| 16 May 2026 |
 K900 | Then yeah it's either 5950X or wait for 5800X3D re-release probably at Computex? | 11:11:21 |
 magic_rb | Yeah i wish they actually released that one | 11:11:38 |
| K900 | Or try to hunt Aliexpress for 5700X3D | 11:11:44 |
| K900 | But I think they stopped making those also | 11:11:51 |
| K900 | Or at least I haven't seen a non-refurbished listing in a bit | 11:12:19 |
 marienz | ...ugh, I thought that re-release had already happened but you're right, not actually available (at least not readily available in the country I'm in) | 11:17:55 |
| magic_rb | 5700x3d is 380 from ali goddamn | 11:21:30 |
| magic_rb | https://www.marktplaats.nl/v/computers-en-software/processors/m2399857761-amd-ryzen-7-5800x3d-processor hm | 11:24:53 |
| magic_rb | How much of a difference in perf should i expect between those two? | 11:25:53 |
| marienz | 5950x noticeably faster in workloads that can use its cores, 5800x3d noticeably faster in things that can make use of its additional cache (which in practice is mostly games) | 11:26:58 |
| K900 | It's leaked a bunch | 11:27:01 |
| K900 | But not actually shipping yet | 11:27:01 |
| K900 | Yeah they're out of stock ~everywhere now | 11:27:02 |
| magic_rb | Good deal no? | 11:28:50 |
| K900 | 5950X is 16 cores, 5800X3D is 8 cores but huge L3 cache | 11:29:00 |
| K900 | The cores are the exact same | 11:29:08 |
| K900 | The X3D literally just glues a die of extra L3 cache on top | 11:29:24 |
| K900 | Of the exact same silicon | 11:29:27 |
| magic_rb | Gah, im tempted to go for the 5950x then, :/ | 11:29:31 |
| K900 | Which is huge for games | 11:29:41 |
| K900 | But kinda eh in productivity loads | 11:29:50 |
| marienz | I don't have the performance data for these paged in but I think that one hit TDP limitations if you use all cores, so it's not actually twice as fast in productivity workloads despite having twice the cores, but it's certainly faster (if your workload can use the cores) | 11:31:54 |
| marienz | err, that one being the 5950x | 11:32:08 |
| magic_rb | 
Download 1000006682.png | 11:34:09 |
| magic_rb | Timed compilation isnt the best metric, but its something | 11:34:22 |
| magic_rb | Two minutes faster on the 5950x | 11:34:56 |
| magic_rb | But then again, i wonder if nix eval doesnt benefit from the larger l3 cache, same with haskell compilation | 11:35:59 |
| marienz | I wouldn't count on it (unless you can find benchmarks that say otherwise) | 11:36:36 |
| K900 | Practically doesn't do much really | 11:36:58 |
| K900 | The hot set is too big | 11:37:01 |
