how do you track your usage? I use a performance program thats pretty good about tracking and it says only about half a gig is from memory leakage (mostly from a dev engine i use to make my games)
Poll 32 vs 64 bit
- Thread starter spacemarine658
- Start date
-
We have updated our Community Code of Conduct. Please read through the new rules for the forum that are an integral part of Paradox Interactive’s User Agreement.
Dude - no its not. You show me any benchmark that shows the CW engine as highly multi-threaded and I will show you the reasons the benchmark is wrong.
It has only very minimal multi-threading. How can you tell? Because when it starts to run slowly and you look at the engine CPU usage it's not 100%. It's roughly (100/number of cores) + a little bit.
For the most part any performance gain from 2 or 4 cores is simply because the engine no longer has to compete with other processes for CPU time.
You're all barking up the wrong tree. If you're thirsty ask for water, not a back massage. If you want performance gains ask for multi-threading, not a 64 bit engine. In either case both would be nice but only one of them actually solves the problem.
This thread is giving me PTOSD flashbacks to Windows XPx64.
A project of such awesome pointlessness it was only surpassed by Win7 32bit
- 2
A project of such awesome pointlessness it was only surpassed by Win7 32bit
Well, at least with Windows 7 32bit, it was installable on the masses of older computers that did not have 64 bit CPUs and easily upgraded via reinstall from the same disc.
This can only haven been written by someone who didn't experience the huge performance difference between single threaded HoI3 and multithreaded HoI3.
HoI3 as it is puts a second core to good use but the return from cores beyond that are minimal. Why that is should be obvious to anyone with some knowledge on the matter.
- 1
Ho Boy. So many technical misconceptions.
Is 'porting' a 32-bit engine over to 64-bit hard. Not really. But the performance gains are minimal, since a piece of code which is designed from the ground up to be 64-bit is very different from one that is designed to be 32-bit, and then ported over.
Is re-writing basically the same piece of software optimized for 64-bit alot of work. Yes. Since you should really build in support to take advantage of the extended memory capacity you have available. That means a complete rework of how you do things, more often than not.
The same basic thing applies w.r.t. multithreading. Adding some basic multithreading support (ie, do this process that is non-critical on another core) is fairly easy. Designing the game from the ground up to take advantage of multi-cores is hard, since to get full use, you need to be able to basically separate tasks out and largely desynchronize them from each other. That takes a lot of thought to do well.
So would I love to see good multithreading and 64-bit (absofragginglutely) but that should only happen as a serious effort, designed to make a great engine. Not a port or a hack.
Is 'porting' a 32-bit engine over to 64-bit hard. Not really. But the performance gains are minimal, since a piece of code which is designed from the ground up to be 64-bit is very different from one that is designed to be 32-bit, and then ported over.
Is re-writing basically the same piece of software optimized for 64-bit alot of work. Yes. Since you should really build in support to take advantage of the extended memory capacity you have available. That means a complete rework of how you do things, more often than not.
The same basic thing applies w.r.t. multithreading. Adding some basic multithreading support (ie, do this process that is non-critical on another core) is fairly easy. Designing the game from the ground up to take advantage of multi-cores is hard, since to get full use, you need to be able to basically separate tasks out and largely desynchronize them from each other. That takes a lot of thought to do well.
So would I love to see good multithreading and 64-bit (absofragginglutely) but that should only happen as a serious effort, designed to make a great engine. Not a port or a hack.
- 4
- 2
Oddly enough, i had no performance issues with HoI3 even without expansions, running on a single core (Athlon 3400+ iirc - not OCed).
IIRC the optimisation was patched in before expansions. HoI3 was excruciatingly slow on release. Ealy patches made a big difference.
I'm going to file the anecdotes in the pile of evidence supporting a highly multi-threaded CW engine right next to the FPS benchmark with 2/4 cores switched off. In the mean time I'm going to stick with watching my CPU shuffle along at 34% while EU4/HOI3 stutters at full speed as a much better indication of it's MT abilities.
The only thing that is obvious is that a large number of people don't read very well. There are a large number of thorough posts through this whole thread that spell out why 64bits will make no significant difference yet people still think it's essential. Multi-threading would make a large difference to the performance but people would rather moan about something PDS have said they aren't going to do than ask about something that would actually help.
- 1
Your core point about maximizing performance is solid, but the argument around it I can only class as sensationalist at best. When running EU4 at speed 5, it takes about 33% of my friends CPU time: http://i.imgur.com/ouacHzQ.png. He's got 8 logical cores. 33% is equivalent to about 2 and a half logical cores. There's clearly quite a bit of multi-threading involved. Is it as extensive as it could perhaps be? No. Is it a significant aspect of the performance? Yes. Is barely, and not in any meaningful way an accurate statement? No. Especially considering the 25% performance increase PDX were able to achieve in For The Motherland via multi-threading.
Most people don't understand anything about computer architecture, so they have a misguided sense of what is important for performance. I got eviscerated for saying that the last thread ~_~
- 3
Ok.
- For the most part you're simply measuring engine speed which is not directly equal to it's ability to multi-thread. All you've shown is that faster CPU's are faster and if the cores are equal more is better for the computer as a whole. As I mentioned earlier, if you move from 1-2 cores to 4 then the engine no longer has to share with other processes/applications which will result in performance increases for even completely single threaded applications.
- Your screenshot agrees with me, not you. If you look at your average CPU usage (excluding the spike) then you're at about 37%. Take away ~10% for background tasks and you are at.......wait for it...... 27%. Which is about 1 cores worth + a little bit which suggests that the engine is (mostly) single threaded.
- 1
oh my... this again...
Will paradox games be 64bit? Yes, probably when they build their next engine version. Before that, probably no.
Why? Because 64bits offer no advantage unless your program is designed to use more than 4GB of memory. Their engine is not designed for that. Usually paradox games use less than 1GB.
About multithreading:
They added multithreading with some of the HOI3 updates. It improved performance dramatically. Especially if your CPU had less single core power (HOI3 in my old laptop went from unplayable to quite fast in one update). Now the engine theoretically uses as many cores as you have but since it is extremely hard to divide the program flow to many threads, the advantage given by multiple processing cores gets smaller and smaller when you add cores. I have found that EU4 runs better with 4 cores compared to 2 cores but i noticed no difference in 8 core and 4 core systems.
For those who are not programmers:
you cannot just divide a program for several cores. You have to specify which things are done where and synchronize the process and mostly the cores end up waiting others to finish their tasks to be able to continue. Multithreading is extremely powerful tool with easily scalable tasks such as image or sound processing (or in any application where you can have a large series of unrelated operations) but much more difficult with games that have to follow certain synchronized logic.
Will paradox games be 64bit? Yes, probably when they build their next engine version. Before that, probably no.
Why? Because 64bits offer no advantage unless your program is designed to use more than 4GB of memory. Their engine is not designed for that. Usually paradox games use less than 1GB.
About multithreading:
They added multithreading with some of the HOI3 updates. It improved performance dramatically. Especially if your CPU had less single core power (HOI3 in my old laptop went from unplayable to quite fast in one update). Now the engine theoretically uses as many cores as you have but since it is extremely hard to divide the program flow to many threads, the advantage given by multiple processing cores gets smaller and smaller when you add cores. I have found that EU4 runs better with 4 cores compared to 2 cores but i noticed no difference in 8 core and 4 core systems.
For those who are not programmers:
you cannot just divide a program for several cores. You have to specify which things are done where and synchronize the process and mostly the cores end up waiting others to finish their tasks to be able to continue. Multithreading is extremely powerful tool with easily scalable tasks such as image or sound processing (or in any application where you can have a large series of unrelated operations) but much more difficult with games that have to follow certain synchronized logic.
- 1
A lot of people in this forum hate admitting when they are ignorant and like pretending they know what they are talking about no matter how painfully obvious it is that they do not.
A perfect example of this was a pair of posters on another thread who insisted that there was 'no such thing' as an 'historical fact'.
Theoretically it can do anything it wants. In reality the engine has the performance of only slightly over a single CPU core on a machine.
Almost all applications have a lot of low hanging fruit in the multi-threading area that don't require large changes to the underlying single threaded engine. Just need to identify the smaller tasks/sub-tasks that can be multi-threaded and do those.
The bigger problem is probably the language. Threading in a modern managed language is easy as. Doing it in C/C++/Delphi is a bit more