Note
This article was last updated on 2024-03-09, the content may be out of date.
Introduction
netcup GmbH sells interesting ARM servers. I bought the VPS 8000 ARM G11 version recently. It has 18 vCore (ARM64), offical website said it’s based on Ampere Altra Max CPUs.
| Products |
VPS 8000 ARM G11 |
| Virtualisation technology |
KVM |
| Processor |
18 vCore (ARM64) |
| RAM |
64GB DDR4 ECC |
| Storage space |
2048 GB SSD |
| Network card |
2500 Mbit/s |
| Traffic |
inclusive |
General Testing
Testing script from Teddysun provides an idea of over all performance
1
|
wget -qO- bench.sh | bash
|
The result is here:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
|
-------------------- A Bench.sh Script By Teddysun -------------------
Version : v2023-10-15
Usage : wget -qO- bench.sh | bash
----------------------------------------------------------------------
CPU Model : CPU model not detected
CPU Cores : 18
AES-NI : ✓ Enabled
VM-x/AMD-V : ✗ Disabled
Total Disk : 2.0 TB (20.7 GB Used)
Total Mem : 62.7 GB (3.1 GB Used)
System uptime : 1 days, 14 hour 48 min
Load average : 0.25, 0.14, 0.10
OS : Debian GNU/Linux 12
Arch : aarch64 (64 Bit)
Kernel : 6.6.8
TCP CC : bbr
Virtualization : KVM
IPv4/IPv6 : ✓ Online / ✓ Online
Organization : AS197540 netcup GmbH
Location : Nürnberg / DE
Region : Bavaria
----------------------------------------------------------------------
I/O Speed(1st run) : 470 MB/s
I/O Speed(2nd run) : 673 MB/s
I/O Speed(3rd run) : 676 MB/s
I/O Speed(average) : 606.3 MB/s
----------------------------------------------------------------------
Node Name Upload Speed Download Speed Latency
Speedtest.net 174.35 Mbps 147.69 Mbps 0.29 ms
Los Angeles, US 626.82 Mbps 1950.34 Mbps 147.10 ms
Dallas, US 697.90 Mbps 2109.19 Mbps 133.37 ms
Montreal, CA 554.36 Mbps 904.78 Mbps 94.53 ms
Paris, FR 2108.91 Mbps 2321.03 Mbps 21.98 ms
Shanghai, CN 640.20 Mbps 1484.39 Mbps 255.22 ms
Hongkong, CN 5.00 Mbps 0.55 Mbps 301.56 ms
Mumbai, IN 1685.98 Mbps 2417.37 Mbps 116.27 ms
Singapore, SG 1592.12 Mbps 2572.71 Mbps 179.95 ms
Tokyo, JP 423.71 Mbps 1122.09 Mbps 217.83 ms
----------------------------------------------------------------------
Finished in : 5 min 15 sec
Timestamp : 2023-12-30 03:55:03 CET
----------------------------------------------------------------------
|
Yet-Another-Bench-Script
1
|
curl -sL yabs.sh | bash
|
The result is very detailed and long, I split the result into parts. First, let’s check the script information:
1
2
3
4
5
6
7
8
9
|
# ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #
# Yet-Another-Bench-Script #
# v2023-11-30 #
# https://github.com/masonr/yet-another-bench-script #
# ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #
Sat Dec 30 04:16:48 AM CET 2023
ARM compatibility is considered *experimental*
|
Then, the basic OS information:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
|
Basic System Information:
---------------------------------
Uptime : 1 days, 15 hours, 15 minutes
Processor : Neoverse-N1
BIOS virt-5.2 CPU @ 2.0GHz
CPU cores : 18 @ ??? MHz
AES-NI : ✔ Enabled
VM-x/AMD-V : ❌ Disabled
RAM : 62.7 GiB
Swap : 0.0 KiB
Disk : 2.0 TiB
Distro : Debian GNU/Linux 12 (bookworm)
Kernel : 6.6.8
VM Type : KVM
IPv4/IPv6 : ✔ Online / ✔ Online
IPv6 Network Information:
---------------------------------
ISP : netcup GmbH
ASN : AS197540 netcup GmbH
Host : NETCUP-GMBH
Location : Karlsruhe, Baden-Wurttemberg (BW)
Country : Germany
|
Then, the Disk IO Test:
1
2
3
4
5
6
7
8
9
10
11
12
13
|
fio Disk Speed Tests (Mixed R/W 50/50) (Partition /dev/vda4):
---------------------------------
Block Size | 4k (IOPS) | 64k (IOPS)
------ | --- ---- | ---- ----
Read | 138.63 MB/s (34.6k) | 261.70 MB/s (4.0k)
Write | 138.54 MB/s (34.6k) | 269.48 MB/s (4.2k)
Total | 277.18 MB/s (69.2k) | 531.19 MB/s (8.2k)
| |
Block Size | 512k (IOPS) | 1m (IOPS)
------ | --- ---- | ---- ----
Read | 298.21 MB/s (582) | 442.51 MB/s (432)
Write | 323.72 MB/s (632) | 493.71 MB/s (482)
Total | 621.93 MB/s (1.2k) | 936.22 MB/s (914)
|
Then, the Network Speed Test:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
|
iperf3 Network Speed Tests (IPv4):
---------------------------------
Provider | Location (Link) | Send Speed | Recv Speed | Ping
----- | ----- | ---- | ---- | ----
Clouvider | London, UK (10G) | 2.45 Gbits/sec | 2.37 Gbits/sec | 20.5 ms
Scaleway | Paris, FR (10G) | 2.46 Gbits/sec | busy | 20.1 ms
NovoServe | North Holland, NL (40G) | 2.47 Gbits/sec | 2.38 Gbits/sec | 9.96 ms
Uztelecom | Tashkent, UZ (10G) | 1.99 Gbits/sec | 996 Mbits/sec | 84.2 ms
Clouvider | NYC, NY, US (10G) | 2.00 Gbits/sec | 542 Mbits/sec | 91.6 ms
Clouvider | Dallas, TX, US (10G) | 1.44 Gbits/sec | 648 Mbits/sec | 128 ms
Clouvider | Los Angeles, CA, US (10G) | 1.16 Gbits/sec | 1.04 Gbits/sec | 147 ms
iperf3 Network Speed Tests (IPv6):
---------------------------------
Provider | Location (Link) | Send Speed | Recv Speed | Ping
----- | ----- | ---- | ---- | ----
Clouvider | London, UK (10G) | 2.44 Gbits/sec | 2.33 Gbits/sec | 20.5 ms
Scaleway | Paris, FR (10G) | busy | busy | 32.2 ms
NovoServe | North Holland, NL (40G) | 2.48 Gbits/sec | 2.35 Gbits/sec | 9.97 ms
Uztelecom | Tashkent, UZ (10G) | 2.10 Gbits/sec | 875 Mbits/sec | 84.4 ms
Clouvider | NYC, NY, US (10G) | 1.78 Gbits/sec | 893 Mbits/sec | 91.6 ms
Clouvider | Dallas, TX, US (10G) | 1.39 Gbits/sec | 852 Mbits/sec | 128 ms
Clouvider | Los Angeles, CA, US (10G) | 1.16 Gbits/sec | 549 Mbits/sec | 147 ms
|
Then, Geekbench 6 Result:
1
2
3
4
5
6
7
|
Geekbench 6 Benchmark Test:
---------------------------------
Test | Value
|
Single Core | 1072
Multi Core | 8650
Full Test | https://browser.geekbench.com/v6/cpu/4192622
|
Finally, I tested Geekbench 5 only, with following script:
1
|
curl -sL yabs.sh | bash -s -- -i -f -n -5
|
Here is Geekbench 6 Result:
1
2
3
4
5
6
7
|
Geekbench 5 Benchmark Test:
---------------------------------
Test | Value
|
Single Core | 798
Multi Core | 10975
Full Test | https://browser.geekbench.com/v5/cpu/22091599
|
Review
Now, let’s discuss the result.
VPS 8000 ARM G11’s Geekbench 6 Multi-Core Score is around 8650, to help you get an idea od its level, I choiced some well known CPUs for comparing.
| CPU |
Single-Core |
Multi-Core Score |
| Intel Core i9-13900K |
2970 |
20095 |
| AMD Ryzen 9 7950X |
2930 |
19224 |
| Intel Core i9-12900K |
2610 |
15428 |
| AMD Ryzen 9 7940HS |
2481 |
11741 |
| AMD Ryzen 7 5800X3D |
2085 |
10922 |
| AMD Ryzen Z1 Extreme |
2252 |
9689 |
| AMD Ryzen 7 7840U |
2099 |
8718 |
| Intel Core i9-9900K |
1664 |
7977 |
My Framework Laptop 13 with AMD Ryzen 7 7840U has a score of
| CPU |
Single-Core |
Multi-Core Score |
| AMD Ryzen 7 7840U |
2256 |
12127 |
This difference is because modern CPUs have dynamic booster mechanism, it can run above the base clock.
One more thing, I though a few records of Ampere Altra Max 2800 MHz (96 cores) on the Geekbench 6 website. The Single-Core Score is 958 and Multi-Core Score is 11149.
This is odd, my 18 vCore server has a Multi-Core Score of 8650, the Altra Max 2800 MHz (96 cores) got only 11149. I guess the Altra Max 2800 MHz (96 cores) records on the Geekbench 6 website is under performanced, since it’s ADLINK Ampere Altra Developer Platform. It may be not tuned to optimized perfermance.
James