@ItoIian@milos_gis Do you have any idea what farms with this density of livestock and crop production do to things like water quality? I am not saying we need more residents either.
@shatrix@The_Only_Signal I’m only pointing out the silly buying comparisons. It’s like comparing a camper van to a scooter because they both reach the same speed.
@shatrix@The_Only_Signal At way higher heat/power, without the 2TB fast SSD, 128GB unified memory, portability, etc. Also OOMs on larger context sizes/higher quants, so you need more than just a 3090.
Still pretty cool to see such an old card perform this well.
@lxztlr Tough call, but we all know it’s the right one. Microsoft keeps saying they’ve changed and will listen to users, but in the end nice people like Scott H. are not the ones setting the direction. The board and shareholders are, and profit is ultimately the main driver.
@lxztlr I knew that. 😄 Agree on the Microsoft politics. Maybe naive, but the .NET team and some other engineering teams feel a lot less obnoxious than the Windows/Office side.
@ivanfioravanti An unbiased version would already be an improvement. And if this is considered an “unbelievable resource,” we’re doomed. Also funny how he’s talking shit about Apple, which he can of course, while expecting people to buy an NVIDIA RTX 6000 Pro.
@vaxryy Right. And it feels wasteful not to fix it upstream where everyone benefits. Custom patches just become tech debt the moment mainline moves on.
@TeksEdge Think you may have your wires crossed. Apple isn’t going anywhere. The >96GB Studio gap is a product-cycle artifact: M5 Studio is incoming, so the Ultra-class SKUs are paused, not discontinued.
@Roadman_Podcast The biggest benefit is staying mentally fit. Long solo rides will demoralize most riders at some point. For effective physical training, though, I don’t understand why more people don’t ride in pairs with adjustable gaps between riders: larger gaps if you are ‘too good’, shorter
@leetllm@MemoryReboot_ Well, I have 5070ti in a decent setup: 9950, 64gb @6000. But is is absolute dogshit for inference compared to my m5 max laptop(!). You need context and 27B is the current gold standard. Why the hate? 10k for a Mac, is also false.