U720 Rev A Performance Versus S720 Rev A Performance

Kane3162 · EVDO User Joined: 03 Dec 2006 Posts: 64

Forcefed,

these test results are without the Y Cable and with diffrent HA addresses.

Please do not use this results until he has a chance to get the same HA address and using the proper Y Cable.

Forcefed · Posted: Thu Dec 14, 2006 10:22 am Post subject:

OK, I keep seeing people mention "HA Addresses"...I am very PC savvy, but what does this mean? It's obviously something EVDO related, but I don't know how this is could have skewed the results?

Sorry if this is a silly question Smile

And I am not worried, the guys at EVDOinfo have been AWESOME to deal with so far, I am sure they will help me get the best solution no matter what.

Cool

Kev

xenophon · EVDO Addict Joined: 30 Aug 2005 Posts: 1973

Scott · Posted: Fri Dec 15, 2006 3:52 am Post subject:

The Novatel Merlin S720 vs Novatel Ovation U720 on Apple and Windows post has been updated back yonder< on page 2.

Semi75 · EVDO Junkie Joined: 25 Feb 2006 Posts: 230

Those pings looks better and inline with the s720 while using the cable to better at times. It looks like pings test just prove the u720 needs a decent amount of power by using the cable/dc sum or pings will be erratic and poor in compared to the pcmcia devices.

Someone correct me if I'm wrong but due to being signal challenged the first set of test on Rev 0 are erratic and can't be used as a decent baseline.

As far as the sum cable goes it will only need to take up two usb ports for the extra power correct? I am assuming modifing another adapter to use three slots would not gain the device any more power?

Kane3162 · EVDO User Joined: 03 Dec 2006 Posts: 64

well, this is tough.

If any siteadmin wants to chime in.

For BEST Latency and Speed which should i go with.

I work (Support-Billing-Corp roughly 200DB querys a sec) and connect to many systems remotely.

illburrow · EVDO Newbie Joined: 15 Dec 2006 Posts: 1

i have both the u720 and the s720. same exact billing address yet different HA's. that aside here are the speeds i've gotten over the last few days :

12/15/2006 4:25 PM GMT 1.10 Mb/s 0.10 Mb/s 313 ms Denver, CO ~ 850 mi

12/15/2006 4:23 PM GMT 0.80 Mb/s 0.10 Mb/s 302 ms Denver, CO ~ 700 mi

12/14/2006 10:53 PM GMT 1.30 Mb/s 0.10 Mb/s 277 ms Washington, DC ~ 1250 mi

12/14/2006 10:52 PM GMT 1.10 Mb/s 0.10 Mb/s 315 ms Liberty Lake, WA ~ 1600 mi

12/14/2006 8:54 AM GMT 1.30 Mb/s 0.10 Mb/s 251 ms Houston, TX ~ 150 mi

Pinging www.l.google.com [72.14.203.99] with 32 bytes of data:

Reply from 72.14.203.99: bytes=32 time=156ms TTL=235
Reply from 72.14.203.99: bytes=32 time=141ms TTL=235
Reply from 72.14.203.99: bytes=32 time=230ms TTL=235
Reply from 72.14.203.99: bytes=32 time=218ms TTL=235

Ping statistics for 72.14.203.99:
Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
Minimum = 141ms, Maximum = 230ms, Average = 186ms

ill post the s720's results tonight.

edit : forgot to add that this was using the y cable

Scott · Posted: Fri Dec 15, 2006 12:58 pm Post subject:

What all the numbers [may] mean:

HA address would appear to have a bigger effect on latency (and bandwidth) than the USB extension/DC sum cable does (unless DC summing lets you connect to Rev-A).

Your route to whatever service/host you're using and that host will be a large part of your latency too though. Compare the average latency between gateway/LA server/IL server... there is a significant difference.

Similar to antennas, the USB extension/DC sum cable will not significantly lower latency in a good Rev-A signal area, nor will it increase bandwidth if you have good signal.

Extending and power summing should help in marginal areas, or weak signal areas, be they 1xRTT, 1xEVDO Rev-0, or 1xEVDO Rev-A.

An extension-only cable (lack of two USB ports side-by-side) may help lower signal-to-noise by getting the radio away from computer's EMI/RFI signature, but if it lowers DC power too much, that may actually increase latency and reduce transfer speeds.

Semi75 · EVDO Junkie Joined: 25 Feb 2006 Posts: 230

So basically by looking at all the tests with the HA being the same as well as both devices having a good signal there is virtually little to no difference in the the two devices in terms of latency.

I was reading it as the additional power through the y cable even in a decent signal area would improve latency. After relooking at the data again you are correct and that makes sense and explains the problems w/ the Mac using the extender only. Thanks for posting,

xenophon · EVDO Addict Joined: 30 Aug 2005 Posts: 1973

Good to see that the U and S720 have same latency if using the same HA.

After you posted the updates, I did some ping/traceroute testing using HAs to each coast then pinged a server on opposite coast and can see that the HA does play a significant role to latency, moreso than bandwidth.

The bottom line is that setting the HA to the region nearest to device is best.

I wonder how many cards out there are not set to an HA in same region the user lives? We have an issue that we buy hundreds of cards for company employees and they are set to the same region, where purchased. But we ship the cards to our users across the county, so many do not have the HA set near where they are.

Scott · Posted: Fri Dec 15, 2006 2:09 pm Post subject:

I did notice something interesting. While it didn't greatly impact latency, identical HA addresses on devices with different mobile numbers are assigned different router addresses.

This could be significant for bandwidth since the router appears to always be the first TCP/IP hop. I configured a number of different HA addresses, but the device still would be assigned the same router.

Alex · Posted: Fri Dec 15, 2006 2:14 pm Post subject:

just adding a thought i have about why devices may be assigned to certain HA's by default...

it could be that Sprint is 'load balancing', in a way.

or it could be they know that the nearest HA to a user has insufficient bandwidth because the area is oversold.

if that were the case, it makes sense to me to have sprint assign an HA that is going to provide better bandwidth, than latency.

just thinking out loud... don't take this as being factual evidence presented.

xenophon · EVDO Addict Joined: 30 Aug 2005 Posts: 1973

WEB_SRFR · Posted: Mon Dec 18, 2006 7:43 am Post subject:

How do these two cards compare as far as power consumption performance is concerned?

Forcefed · Posted: Tue Dec 19, 2006 2:39 am Post subject: