Pandora Wi-Fi On 350kilobytes/s


Status
Not open for further replies.
Dead1nside said:
You don't have to have your own full blown file server to be able to properly test the intranet transfer speeds, just a file share will work fine. I'm sure MWeston knows that already though.

SAMBA (windows shares) it is often quite slower then plain FTP (depends on cpu etc etc).
FTP is about the best I can think of, simply due the simplicity of the protocol, we are trying to get an idea of the wifi performance, not about any protocol implementation :)
 
Last edited by a moderator:
One question about the info posted in the blog

"Speed can probably still be improved - Michael just found out about that yesterday, he will try to tweak more (or even combine the two mods) during the next days, until he leaves for UK to start assembling (he will do the modification of each board while assembling)."

So, one mod is the resistor. which is the other? maybe software tweak?

And, could there be a simmilar problem with bluetooth? as far as I read here, the BT module is been barely used, only some basic testing connections by ED with his mobile phone. So, how could we be sure BT will work properly? (headphones, gps etc). is there a BT manager available for the first release?

great news about the wifi, that speed is ok for me :)
 
Polossatik said:
Dead1nside said:
You don't have to have your own full blown file server to be able to properly test the intranet transfer speeds, just a file share will work fine. I'm sure MWeston knows that already though.
SAMBA (windows shares) it is often quite slower then plain FTP (depends on cpu etc etc).
FTP is about the best I can think of, simply due the simplicity of the protocol, we are trying to get an idea of the wifi performance, not about any protocol implementation :)
If he's got a linux system, he could just rsync something over, no need to faff about with any kind of file server.
And if he needs it setting up an ssh server on most linux distros usually just means installing it.
 
Last edited by a moderator:
Aninhumer said:
Polossatik said:
Dead1nside said:
You don't have to have your own full blown file server to be able to properly test the intranet transfer speeds, just a file share will work fine. I'm sure MWeston knows that already though.
SAMBA (windows shares) it is often quite slower then plain FTP (depends on cpu etc etc).
FTP is about the best I can think of, simply due the simplicity of the protocol, we are trying to get an idea of the wifi performance, not about any protocol implementation :)
If he's got a linux system, he could just rsync something over, no need to faff about with any kind of file server.
And if he needs it setting up an ssh server on most linux distros usually just means installing it.

Most people who refer to "file share" mean windows sharing/SAMBA.
You want to have your results influenced in the least possible way by protocol stuff.
If you "other" pc is a windows system then it's better to use/install on that windows box a ftp server. Samba is slow.
For linux "server" same goes imho (ok you can install it from repos if needed...not the point), it's such a simple protocol and gives most of the time a nice direct feedback in the ftp client about the speed that was seen.

Maybe the same goes for Rsync,I donno. I simply use that to back things up, never looked at the performance actually.
 
Last edited by a moderator:
Are my posts invisible or are people just ignoring them?

iperf is the easiest/best solution for doing throughput tests on a network, as that's what it's designed to do. Once it's compiled for the pandora, it'll just be a case of doing a single command on the pandora to launch the server, and then from any other PC on the network you can do some other command to connect and do the tests. There's even a GUI available that can be used on the PC (linux or windows) to make a pretty graph and to tweak all of the different settings.
 
Last edited by a moderator:
chame said:
So, one mod is the resistor. which is the other? maybe software tweak?
The blog mentions changing the board slightly to fix the trace that was changed. ED seems to be suggesting that the new board revision may go through to fix the trace, but also the resistor put on.
Although it'll probably just be the resistor, in my opinion. No point in changing boards if it's not needed, eh?
 
Last edited by a moderator:
Javacat said:
Are my posts invisible or are people just ignoring them?

iperf is the easiest/best solution for doing throughput tests on a network, as that's what it's designed to do. Once it's compiled for the pandora, it'll just be a case of doing a single command on the pandora to launch the server, and then from any other PC on the network you can do some other command to connect and do the tests. There's even a GUI available that can be used on the PC (linux or windows) to make a pretty graph and to tweak all of the different settings.

No, I'm not ignoring it, just ignore those who ignore your posts as those are most likely trolls.
My posts are also overlooked by far too often around this place. Looks like answering specific persons (Either using "Nickname:" or using the quote tag) is uncommon here which sucks as anybody is just throwing in random replies at times.
Personally I didn't post iperf yet because the team has seen the posts by that asian person who benchmarked his beagleboard with the working wifi chip. Additionaly I'd assume that the team is able to do a slight bit of research themself by typing "network benchmark" into google which gives tons of other good options.
 
Last edited by a moderator:
JayFoxRox said:
Personally I didn't post iperf yet because the team has seen the posts by that asian person who benchmarked his beagleboard with the working wifi chip. Additionaly I'd assume that the team is able to do a slight bit of research themself by typing "network benchmark" into google which gives tons of other good options.

Well, yeah, though the only one who mods those things is MichaelW, and currently he is trying to squeeze out more speed instead. His normal PC does still download a tad faster than the Pandora does, so there is room of improvement. notaz also wants to do the mod, and as soon as he did, he wants to do proper speed tests and maybe tweak the driver.

Michael is a hardware genius, but not used to do any hardware benchmarking using software, so he would need time to find out and set up everything.
Instead of doing this, he is rather working on the hardware until he hits the "won't get faster because of my internet speed"-limit.
 
Last edited:
JayFoxRox said:
No, I'm not ignoring it, just ignore those who ignore your posts as those are most likely trolls.
My posts are also overlooked by far too often around this place. Looks like answering specific persons (Either using "Nickname:" or using the quote tag) is uncommon here which sucks as anybody is just throwing in random replies at times.
Personally I didn't post iperf yet because the team has seen the posts by that asian person who benchmarked his beagleboard with the working wifi chip. Additionaly I'd assume that the team is able to do a slight bit of research themself by typing "network benchmark" into google which gives tons of other good options.
I didn't mean to make my post sound flamey/angry, but thanks :p

I'd assumed that they were wanting to proper benchmarking now, as that's what most of the other posts in this thread seem to be about, but guess they aren't ready yet from what you and ED say. I'm just eager for some decent local performance results, as that's the deciding factor for me as to whether I get a Pandora or not :p

I'll just have to wait and see what MW and notaz come up with next!
 
Last edited by a moderator:
But why can't MWeston set up a simple samba share on his PC, and then connect to it with the pandora, grab a file look at the speed? That should give a better maximum speed then trying to grab something over the internet. If he's using windows all he has to do is right click a folder and say 'share.' Under Linux, it's a little more time consuming, but the config files are available online, so that too should be easy. Macs just needs the share option enabled in the System Settings.
 
EvilDragon said:
JayFoxRox said:
Personally I didn't post iperf yet because the team has seen the posts by that asian person who benchmarked his beagleboard with the working wifi chip. Additionaly I'd assume that the team is able to do a slight bit of research themself by typing "network benchmark" into google which gives tons of other good options.

Well, yeah, though the only one who mods those things is MichaelW, and currently he is trying to squeeze out more speed instead. His normal PC does still download a tad faster than the Pandora does, so there is room of improvement. notaz also wants to do the mod, and as soon as he did, he wants to do proper speed tests and maybe tweak the driver.

Michael is a hardware genius, but not used to do any hardware benchmarking using software, so he would need time to find out and set up everything.
Instead of doing this, he is rather working on the hardware until he hits the "won't get faster because of my internet speed"-limit.

This seems like the wrong approach. Take the internet out of the equation and test transfer speeds on a local lan.
 
Last edited by a moderator:
EvilDragon said:
JayFoxRox said:
Personally I didn't post iperf yet because the team has seen the posts by that asian person who benchmarked his beagleboard with the working wifi chip. Additionaly I'd assume that the team is able to do a slight bit of research themself by typing "network benchmark" into google which gives tons of other good options.

Well, yeah, though the only one who mods those things is MichaelW, and currently he is trying to squeeze out more speed instead. His normal PC does still download a tad faster than the Pandora does, so there is room of improvement. notaz also wants to do the mod, and as soon as he did, he wants to do proper speed tests and maybe tweak the driver.

Michael is a hardware genius, but not used to do any hardware benchmarking using software, so he would need time to find out and set up everything.
Instead of doing this, he is rather working on the hardware until he hits the "won't get faster because of my internet speed"-limit.

why dont he just run basic http server on his computer, and see what speed he can get over the local network?
say if he has windows, get xerver (java based, iirc)

then just use something like
wget <ip>/fodler/somefile.ext

watch the speeds
 
Last edited by a moderator:
I just want the Pandora to be able to utilize my WiFi to ITS fullest extent.
 
Polossatik said:
Aninhumer said:
Polossatik said:
Dead1nside said:
You don't have to have your own full blown file server to be able to properly test the intranet transfer speeds, just a file share will work fine. I'm sure MWeston knows that already though.
SAMBA (windows shares) it is often quite slower then plain FTP (depends on cpu etc etc).
FTP is about the best I can think of, simply due the simplicity of the protocol, we are trying to get an idea of the wifi performance, not about any protocol implementation :)
If he's got a linux system, he could just rsync something over, no need to faff about with any kind of file server.
And if he needs it setting up an ssh server on most linux distros usually just means installing it.

Most people who refer to "file share" mean windows sharing/SAMBA.
You want to have your results influenced in the least possible way by protocol stuff.
If you "other" pc is a windows system then it's better to use/install on that windows box a ftp server. Samba is slow.
For linux "server" same goes imho (ok you can install it from repos if needed...not the point), it's such a simple protocol and gives most of the time a nice direct feedback in the ftp client about the speed that was seen.

Maybe the same goes for Rsync,I donno. I simply use that to back things up, never looked at the performance actually.

Are you guys insane?

why not use stupid plain PING?
Read teh man page and you will see you can set different block sizes ;)
Further, it is an IP-Protocol so it is even easier on the CU then FTP because it doesn't envolve any "Layer 4" aka RCP or UDP stuff.
And the acces to HD and/or SD i snot counted as well. If you FTP the data has to be read and has to be written.

However, I have no clue how reliable PING to a Windows computer is. :p

And last but not least, if you want to put a bit more effort into it, use netperf.
 
Last edited by a moderator:
Creature XL said:
Are you guys insane?

why not use stupid plain PING?
Read teh man page and you will see you can set different block sizes ;)
Further, it is an IP-Protocol so it is even easier on the CU then FTP because it doesn't envolve any "Layer 4" aka RCP or UDP stuff.
And the acces to HD and/or SD i snot counted as well. If you FTP the data has to be read and has to be written.

However, I have no clue how reliable PING to a Windows computer is. :p

And last but not least, if you want to put a bit more effort into it, use netperf.

I looked at the man page, but I don't understand. How do you measure throughput using ping? Could you give an example?
CPU or or read/write speed to the HD and SD cards would certainly not be a bottleneck for transferring files over a wireless connection.
 
Last edited by a moderator:
So what's the latest status on speeds? The last numbers I saw was the 606KB/s, which is nicely fast, but that was in the wireless-only image. The everything image, I heard a 66KB/s speed, but that was paired with the 350KB/S in the stripped image.

While the extra raw speed in the reduced image is nice, I can say that I probably won't ever be using that myself, so I'm more concerned about the 1/5 speed (has that improved, yet?) that happens in the full system. Do we have any explanation for these vast differences? People have mentioned that it migtht be peak vs. stable, but I doubt that it has to do with MWeston's outside internet speeds. Is this one likely to be a software issue, or is it possible that it's still some hardware problem that's affecting the speed under different circumstances?

I'm OK with the 600 KB/s, though more is always better. But I personally feel that 60 KB/s would be a bit of a letdown, unless it can be fixed later in software. I'd still be satisfied with my Pandora (sooner, rather than later) at those rates, but I would greatly prefer otherwise.
 
Something like netcat should suffice. Very easy to use and minimum overhead. Also it's included in many linux flavors (man nc).
 
benji_stein said:
So what's the latest status on speeds? The last numbers I saw was the 606KB/s, which is nicely fast, but that was in the wireless-only image. The everything image, I heard a 66KB/s speed, but that was paired with the 350KB/S in the stripped image.

While the extra raw speed in the reduced image is nice, I can say that I probably won't ever be using that myself, so I'm more concerned about the 1/5 speed (has that improved, yet?) that happens in the full system. Do we have any explanation for these vast differences? People have mentioned that it migtht be peak vs. stable, but I doubt that it has to do with MWeston's outside internet speeds. Is this one likely to be a software issue, or is it possible that it's still some hardware problem that's affecting the speed under different circumstances?

I'm OK with the 600 KB/s, though more is always better. But I personally feel that 60 KB/s would be a bit of a letdown, unless it can be fixed later in software. I'd still be satisfied with my Pandora (sooner, rather than later) at those rates, but I would greatly prefer otherwise.

I wouldn't read too much into these numbers. MWeston is trying this against websites on the internet. These values likely do not represent the max speeds that the wifi chip can handle. That's why people are asking for LAN testing to be done.
 
Last edited by a moderator:
Status
Not open for further replies.
Back
Top