netcup Turin root servers

akrdev

This time payment is yearly for root server with 12 month contract. Earlier they used to raise the bill every two months.

Slewrate

anyone compared Turin to Grado series?

honovida

@kse said:

@honovida said:

@Eplekake said:
If only VM-x/AMD-V was enabled....

What is it

virtualization support to run KVM in a KVM virtual machine. also called nested virtualization

Thank you

pbx

@pbx said: When we don't choose the location, do you guys know which location will the server more likely be? They probably have a lot of capacity left in NUE but it might be cheaper for them to provision servers in Austria (as it's their own DC there)...

Just tried, the RS got provisioned in DE.

Yabs :

# ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #
#              Yet-Another-Bench-Script              #
#                     v2025-04-20                    #
# https://github.com/masonr/yet-another-bench-script #
# ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #

Thu Sep 18 11:24:55 PM CEST 2025

Basic System Information:
---------------------------------
Uptime     : 0 days, 0 hours, 4 minutes
Processor  : AMD EPYC 9645 96-Core Processor
CPU cores  : 4 @ 2296.550 MHz
AES-NI     : ✔ Enabled
VM-x/AMD-V : ❌ Disabled
RAM        : 7.8 GiB
Swap       : 0.0 KiB
Disk       : 251.9 GiB
Distro     : Debian GNU/Linux 12 (bookworm)
Kernel     : 6.1.0-39-amd64
VM Type    : KVM
IPv4/IPv6  : ✔ Online / ✔ Online

IPv6 Network Information:
---------------------------------
ISP        : ANEXIA Internetdienstleistungs GmbH
ASN        : AS47147 ANEXIA Internetdienstleistungs GmbH
Host       : Netcup KVM
Location   : Nuremberg, Bavaria (BY)
Country    : Germany

fio Disk Speed Tests (Mixed R/W 50/50) (Partition /dev/vda4):
---------------------------------
Block Size | 4k            (IOPS) | 64k           (IOPS)
  ------   | ---            ----  | ----           ---- 
Read       | 322.80 MB/s  (80.7k) | 1.60 GB/s    (25.0k)
Write      | 323.66 MB/s  (80.9k) | 1.60 GB/s    (25.1k)
Total      | 646.47 MB/s (161.6k) | 3.20 GB/s    (50.1k)
           |                      |                     
Block Size | 512k          (IOPS) | 1m            (IOPS)
  ------   | ---            ----  | ----           ---- 
Read       | 1.45 GB/s     (2.8k) | 1.63 GB/s     (1.5k)
Write      | 1.53 GB/s     (2.9k) | 1.74 GB/s     (1.7k)
Total      | 2.98 GB/s     (5.8k) | 3.38 GB/s     (3.3k)

iperf3 Network Speed Tests (IPv4):
---------------------------------
Provider        | Location (Link)           | Send Speed      | Recv Speed      | Ping           
-----           | -----                     | ----            | ----            | ----           
Clouvider       | London, UK (10G)          | 2.71 Gbits/sec  | 2.36 Gbits/sec  | 26.1 ms        
Eranium         | Amsterdam, NL (100G)      | 1.50 Gbits/sec  | 2.34 Gbits/sec  | 20.8 ms        
Uztelecom       | Tashkent, UZ (10G)        | 1.49 Gbits/sec  | 1.50 Gbits/sec  | 91.9 ms        
Leaseweb        | Singapore, SG (10G)       | 1.43 Gbits/sec  | 1.31 Gbits/sec  | 156 ms         
Clouvider       | Los Angeles, CA, US (10G) | 1.12 Gbits/sec  | 594 Mbits/sec   | 150 ms         
Leaseweb        | NYC, NY, US (10G)         | 1.40 Gbits/sec  | 2.14 Gbits/sec  | 92.5 ms        
Edgoo           | Sao Paulo, BR (1G)        | 151 Mbits/sec   | 106 Mbits/sec   | 228 ms         

iperf3 Network Speed Tests (IPv6):
---------------------------------
Provider        | Location (Link)           | Send Speed      | Recv Speed      | Ping           
-----           | -----                     | ----            | ----            | ----           
Clouvider       | London, UK (10G)          | 2.70 Gbits/sec  | 2.28 Gbits/sec  | 26.1 ms        
Eranium         | Amsterdam, NL (100G)      | 2.68 Gbits/sec  | 2.32 Gbits/sec  | 20.7 ms        
Uztelecom       | Tashkent, UZ (10G)        | 2.40 Gbits/sec  | 1.89 Gbits/sec  | 92.0 ms        
Leaseweb        | Singapore, SG (10G)       | 546 Mbits/sec   | 1.63 Gbits/sec  | 156 ms         
Clouvider       | Los Angeles, CA, US (10G) | 872 Mbits/sec   | 815 Mbits/sec   | 149 ms         
Leaseweb        | NYC, NY, US (10G)         | 2.44 Gbits/sec  | 1.93 Gbits/sec  | 92.5 ms        
Edgoo           | Sao Paulo, BR (1G)        | 183 Mbits/sec   | 151 Mbits/sec   | 256 ms         

Geekbench 6 Benchmark Test:
---------------------------------
Test            | Value                         
                |                               
Single Core     | 2003                          
Multi Core      | 6228                          
Full Test       | https://browser.geekbench.com/v6/cpu/13902487

YABS completed in 12 min 20 sec

gbzret4d

@loay said:

@pbx said:

@itachikonoha said: 8.74 euro for 1 year contract supposing you don't chose location.

When we don't choose the location, do you guys know which location will the server more likely be? They probably have a lot of capacity left in NUE but it might be cheaper for them to provision servers in Austria (as it's their own DC there)...

Most of the time it will be provided in Austria unless it is one of their old gen offers.

Is this true? If so, i will get one.

kse

@gbzret4d said:

@loay said:

@pbx said:

@itachikonoha said: 8.74 euro for 1 year contract supposing you don't chose location.

When we don't choose the location, do you guys know which location will the server more likely be? They probably have a lot of capacity left in NUE but it might be cheaper for them to provision servers in Austria (as it's their own DC there)...

Most of the time it will be provided in Austria unless it is one of their old gen offers.

Is this true? If so, i will get one.

i guess it always depends on how many ressources they have available at which location. i experienced that it always changes. most likely some algorithm to determine where it should be placed to balance the VMs over the different locations

Protocol903

> @pbx said: Geekbench 6 Benchmark Test:

Test | Value
|
Single Core | 2003
Multi Core | 6228
Full Test | https://browser.geekbench.com/v6/cpu/13902487

YABS completed in 12 min 20 sec

this is bad... its just a few points away from being same as G11 performance.

itachikonoha

@Protocol903 said:

> @pbx said: Geekbench 6 Benchmark Test:

Test | Value
|
Single Core | 2003
Multi Core | 6228
Full Test | https://browser.geekbench.com/v6/cpu/13902487

YABS completed in 12 min 20 sec

this is bad... its just a few points away from being same as G11 performance.

One shouldn't buy from netcup if performance is an important parameter.

Netcup is good for middle of anything. It is neither good nor bad.

jackdoggett

@Protocol903 said:

> @pbx said: Geekbench 6 Benchmark Test:

Test | Value
|
Single Core | 2003
Multi Core | 6228
Full Test | https://browser.geekbench.com/v6/cpu/13902487

YABS completed in 12 min 20 sec

this is bad... its just a few points away from being same as G11 performance.

They are using zen 5c CPUs, which is zen 5 with cut down l3 cache. The performance is pretty close to zen 4. The whole point of zen 5c is squeeze more (lower performance and smaller) cores into the same die size so cloud companies can sell more VPS. CPU wise there's no advantage with the G12 root servers vs the G11 root servers which use full zen 4 cores.

pbx

@Protocol903 said: this is bad... its just a few points away from being same as G11 performance.

Yeah, the result is lower than I expected. Pretty close to the performance of a G11 RS. Good to know I guess

Protocol903

@pbx said: Yeah, the result is lower than I expected. Pretty close to the performance of a G11 RS. Good to know I guess

I raised a ticket with them. Lets see how it goes.

Protocol903

Your case has been carefully checked by our technical department. The results you observed, including the single-core values, are within the expected range for the G12 product line. Please note that benchmark scores can differ depending on architecture, workload, and measurement conditions. A newer generation does not always achieve higher results in every individual test compared to an older one.

We can confirm that there are no irregularities on the host system and that the guaranteed resources for your server are fully available.

If you experience performance issues in your actual applications that go beyond such benchmark variations, please let us know together with concrete examples so we can review this more precisely.

bruh

Advin

@jackdoggett said:

@Protocol903 said:

> @pbx said: Geekbench 6 Benchmark Test:

Test | Value
|
Single Core | 2003
Multi Core | 6228
Full Test | https://browser.geekbench.com/v6/cpu/13902487

YABS completed in 12 min 20 sec

this is bad... its just a few points away from being same as G11 performance.

They are using zen 5c CPUs, which is zen 5 with cut down l3 cache. The performance is pretty close to zen 4. The whole point of zen 5c is squeeze more (lower performance and smaller) cores into the same die size so cloud companies can sell more VPS. CPU wise there's no advantage with the G12 root servers vs the G11 root servers which use full zen 4 cores.

Power efficiency is also a big factor.

The AMD EPYC 9645 based on Zen 5c can only boost to 3.7 GHz (3.3 GHz all-core) and has a 320W TDP.
The AMD EPYC 9655 based on Zen 5 can boost to 4.5 GHz (4.1 GHz all-core) and has a 400W TDP.

Both processors have the same number of cores.

It wouldn't surprised me if Netcup switched due to new EPYC processors having more cores, better availability, and around the same power as their older processors. Performance won't be "significantly" better than what Netcup was using with Genoa instances.

That being said, Netcup's price for Turin is still very reasonable, even if it's the more power efficient CPU SKU.

supple

@remy said: Am I cured?

Not much of an upgrade compared to offers from other LET providers or their previous generation.

turin is 512bit AVX512

Protocol903

I simply got a refund for it. Performance continued to dip. Back to my G11 Vienna babe