dslreports logo
site
 
    All Forums Hot Topics Gallery
spc

spacer




how-to block ads


Search Topic:
uniqs
5223
share rss forum feed


xymox1
Premium
join:2008-05-20
Phoenix, AZ

1 edit

Upgrade from dpc3000 to dpc3010 and slower upstream speeds

[*** UPDATE ***] the problem I describe below turned out to be old firmware, so read it knowing the problem is solved. I have however gotten into doing performance mods and testing of all the top modems later in this thread.
_________________________________________________________
Original posting below.
________________________________________________________

Strange...

For over a year every time i do a speedtest.net test I get 63-65Mb/s down and 5.5 Mb/s up..

I wanted to be ready for the future and decided to get a 8x4 modem rather then my 4x4.

The weirdest thing appeared to happen.

My upload ( upstream ) speed dropped. Now I barely get 3-5 to 4. My levels look fine.

Could this be a bad modem ?

Ive ordered up a replacement and I also ordered up a SB6180.

Very strange that dpc3100 is not as fast upstream as a dpc3000. It must be the modem is defective.

The upstream is the hard part as the modem acts as a transmitter. So I could see that my trouble might be a defective modem



xymox1
Premium
join:2008-05-20
Phoenix, AZ

1 edit

Well well well...

So I changed back to my DPC3000 and I am right back to my upstream rate of 5.5M/s using all my normal sites in speedtest.net.

I then changed to the DCP3010 again to make sure and sure enough I could only get a max of 4.3Mb/s...

I tested using 3 different servers at speedtest, I tested each twice. ALL the reults were the same.

The DPC3010 topped out at 4.3
The DPC3000 was always 5.5 which is the cap

I also got slightly better and more consistent download speeds from the DPC3000.

OK so maybe I got a bad DPC3010. I get a new one monday. If I find out that the new dpc3010 is again slower then the dpc3000 then that will be a interesting find and indicate that the DPC3000 is the better device over the DPC3010.

The transmit / rcv levels for both modems showed virtually the exact same numbers.

Anyone have any explanation ?

Im also going to get a SB6180 later in the week and will test that as well.

For reference my levels are +3 down and 42db up power. The channels dont vary much in levels.

Every single time with the DPC3000



xymox1
Premium
join:2008-05-20
Phoenix, AZ

Hmmm......

Further playing. I cant be sure of course until I get my new dpc3010 on monday, but I am starting to suspect that maybe the DPC3000 might be better, at least in my cox cable area using the current Ultimate tier. This might change once they roll out 100Mb tho.

The dpc3010 is a completly different device based on a whole different Broadcom chip set. The ethernet controller ship is built in for example.

Every single chip maker is different in the 3010.

The dpc3000 uses a Texas Instruments chip vs the 3010 which uses a Broadcom.

Even the radio chips are different manufacturers.

So these are very different devices.

The upstream rate I think is cap by the modem itself, which means the chip in the unit has to do some form of rate limiting. This rate limiting might be different between Texas Instruments and Broadcom. In fact the whole way it handles the data stream might be different.

So while the upstream channels are capable of way faster then 5.5Mb/s the cap rate limiting methods might produce different results at these low speeds with different chips.

When I get my new modem, if I verify it does the same 4Mb/s vs 5.5mb/s on the dpc3000 that would be very interesting.

It will also be interesting to see how the SB6180 does on this.



xymox1
Premium
join:2008-05-20
Phoenix, AZ

HMmmmmmmm....

Turns out that Texas Instruments holds a key patent on ACK filtering which has a big effect when a upstream cap is doing rate limiting. So possibly the TI based chips produce better results when rate limiting a upstream channel ?

»www.google.com/patents/US2004010···&f=false

Background »www.ti.com/lit/wp/spay014/spay014.pdf



Anonguy

@cox.net

Are you having Cox upgrade the firmware at activation?



xymox1
Premium
join:2008-05-20
Phoenix, AZ
reply to xymox1

Hmmm.... No i figured the system would just see it was old and upgrade it. I dont have it hooked up now so i can look at firmware version.

Excellent point tho. When I activate the new one monday i will make sure this is done.


lilstone87

join:2009-04-09
Portsmouth, VA
kudos:3
Reviews:
·Cox HSI

said by xymox1:

Hmmm.... No i figured the system would just see it was old and upgrade it. I dont have it hooked up now so i can look at firmware version.

Excellent point tho. When I activate the new one monday i will make sure this is done.

Well it's possible the rep you will talk to when having the modem provisioned, won't have access to upgrade the firmware to the latest one cox is using. However if the firmware on the modem, doesn't match what cox is using for that modem in your local cox area. Between 2-4am most days cox runs a modem check, which will push a firmware update to modem's running firmwares, that are not current with the system.


xymox1
Premium
join:2008-05-20
Phoenix, AZ
reply to xymox1

I have done more baseline testing using some different methods so I can better test performance when I get the new modems.

I have a server at a hosting company and it has a high bandwidth connection. Last night I uploaded a 20GB file. This allows me to test a really long transfer and get a better idea of transfer rates. Im now measuring the speeds using my PC based router which runs Vyatta based on Linux. This allows me to use all the linux tools so I can measure the data rates very precisely live and with averages.

The upload rate averaged for the entire download was 5.4Mb/s which agreed well with the speedtest.net testing and makes sense as its the cap rate on my Ultimate tier of service.

It would be interesting to try transfers with different packet sizes as smaller packets require more processing overhead and might be different with different modems and chip sets.

Im also going to test download speeds, but I think the capping of upload speeds seems to be more important.

Im also going to look at latency and jitter. I have multiping and I set it up doing 0.1 ping interval, 10 pings a second, and then chart the results. I ping the cox gateway for my connection. I do this with no transfer going on and also with a full transfer of data that maxes out my upstream cap.

So im going to do all these tests for the new DPC3010 I am going to get, I already have all these tests with the DPC3000 and when I get my SB6180 later in the week I will do all tests on that as well.

Im also going to open them all up and look at chip sets. I know what the DPC3010 and 3000 use already. i dont know about the Motorola yet.



xymox1
Premium
join:2008-05-20
Phoenix, AZ
reply to xymox1

quote:
Well it's possible the rep you will talk to when having the modem provisioned, won't have access to upgrade the firmware to the latest one cox is using. However if the firmware on the modem, doesn't match what cox is using for that modem in your local cox area. Between 2-4am most days cox runs a modem check, which will push a firmware update to modem's running firmwares, that are not current with the system.
Interesting. I will check on this. How do I find out what current firmware versions are for cox on my modem ?


xymox1
Premium
join:2008-05-20
Phoenix, AZ

I did LOTS of testing over the weekend on the DPC3000. It does a rock solid 5.5 Mb/s up on all sorts of tests.. Over and over again. Never anything short of 5.5Mb/s

I just switched back to the orginal DPC3010 to revisit it and do some tests..

EVERY TIME I test I get 4.3-4.5. NEVER anything over 4.5. I tested using very large file transfers and speedtest.net and everything was in agreement. The DPC3000 was clearly faster upload.

Upload is very important. Its not so much overall speed but latency and jitter that count. However the speed difference is showing something is wrong.

Im running multiping using 10 pings a second to the cox gateway now. Its appearing that there might be a tiny decrease and maybe less jitter using the DPC3010. I will have to chart overnite and then compare with my charts for the DPC3000. I will post charts here in the next few days.

I get my new dpc3010 tomorrow. I wanted to get a good baseline set of measurements on this one first.

As far as firmware revisions.. I am also going to see if it updates firmware overnite.

Bootloader Revision: 2.3.0_R1
Current Software Revision: dpc3010-v302r12901-091119a
Firmware Name: dpc3010-v302r12901-091119a.bin
Firmware Build Time: Nov 19 2009 22:12:51

My signals

Channel 1: 7.8 dBmv 41.1 dBmv
Channel 2: 7.6 dBmv 41.0 dBmv
Channel 3: 7.5 dBmv 40.0 dBmv
Channel 4: 8.1 dBmv 39.9 dBmv
Channel 5: 8.0 dBmv 39.7 dBmv
Channel 6: 7.2 dBmv 38.6 dBmv
Channel 7: 0.0 dBmv 0.0 dBmv
Channel 8: 0.0 dBmv 0.0 dBmv

Channel 1: 42.0 dBmv
Channel 2: 42.0 dBmv
Channel 3: 42.0 dBmv
Channel 4: 0.0 dBmv


lilstone87

join:2009-04-09
Portsmouth, VA
kudos:3
Reviews:
·Cox HSI

Well I can tell you, you have a old firmware, and should update over night. As you see that firmware is a 2009 build, and I am pretty sure cox has a 2011/2012 firmware that should be pushed to your modem over night. I suggest doing more testing after the firmware is updated, and if it doesn't update overnight. I am sure you can PM CoxTech1 See Profile or odog See Profile with the modem's mac address, and they will get you on the latest firmware cox is using for that modem.



xymox1
Premium
join:2008-05-20
Phoenix, AZ
reply to xymox1

Yes, I saw most people here in phx has 2011 firmware..

VERY interesting tho. I am now very sure im getting 3.8 Mb/s up. Ive tested using a number of different methods and they are coming up 3.8-4.1,, The same tests constantally returned 5.5 with the DPC3000..

However i can also show clearly a 1ms drop in latency. About the same jitter. However 1ms is alot in this case. DPC3000 = avg 7.5 DPC3010 = avg 6.5..

But 4mb/s VS 5.5Mb/s is alot.. Im sure broadcom and texas instruments have different stratigies in managing cap rates. One might have lower latency but lower speeds and one might have higher latency and higher speeds..

But YES, I will await a modem firmware upgrade.. I will PM a cox rep to be sure..



xymox1
Premium
join:2008-05-20
Phoenix, AZ
reply to xymox1

Yes I have had some very good responses to issues long ago by posting here and working with CoxTech1. I have PM'ed him and DPC3010 awaits a upgrade..


lilstone87

join:2009-04-09
Portsmouth, VA
kudos:3
Reviews:
·Cox HSI

said by xymox1:

Yes I have had some very good responses to issues long ago by posting here and working with CoxTech1. I have PM'ed him and DPC3010 awaits a upgrade..

I really don't think another Cisco DPC3010 is gonna matter. But the modem running a old firmware, can cause a negative effect like the problem your seeing. So the key here will be getting that modem on the latest firmware, then doing some testing again. between 2-4am most nights cox runs a modem check, and updates the firmware on modem's not running correct firmware. So as long as you leave the modem online over night, in the morning it should be updated.

I will say I own a Cisco DPC3010 myself, and hated the very vanilla modem status page, and no access to modem log for troubleshooting. So I currently use a Motorola SB6180, as the modem side of things show a lot more info. My first D3 modem was a Motorola SB6120(still own), and I have also used Zoom 5341/5341J, Cisco DPQ3212(phone/internet modem), and Cisco DPC3010. So I have used a lot of D3 modem's, and Motorola is the one I like as I have more access to modem info when there is a problem. As the Cisco modem's are locked down once online, and very restricted with info.


odog
Cable Centric Vendor Biased
Premium,VIP
join:2001-08-05
Atlanta, GA
kudos:11
Reviews:
·Comcast
reply to xymox1

First let the firmware upgrade on the 3010, it is running very old "out of box" code and definitely needs it. Second don't use a 20GB file to test upload, that's just kind of pointless. A 100MB file will work fine and show the small powerboost peak and normal sustained rate plateau.

TI and Broadcom all have their own version of the TCP ack suppression tech. Broadcom's is called Propane, TI has TurboDox. In the vast majority of situations, they have minimal impact at best. They are often disabled, and or require both the modem and US receiver use the same tech.



CoxVegas

join:2011-07-25
Las Vegas, NV
kudos:10
reply to xymox1

I got with xymox1 and upgraded the firmware on the modem for him.



xymox1
Premium
join:2008-05-20
Phoenix, AZ
reply to xymox1

CoxVegas took care of my firmware update very quickly. I have always been impressed with the speed and competence showed by the Cox reps who man these pages.

I need to do full tests but the firmware upgrade to my orginal DPC3010 seems to have fixed the issue. So all my complaining might have been for naught. Well, finding the slightly lower latency thru the DPC3010 is important and finding virtually no difference in performance using DOCSIS 6x3 vs 4x3 with my current speeds in phoenix is important. The modems are almost identical in the 65M/s ultimate tier.



( in case you wonder, the San Jose - Smug Mug server is the fastest and most repeatable test from my location in phx )

I will do more testing once my second one arrives later today. Im still going to use that one as it came right from someone who got me a recent production unit manufactured not long ago.

I will also be getting the SB6180 and will do testing with that one as well.

Im going to really test these to determine if there is any difference in real performance. Im also going to look them over from a engineering standpoint as my real life job is a electrical engineer.

One of the things thats a issue for both the DPC3000 and 3010 is heat. While the build quality is quite good bot the processor chip and the radio get overly hot. Using a good temp probe and measuring the chip on both they both sit right at the very top edge of thier thermal range. This is bad. I modded my 3000 with a small fan long ago. The 3010 seems to radiate even more thermal energy but I have not tested yet. It will be interesting to see the Motorola design and see how it handles the intense processing requirements thermally.

I wonder what change in firmware caused the upload speed to be better ? Too bad there are no public version notes..


lilstone87

join:2009-04-09
Portsmouth, VA
kudos:3
Reviews:
·Cox HSI

Trust me there are many changes from the 2009 build you had on it to the 2012(?) build that was pushed to the modem. I mean just think of it like a router, one firmware can fix, or break many things. The reason ISP's handle the testing of firmwares for modem's on there network, and why it takes a while for new firmwares to be released. No firmware is perfect, as there is always some kind of bugs, you just hope they are minor one's that you can work around.

Also thinking about it, when I first started using my Cisco DPC3010. Which now is a backup, and collects dust. I had problem's with upload speed being rather low, and once cox pushed a newer firmware thing's were better. Which to me seemed like the CMTS, and modem worked together better after the newer firmware. I simply don't use the DPC3010 any more, because as I said before. I enjoy having access to more modem info, and the D3 motorola modem's provide that for troubleshooting.



xymox1
Premium
join:2008-05-20
Phoenix, AZ
reply to xymox1

Indeed that was my thinking that there is some protocol harmony occuring now with the new firmware. Maybe not so much fixing a bug as implementing some newer scheme in rate limiting that was indeed working with a CMTS. In fact as these are Cox branded firmware files maybe its specific to Cox. It would still be really fun to read the release notes for the firmware revisions.

The firmware that solved the problem
Bootloader Revision: 2.3.0_R1
Current Software Revision: DPC3010-v302r12901-110516a-COX
Firmware Name: dpc3010-v302r12901-110516a-COX.bin
Firmware Build Time: May 16 2011 17:09:15

Once my newest DPC3010 arrives I will be testing and I also plan on doing some mods to it like a proper fan and heatsink once I have tested it in its stock state.



xymox1
Premium
join:2008-05-20
Phoenix, AZ
reply to xymox1

Well the "new" dpc3010 i got was actually a older manufacturing date then the one I had already But thats just fine as I had already gotten the new firmware on the orginal one.

I went ahead and did some mods. Changed out the heat sink on the broadcom QAMLink BCM3380DKFSBG TN1136. It was a crummy heat sink. I used a real one with more surface area. I also mounted a small fan directly above the heat sink in the case. This dramatically lowered the chip temp.

I also did a number of mods to the power supply rails that feed the various sections. I added 6X the uF of the power supply caps and used tantalum caps. This had the effect of dramatically lowering the noise on the power rails to the various sections. The caps I put in cost more then the modem ! and of course thats why they could not use this quality of part. The power supply rails are now very clean and have very little noise.

The S/N on the status page gained 1db, which was suprizing as I would not have expected to see much performance increase but there was

The mods were meant to make the modem far more stable and happy, not so much increase performance.

I like that the DPC3010 has a shielded ethernet connector for use with shielded interconnects. I use cat 7a and these cables are always shielded.

Off to do more testing. First tests show rock solid 65/5.8. The 5.8 is surprizing and its consistent. It was only showing 5.5 before the mods so the mods might have made the modem a tiny bit better performance.

I will have pics later tonight.

Very consistent.



xymox1
Premium
join:2008-05-20
Phoenix, AZ

1 edit

Some pics...

The DPC3010.




RF section...




I changed the heat sink to a better one, used silver based heat sink compound. You can see the 3 2200uF caps and the silver wet tantalum 560uF cap..




A bunch of big 100uF solid tant added here.




There was obvious cost savings on the board. There were places for caps but these were not populated as im sure cost was a factor.

The fan is fugly, but very functional. I used Silenx. Its powered from inside the modem.



xymox1
Premium
join:2008-05-20
Phoenix, AZ

3 edits

While the DPC3000 is now not a consideration I figured I would put up test results just in case anyone ever cares..

The speeds at 65Mb/s / 5.5Mb/s were identical. The 3010 might have edged out the 3000 every so slightly un upload speeds. I think this is because the Broadcom CPU is faster then the TI in the 3000 and therefore can do the cap rate limiting in a more nimble fashion. Simply the CPU can respond faster doing things like ACK.

The differences however are clear in latency and jitter. Mostly jitter.. I took 2 samples from Multiping and put them together for easy comparison. The 3010 clearly has less jitter which results in a lowered averaged latency. This shows 2 15 min long samples.



So the 3010 is better at a smoother connection with less jitter. This is important for gamers and maybe VoiP.

Ive now confirmed the above mods did make a difference, somehow I can get slightly higher upload speeds. Lower noise floor ? Who knows exactly. But before the mods I 100% of the time got 5.5Mb/s up and now I 100% of the time get 5.8 and sometimes as high as 6. Not that big a deal but surprizing considering the mods...



Now I await the SB6180. It will be fun to test that and see how that performs...

The test setup is:
PC connected directly to modem. No router. No other software execpt Multiping. Multiping set to 0.1 ping interval.
PC Intel i7 X990 @ 4.5Ghz 6 cores, ASUS Rampage Extreme IV motherboard, triple channel high performance ram, Adaptec SATA raid controller doing RAID 0 with 2 SSD drives. Intel MB based Ethernet - gigabit. Capable of wireline speed.

For large transfer tests I use a PC based router running Vyatta. The router is capable of 1 million packets per second and does wireline speeds with any packet larger then 128 Bytes.



xymox1
Premium
join:2008-05-20
Phoenix, AZ

OK I got my SB6180 online. Im starting tests. I have requested that my firmware be updated.

Right outta the box its as fast as the DPC3010. While I need to do more tests I cant find any differences in performance..

After looking at the board I know why.. They are IDENTICAL... Yea thats right, IDENTICAL.. While the layout of the boards is a little different, EVERY RESISTOR, CAP AND SEMI is -EXACTLY THE SAME -. In fact THE PC BOARD LAYOUT PART NUMBERS ARE EXACTLY THE SAME. R203, C305, this are EXACTLY the SAME on both for every part and the vaules are the same. There are some minor differences. The DPC3010 has a USB port while the SB6180 has a PROM for storing its far more extensive GUI menu structure.

The layout is different with the SB6180 looking like it might have a better layout and topology for its RF section.

The internal power supply is different between the two in one respect. The SB6180 lacks any caps for filtering on the rails. While this will clearly be way more noisy in the long term with as much heat as comes from the BCM3380 it will slowly over years dry out electrolytic caps so not using them and making it work without them is a good idea long term IF you have a lot of heat. In my case with my fan this is not a consideration and if I decide to keep the SB6180 then I will be adding lots of caps and making the power rails MUCH quieter.

The DPC3010 was made in China and the SB6180 made in Korea.

So how in the world did both these modems end up with exactly the same schematic, parts, part designations on the PC board ? THEY HAD TO HAVE COPIED.. I feel they simply copied the Broadcom suggested EVERYTHING. Copy pasted the whole schematic right from a Broadcom datasheet.

So I dont expect to find any differences in performance between the 2 as they are exactly the same in every visible way. Its possible tho that the firmware is different and thats important, so I will do all my same tests...

There are a few differences worth note. The RF section in the SB6180 *might* be better. But its close. The SB6180 has a better heatsink and this will help long term lifetime. The DPC3010 has a better power supply internally but after 5 years of use might not be as good.

I am awaiting my firmware grade, once its upgraded i will start longer term and more extensive tests.. I may not report the results until next week..



xymox1
Premium
join:2008-05-20
Phoenix, AZ

HMMMmMmmmmmmmmm.....

I was just noticing that the DPC3010 reports a S/N of 39-40 but the motorola is reporting a S/N 29-31... 10db is a LOT of difference in S/N

As its coming from the same bcm3380 I would think its calculating the S/N the same way.

10db is alot. Maybe its the SB6180. Its a used one. I suppose I will have to order up a new one. I like the 6180 simply because of the GUI as no other performance measurements seem to be any different. However 10db of S/N is a whole lot.

The 3010 and 3000 have the same levels, the 6180 has 1db higher for downstream level and shows a lower upstream power. As the RF sections are quite different in layout and shielding it makes sense they have different numbers. 10 db of S/N however is a huge difference.

Maybe its just a bad used modem.


lilstone87

join:2009-04-09
Portsmouth, VA
kudos:3
Reviews:
·Cox HSI

Have you got the firmware updated yet? Either there is some kind of bug, or the modem has issue's. I would think more of a bug, reason I ask if the firmware has been updated yet. If it hasn't leave it online overnight for the firmware to be updated. Also have you reset the modem to defaults threw the modem page, and reset the modem? To see if the snr will correct itself? I seen my D3 eMTA do this recently, showing downstream snr's at around 30-31db, and a simple modem reboot fixed this error.



xymox1
Premium
join:2008-05-20
Phoenix, AZ

I got the ever competant CoxVegas rep to update it for me. Yes the firmware is updated.

I think its the modem. It was used and it looked pretty used looking at the inside. Degraded S/N woulld be a normal sign of some stress on the electrical input from the RF, say a surge or lightning. Funny they replaced the case but the modem itself was in very marginal shape. Thats what I get for buying a used modem off ebay. Well it was $40 and I can play with it to learn things and see how it performed.

It performed great even with the degraded S/N. Maybe a tiny bit faster then a DPC3010. Maybe. I will get a brand new one as I like the detailed stats in the GUI. I also like the ethernet jack, its Amphenol. I like the board layout better. There is more room in the case. I like the way they layed out and isolated the RF sections. I can add some great stability mods and have nice amounts of room.

I have to go outta town for work until next week so i will not be able to get a new SB6180 until after that. I will do a complete set of tests and if its the equal of the DPC3010 will mod the 6180, add a fan and make it my modem.

Overall tho its safe to say these 2 modem are almost identical. I doubt anyone would see any difference in real use at all. The only suggestion I have is to make sure they get lots of ventilation. Heat can kill these boxes quite easily.



digiblur
Premium
join:2002-06-03
Louisiana
reply to xymox1

Wow! I will have to say this is the most anal retentive thread about a modem I have seen. I am picturing all any every electronic on your home having mods done to them. Nice job!!



xymox1
Premium
join:2008-05-20
Phoenix, AZ

hehehe...

Well yes, I do mod a lot of stuff. Usually only analog based stuff as digital stuff does not really benefit from mods. A cable modem is a very analog device. Its receivers and transmitter are very advanced analog devices.

Also i figured google will find this thread and others may want to know the differences in the SB6180 and DPC3010.

Its really just a blog as I do what I was going to do anyway in getting the best network connection I can make.

The BCM3380 is capable of 500,000 packets per second so I want to be sure my whole network can do those speeds. My router can do 1 million pps and wireline speeds along with the computer, so my weak link is the cable modem and what speeds my ISP can provide.

Im so crazy i run a fast Debain based server just for running my recursive caching DNS server.

My network as a whole is lightning fast. I want my cable modem to be as fast as humanly possible. ALSO fast does not mean download and upload speed. There are all sorts of things that go on with communications, error correction, protocols and all sorts of stuff that effect how fast each TCP/UDP connection is setup and broken down and how all that is done over DOCSIS. All that effects perceived "speed" of the connection.



huddster

@cox.net

Very cool work. Would love to buy a modded out SB8160 from you, if you have time to make one. huddh (you know the rest since I'm a cox.net customer). Thanks.



xymox1
Premium
join:2008-05-20
Phoenix, AZ
reply to xymox1

Im not sure the mods make any difference. If they do, its a very small difference. Also im not sure what long term effect the mods might have.

So... I cant find any difference at all between a SB6180 and a DPC3010 performance wise. Well,,, with my limited tests. You would need very sophisticated test gear to fully test these devices. Like what Cable Labs has. Some independent testing lab should test modems and publish specs so we can all see differences. NO ONE does this. The specs are all secret.

Ive decided on the DPC3010. The reason is that Cox seems to be supporting them more as its all they sell and install in my area. Also in real life im not going to be looking at any of the GUI's that the 6180 has. If I do ever want to look I have a 6180 I could hook up. I also like the DPC3010 power supply and board layout better. The 6180 *might* have a better RF section tho. I have no way to test or measure this.

Ive done further mods to mine and I think I have improved performance when signals are really bad. I would get weird results below -12db before mods and now I can get down to -18 down and 57 up and still get what seems pretty normal performance. Sometimes at those levels tho I only get 2 upstream channels rather then 3.

Ive got some attenuators coming and im going to more accurately test my performance at various levels. I also have a unmodded dpc3010 to use for comparisons.

There is no doubt the modded modem works much better on bad connections. But levels of -15 and 56 db should be corrected in the wiring or with the cable co.

It does show however that far less noise on the supply rails to the BCM3380 does improve performance. However mods make no difference whatsoever with normal signal levels. Someone back there knew more filtering of the supplies did matter as there are places for much better filtering on the PCB but these are not populated. No doubt a cost saving step.

The one mod i would recommend to EVERYONE using one of these modems is a small fan. These things run way too hot and a small fan makes a huge difference. Also a better 12V power supply is a good idea. I use a 12V 3A with good regulation.

I will update this more once I get my attenuators and do some more testing.