AndresFreundTec on Nostr: Well, color me very confused. In a CPU bound workload two cores on the same socket ...
Well, color me very confused. In a CPU bound workload two cores on the same socket have substantially different performance (32% slowdown). If I just migrate the running process between the cores, performance changes immediately.
This is on 2x Xeon 5215 system.
I checked that it's not thermals, cpu frequency/boost and the system is idle.
Here's the odd part: The biggest difference evident in perf counters is
a 2.5x difference in icache_64b.iftag_stall, with ~same icache_64b.iftag_miss.
Published at
2024-05-26 08:55:32Event JSON
{
"id": "842aced5429deed88093ae4b2e4327fc88268c58aa8fdc63c74b49d46be647a4",
"pubkey": "f92b50f90934c1b547d709b076927c834559258b00ef49ed037607035b62a4f4",
"created_at": 1716713732,
"kind": 1,
"tags": [
[
"proxy",
"https://mastodon.social/users/AndresFreundTec/statuses/112506551156902614",
"activitypub"
]
],
"content": "Well, color me very confused. In a CPU bound workload two cores on the same socket have substantially different performance (32% slowdown). If I just migrate the running process between the cores, performance changes immediately.\n\nThis is on 2x Xeon 5215 system.\n\nI checked that it's not thermals, cpu frequency/boost and the system is idle.\n\nHere's the odd part: The biggest difference evident in perf counters is\na 2.5x difference in icache_64b.iftag_stall, with ~same icache_64b.iftag_miss.",
"sig": "65f51410aaa7d53466d9f61006e45868f45bd1c98f27b86085b22112c3d29c667f99d16bb194c2c01f17ef20b4af40d551ef4e99f8653889a4b565ff7190ed2c"
}