All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
[benchmark comparison] netcup VPS 200 G8 vs Hetzner CX 11
I have always been curious to see how the two cheapest options for a VPS offered by netcup and Hetzner compare with each other. So I went for it and purchased both to run some benchmarks. Here are their characteristics:
netcup VPS 200 G8
1 vCore
2 GB RAM
20 GB SSD (RAID 10)
40 TB traffic
2.69EUR/month
Hetzner CX 11
1 vCore
2 GB RAM
20 GB NVMe
20 TB traffic
2.96EUR/month
Prices are with my local VAT. There is a difference in the way the two providers bill. Both VPSs are hourly VPS with a max monthly cap. Netcup bills once every 6 months and asks you to anticipate 6 months of usage (16.14 EUR). If and whenever you decide to cancel your VPS you get refunded prorated. Hetzner bills once each month asking you to pay for the usage.
Let's benchmark them. I installed a fresh Debian 10 on both and run iostat several times, at the same time, while they were idling. CPU steal for both was nothing or close to nothing (0.02 for netcup, 0.00 for Hetzner).
CPU flags
netcup
flags : fpu de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pse36 clflush mmx fxsr sse sse2 syscall nx lm rep_good nopl xtopology cpuid tsc_known_freq pni pclmulqdq ssse3 cx16 pcid sse4_1 sse4_2 x2apic popcnt aes xsave hypervisor lahf_lm cpuid_fault invpcid_single pti ssbd ibrs ibpb invpcid md_clear
Hetzner
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch cpuid_fault invpcid_single pti ssbd ibrs ibpb fsgsbase bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx avx512f avx512dq rdseed adx smap clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 arat
bugs : cpu_meltdown spectre_v1 spectre_v2 spec_store_bypass l1tf mds
nench.sh
Netcup
-------------------------------------------------
nench.sh v2019.06.29 -- https://git.io/nench.sh
benchmark timestamp: 2019-07-15 19:50:14 UTC
-------------------------------------------------
Processor: QEMU Virtual CPU version 2.5+
CPU cores: 1
Frequency: 2399.994 MHz
RAM: 1.9Gi
Swap: -
Kernel: Linux 4.19.0-5-amd64 x86_64
Disks:
sda 20G HDD
CPU: SHA256-hashing 500 MB
3.320 seconds
CPU: bzip2-compressing 500 MB
5.269 seconds
CPU: AES-encrypting 500 MB
1.629 seconds
ioping: seek rate
min/avg/max/mdev = 56.0 us / 91.1 us / 7.85 ms / 61.1 us
ioping: sequential read speed
generated 21.7 k requests in 5.00 s, 5.31 GiB, 4.35 k iops, 1.06 GiB/s
dd: sequential write speed
1st run: 611.31 MiB/s
2nd run: 1049.04 MiB/s
3rd run: 1049.04 MiB/s
average: 903.13 MiB/s
IPv4 speedtests
your IPv4: <REDACTED>
Cachefly CDN: 105.15 MiB/s
Leaseweb (NL): 77.91 MiB/s
Softlayer DAL (US): 10.84 MiB/s
Online.net (FR): 57.42 MiB/s
OVH BHS (CA): 21.52 MiB/s
IPv6 speedtests
your IPv6: 2a03:4000:1d:xxxx
Leaseweb (NL): 87.65 MiB/s
Softlayer DAL (US): 0.00 MiB/s
Online.net (FR): 48.83 MiB/s
OVH BHS (CA): 14.47 MiB/s
-------------------------------------------------
Hetzner
-------------------------------------------------
nench.sh v2019.06.29 -- https://git.io/nench.sh
benchmark timestamp: 2019-07-15 19:50:45 UTC
-------------------------------------------------
Processor: Intel Xeon Processor (Skylake, IBRS)
CPU cores: 1
Frequency: 2100.000 MHz
RAM: 1.9Gi
Swap: -
Kernel: Linux 4.19.0-5-amd64 x86_64
Disks:
sda 19.1G HDD
CPU: SHA256-hashing 500 MB
3.578 seconds
CPU: bzip2-compressing 500 MB
5.998 seconds
CPU: AES-encrypting 500 MB
1.253 seconds
ioping: seek rate
min/avg/max/mdev = 37.4 us / 73.6 us / 11.4 ms / 115.8 us
ioping: sequential read speed
generated 16.7 k requests in 5.00 s, 4.08 GiB, 3.34 k iops, 835.4 MiB/s
dd: sequential write speed
1st run: 355.72 MiB/s
2nd run: 325.20 MiB/s
3rd run: 391.96 MiB/s
average: 357.63 MiB/s
IPv4 speedtests
your IPv4: <REDACTED>
Cachefly CDN: 99.21 MiB/s
Leaseweb (NL): 171.83 MiB/s
Softlayer DAL (US): 12.88 MiB/s
Online.net (FR): 141.08 MiB/s
OVH BHS (CA): 23.75 MiB/s
IPv6 speedtests
your IPv6: 2a01:4f8:c2c:xxxx
Leaseweb (NL): 163.30 MiB/s
Softlayer DAL (US): 0.00 MiB/s
Online.net (FR): 152.83 MiB/s
OVH BHS (CA): 23.07 MiB/s
-------------------------------------------------
Geekbench
Netcup: https://browser.geekbench.com/v4/cpu/13909443
Hetzner: https://browser.geekbench.com/v4/cpu/13909455
Serverscope / disk test
The fio
part of Serverscope does not currently work with Debian 10 (fio
fails to build).
So I inspected its source code and run its fio
tests manually with the Debian version of it.
Netcup
random read: IOPS=22.3k, BW=87.3MiB/s (91.5MB/s)(5238MiB/60004msec)
random direct read: IOPS=22.8k, BW=88.0MiB/s (93.3MB/s)(5339MiB/60005msec)
random write: IOPS=122k, BW=478MiB/s (501MB/s)(28.0GiB/60013msec); 0 zone resets
random direct write: IOPS=8664, BW=33.8MiB/s (35.5MB/s)(2031MiB/60009msec); 0 zone resets
Hetzner
random read: IOPS=38.9k, BW=152MiB/s (159MB/s)(9124MiB/60004msec)
random direct read: IOPS=45.3k, BW=177MiB/s (186MB/s)(10.4GiB/60003msec)
random write: IOPS=111k, BW=433MiB/s (455MB/s)(25.4GiB/60003msec); 0 zone resets
random direct write: IOPS=11.6k, BW=45.2MiB/s (47.4MB/s)(2711MiB/60005msec); 0 zone resets
Advantages of netcup:
- cheaper, but weird billing system
- high bandwidth
- better CPU
- peering with DTAG (someone might like this one)
Advantages of Hetzner:
- flexibility in launching and destroying instances and the billing system
- better I/O
Nothing really surprising here, but at least there is empirical data about it now. I leave it to the community for future reference. It would be interesting to see how the higher spec'ed VPSs compare to each other.
Comments
Can you do it for Scaleway's DEV1-S and OVH's VPS SSD 1 too? Considering all of these are < 3 EUR a month range.
Is that at Nuremberg? I always get lower iops/slower CPU at Falkenstein.
I share benchmarks and reviews of systems I am interested to use for something. I am currently not interested in France (or The Netherlands) as a location. But feel free to do it yourself and share with us comparisons are interesting.
This one is at Nuremberg, yes.
Netcup's advantages here don't seem to outweigh the billing pain, at least for this product. Their more interesting products are the root servers imho, because of dedicated cpu.
If only they provided nested virtualization by default. Would be insta buy for me.
@willie What do you think about Contabo.com ? They are like holy shit when it comes to offering RAM + Threads + storage. It is like they are sponsored by DE government.
Check my review in my sig. Ask them to lift the SSD limit and you get a respectable server
@Ympker Nice.
I believe Scaleway uses EPYC 7401P's so single thread performance won't be the greatest. I'm not a fan of obsessing over small differences though. It's more important how loaded the entire box is most of the time.
I think it is 2.49 EUR without EU vat?
If someone have both Hetzner CX41 and Netcup RS 2000 G8, I'd love to see the iops and CPU performance differences.
Yes. You set your location (and VAT) on Hetzner with the horizontal bar on top.
It wouldn't be a fair comparison, though. Netcup has dedicated cores, Hetzner does not and the performance can't be sustained. Netcup has Xeon Gold 6140 cores, Hetzner has (likely) Silver 4114 cores. The difference in IOPS will be similar to the one showed in this benchmark, so around 2x for Hetzner.
@sgheghele Why is the dedicated CPU option more expensive than dedicated machines from them? :P
Also, if you were to pick btw Scaleway DEV1-S and Hetzner which's base VPS which one would you go with?
You mean those from Hetzner? There are different business cases here. If you need the flexibility of the cloud (distributed systems, dynamic load balancing, scaling, ..) you go for the cloud instead of several dedicated servers. If you need sustained and strong CPU usage, you pick their cloud VPS with dedicated CPUs.
I hope you will realize that these "cloud" VPSs are NOT made with us LET people as target customers.
Ok, I see that you are new in here and already got a warning. Here is how to survive here:
For example, if you knew a bit more about these vendors and networking in general, you would realize that asking someone randomly "which do you pick, A or B" would not result in an answer that it is useful for you. To pick me as example, I already told you a couple of posts ago that I do not care about France and The Netherlands as a location. So, I would pick Hetzner no matter what because I can not give a crap about Scaleway's characteristics as it is not located in Germany.
So you love nothing but Germany?
^ don't be a dick, it tends to bring out the wurst in the ever-so-patient people here ... lol
Netcup's dedicated core vps have monthly-only billing, so you have to compare them to dedicated servers, not Hetzner VPS.
nobody loves germany. most germany-haters are germans.
FTFY.