Gigabit internet service and usenetexplorer

[alex]

Does changing the socket buffer size influence the download speed? Say if it to increase it further to 262144 vs 8192 vs default.

Better to restart the program after changing the setting since as long as the connection is alive the value is preserved.

In the code I see by default it is unchecked (it was checked in my settings though).

And running the version above vs the regular version still doesn't make a difference?

[gibbyscott]

The file buffer size is already 500KB, so basically with 700KB chunks it is only couple of writes.

Maybe to try to increase socket buffer size then, they are above the network threads priority option (edit menu->properties->tasks, set socket buffers).

First (if checked) try to uncheck both read write socket buffers to use the default size.

Then check again and them both e.g. to 65536 and check if there will be any change.

Did you run this version above? - 55MB/s, the timer may play a role.

I changed the socket sizes and I am using the above version - download seems to be fluctuating from anywhere from 51MB/s to as high as 64 MB/s. I had to create an exclusion entry in windows defender to exclude the whole usenetexplorer directory.

I just tried a different .nzb import file and the download is now looking like 73MB/s on that one. The specifics to the file are....

6.15k Lines, 768 KB, size, Single Parts/file - the combined parts make a 1.54m Lines, 1.8GB size, 38 of 46 Parts/file

[alex]

Try to check those:

Regular version:

Read/Write Socket buffer size default (unchecked), 8192, 262144, restarts in between.

Then check the same with the version above.

Are you using SSL?

[gibbyscott]

Try to check those:

Regular version:

Read/Write Socket buffer size default (unchecked), 8192, 262144, restarts in between.

Then check the same with the version above.

Are you using SSL?

Yes - using SSL.

The fastest seems to be using the version above with the default (unchecked) - getting about 70MB/s with that. Using 8192 on both versions dragged the download down to about 18MB/s

[alex]

The default setting is right then.

With the regular version (and default buffers) do you notice some speed fluctuation comparing to the version above which gives more stable speed?

So many questions

[gibbyscott]

The default setting is right then.

With the regular version do you notice some speed fluctuation?

So many questions

Yes

I hope there is no cap on my connection... I must have downloaded 100 gigs just tonight in testing! hehe

[alex]

Ok so it makes sense then to reduce the timer interval (actually it was the same from many years ago) and maybe add it to properties.

With smaller part size the difference would be more noticeable then.

You can also try to increase the number of tasks above 20 with the version above, but unlikely there will be a difference.

If CPU is not maxed, the timer frequency is adequate and the disk writing is efficient (since done in 500K chunks), there is not much space left to make it faster.

If you can try to compare when there is larger part size, like 1MB per part with the version above, if the speed stays the same or it is faster.

The other program you mentioned also running many threads corresponding to the number of connections? If to exclude it from scan its speed remains the same or increases? I wonder why it showed higher speed initially.

[gibbyscott]

Ok so it makes sense then to reduce the timer interval (actually it was the same from many years ago) and maybe add it to properties.

With smaller part size the difference would be more noticeable then.

You can also try to increase the number of tasks above 20 with the version above, but unlikely there will be a difference.

If CPU is not maxed, the timer frequency is adequate and the disk writing is efficient (since done in 500K chunks), there is not much space left to make it faster.

If you can try to compare when there is larger part size, like 1MB per part with the version above, if the speed stays the same or it is faster.

The other program you mentioned also running many threads corresponding to the number of connections? If to exclude it from scan its speed remains the same or increases? I wonder why it showed higher speed initially.

Yes I tried the max task increase above 20 and no difference like you expected.

CPU only hits to about 30% so no max. I do notice that larger part size do seem to download faster than smaller part sizes.

NZB get (other program) doesn't have many options to change - it only has an option for number of connections to server - which I set to 20 as that is the max my server will allow.

At least the speeds are better than before with your changes.

[gibbyscott]

Ok so it makes sense then to reduce the timer interval (actually it was the same from many years ago) and maybe add it to properties.

With smaller part size the difference would be more noticeable then.

You can also try to increase the number of tasks above 20 with the version above, but unlikely there will be a difference.

If CPU is not maxed, the timer frequency is adequate and the disk writing is efficient (since done in 500K chunks), there is not much space left to make it faster.

If you can try to compare when there is larger part size, like 1MB per part with the version above, if the speed stays the same or it is faster.

The other program you mentioned also running many threads corresponding to the number of connections? If to exclude it from scan its speed remains the same or increases? I wonder why it showed higher speed initially.

Will you be updating UE with these changes or should I just continue to use the version above? Thanks