[Three-network CMI] DMIT Hong Kong HKG.EB.TINYv2, evening peak speed tests and international connectivity, detailed test

Not many people have used DMIT's Hong Kong EB route, and few have tested it, so let's give it a try.

1. Configuration

Looking Glass test address

2. NodeQuality

bash <(curl -sL https://run.NodeQuality.com)

💻Basic information

Basic System Information:
---------------------------------
Uptime     : 9 days, 14 hours, 1 minutes
Processor  : AMD EPYC 7C13 64-Core Processor
CPU cores  : 1 @ 1996.248 MHz
AES-NI     : ✔ Enabled
VM-x/AMD-V : ❌ Disabled
RAM        : 967.9 MiB
Swap       : 533.0 MiB
Disk       : 21.0 GiB
Distro     : Debian GNU/Linux 12 (bookworm)
Kernel     : 6.12.43+deb13-cloud-amd64
VM Type    : STANDARD PC (Q35 + ICH9, 2009)
IPv4/IPv6  : ✔ Online / ✔ Online

IPv6 Network Information:
---------------------------------
ISP        : US-DMITINC
ASN        : AS906 DMIT Cloud Services
Location   : Hong Kong, Kowloon ()
Country    : Hong Kong

fio Disk Speed Tests (Mixed R/W 50/50) (Partition -):
---------------------------------
Block Size | 4k            (IOPS) | 64k           (IOPS)
  ------   | ---            ----  | ----           ---- 
Read       | 57.78 MB/s   (14.4k) | 472.58 MB/s   (7.3k)
Write      | 57.87 MB/s   (14.4k) | 475.06 MB/s   (7.4k)
Total      | 115.65 MB/s  (28.9k) | 947.65 MB/s  (14.8k)
           |                      |                     
Block Size | 512k          (IOPS) | 1m            (IOPS)
  ------   | ---            ----  | ----           ---- 
Read       | 997.50 MB/s   (1.9k) | 993.91 MB/s    (970)
Write      | 1.05 GB/s     (2.0k) | 1.06 GB/s     (1.0k)
Total      | 2.04 GB/s     (3.9k) | 2.05 GB/s     (2.0k)

Geekbench 5 Benchmark Test:
---------------------------------
Test            | Value                         
                |                               
Single Core     | 679                           
Multi Core      | 702                           
Full Test       | https://browser.geekbench.com/v5/cpu/23909746

 SysBench CPU Test (Fast Mode, 1-Pass @ 5sec)
---------------------------------
 1 Thread Test (Single Core) Score:          2953 Scores
 SysBench Memory Test (Fast Mode, 1-Pass @ 5sec)
---------------------------------
 Single-thread Read Test:          34159.97 MB/s
 Single-thread Write Test:          16952.88 MB/s

🎬IP quality

IPv4	IPv6

🌐Network quality

IPv4	IPv6

📍Return route

Standard CMI. Cross-network return to Telecom and Unicom detours through Shanghai CMI.

IPv4	IPv6

NodeQuality link

3. outbound route (itdog.cn)

Domestic segment: Telecom outbound uses CN2, while Unicom and Mobile use their own backbone networks.
International segment: Telecom goes to HGC; Unicom detours via Japan NTT and then GSL to Hong Kong; Mobile uses CMI.

Unicom outbound detours, so latency and packet loss will not be ideal.

Location	China Telecom	China Unicom	China Mobile
Beijing
Shanghai
Guangzhou

4. TCP-PING (itdog.cn, 2025-11-18 00:00)

During evening peak, Telecom and Unicom both have severe jitter and packet loss.

5. Major Website & DNS Latency Tests

bash <(wget -qO- https://raw.githubusercontent.com/Cd1s/network-latency-tester/main/latency.sh)

Real connection	DNS resolution

6. Proxy speed tests (2025-11-17 23:30-24:00)

The two cross-network carriers cannot really run, and even China Mobile itself cannot achieve good speeds in every region.

*Click a node name to view the speed-test screenshot, including some popular models for comparison.

7. International speed tests - iPerf3 TCP Single-thread (10 seconds per run, 2025-11-17 22:45-23:00)

International connectivity is good as always. The Europe route has the old issue: it does not use RETN, instead going through GSL, so latency is relatively high.

I improved the test method a bit; it can now output test records.

Test records

8. Speedtest Multi-threadspeed test

curl -sL nws.sh | bash

---------------------------------- nws.sh ---------------------------------
      A simple script to bench network performance using speedtest-cli     
---------------------------------------------------------------------------
 Version            : v2025.11.07
 Global Speedtest   : wget -qO- nws.sh | bash
 Region Speedtest   : wget -qO- nws.sh | bash -s -- -r <region>
 Ping & Routing     : wget -qO- nws.sh | bash -s -- -rt <region>
---------------------------------------------------------------------------
 Speedtest.net (Region: GLOBAL)
---------------------------------------------------------------------------
 Location         Latency     Loss    DL Speed       UP Speed       Server      

 ISP: DMIT 

 Nearest          1.15 ms     N/A     725.71 Mbps    762.85 Mbps    1010 - Hong Kong 

 Bangalore, IN    72.99 ms    0.0%    944.92 Mbps    836.00 Mbps    Bharti Airtel Ltd - Bangalore 
 Chennai, IN      80.28 ms    0.0%    965.82 Mbps    863.99 Mbps    RailTel Corporation of India Ltd - Chennai 
 Mumbai, IN       100.25 ms   0.0%    907.64 Mbps    1014.04 Mbps   Melbicom - Mumbai 

 Seattle, US      167.92 ms   N/A     369.53 Mbps    509.05 Mbps    Comcast - Seattle, WA 
 Los Angeles, US  138.74 ms   0.0%    671.49 Mbps    624.52 Mbps    ReliableSite Hosting - Los Angeles, CA 
 Dallas, US       175.78 ms   0.0%    807.74 Mbps    824.78 Mbps    Hivelocity - Dallas, TX 
 Miami, US        210.81 ms   0.0%    638.05 Mbps    771.46 Mbps    Frontier - Miami, FL 
 New York, US     197.19 ms   0.0%    693.10 Mbps    851.52 Mbps    GSL Networks - New York, NY 
 Toronto, CA      191.76 ms   0.0%    693.42 Mbps    442.27 Mbps    Rogers - Toronto, ON 
 Mexico City, MX  187.56 ms   N/A     839.66 Mbps    455.37 Mbps    INFINITUM - Ciudad de México 

 London, UK       187.10 ms   0.0%    684.34 Mbps    712.92 Mbps    VeloxServ Communications - London 
 Amsterdam, NL    183.22 ms   0.0%    671.17 Mbps    720.74 Mbps    31173 Services AB - Amsterdam 
 Paris, FR        186.30 ms   0.0%    708.48 Mbps    875.04 Mbps    Scaleway - Paris 
 Frankfurt, DE    247.81 ms   1.0%    726.12 Mbps    352.37 Mbps    Clouvider Ltd - Frankfurt am Main 
 Warsaw, PL       208.93 ms   0.0%    613.00 Mbps    784.04 Mbps    Play - Warszawa 
 Bucharest, RO    217.76 ms   0.0%    895.04 Mbps    770.32 Mbps    Digi        
 Moscow, RU       121.96 ms   0.0%    952.47 Mbps    822.20 Mbps    Misaka Network, Inc. - Moscow 

 Jeddah, SA       241.33 ms   0.0%    570.68 Mbps    763.12 Mbps    Saudi Telecom Company 
 Dubai, AE        105.70 ms   N/A     736.09 Mbps    128.95 Mbps    e& UAE - Dubai 
 Istanbul, TR     286.36 ms   0.0%    537.17 Mbps    640.99 Mbps    Turkcell - Istanbul 
 Tehran, IR       209.35 ms   0.0%    717.04 Mbps    447.61 Mbps    Irancell - Tehran 
 Cairo, EG        193.20 ms   0.0%    802.45 Mbps    443.73 Mbps    Telecom Egypt - Cairo 

 Tokyo, JP        53.38 ms    0.6%    819.13 Mbps    975.46 Mbps    fdcservers.net - Tokyo 
 Shanghai, CU-CN  475.02 ms   N/A     634.74 Mbps    3.48 Mbps      China Unicom 5G - Shanghai 
 Hong Kong, CN    0.53 ms             992.18 Mbps    FAILED         Misaka Network, Inc. - Hong Kong 
 Singapore, SG    31.48 ms    0.0%    975.02 Mbps    1036.97 Mbps   ViewQwest - Singapore 
 Jakarta, ID      45.41 ms    0.0%    964.96 Mbps    965.03 Mbps    PT Solnet Indonesia - Jakarta 

 Sydney, AU       129.97 ms   0.0%    823.68 Mbps    957.62 Mbps    Aussie Broadband - Sydney 
---------------------------------------------------------------------------
 Avg DL Speed       : 761.41 Mbps
 Avg UL Speed       : 667.46 Mbps

 Total DL Data      : 29.15 GB
 Total UL Data      : 27.45 GB
 Total Data         : 56.61 GB
---------------------------------------------------------------------------
 Duration           : 14 min 27 sec
 System Time        : 17/11/2025 - 19:29:42 CST
 Total Script Runs  : 508934
---------------------------------------------------------------------------
 Result             : https://result.nws.sh/r/1763378095_3M2MWX_GLOBAL.txt
---------------------------------------------------------------------------

9. Summary

Judging from the results, I cannot recommend this model to anyone. Maybe cross-network traffic is rate-limited, maybe the purchased CMI route is not sufficiently “high-Q”, maybe the port capacity is insufficient, or maybe traffic scrubbing is involved. In short, DMIT's current Hong Kong EB route still feels like a placeholder and cannot justify its price. The HKG EB route once had a promotional model (EB.WEEv2), but there has been no follow-up since, and its value is far worse than the Pro-series promotional Victoria plan. For now, we can only wait and see.

10. About the tests

The machines tested in this article, including all machines used for iPerf3 speed tests, were purchased at my own expense. Results may be time-sensitive.
Kernel parameters of the tested machines (/etc/sysctl.conf) are as follows

fs.file-max = 6815744
net.ipv4.tcp_no_metrics_save=1
net.ipv4.tcp_ecn=0
net.ipv4.tcp_frto=0
net.ipv4.tcp_mtu_probing=0
net.ipv4.tcp_rfc1337=0
net.ipv4.tcp_sack=1
net.ipv4.tcp_fack=1
net.ipv4.tcp_window_scaling=1
net.ipv4.tcp_adv_win_scale=1
net.ipv4.tcp_moderate_rcvbuf=1
net.core.rmem_max=33554432
net.core.wmem_max=33554432
net.ipv4.tcp_rmem=4096 87380 33554432
net.ipv4.tcp_wmem=4096 16384 33554432
net.ipv4.udp_rmem_min=8192
net.ipv4.udp_wmem_min=8192
net.ipv4.ip_forward=1
net.ipv4.conf.all.route_localnet=1
net.ipv4.conf.all.forwarding=1
net.ipv4.conf.default.forwarding=1
net.core.default_qdisc=fq
net.ipv4.tcp_congestion_control=bbr
net.ipv6.conf.all.forwarding=1
net.ipv6.conf.default.forwarding=1

11. Acknowledgements

Thanks to “Impart Speed Test“(@Kyoma) for providing public-interest speed-test service.