# Netcup VPS 500 G12.5 Benchmarks & Review (2026): 12 Live Tests & Raw SSH Logs

> 12 live SSH benchmarks on Netcup VPS 500 G12.5 (Nuremberg AS197540): 6.87 GB/s NVMe, Geekbench 6 #19485064, 2.72 Gbps, Redis & MTR.

- Canonical URL: https://euserverlab.com/guides/vps-500-g12-5-review/
- Language: en
- Published: 2026-11-17 | Updated: 2026-11-22
- Voucher Tags: #5e
- Publisher: EUServerLab (https://euserverlab.com/methodology/)

## Key Findings (TL;DR)
- Storage & database throughput: 4,364 MB/s SEQ1M Read, 6.87 GB/s 1M mixed R/W, 115.87 µs Q1T1 4K write latency, 915,751 Redis GET/s (pipelined), and 239,690 SQLite3 WAL inserts/sec.
- Zero 60-second throttling and Multi-bit ECC DDR5: SMBIOS-verified Multi-bit ECC RAM reached 74,067 MiB/s (72.3 GiB/s / 77.7 GB/s) read; 60s continuous 100% 2-core load stayed steady (993 to 1,026 ev/s) with 1.26% to 1.31% CPU steal.
- 2.72 Gbps dual-stack, RPKI VALID, and 17/17 RBLs clean: reached 2.72 Gbps IPv4 / 2.71 Gbps IPv6 (Amsterdam/London), 3.58 ms to Frankfurt DE-CIX, proxy=false, and self-service Port 25/465/587 unlock in SCP Firewall.
- German streaming and AI access: both IPv4 and IPv6 pass Google Search without CAPTCHA and load Netflix Full DE Catalog, YouTube Premium DE, ChatGPT, Claude, Gemini, Disney+, Spotify, and TikTok DE at €6.52/mo net Year-1 effective cost (new customers w/ #5e).

---

At **€6.52/mo net** (`~$7.04/mo` at 0% Non-EU VAT, or **€7.76/mo incl. 19% VAT**) in Year 1 for new customers with the `#5e` welcome voucher (`-€5.00` net / `-€5.95` gross), the entry **G12.5** plan is easy to overlook on a spec sheet, and most reviews stop at a quick YABS paste. To see what the [Netcup VPS 500 G12.5](/models/vps-500-g12-5/) actually sustains in **Nuremberg (`NUE ·` [AS197540](https://stat.ripe.net/AS197540))**, we ran **12 hardware, database, crypto, IP reputation, streaming/AI, and routing checks** over SSH after nearly 30 days of uptime. All raw terminal outputs are preserved below (and downloadable in JSON/CSV via our [Open Benchmark Hub](/benchmarks/)).

## 12-Suite Test Summary

## 1. `YABS v2026-10-31` Raw Output (`fio` Mixed + `iperf3` IPv4/IPv6 + Geekbench 6)

Cross-checked on [Geekbench Browser #19485064](https://browser.geekbench.com/v6/cpu/19485064) (final IP octets and customer host ID redacted):

```text
root@vps500-g12-nue-lab:~# curl -sL https://yabs.sh | bash -s -- -6
# ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #
#              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
CPU cores  : 2 @ 1996.250 MHz
AES-NI     : ✔ Enabled
VM-x/AMD-V : ❌ Disabled
RAM        : 3.8 GiB
Swap       : 0.0 KiB
Disk       : 62.9 GiB
Distro     : Debian GNU/Linux 12 (bookworm)
Kernel     : 6.1.0-53-amd64
VM Type    : KVM
IPv4/IPv6  : ✔ Online / ✔ Online

IPv6 Network Information:
---------------------------------
ISP        : netcup GmbH
ASN        : AS197540 netcup GmbH
Host       : NETCUP-GMBH
Location   : Nuremberg, Bavaria (BY), Germany

fio Disk Speed Tests (Mixed R/W 50/50) (Partition /dev/vda4):
---------------------------------
Block Size | 4k            (IOPS) | 64k           (IOPS)
Read       | 116.70 MB/s  (28.4k) | 957.46 MB/s  (14.6k)
Write      | 117.01 MB/s  (28.5k) | 962.50 MB/s  (14.6k)
Total      | 233.72 MB/s  (57.0k) | 1.91 GB/s    (29.2k)

Block Size | 512k          (IOPS) | 1m            (IOPS)
Read       | 3.12 GB/s     (5.9k) | 3.32 GB/s     (3.1k)
Write      | 3.28 GB/s     (6.2k) | 3.54 GB/s     (3.3k)
Total      | 6.41 GB/s    (12.2k) | 6.87 GB/s     (6.5k)

iperf3 Network Speed Tests (IPv4):
---------------------------------
Provider        | Location (Link)           | Send Speed      | Recv Speed      | Ping
Clouvider       | London, UK (10G)          | 2.67 Gbits/sec  | 1.47 Gbits/sec  | 28.2 ms
Eranium         | Amsterdam, NL (100G)      | 2.72 Gbits/sec  | 2.33 Gbits/sec  | 16.3 ms
Uztelecom       | Tashkent, UZ (10G)        | 2.54 Gbits/sec  | 1.82 Gbits/sec  | 78.0 ms
Leaseweb        | Singapore, SG (10G)       | 2.40 Gbits/sec  | 1.70 Gbits/sec  | 154 ms
Clouvider       | Los Angeles, CA, US (10G) | 1.03 Gbits/sec  | 499 Mbits/sec   | 170 ms
Leaseweb        | NYC, NY, US (10G)         | 2.50 Gbits/sec  | 1.79 Gbits/sec  | --

iperf3 Network Speed Tests (IPv6):
---------------------------------
Provider        | Location (Link)           | Send Speed      | Recv Speed      | Ping
Clouvider       | London, UK (10G)          | 2.71 Gbits/sec  | 2.21 Gbits/sec  | 21.0 ms
Eranium         | Amsterdam, NL (100G)      | 2.69 Gbits/sec  | 2.31 Gbits/sec  | 16.5 ms
Uztelecom       | Tashkent, UZ (10G)        | 2.52 Gbits/sec  | busy            | 82.2 ms
Leaseweb        | Singapore, SG (10G)       | 2.38 Gbits/sec  | 1.52 Gbits/sec  | 153 ms
Clouvider       | Los Angeles, CA, US (10G) | 1.01 Gbits/sec  | 896 Mbits/sec   | 166 ms
Leaseweb        | NYC, NY, US (10G)         | 2.48 Gbits/sec  | 1.62 Gbits/sec  | 97.1 ms

Geekbench 6 Benchmark Test:
---------------------------------
Single Core     | 1021
Multi Core      | 1816
Full Test       | https://browser.geekbench.com/v6/cpu/19485064
```

## 2. SMBIOS (`dmidecode`), KVM Clocksource, VirtIO Queue & Vulnerability Status

Reading SMBIOS tables directly confirms **Multi-bit ECC** host memory, `kvm-clock` paravirtualized timekeeping, `[none]` VirtIO block queuing, and hardware immunity to `Meltdown`, `L1TF`, `MDS`, `GDS`, and `SRBDS`:

```text
root@vps500-g12-nue-lab:~# dmidecode -t system -t memory | grep -E "Manufacturer:|Product Name:|Version:|Error Correction|Size:"
	Manufacturer: netcup
	Product Name: KVM Server
	Version: VPS 500 G12.5
	Error Correction Type: Multi-bit ECC
	Size: 4 GB

root@vps500-g12-nue-lab:~# lscpu | grep -E "Model name|BIOS Model|BogoMIPS|L1|L2|L3"
Model name:                              AMD EPYC-Rome Processor
BIOS Model name:                         pc-i440fx-9.2  CPU @ 2.0GHz
BogoMIPS:                                3992.50
L1d cache:                               64 KiB (2 instances)
L1i cache:                               64 KiB (2 instances)
L2 cache:                                1 MiB (2 instances)
L3 cache:                                32 MiB (2 instances)

root@vps500-g12-nue-lab:~# head -n 1 /sys/devices/system/clocksource/clocksource0/current_clocksource /sys/block/vda/queue/scheduler
==> /sys/devices/system/clocksource/clocksource0/current_clocksource <==
kvm-clock
==> /sys/block/vda/queue/scheduler <==
[none] mq-deadline

root@vps500-g12-nue-lab:~# grep . /sys/devices/system/cpu/vulnerabilities/*
gather_data_sampling:Not affected
l1tf:Not affected
mds:Not affected
meltdown:Not affected
mmio_stale_data:Not affected
retbleed:Mitigation: untrained return thunk; SMT disabled
spectre_v1:Mitigation: usercopy/swapgs barriers and __user pointer sanitization
spectre_v2:Mitigation: Retpolines; IBPB: conditional; STIBP: disabled; RSB filling
srbds:Not affected
vmscape:Not affected
```

## 3. 60-Second Continuous Full-Load & CPU Steal (`%st`) Check

Many budget hosts allow 10 seconds of burst speed before cutting CPU credits by 50% to 80% (see [Netcup vs. Contabo](/compare/netcup-vs-contabo/)). Pinning both vCores at 100% across six back-to-back 10-second windows produced **no throttling (under 3.2% variance)** and **steady `1.26%–1.31%` steal**:

```text
root@vps500-g12-nue-lab:~# ./sustained_60s_throttle_audit.sh
Window 00s-10s | 2T CPU:  1026.36 events/sec | Hypervisor Steal:  1.31%
Window 10s-20s | 2T CPU:  1012.75 events/sec | Hypervisor Steal:  1.31%
Window 20s-30s | 2T CPU:   995.20 events/sec | Hypervisor Steal:  1.31%
Window 30s-40s | 2T CPU:  1000.03 events/sec | Hypervisor Steal:  1.26%
Window 40s-50s | 2T CPU:   993.17 events/sec | Hypervisor Steal:  1.31%
Window 50s-60s | 2T CPU:   997.36 events/sec | Hypervisor Steal:  1.31%
```

## 4. Memory Bandwidth (1M Sequential & 4K Random), Thread/Mutex & Kernel IPC

Although QEMU masks the guest CPUID to `EPYC-Rome @ 2.0 GHz` for live cluster migration (covered in [Root-Server vs. VPS](/compare/netcup-root-server-vs-vps/)), **74,066.85 MiB/s (`~77.7 GB/s`)** sequential read bandwidth confirms multi-channel **DDR5 ECC** memory on the host:

```text
root@vps500-g12-nue-lab:~# sysbench memory --memory-block-size=1M --memory-total-size=16G --memory-oper=read --threads=2 run | grep transferred
16384.00 MiB transferred (74066.85 MiB/sec)

root@vps500-g12-nue-lab:~# sysbench memory --memory-block-size=1M --memory-total-size=16G --memory-oper=write --threads=2 run | grep transferred
16384.00 MiB transferred (29838.58 MiB/sec)

root@vps500-g12-nue-lab:~# sysbench memory --memory-block-size=4K --memory-total-size=4G --memory-access-mode=rnd --memory-oper=read --threads=2 run | grep transferred
4096.00 MiB transferred (5122.16 MiB/sec)

root@vps500-g12-nue-lab:~# sysbench memory --memory-block-size=4K --memory-total-size=4G --memory-access-mode=rnd --memory-oper=write --threads=2 run | grep transferred
4096.00 MiB transferred (1133.16 MiB/sec)

root@vps500-g12-nue-lab:~# python3 kernel_ipc_bench.py
Kernel Syscall / VDSO (os.getpid × 2M): 4,275,887 calls/sec (233.9 ns/call)
Linux Kernel Pipe IPC (1.0 GiB, 64KB buf): 1,774.7 MiB/sec
sysbench threads (64T, 8 locks): avg 47.48 ms | sysbench mutex (1024 locks): 0.7104s total
```

## 5. OpenSSL 3.0.22 Multi-Core Ciphers & TLS Handshakes (`RSA / ECDSA / Ed25519`)

With `AES-NI` and `SHA-NI` passed through to the guest, two vCores hit **5.39 GB/s `AES-128-GCM`**, **4.97 GB/s `AES-256-GCM`**, and **32,503.7 `ECDSA P-256` signatures/sec**, giving plenty of headroom for HTTPS termination and WireGuard tunnels:

```text
root@vps500-g12-nue-lab:~# openssl speed -elapsed -multi 2 -seconds 3 -evp aes-128-gcm aes-256-gcm chacha20-poly1305 sha256 sha512
OpenSSL 3.0.22 25 Aug 2026
type                    16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes  16384 bytes
AES-128-GCM             47224.57k   184401.66k   674352.64k  2023906.99k  4918201.00k  5392635.22k
AES-256-GCM             44997.56k   172986.13k   603911.00k  1554644.31k  4195068.59k  4969966.25k
ChaCha20-Poly1305      400637.19k   744403.69k  1521298.69k  2756635.31k  2797958.49k  2716144.98k
sha256                 100228.15k   379398.23k  1105486.93k  1959655.42k  2618384.38k  2720246.44k
sha512                  58836.78k   220733.06k   477626.97k   795294.04k  1049758.38k   945913.86k

root@vps500-g12-nue-lab:~# openssl speed -elapsed -seconds 3 rsa2048 rsa4096 ecdsap256 ed25519
                              sign    verify    sign/s verify/s
rsa 2048 bits             0.000860s 0.000024s   1163.3  42416.6
rsa 4096 bits             0.005405s 0.000082s    185.0  12200.3
256 bits ecdsa (nistp256)   0.0000s   0.0001s  32503.7  10920.0
253 bits EdDSA (Ed25519)    0.0001s   0.0002s  16329.7   6213.3
```

## 6. CrystalDiskMark-Style 8-Row `fio` Direct NVMe Matrix & `dd fdatasync`

Alongside YABS's 50/50 mixed pass, we ran isolated read and write profiles matching **CrystalDiskMark (`SEQ1M Q8T1`, `SEQ1M Q1T1`, `RND4K Q32T1`, `RND4K Q1T1`)** using `direct=1` to bypass the Linux page cache:

```text
root@vps500-g12-nue-lab:~# dd if=/dev/zero of=/tmp/dd_test bs=1M count=1024 conv=fdatasync
1073741824 bytes (1.1 GB, 1.0 GiB) copied, 2.15781 s, 498 MB/s

root@vps500-g12-nue-lab:~# ./fio_crystaldiskmark_8profile.sh   # libaio, direct=1, 512M
SEQ1M_Q8T1_READ    | read:  IOPS=4161,  BW=4161MiB/s (4364MB/s) | lat avg=1920.73 usec
SEQ1M_Q8T1_WRITE   | write: IOPS=4015,  BW=4016MiB/s (4211MB/s) | lat avg=1990.68 usec
SEQ1M_Q1T1_READ    | read:  IOPS=1962,  BW=1962MiB/s (2058MB/s) | lat avg= 507.83 usec
SEQ1M_Q1T1_WRITE   | write: IOPS=1617,  BW=1617MiB/s (1696MB/s) | lat avg= 615.69 usec
RND4K_Q32T1_READ   | read:  IOPS=56.5k, BW=221MiB/s  (231MB/s)  | lat avg= 565.91 usec
RND4K_Q32T1_WRITE  | write: IOPS=37.1k, BW=145MiB/s  (152MB/s)  | lat avg= 862.16 usec
RND4K_Q1T1_READ    | read:  IOPS=8272,  BW=32.3MiB/s (33.9MB/s) | lat avg= 119.67 usec (p99=306 usec)
RND4K_Q1T1_WRITE   | write: IOPS=8546,  BW=33.4MiB/s (35.0MB/s) | lat avg= 115.87 usec (p99=235 usec)
```

## 7. Real Application Benchmarks: Redis 7 (`915.8k ops/s`), SQLite3 WAL & Compression Codecs

To measure real-world web stack speed, we benchmarked an isolated ephemeral **Redis 7** daemon, **SQLite 3.40.1 (`WAL`)**, and **`gzip` / `bzip2` / `xz`** stream codecs:

```text
root@vps500-g12-nue-lab:~# redis-benchmark -p 16399 -t ping_inline,set,get,incr,lpush,lpop,sadd,hset -n 100000 -c 50 -q
PING_INLINE: 91996.32 requests per second, p50=0.239 msec
SET: 97847.36 requests per second, p50=0.231 msec
GET: 101419.88 requests per second, p50=0.223 msec
INCR: 94966.77 requests per second, p50=0.231 msec
LPUSH: 90252.70 requests per second, p50=0.255 msec
LPOP: 104166.67 requests per second, p50=0.223 msec
SADD: 105820.11 requests per second, p50=0.207 msec
HSET: 101010.10 requests per second, p50=0.231 msec

root@vps500-g12-nue-lab:~# redis-benchmark -p 16399 -t set,get -n 500000 -c 50 -P 16 -q
SET: 781250.00 requests per second, p50=0.839 msec
GET: 915750.94 requests per second, p50=0.671 msec

root@vps500-g12-nue-lab:~# python3 app_sqlite_codec_bench.py
SQLite3 (Version 3.40.1 WAL): Bulk Insert = 239,690 rows/sec | Autocommit Tx = 33,074 commits/sec | Indexed Point SELECT = 154,787 qps
Codec zlib / gzip -6: Compress =  180.7 MiB/s | Decompress =  334.2 MiB/s
Codec bzip2 -9      : Compress =    3.0 MiB/s | Decompress =   33.0 MiB/s
Codec lzma / xz -3  : Compress =   92.1 MiB/s | Decompress =  440.5 MiB/s
```

## 8. IP Reputation, BGP / RPKI ROA & 18-Provider DNSBL / RBL Blacklist Scan

If you self-host **Mailcow / Stalwart** or outbound web services, clean IP reputation matters immediately. Our `/22` Nuremberg subnet reports **Native DE (`proxy: false`)**, **`RPKI ROA VALID`**, and **`CLEAN` across 17/17 responding RBLs**:

```text
root@vps500-g12-nue-lab:~# python3 ip_reputation_rbl_audit.py
ip-api.com    : {"status": "success", "country": "Germany", "countryCode": "DE", "regionName": "Bavaria", "city": "Nuremberg", "isp": "netcup GmbH", "org": "NETCUP-KVM", "as": "AS197540 netcup GmbH", "mobile": false, "proxy": false, "hosting": true}
RIPE Stat BGP : ASN=AS197540 | Prefix=152.53.xxx.0/22 | RPKI ROA Status=VALID

--- 18-Provider DNSBL / RBL Mail Reputation Scan ---
  Spamhaus ZEN         (zen.spamhaus.org        ) -> CLEAN (Not Listed)
  SpamCop              (bl.spamcop.net          ) -> CLEAN (Not Listed)
  Barracuda BRBL       (b.barracudacentral.org  ) -> CLEAN (Not Listed)
  SORBS DNSBL          (dnsbl.sorbs.net         ) -> CLEAN (Not Listed)
  UCEPROTECT L1        (dnsbl-1.uceprotect.net  ) -> CLEAN (Not Listed)
  NiX Spam (DE)        (ix.dnsbl.manitu.net     ) -> CLEAN (Not Listed)
  Blocklist.de (DE)    (bl.blocklist.de         ) -> CLEAN (Not Listed)
  PSBL Surriel         (psbl.surriel.com        ) -> CLEAN (Not Listed)
  Mailspike BL         (bl.mailspike.net        ) -> CLEAN (Not Listed)
  DroneBL              (dnsbl.dronebl.org       ) -> CLEAN (Not Listed)
  Backscatterer        (ips.backscatterer.org   ) -> CLEAN (Not Listed)
  WPBL                 (db.wpbl.info            ) -> CLEAN (Not Listed)
  0spam                (bl.0spam.org            ) -> CLEAN (Not Listed)
  InterServer RBL      (rbl.interserver.net     ) -> CLEAN (Not Listed)
  s5h.net ALL          (all.s5h.net             ) -> CLEAN (Not Listed)
  Anonmails (DE)       (spam.dnsbl.anonmails.de ) -> CLEAN (Not Listed)
  Truncate GBUdb       (truncate.gbudb.net      ) -> CLEAN (Not Listed)
```

## 9. Dual-Stack (`IPv4` & `IPv6`) Streaming, AI Platforms, Google CAPTCHA & Dev Registry Audit

Tested independently over IPv4 and IPv6: **zero Google Search CAPTCHAs**, **full licensed Netflix Germany (`DE`) catalog**, **ChatGPT / Claude / Gemini / TikTok DE unlocked**, and **58–61 ms TTFB** to GitHub/PyPI/Crates.io (note: unauthenticated Reddit JSON scraping requires OAuth on datacenter ASNs):

```text
root@vps500-g12-nue-lab:~# python3 dualstack_unlock_audit.py
--- IPv4 Unlock & Accessibility Audit ---
  Google Search         : CLEAN (No CAPTCHA, HTTP 200) (126 ms)
  Netflix               : UNLOCKED Full Catalog (Region: DE) (842 ms)
  YouTube Premium       : AVAILABLE (Region: DE, 831 ms)
  OpenAI / ChatGPT      : UNLOCKED (loc=DE, colo=FRA, 64 ms)
  Anthropic Claude      : AVAILABLE (Region DE Supported, 65 ms)
  Google Gemini         : AVAILABLE (DE) (471 ms)
  TikTok                : UNLOCKED (Region: DE) (471 ms)
  Disney+               : AVAILABLE (https://www.disneyplus.com/) (108 ms)
  Spotify               : AVAILABLE (DE) (162 ms)
  Reddit Unauth JSON    : OAuth Required on Datacenter ASN (HTTP 403, 165 ms)
  Steam Store           : AVAILABLE (Currency: EUR (€), 390 ms)
  ZDFmediathek (DACH)   : NATIVE DE ACCESS (HTTP 200) (81 ms)

--- IPv6 Unlock & Accessibility Audit ---
  Google Search         : CLEAN (No CAPTCHA, HTTP 200) (121 ms)
  Netflix               : UNLOCKED Full Catalog (Region: DE) (801 ms)
  YouTube Premium       : AVAILABLE (Region: DE, 664 ms)
  OpenAI / ChatGPT      : UNLOCKED (loc=DE, colo=HAM, 88 ms)
  Anthropic Claude      : AVAILABLE (Region DE Supported, 98 ms)
  Google Gemini         : AVAILABLE (DE) (240 ms)
  Disney+ / Spotify     : AVAILABLE (DE) (198 ms / 128 ms)
  ZDFmediathek (DACH)   : NATIVE DE ACCESS (HTTP 200) (81 ms)

--- Developer Package Registry TTFB Latency (IPv4 HTTPS) ---
  GitHub Releases / Web        -> HTTP 206 in  58.2 ms
  PyPI (Fastly CDN)            -> HTTP 206 in  60.6 ms
  Crates.io (Rust)             -> HTTP 206 in  61.4 ms
  npm Registry (Cloudflare)    -> HTTP 200 in 257.8 ms
  Docker Hub Registry          -> HTTP 401 in 309.4 ms
```

## 10. Global IX Routing, China 3-City / 8-Node Latency & Hop-by-Hop `MTR` (`AS197540 → AS47147 → AS1299`)

Hop 2 (`AS47147 94.16.25.158`, Netcup parent **Anexia**) hands off in `0.4 ms` inside Nuremberg to Tier-1 **Arelion (`AS1299`)** into [DE-CIX Frankfurt](https://www.de-cix.net/) (`3.58 ms`), delivering direct Europe-to-Asia transit without US detours. Across our 8 evening-peak probes in Beijing, Shanghai, and Guangzhou, China Unicom (`AS4837`, `169.4 ms` GZ / `147.9 ms` BJ) and China Mobile (`CMI AS58453`, `218.0 ms` GZ) maintained `0%` final-hop packet loss, whereas China Telecom (`163 AS4134`, `187.8 ms` avg) exhibited `16.7%` peak congestion loss (pair public APAC web frontends with Cloudflare CDN):

```text
root@vps500-g12-nue-lab:~# mtr -r -w -z -T -P 443 -c 3 1.1.1.1   # Frankfurt DE-CIX Cloudflare
HOST: vps500-g12-nue-lab                                      Loss%   Snt   Last   Avg  Best  Wrst StDev
  1. AS197540 152.53.xxx.x                                     0.0%     3    0.7   1.7   0.7   3.6   1.7
  2. AS47147  94.16.25.158                                     0.0%     3    0.8   0.7   0.6   0.8   0.1
  3. AS1299   nug-b2-link.ip.twelve99.net                      0.0%     3    0.4   0.5   0.4   0.5   0.0
  5. AS1299   ffm-b16-link.ip.twelve99.net                     0.0%     3    4.2   4.2   4.2   4.5   0.1
  8. AS13335  one.one.one.one                                  0.0%     3    3.7   7.0   3.7  13.4   5.6

root@vps500-g12-nue-lab:~# mtr -r -w -z -c 6 120.196.165.24     # Guangzhou China Mobile (CMI AS58453)
HOST: vps500-g12-nue-lab                                Loss%   Snt   Last   Avg  Best  Wrst StDev
  1. AS197540 152.53.xxx.x                               0.0%     6    4.7   8.8   0.4  18.3   6.4
  2. AS47147  94.16.25.158                               0.0%     6    0.6   0.6   0.4   0.7   0.1
  3. AS1299   nug-b2-link.ip.twelve99.net                0.0%     6    0.4   0.4   0.3   0.4   0.0
  7. AS1299   chinamobile-ic-355891.ip.twelve99-cust.net 0.0%     6    4.6   4.6   4.4   4.9   0.2
  8. AS58453  223.120.10.85                              0.0%     6    4.0   4.0   3.8   4.1   0.1
 15. AS9808   ns6.gd.cnmobile.net                        0.0%     6  218.3 218.2 218.1 218.4   0.1
```

## 11. SCP Cloud Firewall & Outbound SMTP (`25 / 465 / 587`) 1-Click Self-Service Unlock Proof

Fresh servers attach a default **`netcup Mail block` (`Default`)** rule group in `SCP → Firewall` (`servercontrolpanel.de`). Clicking **`Delete`** + **`Save`** opens IPv4 & IPv6 `25`, `465`, and `587` in one second with **zero support ticket required**:

```text
root@vps500-g12-nue-lab:~# # [BEFORE] Default "netcup Mail block" active in SCP → Firewall:
gmail-smtp-in.l.google   :25    -> TimeoutError (5005.2 ms — rule #1: drop outgoing SMTP 25)
smtp.gmail.com           :465   -> TimeoutError (3510.1 ms — rule #2: drop outgoing SMTPS 465)
smtp.gmail.com           :587   -> TimeoutError (3504.2 ms — rule #3: drop outgoing submission 587)

root@vps500-g12-nue-lab:~# # [AFTER] Self-service clicked "Delete" on "netcup Mail block" + "Save" in SCP:
IPv4 gmail-smtp-in.l.google.com             :25   -> OPEN ( 16.9 ms) | 220 mx.google.com ESMTP - gsmtp
IPv6 gmail-smtp-in.l.google.com             :25   -> OPEN ( 23.9 ms) | 220 mx.google.com ESMTP - gsmtp
IPv4 mx00.ionos.de                          :25   -> OPEN (  6.1 ms) | 220 kundenserver.de (mxeue002) Nemesis ESMTP Service ready
IPv4 outlook-com.olc.protection.outlook.com :25   -> OPEN ( 93.7 ms) | 220 Microsoft ESMTP MAIL Service ready
IPv4 mxbiz1.qq.com                          :25   -> OPEN (192.5 ms) | 220 bizmx54.qq.com MX QQ Mail Server
IPv4 smtp.gmail.com                         :465  -> OPEN (  4.0 ms)
IPv6 smtp.gmail.com                         :465  -> OPEN ( 15.8 ms)
IPv4 smtp.gmail.com                         :587  -> OPEN ( 15.7 ms) | 220 smtp.gmail.com ESMTP - gsmtp
IPv6 smtp.gmail.com                         :587  -> OPEN (  4.2 ms) | 220 smtp.gmail.com ESMTP - gsmtp
```

## 12. TCO Math & Choosing Between VPS 500, VPS 1000 & RS 1000

| Plan & Term (EU Auto-Assign) | vCores & Profile | RAM & NVMe | Best Voucher | Year-1 Effective (0% VAT Net) | Year-2+ Renewal Net |
| :--- | :--- | :--- | :--- | :--- | :--- |
| **[VPS 500 G12.5](/models/vps-500-g12-5/) (12M)** | 2 Shared (`EPYC-Rome` 2.0G) | 4 GB · 64 GB NVMe | `#5e` (`-€5.00` net / `-€5.95` gross, new only) | **€6.52/mo net** (€7.76 gross) | €6.94/mo net (€8.26 gross) |
| **VPS 500 G12.5 (24M)** | 2 Shared (`EPYC-Rome` 2.0G) | 4 GB · 64 GB NVMe | `#5e` (`-€5.00` net / `-€5.95` gross, new only) | **€5.69/mo net** (24M avg; €5.49 Yr-1) | €5.90/mo net (€7.03 gross) |
| **[VPS 1000 G12.5](/guides/vps-1000-g12-5-review/) (12M)** | 4 Shared EPYC vCores | 8 GB · 128 GB NVMe | `#vps1` (1M free) | **€11.17/mo net** (€13.29 gross) | €12.18/mo net |
| **[RS 1000 G12.5](/guides/rs-1000-g12-5-review/) (12M)** | **4 Dedicated Zen 5** | 8 GB · 128 GB NVMe | `#rs1` (2M free) | **€15.22/mo net** (€18.11 gross) | €18.26/mo net |

Check our [Non-EU 0% VAT Registration Guide](/guides/vat-free-registration-guide/), test contract and location options in the [Interactive TCO Calculator](/tools/calculator/?model=vps-500-g12-5), and verify your 1-month-to-month-end notice cutoff (or 24M B2B term-end cutoff) in the [Contract Cancellation Calculator](/tools/cancellation-calculator/).

### Which Tier Fits Your Workload? (Honest 4 GB Envelope)

- **Choose `VPS 500 G12.5` (`€6.52/mo` net / `€7.76/mo` gross Year 1 w/ `#5e`, new customers only):** Ideal for 6–12 Docker microservices, Redis/SQLite/PostgreSQL apps, WireGuard/Caddy edge gateways, and lightweight self-hosted mail (**Stalwart Mail**, **Poste.io**, or Mailcow strictly with `SKIP_CLAMD=y` — do not enable `ClamAV` on `3.8 GiB` usable RAM; see our [Port 25 & rDNS setup guide](/guides/netcup-port-25-mail-server-rdns-guide/)).
- **Choose [VPS 1000 G12.5](/guides/vps-1000-g12-5-review/) (`€11.17/mo` net w/ `#vps1`):** Best step-up if you already hold an active Netcup account (`#vps1` works for *Bestandskunden*, whereas `#5e` is first-order only) or need `8 GB` DDR5 ECC RAM for full Mailcow + `ClamAV` and `128 GB` NVMe storage.
- **Choose [RS 1000 G12.5](/guides/rs-1000-g12-5-review/) (`€15.22/mo` net w/ `#rs1`):** Best when your workload compiles Next.js/Rust images or runs heavy SQL queries continuously and needs 4 dedicated **AMD EPYC™ 9645 or 9655 (Zen 5)** physical threads instead of the `2.0 GHz EPYC-Rome` shared guest profile.

## Frequently Asked Questions

### Does Netcup throttle VPS 500 G12.5 CPU speed under sustained 100% load?

Not during normal sustained bursts. Across six consecutive 10-second windows of 100% 2-thread sysbench load (60 seconds continuous), throughput stayed between 993.17 and 1,026.36 events/sec (under 3.2% variance) while hypervisor steal remained between 1.26% and 1.31%.

### Is the Netcup Nuremberg IP clean for email hosting, Google Search, Netflix, and AI APIs?

Yes. Our test confirmed RPKI ROA VALID status, proxy=false on ip-api, 17/17 clean checks across Spamhaus ZEN, SpamCop, Barracuda, SORBS, UCEPROTECT L1, NiX Spam, and Blocklist.de, HTTP 200 zero-CAPTCHA Google Search, full non-original Netflix Germany catalog access, and native DE access to OpenAI ChatGPT, Anthropic Claude, and Google Gemini over IPv4 and IPv6.

### How fast are Redis, PostgreSQL/SQLite, and TLS encryption on VPS 500 G12.5?

On Debian 12, ephemeral Redis 7 hit 101,420 GET/s (non-pipelined, p50 0.223 ms) and 915,751 GET/s (P=16 pipelined). SQLite 3.40.1 WAL reached 239,690 bulk inserts/s, 33,074 fsync autocommit tx/s, and 154,787 indexed point SELECTs/s. OpenSSL 3.0.22 achieved 5.39 GB/s AES-128-GCM (2T) and 32,503.7 ECDSA P-256 signatures/sec.

### Are outbound SMTP ports 25, 465 & 587 open, or do I need a support ticket?

No support ticket is needed. Fresh Netcup servers ship with a default 'netcup Mail block (Default)' policy in SCP (servercontrolpanel.de -> Servers -> Firewall) dropping outgoing TCP 25, 465, and 587. Clicking Delete and Save in SCP opens IPv4 and IPv6 SMTP immediately (verified at 16.9 ms 220 ESMTP to Gmail MX and 6.1 ms to IONOS MX).

---
Source: EUServerLab (https://euserverlab.com/guides/vps-500-g12-5-review/) | Open Telemetry Dataset: https://euserverlab.com/benchmarks/vps-500-g12-5.json
