.31% steal)
DDR5 Host RAM : 74,066.85 MiB/s Seq Read | 29,838.58 MiB/s Seq Write | 5,122.16 MiB/s 4K Rand Read
OpenSSL (2T) : AES-128-GCM 5.39 GB/s | AES-256-GCM 4.97 GB/s | ECDSA P-256 32,504 signs/s
Redis 7 (c=50): 101,420 GET/s (P=1) | 915,751 GET/s & 781,250 SET/s (P=16 Pipelined)
SQLite3 (WAL) : 239,690 Bulk Inserts/s | 33,074 Autocommit Tx/s | 154,787 Point SELECT QPS
IP & Unlock : 17/17 RBLs CLEAN · Google No-CAPTCHA · Netflix Full DE · ChatGPT/Claude/Gemini/TikTok DE ✔