These speeds are far above 1Gbps speeds. A 10Mbps line will read 12.5 Mbps on Speedtest.net at top speed. A 1Gbps line will read 125 Mbps on Speedtest.net at top speed. Speedtest.net measures Mega-Bytes while network ports are in Mega-Bits. 8 Bits = 1 Byte..
That's totally false. Value are in Bits in speedtest.net.
It's even write in the middle of the speed meter image while testing.
For the rest it's ok, 100Mbps = 100/8 for get bytes = 12.5 M/bytes sec.
So Speedtest is as show speed in BITS and not BYTES and as all hosting provider ads network speed of their VPS show speed in Bits
When you said you have a lan at 1Gbps as for your lan network it's 1 Giga bits /sec (so 125 M. bytes/s)
When you said you have 8 mbps internet it's 8 M/Bits second and it's generally ATM speed so it's lower but Telco love put ATM speed for show bigger number.
So it's about 1 Mega BYTES / sec
The speeds you're logging are not accurate. Do some speedtests from servers in Chicago, LA, NY and other locations, then average those for a better idea of your real network speed.
That's true, however that's means you need check quality hosts. When test speedtest.net in USA too many speedtest.net are extremly slow.
There is a lot of many "operator" without more than 1 Gbps link and the speedtest will more look how much free bandwidth they will have on their pipe than the test of the server.
If you look at the Dallas VPS test it's seems totally good test, 22ms, not same network (seems far enough, 22ms was like NY or Atlanta maybe?)
Regarding other comment for quality of test from several time, i doubt a lot it use the same speedtest server except if he don't close his browser
You reply several time in this thread and always show a curl test. You always said it's was slow. First CURL just do 1 concurrent thread download and you speed test was in BYTES not BITS. Showing about 80Mbps from USA to France test if i'm right. Well not bad at all i think. If you do several concurrent download you will get higher speed. It's what do speedtest, start several concurrent download in same time.
For real test it's not a curl or any speedtest.net that's will display quality of the network.
Curl because it's just one download thread, speedtest because you don't known the quality of the test server.
You need test with program like iperf from other datacenter with testing several time of the day and look each time if download speed look stable. But that's beyond most user capacity.
If anyone interrested we can perform some real test and see real result.
To be totally honest, let's said speedtest return 300, 500, 800 Mbps is quite useless because it isn't going show real speed of your application. If you do GSA ser, you will never use more than some 200/300 mbps in peak and on average you are under 100 mbps.
If you look at 1 concurrent thread download on amazon, mega or such, you will NEVER reach 900 Mbps on any hosting just because of distance between the two server.
EDIT SPEED TEST :
Ok just done fast test ON internet from the Storage VPS plan (seedbox) upgrade to 2GB ram (they are good as 200W plan) :
France to UK (easterbourne) : more than 400 Mbps :
http://www.speedtest.net/my-result/3272994601
France to Germany (frankfurt) : more than 400 Mbps :
http://www.speedtest.net/my-result/3273013941 (leaseweb)
France to Luxembourg : more than 400 Mbps :
http://www.speedtest.net/my-result/3273002692
France to USA/Washington : About 370 Mbps DL / 250 UL :
http://www.speedtest.net/my-result/3273027364 (softlayer)
Quality may varies between time of the day (now it's busy it's going heavy time use of internet 7pm to 23pm) and also on destination server backbone.
So to response to those how said it's local network, number from customer said 400Mbps wasn't local network too.
NOTE on COMPETITOR SPEED TEST :
If you saw any competitor show same stats tell me since they only show neighbor test for 90% of them i saw. If i saw next door neighbor datacenter it's 900mbps 24x7 (anyway you can get sometime for near peering same result as well
Note : For do a good speedtest you need at least 2 core, look at cpu, when i done my test with chrome all cpu was use with 1 core at 3.3ghz