Jump to content
Sign in to follow this  
ilovebeer

ArmA II and 64bit

Recommended Posts

I was wondering if anyone heard anything about ArmA II and running it with 64bit operating systems?

Share this post


Link to post
Share on other sites

It would be pretty stupid of them not to support 64bit, especially considering that most new computers within the next year will come with 64 bit Windows as standard.

I've had no problems running ArmA I with 64 bit. I think its just that certain hardware configurations (particularily with ATI cards for some reason) cause problems. Hopefully they'll fix whatever caused those issues.

Share this post


Link to post
Share on other sites
It would be pretty stupid of them not to support 64bit, especially considering that most new computers within the next year will come with 64 bit Windows as standard.

I've had no problems running ArmA I with 64 bit. I think its just that certain hardware configurations (particularily with ATI cards for some reason) cause problems. Hopefully they'll fix whatever caused those issues.

Yeah, i agree. I have 64bit vista 4g RAM and i currently have no issues with ArmA atm.

Share this post


Link to post
Share on other sites

what really interests me is if 64bit is supported! Not if its running on 64bit systems+windows 64bit... since i use 64bit Pro and ArmA runs nicely BUT its still in 32bits ... switching to 64bit would open up some memorystuff i guess... although i dont believe they rewrote the engine to such an extend as to use natively 64bit ... i think this will only happen when ALL CPU and current OS on market will be 64bit only and the world has abandoned 32bit altogether, just like they did with 8bit and such before... but i think it will be awhile till then...

as long as it runs .. keep it going!

Share this post


Link to post
Share on other sites

No, we will not provide a native x64 application for ArmA II. That said, x64 configurations including 8 GB RAM are included in the compatibility testing for ArmA II to make sure it runs well and there are no issues like there were with ArmA.

Share this post


Link to post
Share on other sites
No, we will not provide a native x64 application for ArmA II. That said, x64 configurations including 8 GB RAM are included in the compatibility testing for ArmA II to make sure it runs well and there are no issues like there were with ArmA.

Wow great news Suma! thanks alot.

Share this post


Link to post
Share on other sites
switching to 64bit would open up some memorystuff i guess...

Nope, memory is handled by the OS. The benefits of having native 64-bit software aren't really relevant outside of high-performance stuff like databases and scientific/mathematical apps.

Share this post


Link to post
Share on other sites
No, we will not provide a native x64 application for ArmA II. That said, x64 configurations including 8 GB RAM are included in the compatibility testing for ArmA II to make sure it runs well and there are no issues like there were with ArmA.

So ArmA2 is able to address that memory?

Share this post


Link to post
Share on other sites
No, we will not provide a native x64 application for ArmA II. That said, x64 configurations including 8 GB RAM are included in the compatibility testing for ArmA II to make sure it runs well and there are no issues like there were with ArmA.

So ArmA2 is able to address that memory?

I think it could if it was a 64bit app, but since its not i dont think it will, however im not a programmer so i dont know.

Share this post


Link to post
Share on other sites
No, we will not provide a native x64 application for ArmA II. That said, x64 configurations including 8 GB RAM are included in the compatibility testing for ArmA II to make sure it runs well and there are no issues like there were with ArmA.

So ArmA2 is able to address that memory?

I think it could if it was a 64bit app, but since its not i dont think it will, however im not a programmer so i dont know.

As I pointed out above, memory management is Operating System controlled. At any rate, I'd be kinda worried if ArmA II was consuming more than 4GB of RAM...

Share this post


Link to post
Share on other sites
No, we will not provide a native x64 application for ArmA II. That said, x64 configurations including 8 GB RAM are included in the compatibility testing for ArmA II to make sure it runs well and there are no issues like there were with ArmA.

So ArmA2 is able to address that memory?

I think it could if it was a 64bit app, but since its not i dont think it will, however im not a programmer so i dont know.

As I pointed out above, memory management is Operating System controlled. At any rate, I'd be kinda worried if ArmA II was consuming more than 4GB of RAM...

Well, it would be nice to have extremely high resolution textures though. biggrin_o.gif

It would be a waste to have more than 4 GB RAM memory if it isn't used.

Share this post


Link to post
Share on other sites
It would be a waste to have more than 4 GB RAM memory if it isn't used.

Generally minimizing RAM consumption is considered to be good programming practice... I can't imagine any reason why it would need to use that amount of RAM - even GTA4; the great master of bloat only uses about 1 or 2 GB of RAM. Use your RAM for multitasking. Either way, its not relevant because the OS decides how RAM gets allocated.

Share this post


Link to post
Share on other sites

64bit is not only for more adress space >4GB, but it also give you higher precision in math without clever workarounds. If you ned 64bit precision then you are better off with 64bit os and application. I don't know many non science application that need this kind of precisison, weather forcasting moddeling and, huge sim applications, and alike -hmm is Arma a sim smile_o.gif

@ch_123

According to this article, Arma do allready utilize workarounds related to the 32bit adress space boundary. http://www.bistudio.com/develop....en.html

A transition to 64bit would have get rid of the implementet workaround. For what I know, 64bit could also have other impact not pursused for now, because it's outside the dev envelope.

- bigger world and more objects maybe?

- a weather model hehe - just kidding.

Share this post


Link to post
Share on other sites
64bit is not only for more adress space >4GB, but it also give you higher precision in math without clever workarounds.

This is not true. You can compute with 64b real numbers ("doubles") in 32b without any issues. With some limitations you can even compute with 64b integers - this is easier and somewhat faster in 64b, but this is something which has extremely limited use.

Moreover, all games perform most (or all) of their computation in 32b real numbers (float), because it is a lot faster and the precision is enough for the purpose.

Summed up: the benefits on native 64b are extremely small for a game. (With content creation tools the situation is different - 64b is useful here, because you frequently work with data larger than 4 GB).

Edit: moreover, there are downsides to 64b native: all your pointers are double in size, which usualy means memory consumption grows about 10-40 percent depending on how many pointers are stored by the application. Larger memory footprint also means reduced cache efficiency.

Share this post


Link to post
Share on other sites
64bit is not only for more adress space >4GB, but it also give you higher precision in math without clever workarounds.

This is not true. You can compute with 64b real numbers ("doubles") in 32b without any issues.

As I said ..Without .. workarounds.

[code #ref)64-bit add:

64bit-bit system:

fetch A

Fetch B

add64 A,B

32-bit system:

Fetch A1 //32 LSB

Fetch A2 //32 MSB

Fetch B1/ /32 LSB

Fetch B2//32 MSB

Add32 A1,B1 //add the "first half"

Add32_r A2/B2 //add the "top half" _r[#ref]

64bit math could also be done in a 16bit computer/os, and then the compiler need to use more clever workaround. What is not true about that? The fact that this is hidden for the develope dont change anything.

I did continue to say that I do not know about many applications that need the true 64bit match computing precision. And if ARMA 1 or II or what ever dont use 64bit math calculations in any significant way, yeah sure 64bit ARMA(n) is a waste.

I also remebmer the transision from 8 to 16 and from 16 to 32bit. And the bigger memory footprint was always a big issue, until you got enough memory, better hw, and then nobody did care any more. Same will happen now for sure.

Do we need 128bit computing? Sure but not in my liftetime maybe yay.gif

Share this post


Link to post
Share on other sites
64bit is not only for more adress space >4GB, but it also give you higher precision in math without clever workarounds. If you ned 64bit precision then you are better off with 64bit os and application. I don't know many non science application that need this kind of precisison, weather forcasting moddeling and, huge sim applications, and alike -hmm is Arma a sim smile_o.gif

so that precision is the reason why in Americas army for example the fps limiter is inaccurate with 32-bit but totally accurate with 64-bit?

Share this post


Link to post
Share on other sites
64bit is not only for more adress space >4GB, but it also give you higher precision in math without clever workarounds.

This is not true. You can compute with 64b real numbers ("doubles") in 32b without any issues.

As I said ..Without .. workarounds...

You can do both 64bit integer and floating point operations in 32bit x86 using SSE(2). And you can do multiple operations with a single instruction as well which would be faster than using "regular" 64bit native instructions. Though due to the nature of the x86 processors today a lesser number of x86 instructions does not necessary equal faster code or less executed code due to the internal CISC->RISC translation, OOOE, different pipelines, etc. Also the x87 math co-processor(today just the FPU) has always worked with 80bit precision and has 80bit registers.

x86 is a really ugly mess today, trying to come up with easy rules for what is fast and what isn't seems quite impossible wink_o.gif I believe that in practice what you really only get from native 64bit (x86-64) is ability to address more than 4GB of memory and access to 64bits worth of virtual memory. The performance differences should balance themselves out, with the added registers improving performance in some cases and the added overhead from 64bit pointers decreasing it in others.

Share this post


Link to post
Share on other sites

I rest my case, I did not know that the x86 arcitecthure/enviorment was uable to benefit from the wider 64-registers. Maybe thats good for Sun, IBM and Cray, or have AMD/Intel made a solution that in the longterm is the most valid approach.

Share this post


Link to post
Share on other sites

Its more a result of the lack of any kind of long term planning, instead just extending the age old 8086 technology (from the seventies!) with whatever is cool at the moment. I guess Intel's plan was to both bury x86 and do the 64bit move together with Itanium, but that project didn't quite succeed. Then AMD came up with the 64bit extensions for x86 which Intel adopted as well.

Share this post


Link to post
Share on other sites

There is also another big disadvantage people failed to mention so far - a native 64bit support suffers from the same problem as DX10 - not everyone has the hardware/software to support it.

While the hardware part is not really a problem here because x86 CPUs are too slow to handle the game anyway, quite a lot people still uses a 32bit operating system.

Therefore, just as with DX10, you still gotta have the game backwards compatible.

Share this post


Link to post
Share on other sites
Moreover, all games perform most (or all) of their computation in 32b real numbers (float), because it is a lot faster and the precision is enough for the purpose.

That's what I always wanted to ask; do you really use 32bit floating points numbers in ArmA for everything?

In OFP you clearly used 32bit, there were clearly floating point precision issues when coordinates were bigger than about 15km. 32bit FP loose precision in the millimeter-range when bigger than approx 10km. One would think that error in the millimeter-range is not a problem, but since those errors sum up, it is definitely a problem. Especially since differences in the centimeter-range may decide about hit or not hit.

Since people expect even bigger terrain, and Czernarus looks quite big, how do you solve those precision issues?

Do you partition the space, so that you can use locally 32bit FP? Or do you use 64bit FP for positions only, and 32bit FP for the rest?

Share this post


Link to post
Share on other sites

Well, Chernarus is actually smaller then Sahrani (400km^2 vs 200km^2).

However a big part of Sahrani's 400km^2 is water.

Share this post


Link to post
Share on other sites
Well, Chernarus is actually smaller then Sahrani (400km^2 vs 200km^2).

However a big part of Sahrani's 400km^2 is water.

Well, making the game world small is another solution. biggrin_o.gif

Also by translating [0.0, 0.0] to the center of the map, you'll get twice the range for the same precision (you can use the range [-10.0e5 .. 10.0e5, -10.0e5 .. 10.0e5] without large precision issues). Wonder if this is already implemented...

Share this post


Link to post
Share on other sites

I'll say this again....it would be nice to have both 32 and 64 exe's as Crysis offers.

Those with 32bit pc's would not have any issues.

Those with all 64bit pc's, would get more out of their system, and believe me, there is a big difference running Crysis in 32 compared to 64.

Since Suma stated 64 will not be, maybe we can get it in future releases.

Share this post


Link to post
Share on other sites

My only problem is Armas huge use of the virtual memory in x64. I always get pop-ups telling me to increase the page file. No matter how big it was I kept getting those errors.

Share this post


Link to post
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
Sign in to follow this  

×