Performance "FIX" for Hexacores

[DrNikita]

Hi there,

As with most people here I have also been experiencing performance problems with my fairly high-end rig, But I have found a way to increase the performance a bit.

First of all here are my Rig's relevant specs:

i7 3930k (6-core / 12 with hyperthreading)
two Nvidia 680 in SLI *running Rebirth in single-GPU mode*
16 gig RAM
Windows 8 Prof. 64bit
SSD


So first things first. I play with VSYNC turned on (via NVIDIA control panel NOT via in-game settings). I can't stand screen tearing so vsync needs to be turned on.

I run the game at 1080p with max settings, i.e LOD / VIEW and EFFECT distance 100 and shaders at high. I found that changing the LOD/VIEW/EFFECT sliders had absolutely no impact on the frame rates so I left it cranked up.

The frame rates drop from a steady 60fps to 7-10 fps in some zones and during highway travel. On average the frame rates hover between 20 and 30 fps. That should be playable. However, even though the 30 fps should appear smooth there was severe "stuttering" which made the game unplayable.

I monitor my CPU activity on a second screen and found that one core was always close to 100% activity, a second core at around 50% activity with intermittent spikes of complete inactivity. The 10 remaining "cores" showed intermittent spikes of activity, but never more than 20% on any core.

So here are the things I have tried to improve performance. I found some posts in these forums suggesting to run the game in administrator mode. So I have done that, but with no impact on frame rates.

Next I was wondering, if perhaps my graphics card was pestering the CPU for more frames all the time (since the GPU activity most of the time hovered around 40-50%), so I cranked up the antialiasing settings in the nvidia control panel in order to make the GPU work a little harder.

I set it to 8xCSAA + 4x(supersample)

This brought the GPU activity up to 70-80% and increased my frame rate by around 10 fps.

Finally I managed to get rid of the annoying stuttering by limiting the XRebirth.exe to specific cores.

Considering that it seems modern CPUs struggle more with the game than older quad-cores. I assume that EGOSOFT made the game with quad cores in mind. Perhaps even CPUs without hyperthreading. So, I have disabled all but 4 physical cores via the "affinity" control in the task manager.

The i7 3930k has 6 physical cores or twelve with hyperthreading, so
here is the core setup that I chose: A ("+") denotes the active affinity while ("-") is a designates a deactivated core for the the process.

- CORE0
- CORE1 *hyperthreaded*

+ CORE2
- CORE3 *hyperthreaded*

+ CORE4
- CORE5 *hyperthreaded*

+ CORE6
- CORE7 *hyperthreaded*

+ CORE8
- CORE9 *hyperthreaded*

- CORE10
- CORE11 *hyperthreaded*

For good measure I also set the "priority" of the XRebirth.exe to "above normal".

These changes resulted in all of the four activated cores ( "+" ) to run at about 80-90% continuous activity. Without inactivity spikes! This eliminated the "stuttering" in-game.

I have tried just deactivating the *hyperthreaded* cores and allowing the XRebirth.exe to use all 6 cores. However, the activity level of all cores again drops and you will find that 1 of the cores works really hard while the activity level of the others seem to spike up and down all the time. As a result the "stuttering" returns.

With these changes , i.e with only 4 physical cores working on the XRebirth threads and forcing the GPU to waste resources for supersampling etc I now never drop below 25 fps. Most of the time the game runs between 30-40 fps in Albion and at a steady 60fps outside of Albion.

I have a suspicion that EGOSOFT didn't actually test the game on any CPU with hyperthreading or more than 4 cores. I also suspect that their internal FPS target was a minimum of 30fps on a system with recommended system specs. 30fps seems to be the standart target for "console games" .
Irritatingly, I firmly believe that it is IMPOSSIBLE to play the game on the designated "minimum system requirements" regardless of how you adjust the graphical settings.

I have yet to see any statement from EGOSOFT about what FPS you are supposed to EXPECT with a "minimum spec system" and what FPS they think a "recommended system" is supposed to have.

Finally, while I have managed to tweak my system up to a point where I can actually play the game "performance" wise (disregarding all the other game breaking bugs for now). 30fps on a high-end PC system is unacceptable. While 30fps is enough for me, since I remember happily playing Quake 1 at 15 fps back in the day (before 3dfx Voodoo cards) and because I am used to 30fps on my PS3, I can relate to how offended many of the real hardcore PC gamers feel. Especially, since it seems EGOSOFT actually blatantly LIED to us about their multithreading optimizations. "Bernd" has stated on the steam forums that the game really only has 2 main threads and can only even take advantage of 4 cores. That is not what they told us in this interview, which you can see on YouTube.

Here is the link (the relevant part is at 5min 20sec.) The video should jump to the correct position automatically. But in case it doesn't just skip to 5:20min :


Making of X Rebirth Interview http://www.youtube.com/watch?v=dt4B4e3i ... be&t=5m20s

Well I hope these tips will help some of you get more stable performance. But honestly we shouldn't be asked to make all these tweaks to begin with. I have to say EGOSOFTS stance on the matter is nothing short of infuriating. They knew exactly the performance is broken and didn't man up to it. Instead they keep giving us standart responses like....."Oh your RAM" is unbalanced or "have you tried turning down the graphics settings" ? Riiiight and .... "Oh in a game as complex as XRebirth you can't expect the same sort of framerates as in modern shooters."

Well EGOSOFT? What kind of framerates should we expect with your designated "minimum specs" vs. "recommended spec" ?
I think that is a fair question , don't you ? I mean... c'mon EGOSOFT you made the game you should be able to tell us what your target FPS was.

[banderlog]

I have had similiar findings running a phenom x2 hexacore - when I start the game I notice that it (or well, probably OS) choses which core to put the main thread on and it stabilizes around 80-90% usage. I have tried to use this info to overclock this specific core after rebirth is started (using AMDs overdrive app) but haven't noticed much difference with an overclock of +600MHz on the core in question.

But I certainly will try out the graphic card loading trick, that seems like a really interesting finding!

By the way, it is pretty clear from observation that the game is taking advantage of 4 cores (but arguably not more than that) - you can check it by limiting rebirth to one core, check fps, add another and see an increase. This increase (seemingly) stops after 4 cores.

The load as you say is extremely assymetrical - wouldn't it make sense to make multiple "heavy simulation" threads? Not really an application level person myself but it seems like it wouldn't be impossible.

The game obviously has a ton of potential, the framework seems to be there, so let's not get disheartened with a rough start. There aren't that many pioneers that had a smooth sailing in the infancy of their respective concepts or discoveries!

[BigBANGtheory]

I have a suspicion that EGOSOFT didn't actually test the game on any CPU with hyperthreading or more than 4 cores. I also suspect that their internal FPS target was a minimum of 30fps on a system with recommended system specs. 30fps seems to be the standart target for "console games".

I seem to recall that Egosoft used i7's in development which of course do have HT.

The minimum fps statement in relation to consoles... hmm well there is a pattern of facts forming not in the PC's favour is all I can say, I doubt you'll ever get a straight answer that is credible.

[ziporama]

See http://forum.egosoft.com/viewtopic.php? ... 17#4258217

From what I can tell, XR doesn't like GPUs waiting for the CPU, no matter how many cores it has, so slowing down the GPU by giving it more "stuff" to do such as, higher resolution, more shaders or more post-processing, goes a long way to removing stutters (in my humble experience).

The drawback is you get lower average FPS. The advantage is you get prettier graphics and much less stutter (mine's is in fact, gone).

Since posting that, I also found out that I need to adjust settings whenever GPU perf increases (I recently fine tuned my GPU overclocking for another game) and XR started stuttering again so I cranked up AA yet another notch.

Anything above 24 FPS is playable for me, noting here that my feeble eyes can't really tell the difference between 24 and 100, but that's because I need new glasses or a better monitor.

Z.

[DrNikita]

The load as you say is extremely assymetrical - wouldn't it make sense to make multiple "heavy simulation" threads? Not really an application level person myself but it seems like it wouldn't be impossible.

One thing I want to stress again is that when I limit the affinity of XRebirth.exe to 4 physical cores only, then the CPU load suddenly becomes extremely symmetrical. All 4 cores seem to be chugging away at the same very constant rate. But ONLY, if I keep my GPU busy with supersampling / higher rez / etc. If I turn the graphics settings down in the nVidia control panel the 4 cores go out of whack again.

[mcdjfp]

This is interesting. I have a i7-3960X which is also 6 core (12 with hyperthreading). I only have a single GTX 680, and 32 GB of memory. I have had little to no problem with frame rates.

The first thing I did was crank up almost all of the graphics settings and pull out my active shutter 3D glasses. I have been wondering why I have not been having problems when other similar systems were. I gave my system the overload GPU treatment immediately and never had a chance to notice the problems. Maybe I should bump up the AA the only item not maxed.

[DrNikita]

This is interesting. I have a i7-3960X which is also 6 core (12 with hyperthreading). I only have a single GTX 680, and 32 GB of memory. I have had little to no problem with frame rates.

The first thing I did was crank up almost all of the graphics settings and pull out my active shutter 3D glasses. I have been wondering why I have not been having problems when other similar systems were. I gave my system the overload GPU treatment immediately and never had a chance to notice the problems. Maybe I should bump up the AA the only item not maxed.

When you say you have little to no problem with frame rates, what kinda of frame rates are we talking about ?

[mcdjfp]

I have not checked the exact value since I was not having problems, but the frame rate (with 1 exception that occurred for 10 seconds or so on load in an almost clear zone) has always been high enough to appear smooth, even when turning quickly.

[CJBok]

Start the game with parameter -showfps. That will always show your FPS in lower-right corner.

[kayahr]

Well EGOSOFT? What kind of framerates should we expect with your designated "minimum specs" vs. "recommended spec" ?

Before anyone can answer this question someone must define what really IS the "minimum" and "recommended" spec. I mean, the recommended GPU is a "Nvidia GT500 Series". What does that mean? A GTX590? Or a GT510? That's quite a difference...

[DrNikita]

I have not checked the exact value since I was not having problems, but the frame rate (with 1 exception that occurred for 10 seconds or so on load in an almost clear zone) has always been high enough to appear smooth, even when turning quickly.

Well smooth is a subjective thing. A year ago I would've said 30fps is smooth. Then I got myself a better computer, and now I notice it every time the framerate dips below 60fps. - 30fps is still smooth . Almost all of the games of this last console generation run at 24-30 fps. So it is "smooth" enough for gaming. So try the "-showfps" thing and let us know. Cause if your fps in dense areas is at below 30fps then you are in the same boat as most people that are complaining here. But if you are at a stable 60fps then perhaps there is hope for the rest of us yet

[DrNikita]

Well EGOSOFT? What kind of framerates should we expect with your designated "minimum specs" vs. "recommended spec" ?

Before anyone can answer this question someone must define what really IS the "minimum" and "recommended" spec. I mean, the recommended GPU is a "Nvidia GT500 Series". What does that mean? A GTX590? Or a GT510? That's quite a difference...

Good point.

[quintupularity]

Casual comparisons of Nvidia cards is fruitless, within a series or between them. I have a GTX 660 TI which replaced my GTX 460, and they produce essentially the same frame rates. The only way to know how your card matches up to others is to look up a few reviews. Most tech sites have nice, colorful comparison charts. The overall trend is that the higher the series the better the card, but there is a lot of mixing during the climb to the top.

Therefore its not an easy thing for a developer to state unequivocally which card is the minimum, and which is awesome. A lot of people want simple answers before they buy, so developers provide one. But these days the truth is you just have to do some homework. Use recommendations as a guide, but most importantly KNOW YOUR SYSTEM.

It also helps when the game, you know, works and stuff.

[DrNikita]

Casual comparisons of Nvidia cards is fruitless, within a series or between them. I have a GTX 660 TI which replaced my GTX 460, and they produce essentially the same frame rates. The only way to know how your card matches up to others is to look up a few reviews. Most tech sites have nice, colorful comparison charts. The overall trend is that the higher the series the better the card, but there is a lot of mixing during the climb to the top.

Therefore its not an easy thing for a developer to state unequivocally which card is the minimum, and which is awesome. A lot of people want simple answers before they buy, so developers provide one. But these days the truth is you just have to do some homework. Use recommendations as a guide, but most importantly KNOW YOUR SYSTEM.

It also helps when the game, you know, works and stuff.

Perhaps, but they should at least be able to give us an estimate of frames they expect with their "recommended system specs" . I mean they must know if they recommend that you have a system that runs the game at 20 or 30 or 60fps....... Plus the graphics cards seem mostly irrelevant in this game. It seems you are CPU bound even on mid range graphics cards.

[kayahr]

Casual comparisons of Nvidia cards is fruitless, within a series or between them.

It makes no sense comparing the NVIdia "numbers", yes. You have to compare the specs. But that's not possible when Egosoft just recommends a "NVidia 500 series". When they would say "NVidia GTX660 or better" instead then gamers can really compare their own hardware with this spec. Sure, you have to know that a GTX590 is actually "better" than a GTX660. But that's the usual NVidia problem and not a problem of Egosoft.

[quintupularity]

Perhaps, but they should at least be able to give us an estimate of frames they expect with their "recommended system specs" .

Is a framerate estimate ever announced in any game by any company? Not to my memory. I'd appreciate it if developers did this, but given the near infinite combination of hardware and software out there its an unwinnable strategy. There is no simple "back of the box" publication that would work

A good solution could be that every new game publishes a webpage where the techs list what systems they tested it on, what the settings were in game, etc. Lots of detail in there to chew on.

But in Egosofts case, we know and they know it was an unfinished product. Its a moot point. For now we have to decide on our own whether X:R is worth the wait.

[DrNikita]

Perhaps, but they should at least be able to give us an estimate of frames they expect with their "recommended system specs" .

Is a framerate estimate ever announced in any game by any company? Not to my memory. I'd appreciate it if developers did this, but given the near infinite combination of hardware and software out there its an unwinnable strategy. There is no simple "back of the box" publication that would work

A good solution could be that every new game publishes a webpage where the techs list what systems they tested it on, what the settings were in game, etc. Lots of detail in there to chew on.

But in Egosofts case, we know and they know it was an unfinished product. Its a moot point. For now we have to decide on our own whether X:R is worth the wait.

Every game developer has a FPS target for their "recommended hardware". IF you ask them what it is they will normally tell you. Industry standart for consoles is 30fps. For PC games it varies more, but while devs won't write it on the box. You will find that they are very forthcoming with the information when asked on the forums. Honestly if EGOSOFT would just say ..."guys our target for recommended systems is 30fps" then a lot of people would be fine with it, consider that "smooth" and move on. Just because the game was designed that way. But since nothing is being said everyone is just comparing XRebirth to other PC games with similar looks and are wondering why there is such a large discrepancy in performance. So everyone is assuming that Rebirth is broken. I don't think it is broken, but I suspect the devs are inexperienced with multithreading and made some fundamentally poor decisions in the way the engine / AI is set up. IF that is the case then the low performance cannot easily be fixed. Then the only way of ever playing this game properly is hoping that new CPUs that will be out in a couple of years that have individual cores with much higher performance, rather than just more cores. All that being said, I would like to know what kind of framerates EGOSOFT is getting on their own test machines. And what PC setup get's the best framerates. In short I wanna know their FPS target.

[quintupularity]

I'm with you on this. Lots of companies respond to hostile forums with a hunker down attitude. I'd like to see them in constant communication. It demonstrates curiosity and shows that they are interested in what their customers think.

Not that I blame them, necessarily. Anytime a rep does peek his head in, instantly the angry chumps jump in and derail the conversation with spittle and vile and irrelevancies to the topic at hand. Everything they say will be picked over, speculated about and later on will surely be taken out of context and thrown back in their faces. Let the quote mining begin! Its a challenging environment for them to navigate.

Nevertheless, more communication is almost always better than less. I think they need to suck it up and come talk to us. This forum needs more official communication, a lot more.

[DrNikita]

I'm with you on this. Lots of companies respond to hostile forums with a hunker down attitude. I'd like to see them in constant communication. It demonstrates curiosity and shows that they are interested in what their customers think.

Not that I blame them, necessarily. Anytime a rep does peek his head in, instantly the angry chumps jump in and derail the conversation with spittle and vile and irrelevancies to the topic at hand. Everything they say will be picked over, speculated about and later on will surely be taken out of context and thrown back in their faces. Let the quote mining begin! Its a challenging environment for them to navigate.

Nevertheless, more communication is almost always better than less. I think they need to suck it up and come talk to us. This forum needs more official communication, a lot more.

Yeah I completely agree. If they would just communicate with the people on the forums properly rather than take the stance of "Nope there is nothing wrong and everything is a success" they would garner a lot more good will. I mean the only reason there are so many post on these forums to begin with is because people care about the game. What seems to infuriate most is the complete lack of respect that EGOSOFT'S (Bernd's) interviews have demonstrated towards their buyers. For some reason they seem to think that saving face is the most important thing here, yet they fail to realise that their credibility diminishes every time they let Bernd give an interview. Honestly, I think they should just not let Bernd talk to the community at all and have the actual devs come to the forums and use the community to work out the kinks now.