It turns out that by disabling hyperthreading my rendering stress test is running 25% faster (1M cubes with HT, 1.25M cubes without HT)
In some MT tests I was doing, I found the same to be true. Disabling HT led to better performance. It's hard to say whether that remains true during "real load" and not stress tests.
-
-
I accidentally stumbled upon this at work and tried in my home stress test too. So it makes difference even in real load. Some function took 0.5ms when singlethreaded, suddenly goes to more than 1ms when MT with HT, without HT back to 0.5ms
-
HT is kind of a lie. If one thread is not waiting on external resource (hdd, gpu, whatever) then the other hyperthread won't give you more througput.
- Näytä vastaukset
Uusi keskustelu -
Lataaminen näyttää kestävän hetken.
Twitter saattaa olla ruuhkautunut tai ongelma on muuten hetkellinen. Yritä uudelleen tai käy Twitterin tilasivulla saadaksesi lisätietoja.