[Rickers]

I have a 1900ACv1. 59045 was last stable for me. The next one I used for a while was 60717. Anything in between those two, I deleted since I had issues.

With 59045, I get the same dl speed as with the cable router (675Mb/sec). Once, I go to 60717, the speed is cut in half. It almost looks like it's capped at 300Mb/sec.

Another strange thing is for builds >=60717, if I load them and reset to defaults, then I immediately do a speed test (at speedtest.net) without logging in or changing anything, the speed test will fail with a socket error. With 59045, I get 675Mb/sec. Maybe the current defaults are not appropriate for >=60717.

Anyone else seeing issues like this?

-Rick

[Rickers]

In build 59820, download speed is okay. In the next build, 59887, it's down 50%...that's also the build where the fan was first running all the time.

[juicetoon]

Yep - confirmed this by rolling back. SFE is completely broken in the new builds since 59045. It seems Qualcomm folks have more options with SFE, NSS-ECM and others. QoS is also trash compared to what it was years ago also.

Unsure if it's general ddwrt development in the bin, or all focus is on Qualcomm and Marvell downstream is just an afterthought.

Will also stick to 59045 for now I reckon, and may also try to dig up a build where QoS is working like it used to.

[kernel-panic69]

QoS and SFE are mutually exclusive. IOW, you cannot enable both at the same time. Also, Marvell got DSA prior to the now lower-end QCA devices (R7800 and similar hardware). It could be a DSA issue. It could be an issue with underlying interface order. The developer has at least two different Marvell devices in hand.

[juicetoon]

I understand they are mutually exlusive, if you switch on QoS SFE is off.

This still doesn't change the fact that SFE is completely broken as of now and has been for 6 months.

Nor that QoS WTB FQ_CODEL performs considerably worse than it did years ago.

And if I cannot get satisfactory bufferbloat I might as well get my paid for speeds & SFE is the only hope to get ~900 on a gigabit connection on Marvell at least.

I currently get ~400 without SFE on the latest builds, with terrible bufferbloat issues. Likely going to jump ship soon.

[juicetoon]

The reality is that ddwrt is a closed project operated and controlled by a single company with code written by one seemingly burnt out brainslayer (I understand the burnout), where the only meaningful interaction done between maintainers, contributors and community is done in the Deutsch threads.

A project with as many moving parts as this one really should be opened up where expertise and burdens be shared.

[kernel-panic69]

It seems you share the same thoughts as someone else. Yet you both don't bother trying to contribute as others have because you don't think you can. Well, that's just plain nonsense. Sounds more like you don't want to deal with a neurodivergent German. Which is fine.

[ygor]

with firmware r61848, browsing is little slower top open an new url, and some url are not accessible anymore, like omgubuntu or the steam shop in steam app, and other webstites works, looks to be related to tls handshake size, can marvell user test ?

i'm using wrt3200acm with r61848

[kernel-panic69]

It's probably related to what you noted in the other thread, @ygor. I have my theories as to what the problem is, but it would require proof of concept.

[ygor]

yes, issue is solved, it was due probably to a change between my old release and the last one, the setup/networking page changed a lot, after reconfiguring that part now all works

it was indeed an issue with the config after upgrading and not a firmware issue

[mohi]

@kernel-panic69 Do you have an updated list of safe releases afer r53633, r55615, and r55630? ideally with WRT3200

[kernel-panic69]

I'm on the current release (07-09-2025-r61915). I don't use SFE because I don't have a requirement for it. Now back to ignoring, "It's b'oken, fix!" nonsense that doesn't include serial console logs.