Following up on the previous off-peak test:[Hong Kong CMI] Quick test of YxVM VOL (throttled version) and ISIF CM

Some forum members wanted to see evening-peak results, so I retested it. Thanks to @vpsex and @Hconps for providing some equipment for testing yct005

In summary, VOL return speeds to China Mobile have, surprisingly, improved compared with the previous test, while everything else is basically the same. Overall, ISIF CM currently offers a slightly better browsing experience than VOL

Plan 🇭🇰 YxVM Volume 🇭🇰 ISIF HKG-Bs-CM
Hardware 1vCPU, 0.5GB RAM, 10 GB SSD 1vCPU, 768MB RAM, 20 GB SSD
Traffic 1TB (out) 2000GB (in+out)
Route China Mobile via Lumen, Telecom/Unicom via CMI China Mobile via Lumen, Telecom/Unicom via CMI
Network 1 x IPv4 + 1 x /64 IPv6 1 x IPv4 + 1 x /48 IPv6
Cost Monthly USD$ 3.00 Monthly EUR€ 6.00

🌐 Network Quality

🇭🇰 YxVM VOL 🇭🇰 ISIF CM

⏬ iPerf3 single-thread TCP speed test (single test, 30 seconds each, 2025-08-19 21:00-23:00)

To reduce the impact of QoS policies on domestic IDC test nodes, this test includes a control group also located in Hong Kong, using 163pp/4837/CMI routes with 1Gbps peak bandwidth

As with the previous post, during evening peak, VOL CMI return to Telecom/Unicom is still capped below 100Mbps single-thread, and Lumen->CMI return to Mobile remains lower than ISIF
9929 and CN2 home broadband should be able to pull both VOL and ISIF CM well
But judging from East China home broadband, neither YxVM VOL nor ISIF CM performs well on Mobile or Telecom

Node Download from YxVM Retransmits Average latency Download from ISIF Retransmits Average latency Download from control group Retransmits Average latency
East China Mobile 🇨🇳 home broadband @1Gbps 95.5Mbps 26510 - 143Mbps 16886 - 402Mbps 1301 -
Guangdong Mobile 🇨🇳 Yunxi Huanjing @2Gbps 320Mbps 49669 20.681ms 489Mbps 4609 22.878ms 683Mbps 9655 19.315ms
Wuhan Mobile 🇨🇳 Yunxi Huanjing @2Gbps 104Mbps 34314 49.941ms 201Mbps 944 54.085ms 546Mps 1 52.029ms
East China Telecom 🇨🇳 home broadband @1Gbps 57.0Mbps 18668 86.239ms 55.8Mbps 18396 28.737ms 815Mbps 1 31.688ms
Shenzhen Telecom 🇨🇳 Yi'an Cloud @1Gbps 78.5Mbps 31091 36.783ms 166Mbps 51225 11.157ms 871Mbps 0 11.138ms
Wuhan Telecom 🇨🇳 Yi'an Cloud @1Gbps 81.4Mbps 36470 48.888ms 119Mbps 22860 43.890ms 550Mbps 0 45.643ms
Ningbo Telecom 🇨🇳 Wuyu Cloud @500Mbps 89.7Mbps 28956 39.256ms 478Mbps 25 37.765ms 482Mbps 114 39.122ms
Ningbo Telecom 🇨🇳 Seagull Cloud @800Mbps 95.9Mbps 27894 38.551ms 524Mbps 3 35.870ms 643Mbps 0 32.805ms
Shanghai TelecomCN2 🇨🇳 Liqun @300Mbps 172Mbps 62570 30.626ms 273Mbps 4957 32.232ms 285Mbps 6600 33.092ms
Qingdao Unicom 🇨🇳 Yunxi Huanjing @2Gbps 95.4Mbps 25940 83.818ms 464Mbps 7 43.338ms 519Mbps 0 45.032ms
Beijing Unicom 9929 🇨🇳 Kuxue Cloud @1Gbps 267Mbps 3778 40.743ms 379Mbps 1 46.581ms 350Mbps 25 41.237ms

🇨🇳 East China Mobile home broadband Test 1 @1Gbps

🇭🇰 YxVM VOL 🇭🇰 ISIF CM 🇭🇰 Control group

🇨🇳 East China Mobile home broadband Test 2 @1Gbps

🇭🇰 YxVM VOL 🇭🇰 ISIF CM 🇭🇰 Control group

🇨🇳 Guangdong Mobile Yunxi Huanjing @2Gbps

🇭🇰 YxVM VOL 🇭🇰 ISIF CM 🇭🇰 Control group

🇨🇳 Wuhan Mobile Yunxi Huanjing @2Gbps

🇭🇰 YxVM VOL 🇭🇰 ISIF CM 🇭🇰 Control group

🇨🇳 East China Telecom home broadband Test 1 @1Gbps

🇭🇰 YxVM VOL 🇭🇰 ISIF CM 🇭🇰 Control group

🇨🇳 East China Telecom home broadband Test 2 @1Gbps

🇭🇰 YxVM VOL 🇭🇰 ISIF CM 🇭🇰 Control group

🇨🇳 Shenzhen Telecom Yi'an Cloud @1Gbps

🇭🇰 YxVM VOL 🇭🇰 ISIF CM 🇭🇰 Control group

🇨🇳 Wuhan Telecom Yi'an Cloud @1Gbps

🇭🇰 YxVM VOL 🇭🇰 ISIF CM 🇭🇰 Control group

🇨🇳 Ningbo Telecom Wuyu Cloud @500Mbps

🇭🇰 YxVM VOL 🇭🇰 ISIF CM 🇭🇰 Control group

🇨🇳 Ningbo Telecom Seagull Cloud @800Mbps

🇭🇰 YxVM VOL 🇭🇰 ISIF CM 🇭🇰 Control group

🇨🇳 Shanghai TelecomCN2 Liqun @300Mbps

🇭🇰 YxVM VOL 🇭🇰 ISIF CM 🇭🇰 Control group

🇨🇳 Qingdao Unicom Yunxi Huanjing @2Gbps

🇭🇰 YxVM VOL 🇭🇰 ISIF CM 🇭🇰 Control group

🇨🇳 Beijing Unicom 9929 Kuxue Cloud @1Gbps

🇭🇰 YxVM VOL 🇭🇰 ISIF CM 🇭🇰 Control group

🔗 About The Tests

Many thanks to @vpsex and @Hconps for helping with testing. Apart from that, all machines used in this article were purchased at my own expense; TCP tuning was limited to enabling BBR+FQ, with all other parameters left at defaults; results may be time-sensitive