How about racknerd? The racknerd high-performance amd platform VPS evaluation has the most authentic morning and evening data comparison!

How about racknerd ？ yesterday racknerd I have got a high-performance VPS of the newly launched AMD platform, and today I plan to run a little evaluation data to share with you; In order to be more realistic, I will post the data of the day and the night peak (the domestic backbone network at the night peak is seriously blocked, at least at present, only the CN2 GIA network can save telecom and mobile users). Let's judge for yourself!

Official website: https://www.racknerd.com

Two high-performance VPS promoting AMD platform: KVM virtual, Ryzen 9 3900X, DDR4, NVMe SSD, 1Gbps bandwidth, with one IPv4, solusvm management

Description: The webmaster is in Guangzhou, a telecom user. One test was conducted at about 7:00 p.m. on August 21, and the other test was conducted at about 8:00 a.m. on August 22. You can find evidence in the following two pictures!

At about 7:00 p.m. on August 21:

At about 8:00 a.m. on August 22:

By comparison, generally speaking: (1) The hard disk I/O is actually very strong, with 750MB/S in the evening peak and 658MB/S in the morning, which indicates that someone has already checked in, and there should be a certain load before such a result can be caused; (2) In the evening peak, the uplink bandwidth of telecom is not very good; In the daytime, the uplink bandwidth of telecommunications will fly; China Unicom and China Mobile have little influence, at least according to the measured data.

Let's also look at the comparison of PING values of multiple nodes in the country at night peak and idle hours:

7pm

8:00 am:

According to the average data of more than 140 nodes across the country, about 30 milliseconds are less in the daytime than in the evening, which is roughly equivalent to a network traversing China horizontally or vertically. Hey, the evening peak, the backbone network

The original IP address of the MC machine room can be viewed from HBO, but this is not a good thing

My VPS is 2 cores, which is actually 2 threads. We can see the performance test of single thread and double thread (equivalent to the actual physical 1 core) in the following figure. I think it is not actually measured that this is for my exclusive CPU... The data of the read and write test of memory and hard disk is also tough. DDR4 and NVMe SSDs now seem to be quite different from DDR3 and conventional SSDs. See the figure for yourself:

In the past domestic route tests, Jiangsu Telecom, China Unicom and China Mobile were tested in the late peak, among which the packet loss rate of China Telecom was quite high, and China Unicom and China Mobile were relatively careless:

The return route (evening peak test), the three networks are straightened, which is absolutely no problem!

To domestic telecommunications:

To China Unicom:

Domestic mobile:

Probe: http://173.82.97.195/p.php

Description ：

The VPS during the test did not pay attention to the renewal problem and was cancelled. The difference between the current IP and the test period is because it was newly opened later!

Finally, make a comparison of download speed measurement:

For the first day, Guangzhou Telecom actually reached 17MB/S at the highest speed. The screenshot is relatively stable:

http://173.82.97.195/1gb.bin

At 22:51 on August 22:

Maybe you need to run a Unix bench for comparison?