The Comunity Hardware Poll

[Sgt. Powderhound]

nd do some more experimenting with viewdistances.


Edit: Nobody in the community has the fermi??? No wonder nVidea was in the red last quarter. 

ISn't fermi still vaporware?

Nope, Fermi is the GTX4xx series

[NTH]

ISn't fermi still vaporware?

No idea. never tried to buy it.

I checked out the video memory watcher and if the memory usage it displays is correct, my theory that the full memory limits the fps is wrong 
My testmap only uses 320 MB of 1 GB and framrates drop while moving through the forest.
My next guess is, that geometry projection is only done when the player-position changes. So when you stand still the projected triangles are in the memory and the card only has to fill them with pixels. When you start moving, the triangles have to be recalculated -> drop of fps.
I haven't tried what happens if I pump the 1GB full with textures but I don't have the time for that ATM.

So what we need is a program that monitors the GPU of the Videocard.
Also could it be server - client interaction causing the lag. Meaning the many objects and your position from it needs to be calculated to server and send to your client.
You could test this if more people are connected to your server.

[Knitschi]

Temporary Summary of the Vote:

85% of the voters have a card of the HD 4000 / GF 8000 series or better.
35% of the voters have a card of the HD 5000 / GF 200 series or better.
86% of the voters have 512 MB video memory or more.
49% of the voters have 1GB videio memory or more.

Equipped with the powerful Video Memory Watcher I did some testing of the 27 official maps. The objective was to find out how much video memory and main memory each map uses. The results are shown in the following table:


I shut down FH2 after each test to clear the memory. To get all the data in the video memory I flew in freelook mode over the important parts of the maps. It is possible that the actual video memory need is a few MB larger than the one listed in the table because I did not "collect" all objects. The numbers for the main memory were taken from the task-manager.

I also tested Operation Cobra on low texture and geometry settings. It needed 0.76 GB RAM and 277MB video memory which is less then half of the high settings.

As most of the people have a HD 4000 / GF 8000 or better, I would be interested, what framerates are achieved with the midrange cards of those series (GF 8600/HD 4700). If they run the game at highest settings with 100 fps one could think about pimping up some of the lighter maps.
It would be nice if some owners of those cards can post the typical framerates they get on the most demanding maps (Cobra, Totalize) and on the smallest maps (Hyacinth,PDH,Sfakia).


@NTH:

Also could it be server - client interaction causing the lag. Meaning the many objects and your position from it needs to be calculated to server and send to your client.
You could test this if more people are connected to your server.

I only test on local server without bots. The server - client interaction will probably bring some extra problems with it.

[kummitus]

gotta say it atleast looks quite a lot 

[NTH]

That's some nice and thorough tests you did there.

Video Memory usage is textures being loaded, right ? With texture settings on high off course needing more video memory or it will use Main memory, which require swapping. This is good info for our players.

About the local server, it emulates server - client interaction, so your PC will have to bear the brunt of being a server and client at the same time. It will be interesting to see what a dedicated server will do with your map.

[NoCoolOnesLeft]

Bardia uses more than Pointe du? Bloody hell.

I know its not by much but I found that quite interesting, what with the general chaos, boats, smoke and fly-overs.

[Ts4EVER]

I lag more on PdH, but that is because of the smoke effects.

[Tim270]

Only looking at the range of GPU's is not going to give you a real view of performance.

CPU and RAM are also very important factors. I could have ticket I have a GTX470, It would look like Im playing on all high etc when I could have a low end dual core or even P4 that is holding the rig back.

[Strat_84]

Only looking at the range of GPU's is not going to give you a real view of performance.

CPU and RAM are also very important factors. I could have ticket I have a GTX470, It would look like Im playing on all high etc when I could have a low end dual core or even P4 that is holding the rig back.

This is a FPS you know, CPU doesn't really matter (at least if it's not a 10 years old one).

About RAM, having 2 Gb is a standard nowadays and it looks quite sufficient for BF2, even though it's not a bad thing to have some more available. 

[Fenring]

awesome chart Knitschi thank you very much.

[fh_spitfire]

Good work there Knitschi.

BTW, CPU does matter. For me changing from Pentium D to c2d was like stepping into heaven (for example Giarabub: unplayable on max settings (~14FPS) turned to ~70FPS. Same GPU, same RAM.

[kummitus]

same here, single core athlon, unplayable, dual core, on high.

though I guess there's not much difference on 2 or more cores, just that other specs of the cpu.

[Hockeywarrior]

Good work there Knitschi.

BTW, CPU does matter. For me changing from Pentium D to c2d was like stepping into heaven (for example Giarabub: unplayable on max settings (~14FPS) turned to ~70FPS. Same GPU, same RAM.

YES processor DOES matter alot more than I originally thought. My old PC was a 2.2ghz Athlon dual core and I had playable framerate, but on some maps it would dip down into the 20's which gave me a pretty big disadvantage. And this PC had a GTX 260 graphics card and 4 gigs of RAM.

Ever since I got my new i7 quad core (2.8 ghz), with the same video card and RAM as before, I've been getting FPS well into the 60s no problem in all situations. I know BF2 doesn't recognize 4 cores, but the processor clearly did make the difference.

[Kelmola]

I know BF2 doesn't recognize 4 cores, but the processor clearly did make the difference.

BF2 may not recognize more than a single core, but if you can run all your device drivers, Windows background processes, copy protections installed by other games, firewall, antivirus, PunkBuster, TeamSpeak etc. on the other core, it gets a single core all by itself which it didn't get before.

[Ajs47951]

BF2 does use all 4 cores, in the graphs below the last 2 seconds are at desktop just as the game exits, so not quite idling yet. Not efficient but still being used. This is why core speed is important. 50% of 2200mhz gets trumped by 50% of 2600mhz.