tain.totalcodex.net

Ah interesting, that's good to hear though. Guess it was somehow thrashing the shared VRAM. I'd be surprised by that behavior on Windows but I'm not exactly sure how MacOS handles VRAM to be honest.

So is performance now similar to 1.7.2? Somewhat strange that the different VRAM setting worked fine in 1.7.2, but regardless, it's good that you figured out that it might be related.

this seems to work for me as well as far as i can tell

Ever since 1.7 and detail textures came out I told everyone that for Macs the optimal setting is roughly 50-75% of your total VRAM.

This wasn't some mystery why their performance was crap. It makes no sense to put a detail texture cache rate higher than what your machine even has on it's card.

Is it just me or is it drafty in here all of a sudden?

I don't know about drafty, but it feels hollow and unhelpful in here for some reason.

Don't get your panties in a bunch princess.

Fine, I'll bite.

First of all, the texture cache is not for detail textures alone. It deals with many (most?) textures and sprites the game uses. Second, the texture cache setting they had gave them different results on 1.7.2. Even the devs aren't sure why altering the texture cache allows them to work around their issue. Third, as Mel and punk have told you... if you have nothing useful to add please let others deal with the problem. You aren't helping.

GodzFire wrote:This wasn't some mystery why their performance was crap. It makes no sense to put a detail texture cache rate higher than what your machine even has on it's card.

If that were true then they should have been getting poor performance on 1.7.2 as well, since they were using a setting much, much higher than their card capacity there too.

Whether setting the cache higher than the card's capacity makes sense or not depends on a lot of factors, like how the OS and driver implement their caching systems, and how much regular RAM the system has. In general our expectation is that only older machines with very small amounts of RAM should benefit from setting their texture cache so low.

These machines showing the performance issue don't seem to fit that description, especially Spa's, so it really is a mystery what is going on here. Unfortunately without a system physically available to the devs for testing, this kind of thing is very difficult to track down the cause for.

Things work differently for Macs and OSX. I don't pretend to know the ins and outs as to why, I just know my experience having used them for all these years with Myth and dealing with other people who are using OSX as well.

Setting it at that % of size in relation to the computer's video card RAM just seems to be the best bang for the buck.

GodzFire wrote: Setting it at that % of size in relation to the computer's video card RAM just seems to be the best bang for the buck.

Sure, but that wouldn't have solved Industry's problem for instance. What size do you set it to in a shared/integrated setup then? Obviously OSX is doing something a bit bizarre with the memory regions, but that's all opaque, so it's not like a general rule could address all of these situations. If there was one rule, we wouldn't even have the option - we'd just have myth implement it.

Pyro wrote:I don't know about drafty, but it feels hollow and unhelpful in here for some reason.

Drafty like when a ghost walks into a room.

Yeah, I understood. I was just going a different direction with it.