So I have worked up a matrix of information regarding scaling options, windowed vs full screen and at various resolutions. I will run more tests to average and smooth the numbers out. My results are confusing me at the moment, so I need to isolate more variables before I proceed. Right now I used a fixed camera anchor to give me constant camera location and height, but as game progresses, it becomes more of an apples to oranges comparison (ie. rendering more objects). Also, I find that the FPS will dramatically lower while panning around the environment, then the textures will load once you settle on a point and the FPS will stabilize. In a windowed screen, I notice a roughly 83% drop in FPS while moving around while operating in a 2560X1440 environment, where I am getting a stable 30 FPS while camera is stationary. At same position, running a 1920X1080 environment, I get roughly same stable 30 FPS, but while panning I only see a 50% reduction in FPS. The above was observed with AA and HDR off and preset video set to 'low', and no scaling.
In general it's easy to tell what's causing your slowdown in PA, same as it was in SupCom. If it runs fine zoomed out but slows down when you zoom in, it's the graphics card. If zoom level has little to no impact on FPS, it's your processor. And if units are jumping from one place to another independent of the frame rate, it's the network. I'm hoping there's a lot of optimisation to come here, as it currently seems inevitable in larger games.
You can also check FPS (ctrl+p) with and without UI using Ctrl+U. There is known problem that UI update decrease FPS a lot.
So i quickly tried switching between UI (on and off) and there is a gain of ~10 FPS when I turn off the UI. So the UI does seem to have an effect on the FPS. Is that a result of the way the UI is refreshed? I'm guessing that even though the majority of the UI looks static, that it is constantly refreshing in line with the refresh rate? so as balances for metal and energy fluctuate they can update on screen, but I am guessing the UI is being redrawn along with updated balance. Any insight into this?
Oh come on, what a load of BS. I have a 5970 in daily use and two 5770 in number crunchers. PA doesnt work with CF, tried it. And for technical reasons theres a good chance it will never do so. Tried forcing CF with various profiles which did generate load on both GPUs but resulted in shockingly low fps, much worse than on one. In terms of games, which most times dont make use of eight cores, even a 2600k running stock clocks will absolutely destroy the Xeons. And for other multithreaded applications, something like a 4960 will probably be a good bit faster too. Retire it, at least for gaming. I own an old dual G5 2,7 GHz but after years of good service its time was over. As your MacPro, it wasnt exactly cheap back then . Btw, when put to good work, did the newer Mac Pros still sound like a jetfighter at takeoff? P.S. Unsurprisingly quadfire with two 6990s doesnt work either. Edit: CF of course works with other, supported, games and brings a nice plus in performance. Thats what I bought it for .
Haha! I may very well have to retire the beast! My Mac pro doesn't sound that bad, I have it right under my desk and I cannot hear it unless the GPU fan kicks on high, but even then it is not to loud. My old Apple X-blade server use to sound like an F-14 on takeoff! Now that thing was loud. Darn thing used the IBM processors, which pretty much means it makes a cool coffee table As far as CF. I think it is just like hyperthreaded applications. if something is written to take advantage of it, there will probably be an increase in performance. I didn't expect to see CF performance gains for PA, atleast not in a Beta. I imagine they have a million and one things to consider and get up and running before that is even considered. All that being said, When I play a round in PA, my GPU doesn't go beyond 60& utilization ( monitored in GPUz) and my processor load doesn't go above 23% utilization (system wide) though my 8th core does look to be somewhere near the 80% area. So I know my system is old, but I know there is more performance to coax from this old thing yet! besides, I have a few months to kill before the new mac pro's get out into the wild. Perhaps ill be playing PA on a shiny black trashcan of a computer
Yeah, black/grey boxes ftw. You need watercooled Titan quad SLI. Ive heared that once you get that, the universe awards you with money, fame and attractive women. Dont buy AMD though, Im still waiting for their well shaped redhead to be delivered for the 6990 quadfire thats lying around here . Stuff it into the oldest and most rotten 1990s fullsized PC tower you can find . Well, back to topic: Your system loads sound an awful lot like mine. Ive taken a look at your other thread. Just sitting in the menue screen makes it use 30%(!!!!) of my two GPUs which means 60% of one GPU. Doing nothing. CPU useage around 20% on my i7 920. The rest like joining or lobby is like yours. Playing a bit, building a small base all keeps one core at 80% and the rest around 10. Only one GPU is active at 70%., second one idle. It will be very interesting to see how much performance Uber can squeeze out with future optimizations. Have to find a way displaying actual iops in RAM and VRAM. Regarding the CF, I have asked and read on the forums about it and the consens seems to be that PA doesnt support a fullscreen mode that currently CF needs to function properly and that this is not planned to change anytime soon. So dont expect it at or some time after release. So yea, your basically right on this, the application has to bring basic support for it. If I judged that right general opinion is something in the line of not putting SLI/CF support on high priority as most people dont have it. I fully support that btw., bad luck for me as Im left with half my GPU power but the game has much more important issues to sort out.
FTW indeed! As for RAM, I am noticing that my system will get to about 4 Gigs of RAM, and that is about it. Doesn't seem to matter the size of the system or how long I play. Just throwing that out there to see if that is in line with other systems. I imagine that is a design of the game at the moment. Not sure how much RAM would even be needed to run a game like this, perhaps not much!? And I couldn't agree more with your sentimentality's towards CF/SLI. It HAS to be low priority for a game like this, especially in Beta. And your reasoning is valid, most people just don't have it. In all honesty, I see SLI and CF as being like someone in the past trying to see the future tech trends. They saw what was available with current processing and performance, and thought the only way to go up was to add more cards. Now single cards are routinely breaking new ground, without the need of multiples.
Just one thing to note while talking about crossfire or sli, JAVA does Not support it, so litterately any game or application running anything java based will never ever be able to run Sli or Crossfire mode with your graphics cards, this is the one negative about using java as your game engine, this is why Minecraft cannot run multiple graphics cards as well (yes when using Shader modes and texture mods you need a lot of GPU power) a long with other games.... So if you upgrade, upgrade to a card that thats powerful.
This game doesn't use JAVA at all, but JavaScript used for UI (NOT FOR GAME!). Java and Javascript are similar like Car and Carpet are similar. (c) Somebody PS: Also this game can support SLI, but before it's need to be optimized for it. But at this moment it's can't use CrossFire because AMD don't support it for windowed applications and this game don't have traditional fullscreen.
RAM usage should definitely go above 4GB, at least on PC. for instance I was playing a game on a two-body system, both large mountain planets, and my usage was around 6.2GB. One of the people I was playing with said his was using 12GB (I only have 8GB total so I don't know if mine would have used more if it could).
So SLI is back in... I suppose the safest answer at the moment is to wait until the game is closer to final release to make any real decisions. Benchmarks for this game at this time will not be a great indicator I think of how the final game will run and how well it will run on various hardware configurations. It's a fluid situation.
I will try a larger system in a bit and report back. I'm pretty sure, now that I think of it, that the systems I played while monitoring the RAM usage were pretty similar as far as size and various other traits are concerned.
You sir deserve a cookie. This thread has answered some of my long-standing questions! much thanks for pointing me to this thread. I may now play with the slight UI mod...mwahahaha! (evil laugh)
Tried building and loading a 9 planet system, mostly just 4 main planets with 5 moons sprinkled in there. I got to roughly 8 GB of RAM usage, so I am guessing typical 4GB RAM consumption is based off of typical 2 planet system at average size. While in game, memory consumption only fluctuated roughly 250 megabytes from start to finish of a 50 minute game on a 2 planet system.
