Does more RAM speed up CMAK?

[Carl Puppchen]

I have been playing the Operation "Tobruk - the Easter Battles" via email with a friend for a few weeks. This is a huge operation for the first german assault on Tobruk. Whenever it is my turn to create a movie, it takes a long time, longer than I ever noticed in the past (but that is probably because I usually don't play scenarios this large).

I have 512 meg of RAM and a pretty new video card. I timed it and it took about 5 minutes 17 seconds of the blue bar creeping forward to create a movie.

For xmas I received some Best Buy gift cards. So I picked up an extra 256 meg of RAM and made sure my PC could see it. Then I retried those turns.

To make it more complex, since the first time I ran the 5:17 test I upgraded to CMAK 1.01, so I couldn't re-run that same orders phase, which means my test isn't really "apples to apples".

In any case, I re ran my latest movie turn and it took about 4:00 minutes, which was better. The down side was that the movie from the 5:17 turn was over 1 meg and the movie from the newer turn was about 800k, so maybe there was just less action.

When I run the turns and the blue bar is creeping up I don't see my PC swapping to disk so I assume it is a RAM issue. But I am not technical enough to know for sure, that is why I am posting on this forum.

Best of luck and thanks to Battlefront for creating a great game like CMAK! You can bet your life I wouldn't wait 5 minutes + for a turn to process any where else.

[Michael Emrys]

This is an interesting question. I'm no tech whiz, but while we wait for one to show up I'm willing to BS this subject for a while.

My guess would be that 512M was already plenty for running CM, assuming you don't already have some other RAM-hungry apps running in the background. More likely it's a CPU issue. If your current turn has a lot of projectile fire going on, especially AT, then your CPU is churning out the calculations to beat the band. That takes time. Other factors besides CPU speed effect that as well, such as cache size and speed, bus speed, etc. The more units you have on the board may effect that as well as the CPU has to draw the frames for the video card to hang textures on, as well as calculating the behavior of those units.

Michael

[junk2drive]

i only have an amd duron bargin 1.2 with generic 512ram and an ati radeon 7000 64mb ddr. so far even big ops in bad weather on cmbb or ak the thinking phase has been really quick compared to say enemy turn phase in sp series. i tried cmbo demo way back with onboard graphics and it was horrible. might have somrthing to do with having an 80 gig 7200 rpm hard drive, three partitions, so virtual memory has lots of room to swap files. winxp doesnt like less than 512. lots of programs these days auto check for updates and slow you down secretly. antivirus especially.

pretty new video card doesnt mean its powerfull enough, or that its set up right.

[Carl Puppchen]

Thanks for the thoughts so far.

I did some research in the forum and upgraded my directX drivers to 9.0.

Then I went out and upgraded my NVIDIA GE Force4 MX 440 graphics card up to the latest driver (5.30 I think).

And then I re-ran my movie...

and it took exactly 4 minutes!

This is just an idea but maybe we should set up a "reference" turn in a scenario that people could download and run like a PBEM turn and if we said "30 seconds" was a good time, you could benchmark and see if your machine was up to snuff.

Or if anyone wants to try running the turn that I am talking about, I could email it to you, just open it up and hit "go" (don't plot any orders) - see how long it takes vs. 4 minutes on my system and then I could try to figure out what my system is missing in comparison (or maybe my system is doing OK, I don't know).

I am not saying 4-5 minutes makes a game too long to play - I played "our backs to the Volga" and that game could take forever to run a turn... I just want to know if I am getting the most bang for my buck out of my system and any ideas of where I could tweak something to improve performance is appreciated.

I would assume key elements would be the graphics card, processor (P4 in my case) and the RAM. But that is just speculation

[ February 06, 2004, 10:31 PM: Message edited by: Carl Puppchen ]

[Michael Emrys]

Originally posted by Carl Puppchen:

This is just an idea but maybe we should set up a "reference" turn in a scenario that people could download and run like a PBEM turn and if we said "30 seconds" was a good time, you could benchmark and see if your machine was up to snuff.

Or if anyone wants to try running the turn that I am talking about, I could email it to you, just open it up and hit "go" (don't plot any orders) - see how long it takes vs. 4 minutes on my system and then I could try to figure out what my system is missing in comparison (or maybe my system is doing OK, I don't know).

Okay. I'm game. My addy is in my profile.

I'm on a Mac, so the comparison will be inexact, but might be interesting anyway.

Send it on!

Michael

[voidhawk]

I don't think such a "reference turn" is going to work well. Each time a new movie is calculated is will be different from the previous one because of random factors built into the game. For instance, if a tank gets knocked out in one movie but not another then of course it will not be able to fire at other targets, etc.

So, you'd have to come up with a turn that always ended up with the same sequence of events and the same outcome - that would be an interesting challenge! Perhaps some long range duel without the possibility of either side inflicting damage on the other side.

At any rate, CPU speed, cache and memory bandwidth will be the main factors in movie calculation speed. The video card has nothing to do with it; it only comes into play when viewing the movie. Unless you have so little RAM that the OS uses disk space for virtual memory I can't see that RAM size is going to make a difference either. For me upgrading from a P3 with 100MHz memory to a P4 with dual DDR400 memory made a big difference

[Michael Emrys]

Originally posted by voidhawk:

I don't think such a "reference turn" is going to work well. Each time a new movie is calculated is will be different from the previous one because of random factors built into the game.

You may be exaggerating. Although you are right about random differences each time your computer calculates the same set of orders, the difference in the time it takes to do so may well be minor. Easy to test. In a game, give your orders then save that turn. Hit Go and clock how long the calculations take. Then abort out of that game, select the same turn and do it all over again. Compare times. I'd be willing to bet the difference won't be more than two or three percent.

Michael

[foamy]

In my own personal experience with this slowness problem the CPU speed had a significant effect.

I don't have statistical data or real apples to apples comparisons, but I am absolutley sure that when I upgraded my CPU from an 850 to 1695 Ghz operating frequency the game speed increased enormously.

I did nothing else to memory or anything else and now even the largest of Tank battles run in less than 2 minutes.

[foamy]

Ghz !!, I mean Mhz of course..

[Jarmo]

I did some testing with CMBO a good while back.

To me it seemed the RAM had no impact on calculation speed,

at least not once CM had the "suggested amount", with mac it

was something like 60+ MB's or so, now a bit more.

So I'd guess 128MB's would be in the limit, 256 probably better.

But after that I wouldnt expect an increase in speed.

But I was wrong once in the 80's, might happen again one day...

[Carl Puppchen]

I sent the turn to Jason C but haven't heard anything back yet. My PBEM opponent who has a 6 month newer PC than mine is running the turn at about 3 minutes 12 seconds, so 4-5 minutes isn't so bad.

[Frenchy]

Carl - what's your CPU speed? That more than anything I believe will affect how long it takes CMAK to process a turn for emailing.

[Jack Carr]

I could be wrong, but I think once you get over 512mb of RAM it's generally overkill. WHat operating system are you running? I think Windows98 can't use over a certain amount anyway, not sure what that line in the sand is though.

[Michael Emrys]

Originally posted by Carl Puppchen:

I sent the turn to Jason C but haven't heard anything back yet.

You sent it to me too, but you must not be updated to 1.01 yet, because when I tried to run it I got an error message that said our versions aren't compatible.

Michael

[Carl Puppchen]

I am running a 2.0 ghz Pentium 4, under windows XP SP 2.

Sorry I sent the wrong turn to Michael - I re-sent the benchmark turn after conversion to CMAK 1.01. I had you confused w/Jason because I was reading a lot of his posts - both your posts are very helpful, btw.

My current time on the benchmark turn is 4 min 30 seconds.

To date my experience has been that more RAM doesn't help, assuming you have a good amount of RAM to begin with.

[Hpt. Lisse]

No question, CPU speed/FSB is the hurdle, with Athlon systems winning out (their Floating Point integer calculations are superior to Pentium models, though recently Intel has closed that gap).

Back in the day, I once did a monster CMBO battle, defending as Axis in a long valley map against a British assault. My rig at the time was a 450 Mhz AMD K6-2, and the longest turns took approximately 16 minutes to render. I couldn't believe that thing didn't crash out.

On a side note, your video card is a dog. Get yourself an early birthday present in the form of a FX5900 non-ultra or a 9600XT.

Hpt. Lisse

[ February 11, 2004, 12:22 AM: Message edited by: Hpt. Lisse ]

[Jack Carr]

Windows XP should be able to use alot of RAM should the application require it. 512MB is quite a good amount and I wonder if the CM games would ever be able to use that much. My rig has 1GB of RAM. Overkill for the application but it's there if it ever does call on it.

[Michael Emrys]

Okay, Carl, I ran your test turn and here are the results.

But first, I ought to give you the specs on my Mac. It's a G4 733MHz SP, 512MB RAM.

The first time I tried to crunch the turn, the game quit after 4:30 minutes. The blue bar was about halfway across at that time.

So I allotted more RAM to the game, bringing it up to a total of 120MB and ran it again. This time it ran to completion and took exactly 15 minutes to complete.

Supposing that it would have completed the first time, and projecting from the time it took to go halfway, it might have taken 9 minutes. Why the apparent large discrepency between the two times I cannot explain. It might be that differences in random events in the two runnings might have caused the TacAI to behave differently and the battle to develop differently and more complexly.

To borrow a phrase from Jason C., I hope this is of some interest.

Michael

[Carl Puppchen]

Thanks for taking the time to test it. I think I will follow another poster's advice and go out and buy a new NVIDIA graphics card and see if that speeds up my performance at all.

Given that the turn took you 9 minutes (approx), I won't do any more complaining about the processing power of my PC :cool:

Thus the conclusion seems to be, once again, that incremental RAM doesn't speed up performance, assuming that you have a decent amount in the first place (i.e. above system minimums).

[ckct]

The best computer upgrade depends on the slowdown.

If your camera or the movie is choppy, is the video card.

If your calculation speed is slow, is your CPU.