| 1 Jun 2025 |
 K900 | So it can't POSSIBLY advertise itself correctly | 12:02:41 |
| K900 | Becuase the EEPROM does not know what the fuck the rest of the thing is doing | 12:02:52 |
| K900 | In reply to @emilazy:matrix.org
don't I really wish I could advertise both 2500BASE-X and 10000BASE-X to the thing I'm plugged into? That's the fun part you don't | 12:03:11 |
 emily | : ( | 12:03:22 |
| K900 | The "right" way is to just do SGMII | 12:03:23 |
| K900 | And UXSGMII | 12:03:27 |
| emily | okay but like | 12:03:32 |
| K900 | Which both have in-band negotiation | 12:03:33 |
| K900 | B u t | 12:03:39 |
| emily | switches that only take 10G and nothing else are definitely a thing | 12:03:40 |
| K900 | SGMII ends at 1G | 12:03:45 |
| emily | right | 12:03:51 |
| K900 | And USXGMII starts at 5G | 12:03:53 |
| K900 | In reply to @emilazy:matrix.org
switches that only take 10G and nothing else are definitely a thing They very likely just take USXGMII | 12:04:10 |
| emily | right | 12:04:31 |
| K900 | And the reason people say they only take 10G is because 5G SFP modules are bigfoot | 12:04:33 |
| emily | so I guess, 2.5G was forced into SGMII? | 12:04:40 |
| emily | so if you are making such an SFP+, you can advertise SGMII with 2.5 and USXGMII with 5 and 10 | 12:04:52 |
| emily | and it'll work with both 2500BASE-X and 10000BASE-X only things? | 12:04:59 |
| emily | isn't that… autoneg? | 12:05:03 |
| K900 | No that's the fun part | 12:05:06 |
| K900 | It wasn't | 12:05:09 |
| K900 | Ever | 12:05:10 |
| emily | maybe I have no clue what autoneg is! | 12:05:10 |
| K900 | The only IEEE standard way of doing 2.5G specifically over SFP is 2500base-x | 12:05:30 |
| emily | uh. I thought SGMII was like an umbrella term for 1000BASE-X. | 12:05:50 |
| emily | and USXGMII is 10000BASE-X and 5000BASE-X. | 12:05:59 |
| emily | are they separate things entirely? | 12:06:04 |
| K900 | 2.5G SGMII isn't real | 12:06:07 |
| K900 | It's a weird vendor thing | 12:06:14 |
