Jump to content

Hardware vs. part count


Recommended Posts

Sure, but that does not explain why my CPU is about half loaded. With a plane of 353 parts, I get 8 frames per second, which is of course quite slow. Most people would think this is predictable due to the number of parts. However:

I would not think it was predictable at all, and I fear you will not like my answer to any of this.

As you can see, there is nothing obviously bottlenecking performance. All parts have actually plenty of performance to spare, but still the frame rate is dramatically low. Performance should be higher, but for some reason is not. KSP just seems to be incredibly inefficient in some way.

For you to be getting such poor performance, it's not the game bottlenecking you, your cpu is faster than my old amd fx 8120, and that cpu could handle a 350 part ship with ease.

Somewhere, something in your pc is not properly set up or tweaked, either that, or the other thing I would check, are temperatures to make sure your cpu is not throttling back.

You should be getting MUCH better performance than that.

Also, do not trust windows task manager to be entirely correct about the load on your cpu, KSP only uses one core for the physics calculations, this is an inherent design in the unity game engine.

Link to comment
Share on other sites

Sure, but that does not explain why my CPU is about half loaded. With a plane of 353 parts, I get 8 frames per second, which is of course quite slow. Most people would think this is predictable due to the number of parts. However:

CPU - i5 2500K: total load ~30%, most loaded core sometimes at ~70%, often at ~45%

GPU - GTS 450 1 GB DDR5: load about 30% too, 838 MB of 1 GB VRAM used, memory controller at 21%

RAM - 8 GB DDR3: 6,45 GB of 8 GB used

As you can see, there is nothing obviously bottlenecking performance. All parts have actually plenty of performance to spare, but still the frame rate is dramatically low. Performance should be higher, but for some reason is not. KSP just seems to be incredibly inefficient in some way.

That is normal. It shows that way in the system monitor, because OS is switching cores on the fly. All cores will get some tasks, none of the will get 100% utilized.

Link to comment
Share on other sites

KSP is single threaded and 32-bit. One core and ~2GB of RAM.

Well, KSP is not singlethreaded at all, at least not with mods. I could show you the graphs, but I can just tell you too. Load idle: 0-2%. The load with KSP is about 30-50% on all four cores. People keep telling me KSP is supposed to be single threaded, but I am not making this up. Some mods are offloading work to different cores, but I think KSP has gotten better at multithreading too.

Link to comment
Share on other sites

Well, KSP is not singlethreaded at all, at least not with mods. I could show you the graphs, but I can just tell you too. Load idle: 0-2%. The load with KSP is about 30-50% on all four cores. People keep telling me KSP is supposed to be single threaded, but I am not making this up. Some mods are offloading work to different cores, but I think KSP has gotten better at multithreading too.

The part of the game that is the performance limiting factor, the physics calculations, are in fact only running on one core.

Link to comment
Share on other sites

I would not think it was predictable at all, and I fear you will not like my answer to any of this.

For you to be getting such poor performance, it's not the game bottlenecking you, your cpu is faster than my old amd fx 8120, and that cpu could handle a 350 part ship with ease.

I am using FAR, RemoteTech and a number of other mods. This does add to the calculations, but it might be interesting to see whether that makes all that difference.

Somewhere, something in your pc is not properly set up or tweaked, either that, or the other thing I would check, are temperatures to make sure your cpu is not throttling back.

Temperatures are more than okay, something like 40 degrees Celsius for the CPU and 50 for the GPU - I have massive after market coolers since I loathe noise. The PC is tweaked out and - very likely - properly set up. I am not having trouble with any game or application other than KSP and - although it is always dangerous to say this about yourself - I kind of know what I am doing.

This, of course, does not account for the lack of performance without obvious causes.

Also, do not trust windows task manager to be entirely correct about the load on your cpu, KSP only uses one core for the physics calculations, this is an inherent design in the unity game engine.

The readings are correct and is really is KSP that is causing them. The mods probably make a difference.

Link to comment
Share on other sites

The part of the game that is the performance limiting factor, the physics calculations, are in fact only running on one core.

In that case one core should be maxed out, which is not the case. If Windows were switching rapidly between cores (which I think it is not) the maximum load should be 25%, which is also not the case.

Edited by Camacha
Link to comment
Share on other sites

In that case one core should be maxed out, which is not the case. If Windows were switching rapidly between cores (which I think it is not) the maximum load should be 25%, which is also not the case.

I don't think that's how multi-core processors work. They are constantly balancing the load across all of the cores, there are probably lots of reason for this, but one of them is even heat dissipation.

Link to comment
Share on other sites

Temperatures are more than okay, something like 40 degrees Celsius for the CPU and 50 for the GPU - I have massive after market coolers since I loathe noise. The PC is tweaked out and - very likely - properly set up. I am not having trouble with any game or application other than KSP and - although it is always dangerous to say this about yourself - I kind of know what I am doing.

This, of course, does not account for the lack of performance without obvious causes.

You're not going to like this either, but, I restate, also as someone who has been doing this for a very long time now, something is not properly tweaked, there is no reason you should have such low performance with the hardware you have otherwise.

I don't think that's how multi-core processors work. They are constantly balancing the load across all of the cores, there are probably lots of reason for this, but one of them is even heat dissipation.

This.

Link to comment
Share on other sites

I don't think that's how multi-core processors work. They are constantly balancing the load across all of the cores, there are probably lots of reason for this, but one of them is even heat dissipation.

There is a lot of load balancing going on, but when a program is thoroughly single threaded, that is very visible in the Task Manager (and other programs intended to monitor CPU usage). For instance, Cities XL is a game that is notorious for being single threaded and it shows beautifully (of horribly :)) in the Task Manager. This is not the case with (my) KSP. When a program is heavily single threaded, this should show, dispite any switching between cores. I can say with a high degree of certainty that CPU performance and/or the use of one thread is not the cause of the bottleneck here.

As a test, I ran KSP with the affinity set to a single core and then switched the affinity to two cores. My frame ratdoubled from 20 to 40 when I did this (with a random craft). The single core is fully loaded and thus maxed out, with two cores one is nearly fully loaded and the second one about a third to a half. Adding extra cores does not make any difference, except for how the load is distributed.

You're not going to like this either, but, I restate, also as someone who has been doing this for a very long time now, something is not properly tweaked, there is no reason you should have such low performance with the hardware you have otherwise.

I can safely say I have been doing this for a while now too and - although this is a bit dangerous to say about yourself - I kind of know what I am doing. Of course, there is always room to learn. Without any further testing it is hard to say anything sensible. All we know for sure is that there is a bottleneck somewhere, somehow, and that a lot of the usual suspects have been eliminated (raw processing power, temperature, disk I/O, RAM/VRAM et cetera).

KSP does not seem to like a secondary installation on a different location, when I have some spare time I will have to fiddle around with this some more.

Edited by Camacha
Link to comment
Share on other sites

- although this is a bit dangerous to say about yourself - I kind of know what I am doing.

I don't know why you keep saying it's dangerous to say, let me not so humbly proclaim I build monster pc's and I know what I'm doing.

Your cpu should give you far more than 8 fps on a 350 part ship, something on your system is not working as it should, I'm not saying it's a "fault" of yours, I'm saying you're not getting what you should for the hardware you have.

And I do not have physical access to your pc so I can't sit here and figure out what exactly it may be.

I have every version of the game from .16 onward "installed", with multiple copies of each "installed" too, so I wouldn't think that to be a problem either.

What is your video card again?

Link to comment
Share on other sites

Speaking of overclocking, does anyone have any tips for OC'ing an old AMD Phenom-II X4 940? 3.0Ghz by itself. Used to be able to do it from within Windows with Catalyst, but it seems they've removed that ability recently. I know for a fact that I can get it up to 3.3Ghz safely and stably with the stock cooling unit. I've pushed it to 3.5 before, I just don't know how far I can push it. But how would I go about doing it?

Link to comment
Share on other sites

I don't know why you keep saying it's dangerous to say, let me not so humbly proclaim I build monster pc's and I know what I'm doing.

As a pilot, I learned the day you become complacent is the day you become dangerous and reckless and should be the day you quit. The best pilots are aces, yet humble, and always open for new information, insights and knowledge and very aware of their own fallibility. I feel this is true for pretty much anything else in life too. Know that you know, but also know there is always a lot more you do not know.

Your cpu should give you far more than 8 fps on a 350 part ship, something on your system is not working as it should, I'm not saying it's a "fault" of yours, I'm saying you're not getting what you should for the hardware you have.

I think it is a bit too early for that conclusion. As stated a couple of times before, a number of mods are influencing these results as well are other factors. I need to do some testing to see whether removing those changes anything, unfortunately I am very pressed for time at the moment.

Pretty much the only way to find the bottleneck is testing, testing and more testing, systematically excluding possible causes :)

What is your video card again?

It is an ASUS GTS 450 with 1 GB GDDR5 with a custom cooler.

Link to comment
Share on other sites

As a pilot, I learned the day you become complacent is the day you become dangerous and reckless and should be the day you quit. The best pilots are aces, yet humble, and always open for new information, insights and knowledge and very aware of their own fallibility. I feel this is true for pretty much anything else in life too. Know that you know, but also know there is always a lot more you do not know.

This..

From having been involved in pretty much everything from industrial maintenance to CIO level IT, I can pretty much attest to the fact that people who brag about being 'experts', seldom are. And generally the louder they are about it, the less they actually know.

Link to comment
Share on other sites

This..

From having been involved in pretty much everything from industrial maintenance to CIO level IT, I can pretty much attest to the fact that people who brag about being 'experts', seldom are. And generally the louder they are about it, the less they actually know.

I disagree, and have proven myself many times. Humble has it's time and place, this isn't one of them.

Link to comment
Share on other sites

I run KSP with a...

i3-2330m and 8 gigs of ram shared with the video card, an Intel HD 3000.

Obviously i dropped the quality accordingly, but i still don't get how i can play a 720p youtube video at the same time without much of a framerate change.

Actualy the only thing bothering me is the cuts in the sound.

edit:(obviously a laptop)

Edited by Mokmo
Link to comment
Share on other sites

I run KSP with a...

Actualy the only thing bothering me is the cuts in the sound.

The cuts in the sound are not linked to performance, they appear to be a different issue. There was a topic on it BTC (before the crash), but I think that got fed to the Kraken.

Playing Youtube video at the same time as playing without much impact probably has to do with hardware acceleration and dedicated circuitry taking the load away from the general processing bits of your CPU :)

Link to comment
Share on other sites

Speaking of overclocking, does anyone have any tips for OC'ing an old AMD Phenom-II X4 940? 3.0Ghz by itself. Used to be able to do it from within Windows with Catalyst, but it seems they've removed that ability recently. I know for a fact that I can get it up to 3.3Ghz safely and stably with the stock cooling unit. I've pushed it to 3.5 before, I just don't know how far I can push it. But how would I go about doing it?

The main thing to watch out for is the temperature limit on the AMD silicon. According to the website (can't remember exactly where but a quick google search would find it for you) the maximum junction temperature should never exceed 65°C. This is a royal pain in the backside for overclocking the 940. I managed to get mine to 3.8GHz with a Corsair H100 but the junction temperature is just too low to really push them high.

If you still wanted to push it, bump the voltage up one notch (AMD Overdrive is quite good) then increase the multiplier 1x and stress test (OCCT for 10 minutes to start wit or similar). Keep bumping the multiplier till it crashes then bump the voltage till its stable. This can help you find the maximum for your CPU but your long term stability depends on a huge range of clocks (thanks AMD). When you find a setting which you are happy with, load the settings into the bios and then stress test for a few hours. Voltage can send temperatures through the roof.

Northbridge frequency and RAM clock are the two main culprits and I've never found a magical way to work them out except for trial and error. Good luck but watch the temperatures!

Link to comment
Share on other sites

The main thing to watch out for is the temperature limit on the AMD silicon. According to the website (can't remember exactly where but a quick google search would find it for you) the maximum junction temperature should never If you still wanted to push it, bump the voltage up one notch (AMD Overdrive is quite good) then increase the multiplier 1x and stress test (OCCT for 10 minutes to start wit or similar). Keep bumping the multiplier till it crashes then bump the voltage till its stable. This can help you find the maximum for your CPU but your long term stability depends on a huge range of clocks (thanks AMD). When you find a setting which you are happy with, load the settings into the bios and then stress test for a few hours. Voltage can send temperatures through the roof.

Temperature is the most important thing to watch indeed. If you want to run a chip as a daily driver I would also recommend doing a little research what a safe voltage is for 24/7 use. Usually, when temperature is under control, this will be fine anyway, but sometimes a chip can degrade due to too much voltage over an extended period of time. What the number is differs from chip(family) to chip, so a little Googling and research is in order. For instance, most people would say Sandy Bridge would be safe up to about 1,4 volt, maybe a little more.

Intels have about the same temperature range, btw. CPU's just do not like to be cooked :)

and have proven myself many times.

As any pilot has, before the final fatal crash ;)

Edited by Camacha
Link to comment
Share on other sites

Intels have about the same temperature range, btw. CPU's just do not like to be cooked :)

I just wrote a long paracraph on voltages for AMD chips but my browser crashed and I lost it. AMD quotes the max case temperature for its chips which is 65°C from memory, not maximum junction temperature which would be like 95+. As for daily driver voltages, I needed 1.35V(stock) to reach 3.4GHz, and 1.475V to reach 3.8GHz. I wouldn't go higher than 1.45V for daily driver if you have aftermarket cooling. With the stock cooler, I wouldn't even overvolt as it is one of the worst coolers I've every used (even worse thanstock Intel).

On a side note, AMD revised the silicon and package slightly for the 1090T 95W hex core part (NOT the 125W part) and gave it a higher case temp of 71°. If you could find one cheap then it would be a drop in replacement and you could either leave it at stock (3.3 GHz or something) or push it up much higher if you wanted. Food for thought.

Link to comment
Share on other sites

As any pilot has, before the final fatal crash ;)

This is a nonsensical reply, flying a plane is nothing like building a pc, I consistently build pc's that bench and game very speedily for the hardware used, often doing better than people with the next "teir" hardware.

And in the event of any problems, I quickly find and fix them, every single time.

You're comparing apples to bees because..you aint got nothin else brah.

You do have something that isn't working right in your pc though, so...*cough*, and blaming other thing's isn't going to solve it.

To use your very flawed comparison, because hey it goes both ways, maybe you've crashed your plane.

AMD cpu's, depsite what AMD says, I wouldn't take over high 50's c., they tend to start behaving badly after that.

Sandy bridge I would also keep in the high 50's or below, Ivy bridge is optimal at mid to high 60's c, but can operate in the 70's without much trouble...I just wouldn't be happy with that personally.

Those are all full load temps, idle should be ...well as low as possible...but low 30's is normal for most chips despite brand.

Edited by _Aramchek_
Link to comment
Share on other sites

Hi,

I have a 6-core i7 3930k with 32GB of RAM.

I get around 4-6 frames per second with a 1000-piece ship, and 2 frames per second with my 1800+ piece 270-ton fully loaded carrier.

That's the worst I've ever gotten. Most ships <400 pieces don't bring my game down by more than 20%.

Link to comment
Share on other sites

You do have something that isn't working right in your pc though, so...*cough*, and blaming other thing's isn't going to solve it.

I am not blaming anything, as I have stated that more testing needs to be done. Not much useful can be said until then, which our posts seem to confirm... The rest has been said and does not need to get repeated.

Due to time and the lack thereof, the next clean install will probably be 0.20. I will make sure to check out the stock performance and report back.

Edited by Camacha
Link to comment
Share on other sites

Sure, but that does not explain why my CPU is about half loaded. With a plane of 353 parts, I get 8 frames per second, which is of course quite slow. Most people would think this is predictable due to the number of parts. However:

CPU - i5 2500K: total load ~30%, most loaded core sometimes at ~70%, often at ~45%

GPU - GTS 450 1 GB DDR5: load about 30% too, 838 MB of 1 GB VRAM used, memory controller at 21%

RAM - 8 GB DDR3: 6,45 GB of 8 GB used

As you can see, there is nothing obviously bottlenecking performance. All parts have actually plenty of performance to spare, but still the frame rate is dramatically low. Performance should be higher, but for some reason is not. KSP just seems to be incredibly inefficient in some way.

Reporting back, as promised. With a totally new installation (vanilla settings all the way) of 0.20 I am getting about 40 fps on the runway and 25 - 35 in flight (or an attempted flight, as FAR is not installed :)) Some rare dips to 20, but those are short lived ). This is the same Windows installation as before, just KSP has changed. Quite a major difference with the 8 fps as before, which confirms my suspicion that the mods were a major contributor to the relatively low frame rate.

However, still no maximally loaded CPU and/or GPU. CPU load is about 35-40% (again, pretty evenly distributed over four cores) and GPU load is about 50%.

Link to comment
Share on other sites

This thread is quite old. Please consider starting a new thread rather than reviving this one.

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...