Can someone who keeps close tabs on their data speeds with AT&T run a speed test and see what you get?
I just read on wmexperts.com that AT&T has completed their 7.2 HSPA upgrade to the 3G network and I am getting some really high results on speed tests, but I don't really know what I would've gotten a week or two ago.
Can anyone confirm that the network is much faster?
i did this speedtest.net with my laptop connected to my fuze via usb. i had 1 bar of reception. i got a download speed of 1.3mb/sec not bad at all
Full bars 120ms 1.3 down, worse than usual.
Okay, I just got 11834 kbps at mobilespeedtest.com. But I was using the skyfire browser, so is that cheating because their server downloads info and then sends it to my phone?
Okay, I tried it with IE Mobile and got 3213, 2000, and 2991 kbps. With 1-2 bars.
im on energy pheonix rom, on long island. in H data mode.
I ran mobilespeedtest.com in opera with 3 other tabs ope (old browsing session) and got 294 Kbps, 400 Kpbs.
Then in IE Mobile: 377.
Then I clicked the "test again" link on the mobilespeedtest for a 1 MB download, in IE mobile, and it returned 2388 Kpbs, 2197 Kbps.
Then ran the 3 MB test, 2050.
So dont just trust the 100KB test, click the 1mb. Then a 3mb. If your speeds increase as you go up, latency is the problem.
There might not be many of us out there, but for those of us using unlocked and rooted mytouch on the ATT network I've found after some research a way to almost double my Edge speed. This information has been posted before in various pieces on various sites but not directly related to ATT/Edge only. Ive tested this on cyanogen 5.0.8 built will include instructions here that you can give this a shot depending on your rom. Prior to this tweak I ran three download tests, 63/38 84/41 81/41. After, I got 191/102, 133/114, 123/95.
Note that it seems that for me the speed increase takes about 10 seconds to kick in from sleep once the data connection is active.
Under the system directory, look for three files. Open each file and find a section called #GPRS class or #RIL settings. When you find the file containing this, this is the file to edit. Below I'm using system.prop. Also, when editing, do not use notepad or wordpad, I personally use EditPad Lite.
using the android sdk, and connecting your phone to pc ->
adb pull /system/build.prop build.prop
open build.prop with a unix compatible text editor
Change the following sections below:
# Default network type.
# 0 => WCDMA preferred.
ro.telephony.default_network=1
# RIL settings
ro.ril.gprsclass=32
ro.ril.hsxpa=0
ro.ril.hep=1
ro.ril.enable.dtm=1
eliminate all other lines that start with ro.ril if they exist.
then go back to your command prompt, and type
adb remount
adb push build.prop /system/build.prop
reboot phone twice after this.
The idea here is that these lines will force only searching for Edge networks, and improve the full use of multislot data/voice thoroughput on ATT Edge network. I would think technically that this would also maximize data thoroughput for those that want to save battery life from the sucking power of 3G.
I'm posting this because I'd like some feedback from ATT people that can try this. As I'm in metro Atlanta I'm not sure that these settings work for every area, and I've only tested in two spots here within a mile of each other in the three tests. I'm also at this point concerned about battery life outside of my one day testing and other areas.
I'll try this on two different G1's running on AT&T this next week and report back. Thanks
Update 8/31/10: Testing with Super D I see a difference. Was 0.07 mbps Down/ 0.08 mbps up. Now 0.18 mbps Down/ 0.16 mbps Up. Every little bit helps I guess. Thanks for posting this thread.
Which ROM are you using this on? I'll try this out on my phone at the end of the week. Also where did you find this exact same set up. I want to read more about it. I see other posts with different values but not with this same exact set up.
http://forum.xda-developers.com/showpost.php?p=5123531&postcount=46
As I mentioned I'm using Cyanogen 5.0.8. I didn't read where anyone is using this exact setup, I researched the values and put them together based on my best understanding of them.
As of now I'm seeing battery life unaffected to actually improved. Things like facebook and my MLB At Bat are majorly improved, takes much less time for the stats to populate so I'm spending less time in the apps and in turn I'm assuming that lends to battery life. I also used to drop a call in one particular spot and I'm not dropping that call so far. I've been in various parts of the city and the download speed is fairly consistent witht he 130-180k range.
Any word on this working with CM6? There only seems to be these in the RIL section:
ro.ril.hsxpa=1
ro.ril.gprsclass=10
Should the other two values be added?
SRDMizzou said:
Any word on this working with CM6? There only seems to be these in the RIL section:
ro.ril.hsxpa=1
ro.ril.gprsclass=10
Should the other two values be added?
Click to expand...
Click to collapse
As far as what should be added/removed, please more carefully read post one.
In answer to whether it works with CM6, these values directly support the gprs code of the phone so it should in theory work on any android based device.
I appreciate the time you've spent, thanks. I'm trying this asap. I have to test before and after, tho. I'll let you know. I never even considered looking into this, and what I see looks very promising. Thanks for the work.
BEFORE
Test 1: ping- i forgot to look
download- 79
upload-109
Test 2: ping-772
download- 90
upload-79
Test 3: break in between 2&3, for fun. it was asleep for a while.
ping- 767
download-51
upload-95
Test 4, another pretty long break
ping:927
download:23
upload:116
Test 5, no break
ping 16
download 44
upload 100
test 6, no break
ping 39
download 46
upload 111
AFTER
Same physical location, 2 reboots, cable unplugged as before.
Test 1: ping 551
download 33
upload 122
test 2 soon after
ping 541
download 27
upload 104
test 3 with an asleep break in between for a while
ping 428
download 47
upload 105
test 4 right after
ping 429
download 39
upload 102
Test 5 right after
ping 422
download 28
upload 116
Test 6, after a break with screen off/sleeping and a 30 second break with screen on before test
ping 430
download 26
upload 105
In conclusion, no it didn't make any benefit to me. In fact, there was a decrease in speed after applying the mods.
I used a nightly of cyanogenmod 6 from 2 days ago (right before rc3)
I will try donut soon. Eclair sucks. Note that cyanogenmod did not have the last two ro.ril lines, but it did have the telephony one. I added the necessary ones, of course.
i updated my post above.
Hi Jcarrz, thanks for the post. I think something is wrong with your location/configuration if you are getting 100% upload speed over download. It should be the reverse, but even before my config change I was getting double your download speed. Also, a ping time of <100 ms into an edge network is a bit unheard of as far as I know, what speedtest are you using? Did you delete the existing lines? the two most important lines to add are the dtm=1 lines and the grprsclass=32.
One thing about this is that not every city may support class 32 download. Also it may take some time for GPRS to establish 5 timeslots to get the full download speed. Another thing to follow up on my post, as I mentioned 10 seconds or so at least before starting the speedtest, but I've found that even 1-2 minutes after coming from sleep I'm occasionally getting 215k down in some places. But in your case there's definately something else going on....
hongkongpheoy74 said:
Hi Jcarrz, thanks for the post. I think something is wrong with your location/configuration if you are getting 100% upload speed over download. It should be the reverse, but even before my config change I was getting double your download speed. Also, a ping time of <100 ms into an edge network is a bit unheard of as far as I know, what speedtest are you using? Did you delete the existing lines? the two most important lines to add are the dtm=1 lines and the grprsclass=32.
One thing about this is that not every city may support class 32 download. Also it may take some time for GPRS to establish 5 timeslots to get the full download speed. Another thing to follow up on my post, as I mentioned 10 seconds or so at least before starting the speedtest, but I've found that even 1-2 minutes after coming from sleep I'm occasionally getting 215k down in some places. But in your case there's definately something else going on....
Click to expand...
Click to collapse
yea, that ping was pretty crazy. I live really close to los angeles, I don't think a crappy network is the problem. What do you mean by delete existing? i edited two that were there, and added two that weren't. I am using the ookla labs speedtest.net application. I have always had a higher upload than download. Everywhere. I waited for long periods of time between tests. And after waking up from sleep. The highest speed test I have seen is a 250, on htcclay's superfast 1.3.3. I'll have to take a look on the values in that rom.
Thanks for your help, prehaps you could upload the EXACT build.prop for, say, a cm6 build that I could use?
Jcarrz1
ro.ril.hsxpa=2
ro.ril.gprsclass=12
ro.telephony.default_network = 0
I seem to get the best speeds i've seen with this config. Namely the hsxpa=2, because that one, when changed to 1 or 0, dents my speeds.
jcarrz1 said:
ro.ril.hsxpa=2
ro.ril.gprsclass=12
ro.telephony.default_network = 0
I seem to get the best speeds i've seen with this config. Namely the hsxpa=2, because that one, when changed to 1 or 0, dents my speeds.
Click to expand...
Click to collapse
Thanks for reporting back. I think this shows that indeed the edge network in your location just isn't a class 32, or at least it is and is allocating more timelosts for upload than download. Here's how the classes should in theory work. Edge can alot 2-6 timeslots to transfer data. a class 10 or 12 is 5 slots data, 10 says that in theory up to 4 of those slots could be download and 2 could be upload, while 12 is up to 4 down 4 up (but only 5 max in any config). A class 32 can allocate 6 slots data, 32 also says that 5 could be download or 5 could be upload. Try setting the class at 30. This will give you a shot at 6 slots guarantee that no more than one timeslot can be set for upload (5 download 1 upload). If that doesn't work, in theory actually then moving to a class 10 setting may overall benefit you (4 download 2 upload).
hsxpa setting should technically do nothing for you on AT&T unless your phone supports HSPDA 850/1900, I set it on my magic 32B because it forces UTMS99 which is not implemented and therefore should in theory guarantee my radio never finds something to scan. I understand that turning on 2G only doesn't work in all roms, so I set that just in case.
Just so you know, the values in gprsclass can be 10,12,30,31,32,33,34.
I don´t think that ro.ril.hsxpa=0 deactivates 3G. AFAIK it just limits the phone to plain normal 3G (=UMTS) and ensures that it does not attempt to switch to 3.5G (HSDPA/HSUPA) connections.
From what I see in wikipedia, UMTS99 is nothing else than the first release of UMTS, therefore definately being supported.
Better set it to =2, just in case you travel to a country where you have 3G (and 3.5G, etc.) connectivity.
Hellek said:
I don´t think that ro.ril.hsxpa=0 deactivates 3G. AFAIK it just limits the phone to plain normal 3G (=UMTS) and ensures that it does not attempt to switch to 3.5G (HSDPA/HSUPA) connections.
From what I see in wikipedia, UMTS99 is nothing else than the first release of UMTS, therefore definately being supported.
Better set it to =2, just in case you travel to a country where you have 3G (and 3.5G, etc.) connectivity.
Click to expand...
Click to collapse
EDIT: Sorry I just looked up the specs on the phones again and I never noticed UMTS support, still in essense you still want to keep this at hsxpa=0, unless you want 3G on roaming coverage as mentioned. This article is about ATT users and more geared to only ATT bands, ATT does not offer HSDPA over the bands this phone supports so disabling it should keep it from scanning roaming networks for it. I'm removing the other ril values from the phone to keep the phone happy with the fastest edge connection it can get.
So I just tried updating the settings on Superfly TCBY Froyo 1.2. It didn't have the same exact setting lines so I just added the lines manually. All tests done on a fresh reboot, SpeedTest the only app open:
Before
Time of day: 1:10 PM
Run 1
Server: Los Angeles, CA
Latency: 829 ms
Download: 10.3 kB/s
Upload: 9.3 kB/s
Run 2
Server: Los Angeles, CA
Latency: 121 ms
Download: 14.7 kB/s
Upload: 12.1 kB/s
After (2 reboots):
Time: 1:25 PM
Run 1
Server: Los Angeles, CA
Latency: 423 ms
Download: 19.9 kB/s
Upload: 13.7 kB/s
Run 2
Server: Los Angeles, CA
Latency: 394 ms
Download: 18.3 kB/s
Upload: 14.6 kB/s
So it did improve speeds a little. How much is due to the settings is unsure. It could be related to the time of day and traffic during the time of day. I will keep the settings and see how they hold up.
I have been using the normal settings without modifications for Google Nav and it works perfectly fine. I will see how these settings hold. I'll attach my build.prop. Just remove the ".txt"
dawgpound, can you report what city you are in. I do agree that time of day and tower can play into this, this is why my before and after tests were done about 2-3 miles apart from each other. They were also done during daytime hours.
To update on my end, I'm still testing in various areas of atlanta and I'm finding that I actually connect "G" in some areas. I'm assuming UMTS99 since the phone only supports this. The download speeds were terrible, worse than edge for me though. I know this implementation is older and probably not omptimized.
Using Speedtest I'm still not getting any worse than 40k download anywhere in metro atlanta. But, one thing I've found is that in some cases I would get a upload speed greater than download, running Speedtest again even immediately after would produce much higher downlad and upload speeds and maintain these. I think this goes back to my theory that when you initiate the data transfer that it takes a while for the phone to build all the timeslots to the carrier when you enable them in some situations.
Sorry for responding so late. I'm doing the tests in LA. I went to Universal Studios the other day with the modified settings and my phone failed badly when using data. I was getting full signal on my phone with the EDGE sign but it wasn't sending or receiving data most of the day.
Unfortunately, I wasn't able to switch back and forth to see if it was just in the theme park or not. I've noticed in crowded areas, it lags pretty bad compared to the stock settings. I also was connecting to G speeds in some areas but only for a few seconds before switching back to EDGE. Right now, I'm back on stock settings to compare. I will say though, with the modified settings, download speeds increased by average 5 kB/s but upload speed was on avg 5 kB/s less than dl speeds. Stock settings, both download and upload speeds are close to each other.
Hey guys. I LOATHE posting new threads, but this problem has been plaguing me since I bought this phone, and I need to find out whether i'm being a retard, or whether my phone is a dud.
I recently just updated to the latest V21Y baseband using OTA. I have since gone back to Hellsfire ICS.
I've used the standard GPS.CONF. I've used many different kinds, I've tried Google Supl server, i've tried T-mobile. I get locks sometimes in 90 seconds, sometimes in 5 minutes. It's totally unreliable and annoys me that my old Sony X10 gets a lock in 1-2 seconds whereas this thing has caused me to get lost a million times.
I'm in Kuala Lumpur, Malaysia. I've changed the NTP server to the Malaysian equivalent.
Still no cake.
To summarize:
1) It DOES lock....but after 90 seconds - 5 minutes
2) I use GPSTEST by Mike Lockwood in order to get my lock. I've used GPS Status & Toolbox by MobiWIA - EclipSim but that seems to be worse
3) i've tried a variety of different ROMs from Weapon, to Hellsfire, Eaglesblood. Same issues
Baseband: M6600A-SCAUTNZ-2.0.9720T 1 [Mar 11 2012]
Kernel:2.6.32.59 - CM9-faux123-050 [email protected] #39 SMP Preempt Wed May 9 (Does the kernel affect GPS performance?)
CynogenMod Version: 9.0.0 - HellFireSandwich_v1.9
Thanks
Check out this post, it helped me immensely....
http://forum.xda-developers.com/showpost.php?p=29199464&postcount=4
Here is my gps.conf which works pretty well on cm7.2 with 21y bb
NTP_SERVER=xtra1.pancetta.net
XTRA_SERVER_1=http://xtra1.gpsonextra.net/xtra.bin
XTRA_SERVER_2=http://xtra2.gpsonextra.net/xtra.bin
XTRA_SERVER_3=http://xtra3.gpsonextra.net/xtra.bin
# DEBUG LEVELS: 0 - none, 1 - Error, 2 - Warning, 3 - Info
# 4 - Debug, 5 - Verbose
DEBUG_LEVEL = 3
# Intermediate position report, 1=enable, 0=disable
INTERMEDIATE_POS=1
# Accuracy threshold for intermediate positions
# less accurate positions are ignored, 0 for passing all positions
ACCURACY_THRES=1000
################################
##### AGPS server settings #####
################################
# FOR SUPL SUPPORT, set the following
SUPL_HOST=lbs.geo.t-mobile.com
SUPL_CERT=/etc/SuplRootCert
SUPL_PORT=7275
PHONE_TYPE=UMTS
# FOR C2K PDE SUPPORT, set the following
# C2K_HOST=c2k.pde.com or IP
# C2K_PORT=7275
################################
# EXTRA SETTINGS
################################
# Wiper (wifi positioning), 1=enable, 0=disable
ENABLE_WIPER=1
Although not in the same geographical location as you, I've had the same issues on all ROMs and Kernel combinations. I have found that the only combo that gets me a lock withing 1-3 sec. is the new BB on a stock rom. Give it a try!
hi! I was very convinced of the fact that the q is not able to do edge.
Now I am curious about the download/upload-speed one can expect from gprs. If I do a speed test on 2g only (phone info shows 1:gprs) I receive approximately 200 kbps down and around 100-130kbps up. Ping is around 200ms.
According to some internet research the speeds stated above are way to good for gprs only. So may be one is able to verify/or decline my guesses.
What are the possible/average edge speeds and what may one expect from real gprs speeds.
Are those speeds still correct?
http://www.elektronik-kompendium.de/sites/kom/0910141.htm
THX for help!
Sent from my awesome XT897
kecksk0enig said:
hi! I was very convinced of the fact that the q is not able to do edge.
Now I am curious about the download/upload-speed one can expect from gprs. If I do a speed test on 2g only (phone info shows 1:gprs) I receive approximately 200 kbps down and around 100-130kbps up. Ping is around 200ms.
According to some internet research the speeds stated above are way to good for gprs only. So may be one is able to verify/or decline my guesses.
What are the possible/average edge speeds and what may one expect from real gprs speeds.
Are those speeds still correct?
http://www.elektronik-kompendium.de/sites/kom/0910141.htm
THX for help!
Sent from my awesome XT897
Click to expand...
Click to collapse
According to QXDM diagnostic information I observed, the Photon Q modem is using MCS (EDGE Modulation and Coding Schemes) all the way up to MCS-9, while the RIL still reports only GPRS connection.
So the Photon Q actually does EDGE. The only issue is that the RIL is not reporting it. RIL signals GPRS technology (1) while EDGE (2) is in use.
In practice, the limiting factor therefore is just the small size of buffers defined by net.tcp.buffersize.gprs property that are in use when net.tcp.buffersize.edge should be used instead.
So I recommend to set the net.tcp.buffersize.gprs property to the net.tcp.buffersize.edge values to enjoy the full advantage of EDGE speed.
kabaldan said:
...
So I recommend to set the net.tcp.buffersize.gprs property to the net.tcp.buffersize.edge values to enjoy the full advantage of EDGE speed.
Click to expand...
Click to collapse
The for the fast feedback, will you describe how to do that? Is it a build.prop? Will it be included in ongoing cm10 mod?
Sent from my awesome XT897
kecksk0enig said:
The for the fast feedback, will you describe how to do that? Is it a build.prop? Will it be included in ongoing cm10 mod?
Sent from my awesome XT897
Click to expand...
Click to collapse
Assuming I have not misunderstood kabaldan...
You need a rooted system to apply properties. You can get current buffer values (adb shell or su in Terminal)
GPRS:
Code:
getprop net.tcp.buffersize.gprs
EDGE:
Code:
getprop net.tcp.buffersize.edge
To set them:
Code:
setprop net.tcp.buffersize.gprs $(getprop net.tcp.buffersize.edge)
or if that fails
Code:
setprop net.tcp.buffersize.gprs `getprop net.tcp.buffersize.edge`
but it will not survive a reboot. So you should get edge values with getprop and hardcode it as a comma separated sequence into /system/build.prop
e.g.
Code:
su
mount -o remount,rw /system
cp /system/build.prop /system/build.prop.backup
echo net.tcp.buffersize.gprs=$(getprop net.tcp.buffersize.edge) >> /system/build.prop
reboot
The ">>" in the above command is very important!
get/setprop is a part of toolbox which is included in stock rom as well so I'd expect it to work on rooted stock rom (I can't verify as I'm using CM10).
I could notice in *#*#4636#*#* (called "Test", if I remember correctly it was called "Testing phone" in ASA14)
that it changed sometimes to EDGE after disabling the phone (on ASA14). Only for a moment, because the
phone was being disabled. It didn't stay EDGE on reenabling the phone.
I had no connection (because of wrong APN, wasn't aware of) and so I tried different things in "Test" and noticed it.
My device is redmi note 8 pro, and I have flashed corvus, havoc, and evolution x. All three load app which depend on internet slow to load and/or even disconnecting. Especially for call of duty mobile, was the app that loads longer, reconnects and even disconnects (leading to me having to restart and login again). I'm comparing this to stock miui 12 rom, global version. I'm not sure is it related to kernel, but all roms have different kernels and all have this problem. Maybe its a related to mediatek chips? Or is there some thing i can install that improves connectivity? Hope someone can shed light on this. Thank you
Try to increase the time between Wi-Fi network scans
By default, Android scans for nearby Wi-Fi networks every 15 seconds when you have Wi-Fi enabled. This helps when you're moving around between multiple access points, but for the most part, the high interval only serves to drain your battery and slow down over-all performance. If you'd rather have your phone or tablet wait for 5 minutes before it looks for nearby Wi-Fi networks again, add the following line to the end of your Android's system file called build.prop:
Code:
wifi.supplicant_scan_interval=300
Try to increase system's TCP buffers size
That can help increase the speed of a slow Internet connection, especially when using mobile networks. You do so by adding these lines to build.prop file:
Code:
net.tcp.buffersize.default = 4096,87380,256960,4096,16384,256960
net.tcp.buffersize.wifi = 4096.87380,256960,4096,16384,256960
net.tcp.buffersize.umts = 4096,87380,256960,4096,16384,256960
net.tcp.buffersize.gprs = 4096.87380,256960,4096,16384,256960
net.tcp.buffersize.edge = 4096,87380,256960,4096,16384,256960
Also, prescribing Cloudflare’s DNS servers in some cases reduces the ping time. You again alter build.prop accordingly:
Code:
net.rmnet0.dns1 = 1.1.1.1
net.rmnet0.dns2 = 1.0.0.1
net.dns1 = 1.1.1.1
net.dns2 = 1.0.0.1
Wow a reply! Thanks a lot! Wells I'm back on my stock rom now, but basing on your recommendations, are you saying custom roms have settings that are slow/standard? I'm still new to the custom rom thing and a noob at entering codes... But yea I used call of duty mobile as my benchmark since thats my go to game. For miui stock I'd say loads the game in 30 seconds, where as custom roms loads it in 1min plus... I'm using home internet all by myself and speed is good.
Also one more point is that when I turn off the screen, then when turn it back on again (in like a few minutes) and re enter the app, on custom roms, its like its completely disconnected from the internet, it cannot load anymore unless i restart the app... but on miui it loads for around 5 seconds and its reconnected. I think I'm repeating myself here. but yea. thanks!