Sprint/Verizon EVDO and Cingular HSDPA compared

xenophon · EVDO Addict Joined: 30 Aug 2005 Posts: 2003

What's the case in other parts of the city? Last time I was in NYC I got over 1Mbps in lower Manhattan and Brooklyn Heights on Sprint. Could be your laptop if having slowness with both in any area. My brother in B Heights has a Sprint card and has mentioned good performance, including on a train to Philly.

Your situation is the exception, not the rule.

xenophon · EVDO Addict Joined: 30 Aug 2005 Posts: 2003

Here's the latest update. Post #1 was updated with the following... See post 1 for methodology.

Performance measurements are in bits, not bytes.
1000Kbits/sec = ~125KBytes/sec
600Kbits/sec = ~75KBytes/sec

As of April 5, 2006

Sprint
Download Performance
Entries: 142
Average: 609Kbps
Median: 476Kbps
Peak: 1858Kbps
>1Mbps: 16% of tests
>600Kbps: 37%
<400Kbps: 37%
Std Deviation: 413Kbps
Upld Avg: 83Kbps

Verizon
Download Performance
Entries: 109
Average: 554Kbps
Median: 509Kbps
Peak: 1280Kbps
>1Mbps: 7% of tests
>600Kbps: 33%
<400Kbps: 25%
Std Deviation: 254Kbps
Upld Avg: 98Kbps

xenophon · EVDO Addict Joined: 30 Aug 2005 Posts: 2003

Sprint turned off compression proxy servers about a week ago. The dslreports archive shows 2 weeks of data, so I did two compiles of Sprint results... one with 2 weeks of data (1 week includes compression proxies) and another with the last week (no compression proxies) to see how turning off compression proxies may have made a difference.

Also have updated Verizon results. Verizon's compression seems to help upload but perhaps not download rates.

Both Sprint and Verizon have a high % of results (26%-33%) below the marketed range of 400-700Kbps. If I were them, I'd check out those zipcodes.

See post #1 in this thread for methodology of how info is compiled.

__________________________________

Performance measurements are in bits, not bytes.
1000Kbits/sec = ~125KBytes/sec
600Kbits/sec = ~75KBytes/sec

As of April 21, 2006

Sprint - 2 weeks of results
Download Performance
Entries: 146
Average: 603Kbps
Median: 480Kbps
Peak: 1755Kbps
>1Mbps: 15% of tests
>600Kbps: 37%
<400Kbps: 33%
Std Deviation: 390Kbps
Upld Avg: 78Kbps

Sprint - 1 week results after compression proxies turned off
Download Performance
Entries: 71
Average: 661Kbps
Median: 512Kbps
Peak: 1755Kbps
>1Mbps: 21% of tests
>600Kbps: 41%
<400Kbps: 26%
Std Deviation: 433Kbps
Upld Avg: 81Kbps

Verizon- 2 weeks of results
Download Performance
Entries: 125
Average: 587Kbps
Median: 529Kbps
Peak: 1760Kbps
>1Mbps: 11% of tests
>600Kbps: 41%
<400Kbps: 27%
Std Deviation: 320Kbps
Upld Avg: 93Kbps

____________________________________________________

So it looks like Sprint turning off compression proxies may have improved things. It will be interesting to see results for another 2 weeks. Verizon seems to be improving averages and they are both pretty comparable. Both need to improve the % of users getting <400Kbps me thinks, though perhaps some are doing bluetooth connections, which typically cannot go over 300Kbps. Maybe there are alot of fringe area users not near a tower. Could be that there are just some slow/loaded towers out there.

None of this is hard data since none of this is controlled testing but the patterns of months of real-world results probably give us usable info.

Results taken from..
http://www.dslreports.com/archive/myvzw.com (Verizon)
http://www.dslreports.com/archive/spcsdns.net (Sprint)

____________________________________________________

Here is another test method for EVDO phones that can't do Java based tests as done above. It is mostly phones but laptop cards can test too. These results include slower 1xRTT results too so the averages are lower. The site does not test latency correctly. Big problem with this site is that the Opera moble browser does not clear cache so Opera results are too high.

http://www.dslreports.com/mspeed?domains=1

Domain Avgkbps Samples AvgLatency
spcsdns.net 369.9 4801 0.9
myvzw.com 359.3 5532 0.7
vodafone.com.au 191.2 7 1.2
mycingular.net 105.6 5111 1.6
tmodns.net 97.4 5876 1.7

xenophon · EVDO Addict Joined: 30 Aug 2005 Posts: 2003

See post #1 for methodology. I removed the tests I've done in Vegas since it would offset quite a bit as I've done many tests.

__________________________________

Performance measurements are in bits, not bytes.
1000Kbits/sec = ~125KBytes/sec
600Kbits/sec = ~75KBytes/sec

As of May 4, 2006

Sprint
Download Performance
Entries: 130
Average: 627Kbps
Median: 574Kbps
Peak: 1755Kbps (no compression)
>1Mbps: 14% of tests
>600Kbps: 47%
<400Kbps: 26%
Std Deviation: 344Kbps
Upld Avg: 81Kbps

Verizon
Download Performance
Entries: 114
Average: 545Kbps
Median: 439Kbps
Peak: 1563Kbps (w/compression)
>1Mbps: 10% of tests
>600Kbps: 37%
<400Kbps: 42%
Std Deviation: 314Kbps
Upld Avg: 95Kbps (higher with compression results)

Looks like Sprint turning off compression has improved things. Verizon's compression helps upload but doesn't appear to help download compared to Sprint with no compression.

Sprint % <400Kbps shows improvement two reports in a row since turning off compression but is higher than earlier in the year. Verizon % <400Kbps is rather high for this report. Need a couple more results compiled to see if it stays that high over next month or so.

____________________________________________________

Results taken from..
http://www.dslreports.com/archive/myvzw.com (Verizon)
http://www.dslreports.com/archive/spcsdns.net (Sprint)

____________________________________________________

Here is another test method for EVDO phones that can't do Java based tests as done above. It is mostly phones but laptop cards can test too. These results include slower 1xRTT results too so the averages are lower than methodology above. The site does not test latency correctly. Big problem with this site is that the Opera moble browser does not clear cache so Opera results are too high.

http://www.dslreports.com/mspeed?domains=1

Domain Avgkbps Samples AvgLatency
spcsdns.net 409.3 4558 0.8
myvzw.com 355.1 4973 0.8
tmodns.net 101.3 6376 1.7
mycingular.net 100.8 5317 1.6

Michael · Posted: Thu May 04, 2006 12:25 pm Post subject:

So, according to this thread, the speeds were below par, and you decided not to count them to help the benchmarks?

That seems more like a marketing move to make things "look better".
_________________
EVDO :: EVDO News :: EVDO Antennas :: Buy Verizon :: Buy Sprint :: EVDO Amplifier

xenophon · EVDO Addict Joined: 30 Aug 2005 Posts: 2003

I left them out because I did so many tests in Vegas. It would have been to Sprint's advantage if I left them in because it would have raised the avg 15Kbps and lowered Verizon's avg by 5Kbps. Funny that you are making such an accusation as I intentionally left them out to prevent getting flack for raising Sprint's and yet I get flack anyway for something misunderstood.

Look deeper in the thread and you'll see that Sprint's performance picked up later.

Here are the results for Vegas if you'd like to factor those in. It actually would make Sprint look better and Verizon worse. Still want me to include them to make Sprint look even better or is that marketing?
http://www.dslreports.com/archive?zip=89119&dnsdom=&start=Search

xenophon · EVDO Addict Joined: 30 Aug 2005 Posts: 2003

Performance measurements are in bits, not bytes.
1000Kbits/sec = ~125KBytes/sec
600Kbits/sec = ~75KBytes/sec

As of June 12, 2006

Sprint
Download Performance
Entries: 136
Average: 673Kbps
Median: 567Kbps
Peak: 1931Kbps
>1Mbps: 23% of tests
>600Kbps: 46%
<400Kbps: 26%
Std Deviation: 383Kbps
Upld Avg: 79Kbps

Verizon
Download Performance
Entries: 112
Average: 570Kbps
Median: 430Kbps
Peak: 1585Kbps
>1Mbps: 17% of tests
>600Kbps: 39%
<400Kbps: 37%
Std Deviation: 322Kbps
Upld Avg: 93Kbps

Compiled from...
http://www.dslreports.com/archive/spcsdns.net
http://www.dslreports.com/archive/myvzw.com

See post #1 for methodology.

krum · EVDO Newbie Joined: 15 Jun 2006 Posts: 1

Michael · Posted: Thu Jun 15, 2006 7:45 pm Post subject:

I am confused by this:

xenophon · EVDO Addict Joined: 30 Aug 2005 Posts: 2003

Michael, it's based on the Census Bureau 'metro' definitions, not the carriers' definitions. The carriers claim to be in many more 'cities' but they would count, say Joliet and Aurora as a separate city as Chicago. So I took the metro areas as defined by the Census Bureau, which is a less number than what the carriers claim and compared them to the EVDO maps.

http://www.evdoforums.com/viewtopic.php?t=1841

If you look at the list above, Sprint does have EVDO in 150 distinct MSA markets, not the 220+ 'cities' they claim. If you look at the EVDO maps, it's not just one tower but a good chunk of the market for most of them, especially those over 100K. But granted, maybe only 50-75% of the markets are covered currently with near 100% expected by end of the year.

Verizon is indeed in far fewer MSA markets but may have some metro areas more broadly covered. Verizon is especially lacking in markets between 300K-1M compared to Sprint, but apparently they have begun rolling out to a few more smaller markets (ie, Tucson). Verizon is still missing EVDO in a few markets over 1M while Sprint is in all markets over 1M.

As I said in the first post, Sprint is in more markets but Verizon claims 150 million coverage while Sprint claims 152+ million coverage. Looking at the maps and where density is at, Sprint may be quite a bit more conservative than Verizon.

Everyone needs to look at the maps to confirm if the carrier has coverage where they plan to use EVDO.

xenophon · EVDO Addict Joined: 30 Aug 2005 Posts: 2003

xenophon · EVDO Addict Joined: 30 Aug 2005 Posts: 2003

Post #1 has been updated...

- Performance results
- TOS updates
- Roaming info updates
- Better clarity on some definitions

As of July 24, 2006

Sprint
Download Performance
Entries: 120
Average: 605Kbps
Median: 568Kbps
Peak: 1503Kbps
>1Mbps: 14% of tests
>600Kbps: 48%
<400Kbps: 25%
Std Deviation: 298Kbps
Upld Avg: 75Kbps

Verizon
Download Performance
Entries: 102
Average: 519Kbps
Median: 472Kbps
Peak: 1599Kbps
>1Mbps: 6% of tests
>600Kbps: 29%
<400Kbps: 35%
Std Deviation: 283Kbps
Upld Avg: 108Kbps

xenophon · EVDO Addict Joined: 30 Aug 2005 Posts: 2003

As of Aug 16, 2006

Sprint
Download Performance
Entries: 108
Average: 617Kbps
Median: 552Kbps
Peak: 1533Kbps (Rev A tests omitted)
>1Mbps: 12% of tests
>600Kbps: 41%
<400Kbps: 18%
Std Deviation: 304Kbps
Upld Avg: 80Kbps

Verizon
Download Performance
Entries: 110
Average: 549Kbps
Median: 527Kbps
Peak: 1432Kbps (Rev A tests omitted)
>1Mbps: 6% of tests
>600Kbps: 39%
<400Kbps: 33%
Std Deviation: 273Kbps
Upld Avg: 98Kbps

Looks like Sprint has been improving on % <400Kbps last couple of months, now less than 20%. Verizon has 1/3 of results below marketed avg range of 400Kbps+, not great. May explain why they are knocking off heavy users. Wonder if this is due to extending EVDO rollout on VZW 800mhz towers.

With Rev A rollout coming (possibly some towers already in progress), it should help improve Rev 0 devices as the Rev A network expands. Sprint plans 40m pops on Rev A by end of this year. Verizon hasn't been specific but has mentioned rollout should begin later this year.

xenophon · EVDO Addict Joined: 30 Aug 2005 Posts: 2003

Sprint recently released RevA cards and averages seemed to improve even on Rev0 network. Verizon rolled out to more 800mhz markets and averages appeared to drop, especially with a whopping 40% getting below 400Kbps. That's the highest we've seen to date. Sprint has only 24% getting below 400Kbps.

These results can't be taken literally but the consistent trend over many months carries more weight.

As of Sept 29...

Sprint
Download Performance
Entries: 147
Average: 622Kbps
Median: 557Kbps
Peak: 1553Kbps (Rev A tests omitted)
>1Mbps: 17% of tests
>600Kbps: 43%
<400Kbps: 24%
Std Deviation: 314Kbps
Upld Avg: 72Kbps

Verizon
Download Performance
Entries: 126
Average: 534Kbps
Median: 467Kbps
Peak: 1682Kbps (Rev A tests omitted)
>1Mbps: 9% of tests
>600Kbps: 37%
<400Kbps: 40%
Std Deviation: 312Kbps
Upld Avg: 92Kbps

Results from...
http://www.dslreports.com/archive/myvzw.com
http://www.dslreports.com/archive/spcsdns.net