Jump to content
The Dark Mod Forums

Drawcalls and performance issues


Bikerdude

Recommended Posts

In my instance the DC went up from 3100 to 4100, so if I want to contiue to use the hedge model I will have to redesign that garden

 

What kind of impact does that have on fps?

Link to comment
Share on other sites

What kind of impact does that have on fps?

 

No impact at all... at least not on my system. I changed it so it doesn't switch the model to a low res and from a distance and even that doesn't cause an impact to my FPS.

 

rLK2Rl0.jpg

Edited by Goldwell
Link to comment
Share on other sites

In my instance the DC went up from 3100 to 4100, so if I want to contiue to use the hedge model I will have to redesign that garden

What kind of impact does that have on fps?

 

Yikes, over 3000!? Well, if you're already looking at single digit FPS for some systems, 4100 wouldn't be any worse, as it'd still be single digit FPS! LOL

 

In my case the replacement reduced drawcalls 25%, but below 1000 in both cases (686 to 511), which on my system allows for 60 fps either way. The LOD low poly version reduced drawcalls 10% as best I can figure (trying to not get other polygons rendered in the wider field of view from moving back)--which is more significant than it looks since most objects in view aren't changing.

 

But to answer your question Spring, the impact of excessive drawcalls would depend entirely on the individual hardware. For those with very high end hardware, there might be no impact, they'd get over 30 FPS either way. For those with less expensive hardware, it might be unplayable regardless. Those that have "middle" hardware that can handle a significant amount, but are taxed by going over their ability, would see the biggest benefit from lowering the drawcalls a bunch, as they'd be crippled by the too high, and might get teens or twenty-something FPS if it was lowered to be within their capability.

"The measure of a man's character is what he would do if he knew he never would be found out."

- Baron Thomas Babington Macauley

Link to comment
Share on other sites

What kind of impact does that have on fps?

In the worst spot the FPS dropped from 61 to 41, but that is with tweaked "lod_normal_distance" and "lod_1_distance" figures to 650 & 1150. So I set them back to the default of 500/600 and added a "hide_distance" and "dist_check_period" 1500/0.33 the DC/FPS dropped back to what it was before of 3100/51.

its all about the hardware people have.

Unfortunately, in order to move forward we will have to raise the min DC level for TDM up from D3 default of 1500 to 3000 to allow mission authors some wiggle room.

Link to comment
Share on other sites

I wouldn't call my system that high end but I doubt it would make any noticeable impact.

 

I mean I just created 5 rows of them and put them all in one shot and there was no impact on FPS again

 

QT79uh5.jpg

Link to comment
Share on other sites

Unfortunately, in order to move forward we will have to raise the min DC level for TDM up from D3 default of 1500 to 3000 to allow mission authors some wiggle room.

 

Heh, that's not moving forward! Designing smartly and efficiently is progress, being inefficient and suggesting folks spend money on different hardware is moving backward in my opinion. My first map had horrible performance compared to my second.

 

There is no default drawcall value. There are just impacts from designing so that folks have poor performance. That's a choice. I was about to say most recent missions haven't gone near that, but let's actually take a look at those released this year (since this issue has come up before)...

 

Note, I have a wider field of view set, so my drawcall numbers are higher than typical:

Poets: 800-1200

Reputation: 600-1200 (but most typically 1000)

Breaking Out: 300-1500

Noble Home: 800-1400

InnBiz: 500-1600 (from making it, I do know a spot and direction to look which is 1900)

WS3: 300-1100

WS2: 900-1900

Gatehouse: 300-1400

Window: 400-2000 (my first mission, when I was more ignorant)

 

Interestingly consistent results considering I don't believe most people check drawcalls while mapping or have target goals. Note also, those peaks aren't areas that players spend much time, they tend to be large open areas with long sight lines and many, perhaps overlapping, lights.

 

Which reminds me, drawcalls aren't the final measure of performance, fewer lights will result in lower drawcalls, but there can still be an overload of tris being rendered, and shadows, that'll slow things down for certain hardware.

 

PS: 1500 wasn't a default value, it was considered a guide to stay below, with under 1000 a target to shoot for. If you consider 1500 your default, you'll find the performance poor enough to be noticeable and folks might comment. Much higher than that, and you are likely to only have your mission accessible to limited hardware.

 

PPS: For others who wish to create maps with lower drawcalls/better performance, proper visportals, limited sight lines and not overlapping lights are key. Using LOD (level of detail) and func_statics w/noshadows also helps. The wiki article, "Performance: Essential Must-Knows" has further details.

"The measure of a man's character is what he would do if he knew he never would be found out."

- Baron Thomas Babington Macauley

Link to comment
Share on other sites

Unfortunately, in order to move forward we will have to raise the min DC level for TDM up from D3 default of 1500 to 3000 to allow mission authors some wiggle room.

 

Wouldn't D3 defaults be ten years out of date by now?

 

Mappers definitely need to think about performance and should map with it in mind, but I get a little concerned when I hear people talking about having to redesign their maps to fall under a certain number of drawcalls. It reminds me of years ago when there were a number of people pushing mappers to caulk any unseen surface in their map to save performance. When we finally did some tests on it, it turned out there was no detectable performance benefit at all.

Link to comment
Share on other sites

  1. That's not moving forward! Designing smartly and efficiently is progress, being inefficient and suggesting folks spend money on different hardware is moving backward in my opinion. My first map had horrible performance compared to my second.
  2. WS2: 900-1900 & Gatehouse: 300-1400 I don't believe most people check drawcalls while mapping or have target goals, also those peaks aren't areas that players spend much time,
  3. they tend to be large open areas with long sight lines and many, perhaps overlapping, lights.
  4. Which reminds me, drawcalls aren't the final measure of performance, fewer lights will result in lower drawcalls, but there can still be an overload of tris being rendered, and shadows, that'll slow things down for certain hardware.
  5. 1500 wasn't a default value, it was considered a guide to stay below, with under 1000 a target to shoot for. The wiki article, "Performance: Essential Must-Knows" has further details.

  1. There is a point where we as mappers can't make the map perf any better and retain our original artistic intent. Fyi the rec. min. for D3 was a 2.0GHz dual core CPU, 3GB ram, NVidia 9800GT / ATi HD5750(which as Springs pointed out is 10yrs old). The issue some people are using rigs/laptop not far of that spec and its been said from the start TDM would require a higher min-rec-spec.
  2. You'll find there are a bunch of location's in both of those mission where the DC was at 2000-300 for sustained periods of time.
  3. i disagree, there are a bunch of maps where even with the best perfing possible (I know as I have helped or worked on 75% of the released maps) it was impossible to get tge DC down the FPS up. And you find that as time goes on this is going to happen more and more.
  4. Agreed, but it is a very good indication on a well optomized map.
  5. 1000 is nigh on impossible now, even a well detailed room will utterly spank this. And on the subject of DC I only notice a slowdown on my rig (GTX670/Corei5) when it hits 4000 and even then Its very mild because it only stops the game being butter-smooth which I myself prefer. Im not bitching, I usually do everying I can to help fellow authors get good perfing maps, but sometimes the little extra perft is not work amount of effort it takes to get it.

It reminds me of years ago when there were a number of people pushing mappers to caulk any unseen surface in their map to save performance. When we finally did some tests on it, it turned out there was no detectable performance benefit at all.

Caulk is only really required on hidden faces on FS when the DC in that location needs tweaking, otherwise yeah pointless - unless you like to be neat and tidy (or anal in my case)

 

Anyway, this is way off-topic - Springs feel free to move this to another thread.

Link to comment
Share on other sites

  1. 1000 is nigh on impossible now, even a well detailed room will utterly spank this.

Actually the rooms were at the lowest end of those numbers. It's big open outdoor areas with lots of lights that went into the four digits.

 

Gatehouse is a good example, most of the mission features rooms in the low hundreds, rarely breaking 500. Similarly compare WS3 (rooms at 300+) to WS2 (outdoors, with a very high lowest range). Poets is an exception, as I didn't go into the side rooms, and the main room features multiple overlapping lights. Go into the side room and you are looking at 128-137. Go upstairs to that big room with multiple non-overlapping lights and it's 400+/-.

 

And in a room you can typically only view a small portion of it, so it's hard to get many overlapping lights in view compared to a street view.

"The measure of a man's character is what he would do if he knew he never would be found out."

- Baron Thomas Babington Macauley

Link to comment
Share on other sites

PS: 1500 wasn't a default value, it was considered a guide to stay below, with under 1000 a target to shoot for. If you consider 1500 your default, you'll find the performance poor enough to be noticeable and folks might comment. Much higher than that, and you are likely to only have your mission accessible to limited hardware.

 

What's the problem in a mission being aimed at better hardware? Or the benefit in insisting that all maps run on crappy hardware? There are plenty of TDM missions that'll run on old hardware, for those who have it. People still using machines from 200x will generally be grateful to find some maps that run smoothly given their general experience of modern games, not surprised by some that won't. And there's obvious benefit to be had from having high-end maps in the TDM mix too.

Link to comment
Share on other sites

You might be reading beyond what I was saying?

 

What's the problem in a mission being aimed at better hardware? Or the benefit in insisting that all maps run on crappy hardware? There are plenty of TDM missions that'll run on old hardware, for those who have it. People still using machines from 200x will generally be grateful to find some maps that run smoothly given their general experience of modern games, not surprised by some that won't. And there's obvious benefit to be had from having high-end maps in the TDM mix too.

 

I have no problem with missions aimed at better hardware. As I stated, it just impacts who can access your mission, it's a design choice. (And obviously, my maps appreciate better hardware, having higher drawcall values.)

 

I don't know if anyone has insisted of such. My impression is there are plenty of missions that won't run on inexpensive hardware already, but I've never looked in to it. I have higher end hardware, so am relatively clueless in that regard. I also don't know if design changes would improve the performance of those missions.

 

However, suggesting missions should be designed to intentionally limit who can run them does seem silly to me, especially since we have had many high quality missions designed efficiently, which has the added benefit of not excluding the audience.

 

A ramification of encouraging poor performing maps that can only be run on "better hardware" is TDM reputation, if people try it and find they can't run things, or maps perform poorly, they won't recommend it to others, or be interested in designing for it. Similarly if die hard players can no longer play, community suffers again.

 

I don't believe either of those things are likely to take place however, given all the missions released this year typically have hundreds of drawcalls, rather than over a thousand, most places, and again, I don't believe most design around such parameters.

"The measure of a man's character is what he would do if he knew he never would be found out."

- Baron Thomas Babington Macauley

Link to comment
Share on other sites

Did anyone get complaints about performance in this year's missions?

 

If not, maybe that means people have been following lower limits than are necessary.

Link to comment
Share on other sites

If not, maybe that means people have been following lower limits than are necessary.

Well crunch time will be when Melan and I, release 'Penny Dreadfull 2' as several locations hit 3k and in one location only its cracks 4k with the average being 2000-2500 outside.

Link to comment
Share on other sites

Well crunch time will be when Melan and I, release 'Penny Dreadfull 2' as several locations hit 3k and in one location only its cracks 4k with the average being 2000-2500 outside.

 

That's just because all the lights overlap their neighboring lights in that map. If you remove them, drawcalls drop to 300-1800 range, with the average being around 1300 outside.

"The measure of a man's character is what he would do if he knew he never would be found out."

- Baron Thomas Babington Macauley

Link to comment
Share on other sites

That's just because all the lights overlap their neighboring lights in that map. If you remove them, drawcalls drop to 300-1800 range, with the average being around 1300 outside.

Dude, seriously.... <_<

 

Melan will come along and give you what for at some point, but untill he does here is what you are missing -

  • The map in question has big wide open spaces and thats after gentley leaning on Melan to cut down some of the sight lines. - this is the biggest DC culprit
  • The map has a lot of details in any given leaf, more than most other tdm maps.
  • The are hardly any over lapping lights, I have done my best to reduce this but in some case its unavoidable either because reducing the light volue would look shit or it would break Melans original artistic intent.

Link to comment
Share on other sites

  • The map in question has big wide open spaces and thats after gentley leaning on Melan to cut down some of the sight lines. - this is the biggest DC culprit

 

Yups, been there, done that. *points to my maps with their completely open spaces and resulting high drawcalls.

 

As you know, when I was contributing to the map was a while ago, so light tweaks since then I'm obviously ignorant of. As I said before, pretty much every light is overlapping it's neighbor, and I found yesterday's drop in drawcalls with lights removed fascinating. Obviously sight lines didn't change, and although I've been avoiding "looking" too deeply at the map to not spoil my initial play of it, I'm blown away and impressed by the design personally...all the ways sights lines are broken up nicely. I wish my brain could conceive of such convoluted, twisty passages like that.

 

In terms of your comment about artistry, more importantly with the overlapping lights is game-play in my opinion, reducing all the radii would mean larger pools of darkness, altering the chosen feel/sensations/difficulty.

 

But this is distracting from the main topic. I find it enlightening that so far all the maps this year have featured drawcalls predominantly <1000, with peaks 1500+/-. Even this supposedly higher drawcall example map is capable of that too, depending on lighting choices. Ah! Which gives me a potential script idea for performance...

"The measure of a man's character is what he would do if he knew he never would be found out."

- Baron Thomas Babington Macauley

Link to comment
Share on other sites

Let's say this. Draw Calls are not the end all be all of performance optimization.

 

You could, for example, convert a whole map into one large FS or Model and shine one large light on it.

This would result in something lower than 100 drawcalls depending on how many textures were applied.

But performance would not be great because your scene evaluation, fill rate, buffer usage and other performance factors would be blown out.

 

The problem here is that this is subject to the environment. One PC which has a poor GPU but powerful CPU would do a little better to break the scene

into smaller batches thus less load on the GPU "at once" while another PC (for the sake of extremes) with a poor CPU but powerful GPU would want as

few draws as possible since the GPU could handle big gulps of data.

 

On whole lower draws tends to be the more ideal for most configurations but this engine is very strange with it's CPU skinning operations saturating the PCIE bus with triangle data.

Indeed it may be better to instead to talk of systems with slow PCIE bus bandwidth and slow RAM than systems with slow CPU's since that is one of the main bottlenecks.

I would love to see some benches with a mid range CPU\GPU but with PCIE 1.0 vs 2.0 vs 3.0.

 

Yeah, light overlaps are hard to avoid and especially a problem when they afford such visual splendor as with a Melan and Biker mission. :wub:

 

I once speculated over at quake3world about the idea of automating the Strombine style lighting by baking a 3d texture of ambient

radiosity lighting then using some iterative script to slice it into 1D+2D projection images and radius values based on some performance metric

(more overlaps allowed = more slices, etc). Then you would only need to worry about real lights of which you would need fewer since

you would have visually striking ambient shading. I guess the need for this would be made obsolete if either the renderer had better light performance

characteristics or it supported light maps with probes or some other IBL ambient solution. (hence my bountysource tracker). Even if TDM gains a better

renderer for lights, I would still love to see someone build a solution to that concept (slicing 3D texture into 1D+2D textures with variable cube radii ).

 

(I wish Lunaran would make a TDM map to show off his Strombine technique, though Rich_is_Bored's example map is definitely cool.)

Please visit TDM's IndieDB site and help promote the mod:

 

http://www.indiedb.com/mods/the-dark-mod

 

(Yeah, shameless promotion... but traffic is traffic folks...)

Link to comment
Share on other sites

Insightful stuff, and I'd always wondered why some systems handled tris better than others, I always figured it was cpu/gpu without considering the bus as much.

"The measure of a man's character is what he would do if he knew he never would be found out."

- Baron Thomas Babington Macauley

Link to comment
Share on other sites

I only learned about this when I read the PDF about BFG's rendering improvements:

 

http://fabiensanglard.net/doom3_documentation/DOOM-3-BFG-Technical-Note.pdf

 

(Page 5)

 

I've been reading about Doom 3 since before it was out over at Beyond3d and the commentary was always about

how heavy the silhouette calculations were for the CPU. It seems everyone missed the point and the bigger problem

was huge memory copies of skinned vertex data.

Please visit TDM's IndieDB site and help promote the mod:

 

http://www.indiedb.com/mods/the-dark-mod

 

(Yeah, shameless promotion... but traffic is traffic folks...)

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

  • Recent Status Updates

    • Petike the Taffer

      I've finally managed to log in to The Dark Mod Wiki. I'm back in the saddle and before the holidays start in full, I'll be adding a few new FM articles and doing other updates. Written in Stone is already done.
      · 2 replies
    • nbohr1more

      TDM 15th Anniversary Contest is now active! Please declare your participation: https://forums.thedarkmod.com/index.php?/topic/22413-the-dark-mod-15th-anniversary-contest-entry-thread/
       
      · 0 replies
    • JackFarmer

      @TheUnbeholden
      You cannot receive PMs. Could you please be so kind and check your mailbox if it is full (or maybe you switched off the function)?
      · 1 reply
    • OrbWeaver

      I like the new frob highlight but it would nice if it was less "flickery" while moving over objects (especially barred metal doors).
      · 4 replies
    • nbohr1more

      Please vote in the 15th Anniversary Contest Theme Poll
       
      · 0 replies
×
×
  • Create New...