Recent content by Scott Laird

Here's Intel's page for E610 ordering info: Intel® Ethernet Network Adapter E610-XT2 Ordering and Compliance Information | Intel Notice that it lists the M5 "multi" flavor as being cheaper than the "single" version of the E610 by a few dollars. The i226 and i210 (at least) come in regular and...

IIRC the 4550 and 4500 used the same image. jinstall-ex-4500-15.1.R7.9-domestic-signed.tgz shows up on the EX4550 download page, and it looks pretty easy to find.

I *think* the 7020R series stopped at 4.33.

Okay, I've made it through more tests, and I don't think this is a systemic problem with Mellanox NICs. It's more likely to be a firmware bug (or perhaps a feature?) that appeared at some point, as it looks like newer firmware revs have the problem and older ones don't, but I haven't had a...

I'm back testing NTP and PTP performance again, in a more controlled environment now. I haven't made it to the CX4 yet, but I'm seeing interesting things with the CX3 (MCX354A) under test. I ran 6 tests, each 15 minutes long, comparing the time that Chrony set via NTP to a secondary known-good...

It's Cisco, they probably make 7 of them. 2 of them are basically identical with different paint colors. 4 are either obsolete or nearly so, and 1 is weirdly overpriced. Good luck figuring out which one is which. Also, consider running VyOS on a cheap mini-PC. It's a free (mostly, via the...

Mikrotik provides standardized benchmarks for all of their devices on their site; that's one of the couple things that they do very well. *In general*, their switch products (like the CRS line) *can* route, but they're horribly slow at it, and the router products *can* switch, but they're also...

I have 8 mostly-identical M510s, they all seem to behave similarly. I haven't tried a different PCIe CX3 yet.

Also, I have a few more details here. I should probably inventory my collection of CX4 cards; I have to have a spare here somewhere to throw into a test system. I know that I have a spare CX5 and an infinite number of CX3s available.

To give a bit more context, I'm trying to get clocks synced across my network to <1 RTT to improve the accuracy of distributed traces. I'll admit it's a bit of overkill, but Chrony claims that most of my systems are within 200ns of each other except for a couple Realtek outliers (but not all of...

It's the integrated CX3 on the system board on HPE M510 cartridges for moonshoot/EL4000/EL1000 systems.

That's only sort of true. Most newer Intel and Mellanox chips themselves support PPS inputs, but not all of them, and relatively few boards have the PPS/SDP pins available anywhere useful. You can get Intel i210-T1/i225-T1/i226-T1 NICs all day long with SDP inputs, but very few Mellanox NICs...

Yes, but PTP (v1) dates to 2002. *None* of this is really new, and the CX3 claims to have support, as do many of Intel's chips from that era. The CX3 in question is on an HPE EL4000 (M510), to cross the STH streams a bit more :-). The CX4's behavior is the really weird part for me. It's...

I've been running a bunch of experiments with time syncing at home, and I've discovered a handful of weird problems with PTP when using ConnectX-3 or -4 cards on client devices, even though they claim to have a PTP Hardware Clock. I'm running Chrony for syncing via NTP, but I've been having...

First, I've had good luck with MakeMKV for ripping blu-rays. It runs on Windows, Macs, and Linux (although the Linux version is kind of hidden), although, yeah, you have to rip onto the HD before you can do anything else with the file. It sounds like you might be able to extract just the audio...