Instale o Steam
iniciar sessão
|
idioma
简体中文 (Chinês simplificado)
繁體中文 (Chinês tradicional)
日本語 (Japonês)
한국어 (Coreano)
ไทย (Tailandês)
Български (Búlgaro)
Čeština (Tcheco)
Dansk (Dinamarquês)
Deutsch (Alemão)
English (Inglês)
Español-España (Espanhol — Espanha)
Español-Latinoamérica (Espanhol — América Latina)
Ελληνικά (Grego)
Français (Francês)
Italiano (Italiano)
Bahasa Indonesia (Indonésio)
Magyar (Húngaro)
Nederlands (Holandês)
Norsk (Norueguês)
Polski (Polonês)
Português (Portugal)
Română (Romeno)
Русский (Russo)
Suomi (Finlandês)
Svenska (Sueco)
Türkçe (Turco)
Tiếng Việt (Vietnamita)
Українська (Ucraniano)
Relatar um problema com a tradução
On the steam deck with the performance overlay running, there is a hiccup almost precisely every second.
Frankly, it looks like an API call or something set to poll every second or the like.
So I've been attempting to get Isaac running with mods on the Steam Deck, figured I would share some thoughts/findings in case it helps:
1. I'm trying to get a 'basic' set of mods going: the necessary stuff like StageAPI, etc, plus FF, revelations, retribution. Also the encyclopedia obviously.
2. While playing (and without any tweaking on my behalf), the chief issue is micro-stuttering. On the steam deck with the performance overlay enabled, it appears as periodic blips in the waveform. Honestly, it looks like the result of a periodic/polling event handler or the like.
3. Using some basic troubleshooting, I tested disabling all mods and then re-enabling them until the behavior came back. I ultimately traced it to two mods: Revelations Chapter 2 and the Encyclopedia.
4. The Encyclopedia by itself is fine, if I run it on a vanilla Isaac game, the behavior/performance is normal. There 'might' be a small change in performance, based on the waveform, but its honestly imperceptible to my eyes.
5. Revelations is actually the bigger performance hit, but both Revelations and Encyclopedia cause the micro-stutter when enabled in the context of the other mods I have subscribed.
6. With the full performance overlay on, I noticed that the CPU load is extremely variable when operating in the 'micro-stutter' situation. At random one of the cores with be suddenly loaded with a high demand.
7. Using a variety of tools I have been able to improve the performance significantly -- but the problem still remains. Specifically, after extensive tweaking I've found the biggest results with the following changes: I disabled SMT (hyper-threading). For some reason disabling SMT on the deck seems to have nothing but positive impact on the performance, presumably because there's less likelihood of a core randomly getting hit with a big demand and having to ramp up. I also used PowerTools (decky plugin) to peg the CPU clock to 3400. This keeps all the cores at a high clock rate so that the random load never hits an underutilized CPU that has to ramp up. I dropped the GPU clock rate to around 800hz since Isaac never seems to need more than that, thus freeing more power/wattage for the CPU. Lastly, I used some internal Deck tweaks to improve the VRAM to 4GB (from 1GB) and the swap to 16GB (from 1GB).
Of the changes made above, I believe it was disabling SMT and pegging the CPU clock speeds that had the greatest impact on improving Encyclopedia and/or Rev Chapter 2 performance.
The last thing I want to add is something about that 'waveform' the steam deck provides with regards to frame-rate. In a vanilla Isaac game, there are pulses in the waveform that occur rapidly, perhaps just under every second. However, these pulses have no effect on the performance and the game appears to operate gracefully.
However, when I enable either of the offending mods, the waveform changes. The interval between pulses seems longer and the impact is more noticeable.
I'm not a developer, but it makes me wonder if the underlying issue has to do with a regular/polled function or handler that exists in the vanilla game but is getting overrided by something in the mods.
I hope some of this helps. I'd be happy to test further.
Did some additional testing today, there are a lot callbacks to MC_POST_UPDATE that I think could be involved. I've tinkered with a few but I think there are too many to review.
Thanks for the extensive help, I was planning to do some of what you did. Happy to know someone already tested this much.