Guess the bottleneck on my rig

[Scrapper_511]

A custom scenario I'm playing is kicking my rig's butt. Lot's of Russian hardware steamrolling headlong into a village with prepared defenses. Anyhow, framerate gets shot to hell (this scenario only, others run quite well)on my following rig:

Athlon XP3000+ (Barton)
1GB PC2700 RAM
nVidia GeForce 6800GT AGP

That's the best CPU my mobo can handle but I was wondering if adding another Gig of memory would help. The GeForce is relatively new (on my rig)but is the CPU still outpacing it or has the Athlon reached its ceiling?

[JamesT73J]

I'm on a very similar spec to you; I've no issues with SB apart from certain winter maps that are foliage intensive. My solution is to simply switch to TIS which smooths it all out.

If ever I find myself getting annoyed with my PC's performance, I cast my mind back to the original M1TP on my A500, when HEAT rounds took about a month to reach their target :)


James

Edit: I was going to add, ProPE seems to handle large numbers of entities well (the busy maps don't seem to slow it down much) so I'm guessing that isn't your issue. You're probably simply running out of juice on a GPU-intensive map.

[Ssnake]

You can test this by running the original scenario twice - the second time with all detail sliders set to zero, and screen resolution reduced to 800x600. If the frame rate still drops in a similar manner, the graphics card is NOT the bottleneck. More RAM is probably not going to help. I think it may simply be LOS calculations and there is no was around those except faster processors (but then again, with more units in close proximity the number of necessary LOS calculations grows much faster than CPU processing capacities.

A change in scenario design may be a good idea. Can you spawn some of the forces at a later point, e.g. when a similar number of units have been killed already. That alone might help you a lot to keep the frame rates high. In addition you can try and reduce the visibility limit. Finally, map design. If you have lots of trees applied with the spray tool, or if the forests are broken up in many places, the usual tricks to help reduce the amount of details in the landscape don't work well, hence the frame rate will drop. Closing gaps in non-contiguous forests may help as well.

Finally, do you play the scenario primarily from the gunner's or from the commander's point of view, or from the map view? This can have an influence on the frame rate as well.

[ShotMagnet]

Frame-rate dislikes choopers, infantry, and smoke. Alter the availablity of one or more and you may find some happiness.


Shot

[ShoutingDog]

Frame-rate dislikes choopers, infantry, and smoke. Alter the availablity of one or more and you may find some happiness.


ShotBeat me to it...

In the two weeks I've had SB Pro I've noticed that smoke appears to be the biggest frame rate killer - even when you can't see it.

My system is a 3.2GHz P4 with Hyperthreading, 1Gb Ram, ATI X1600 Pro Graphics card with 512Mb onboard RAM. I've yet to find a game that will choke it with 2XFSAA, 8X AF at a 1024X768 res setting.

But SB Pro will with smoke. My framerates run 35 to 55 FPS under most conditions with the settings noted above. If I use the zoom view into trees the FPS will drop to 23 to 25 FPS.

BUT! Not if there's smoke visible anywhere. If the view point is external the FPS drops into the mid 20's. From the TC's position (and the vehicle moving) it'll drop into the high teens and if there's woods around FPS drops into the low teens and this is whether or not I can see the smoke or not.

For some reason my graphics card just doesn't like smoke...

As another poster mentioned when that happens I too just switch to TIS view.

[Lone*star49]

Beat me to it...

In the two weeks I've had SB Pro I've noticed that smoke appears to be the biggest frame rate killer - even when you can't see it.

My system is a 3.2GHz P4 with Hyperthreading, 1Gb Ram, ATI X1600 Pro Graphics card with 512Mb onboard RAM. I've yet to find a game that will choke it with 2XFSAA, 8X AF at a 1024X768 res setting.

...

SD, I have a 3.4 Ghz P4Sp2 with Radeon X1900 512mb, 1Gb Ram DDR3
But my original Asus MB, took a shit from overheating, So I replaced it with a dual MS Encore Processor along with MB.

But, before I had to do that, I read in many a discussion about how to improve framerates in SB, and one, that did work, was that since SB does not use hyperthreading, to disable it, and it did improve framerates FYI.


LS

[ShoutingDog]

...
But, before I had to do that, I read in many a discussion about how to improve framerates in SB, and one, that did work, was that since SB does not use hyperthreading, to disable it, and it did improve framerates FYI.


LS

That's spooky because hyperthreading and it's impact on performance was going to be my next question.

Good to know that turning it off has a positive impact.

[HotTom]

Nils suggestions:

"A change in scenario design may be a good idea. Can you spawn some of the forces at a later point, e.g. when a similar number of units have been killed already. That alone might help you a lot to keep the frame rates high. In addition you can try and reduce the visibility limit. Finally, map design. If you have lots of trees applied with the spray tool, or if the forests are broken up in many places, the usual tricks to help reduce the amount of details in the landscape don't work well, hence the frame rate will drop. Closing gaps in non-contiguous forests may help as well."

I find all of these help frame rates when you are designing a mission.

I've been asked several times why I have units spawn late in the mission, whether it's supposed to be a surprise. It's strictly to help the frame rate and it really works.

Visibility limits also give the FPS a nice boost.

Trees and buildings are real FPS killers. The towns in my terrain maps are very rudimentary for that reason. And I limit smoke for the same reason.

I know that's not much help if you're playing a mission with lots of those items in them already, but it's something for the mission designers to keep in mind.

HT

[ShoutingDog]

Trees and buildings are real FPS killers. The towns in my terrain maps are very rudimentary for that reason. And I limit smoke for the same reason.

HTLimiting the use of buildings and trees is not unreasonable as many realistic battle situations can be imagined and scenarios for same created without either. However, smoke is a fact of life in any realistic contemporary battle especially ones where the multi-spectral stuff is available to the real life forces of technologically advanced nations. Limiting it's use can and I imagine does put a damper on the realism factor.

Any scenario designed as a deliberate assault for example would of necessity have a preparatory artillery barrage followed soon after by smoke to cover the advance of the assaulting force. If frame rate issues limit the use of smoke under these circumstances and thus requires that scenario designers limit their use of it to make the scenario playable then it seems to me that the realism of such scenarios suffers.

SB the game doesn't suffer from not using smoke. SB the sim does (albeit not critically).

[msconfig]

Dust is also a killer, the only battle I've noticed slowdown is one where the map surface generated a lot of dust...

[Scrapper_511]

Thank you, everyone, for your feedback. Regarding the scenario in my opening topic, I created it for pure tank-killing gratuity with little regards to realism, authenticity, and framerates. That said, I do appreciate the scenario design tips but my concern in this topic is certainly hardware related only. I just decided to use this particular scenario as a "benchmark" for my rig to provide an example.

I am thinking of building a new rig, but if my current setup just needs more memory or a faster agp card, then of course I'd rather go this route. This is what I need feedback on.

--

Ssnake, I mostly play from Gunner's position.

Shoutingdog, ~25fps under heavy load is impressive to me. Under the condition I mentioned in my opening topic, it is absolutely unplayable.

[Ssnake]

I'm not sure why smoke would be such a big frame rate killer for some people. I'll kindly ask the beta test team to look into this a bit more.


I find it against your interest however to recommend a hardware upgrade. You already have a pretty good system and hardware upgrades will bring you only limited gains (at the same time, buying the latest and greatest is costly). It therefore makes more sense to invest a few brain cells instead and tailor the scenario to still give you the slugfest that you want while retaining a high frame rate through clever map and mission design.

I mean, even with the fastest PCs out there you can kill the performance totally with excessive line of sight calculations. You simply can't evade this fundamental limitation, so discarding tailored mission design for performance gains as "irrelevant" or "detrimental to realism" deserves honorable mention for the rigid application of righteous motives, but is ultimately stupid because better hardware can't buy you much better performance. Learn to accept the inevitable, and work with what you have. :)

[ShotMagnet]

From personal experience, it's not just smoke. Scenario size, quantity of infantry, and other factors as previously noted all have a synergistic effect. The number of players can also kill FR, depending on whom might be hosting.

Map-size is also an issue. The larger the map the more load appears to be imposed on the board and/or peripherals. Again from personal experience, the initial First Clash missions suffered mightily for the synergy of map, infantry, AFVs, players, and the needs of the host to keep all the 'eggs in the air'.

I'm running an ATI Radeon 9600, pretty good card, and it has troubles with larger missions with lots of people. Anything lesser in terms of memory or processing capacity is going to experience correspondingly more difficulties.

What you have running in the background might also play a part. When you test a mission, observe what is running on you PC (to include malware scanners, e-mail, browsers, etc), then notice what (if anything) happens to frame-rate when those various entities are switched on/off.


Shot