iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
A netcup VPS 1000 G12 tested in Vienna in July 2026 delivered 1,495 sysbench CPU events per second on one thread and 5,974 on four threads. The same run reported about 30 GB/s of memory throughput, 101,000 random 4K read IOPS, and a download rate near 2 Gbit/s from a test server in Germany. These are results from one shared VPS at one point in time—not guaranteed speeds for every customer. They also are not disk benchmarks of today’s VPS 1000 G12.5, which is listed with 128 GB SSD rather than the tested G12’s 256 GB NVMe.
What the July 2026 benchmark measured
Serverküche reported testing a netcup VPS 1000 G12 in the provider’s Vienna data centre in July 2026. The machine ran Debian 13 and had an AMD EPYC-Genoa processor, 4 vCores, 8 GB of RAM, and 256 GB of NVMe storage. The publisher used sysbench 1.0.20 and fio 3.39, alongside 7-Zip, OpenSSL, and curl tests. Its article, published July 28, 2026 and updated September 22, 2026, includes command lines and excerpts of the output.
The figures below are the publisher’s reported results. Serverküche notes that VPS instances share physical CPU resources, so other tenants’ workloads can affect performance. Netcup’s official VPS overview also says that VPS customers do not receive dedicated CPU cores.
| Area | Reported result | What it represents |
|---|---|---|
| CPU | 1,495 events/s on one sysbench thread; 5,974 events/s on four threads | Single-thread and four-thread results from sysbench 1.0.20 |
| 7-Zip | About 27,400 MIPS total | The benchmark’s reported 7-Zip result |
| AES-256-GCM | About 7.7 GB/s on one core | OpenSSL encryption throughput reported for one core |
| Memory | About 30 GB/s | Reported RAM throughput |
| Random disk I/O | 101,000 IOPS read; 67,000 IOPS write | Random 4K operations |
| Sequential disk I/O | 4.3 GB/s read; 3.1 GB/s write | Sequential read and write results |
| Network download | About 246 MB/s, or roughly 2 Gbit/s, from Falkenstein, Germany; around 36 MB/s from a single test in Ashburn, Virginia | Downloads from specific remote test locations, not a general internet-speed guarantee |
| Latency | About 12 ms average ping to 1.1.1.1; about 26 ms over IPv6 | Ping measurements in the reported run |
| CPU steal | 0% during the run | Observed CPU steal time for this test, not a promise about other instances or times |
How fast is that in practical terms?
CPU performance
The one-thread result gives a reference for work that cannot use multiple cores effectively; the four-thread result shows the same benchmark using the plan’s four vCores. The four-thread score is about four times the one-thread score in this particular run. That ratio should not be treated as a scaling guarantee: the test reflects this instance and its conditions, and shared CPU availability can vary.
#1 Best Overall
The reported 7-Zip and AES results add context for compression and encryption workloads, but they are also workload-specific measurements. They do not predict every application’s performance. A database, web server, or build job may respond differently depending on its software, data, concurrency, and storage access pattern.
Memory and storage
About 30 GB/s is the benchmark’s reported memory throughput, not the amount of memory available to applications: the tested VPS had 8 GB RAM. Disk figures are especially tied to the tested generation and device. The random 4K results describe small, scattered reads and writes, while the sequential results describe larger continuous transfers. Neither workload alone captures all real applications.
Network throughput and latency
The roughly 246 MB/s download (about 2 Gbit/s) came from a test server in Falkenstein, Germany. The single Ashburn, Virginia test was much slower, at about 36 MB/s. These results show why a speed test is inseparable from its destination: distance, route, remote-server capacity, and the number of connections can affect throughput. The 12 ms IPv4 and 26 ms IPv6 average ping values were measured to 1.1.1.1 and likewise should not be assumed for a different destination or network path.
Does this benchmark apply to the current VPS 1000 G12.5?
No, not as a direct disk benchmark. The test was of the G12 configuration with 256 GB NVMe. Netcup’s current VPS 1000 listing is G12.5, with 4 x86 vCores and 8 GB RAM, but 128 GB SSD. The CPU and memory results can serve as a historical reference for a similar vCore and RAM configuration; the published NVMe measurements are not measurements of the current G12.5 storage.
Rank #3
- HP MicroServer Gen10 Plus Tower Server for Business with Microsoft Windows Server 2019 OS!
- Intel Xeon E-2224 Quad-Core 3.4GHz 8MB CPU, Up To 4.6GHz Turbo
- 32GB (2 x 16GB) DDR4 PC4-21300 2666MHz Unbuffered Memory
- 16TB (4 x 4TB) 7.2K 6Gb/s SATA 3.5" HDDs in RAID
- Hard drives and memory upgrades included separately NOT installed, installation required.
Netcup’s current overview lists the VPS 1000 G12.5 from €14.50 per month before VAT on a 12-month term. The configured product price varies with contract, location, and network options. Its product listing specifies a 2.5 Gbit/s interface and says traffic is temporarily throttled to 200 Mbps if average traffic over the preceding 24 hours exceeds 2 TB. These are listing terms, not benchmark results; check the official plan pages for the configuration and terms available when ordering.
Netcup’s Generation 12 announcement, dated January 20, 2026, named Nuremberg, Vienna, Amsterdam, Manassas, and Singapore as available locations. The announcement’s launch discount ended February 3, 2026, so it is not a current offer.
Rank #4
How to compare your VPS fairly
A useful comparison needs to match more than the headline score. Serverküche recommends running comparable tests two or three times at different times and using the median, while discarding a warm first run. For CPU and memory, keep the tool versions, thread count, and workload parameters the same. For disk, compare the same random or sequential workload and use direct I/O; for network, use the same destination and note its location.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →- CPU: Compare one-thread and four-thread sysbench results separately. Record the sysbench version and watch for CPU steal time with
vmstatortop. Persistent high steal can indicate contention, while 0% in a single run only describes that observation. - Disk: Keep the workload and direct-I/O setting comparable. Serverküche’s tests use fio with direct I/O;
fio --direct=1bypasses the page cache. Without direct I/O, a result may partly measure cached data in RAM rather than storage performance. - Network: Record the endpoint, its location, and whether the test uses one connection. A single connection may not saturate the network path, and a remote server or long route can limit the result.
- Conditions: Note the data-centre region, test time, operating system, and repeat results. A comparison across different regions or at different levels of host contention may not isolate the provider or plan as the cause of a score difference.
For a purchase comparison, also check storage type and capacity, location, traffic rules, contract length, and total price including VAT. Those product details are separate from a benchmark score.
Best Value
Verdict
The July 2026 test presents a strong snapshot: good reported CPU throughput for four vCores, high memory and storage figures, and a near-2-Gbit/s download from a nearby German server. Its value is as a reproducible reference, not a performance guarantee. In particular, the measured disk results belong to the 256 GB NVMe G12 instance and should not be passed off as results for the currently listed 128 GB SSD G12.5.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
