Source engine performance indicators

In HL1 we had r_speeds to use in the optimisation of our maps, so is there any similar tool available in the source engine? I know valve have said it based optimisations on performance as opposed to polycount, but does that mean we are meant to base our judgements on how the maps run on our own computers using cl_showfps? If anyone knows of other methods of judging map performance in the source engine, please share them here.

Your jumping on the HL2 editing fast. Try posting at the steam forums if no one comes up with the answer here. Im curious to know this command also. :biggrin:

I believe the command is +showbudget

budget?

While I don't quite understand what its telling me yet, it certainly
appears to be the sort of indicator I was looking for :biggrin: I guess a box
room really isn't going to give the feature much opportunity to show
its usefulness!

The box room.... ahhh, our first maps :biggrin:

I've had a dig around in CS:S and found quite a few tools to help mappers judge their level's performance.

First up is, as CJ said, +showbudget. This brings up a panel
that displays how much time is being spent per frame on each main
component of the game - physics, props, geometry, sounds, etc... It
also has an FPS graph that will show spikes when performance drops
significantly.

Next is +showbudget_texture. This displays the amount of memory
used on each type of texture data, such as cubemaps, lightmaps, and
world textures, per frame. There is also a variation on this,
+showbudget_texture_global, that shows the memory use of
textures across the entire map as opposed to the current frame.

We also have a dedicated interface for level performance, accessed
using perfui, but unfortunately it doesn't provide access to
everything we would find handy (such as the above tools). It grants us
portal rendering and occlusion rendering (more information about these
can be found in the
Source
SDK documentation), as well as displaying the fade distances on
props.

Another little feature that some will find handy is the
mat_leafvis command, that draws a red wireframe around the
current leaf that the player occupies. This will be handy in
identifying overly large leafs that grant the player too much overdraw,
as well as areas that have their leaves overly segmented.

So I guess where we once talked about "high r_speeds", we'll now be saying "our budget is too high"? Hmm... that's going to confuse some noobs

Leperous said:
So I guess where we once talked about "high r_speeds", we'll now be saying "our budget is too high"? Hmm... that's going to confuse some noobs

we can always add it to a political thread and say something like "DAMN another asswipe went and made another G dubya" map

Technically, the actual budget or target performance would stay the same, you'd just be over or under it. :smile:

So... does the r_speeds limit go above a few thousand?

/me is asking cuz he really wantz a headz start on HL2 mappin :biggrin:

The point is that r_speeds are no longer used, and performance isn't
only dictated by polycount. As source uses other technologies such as
normal mapping and shaders, you can't tell how well a map will run only
by the polycount. These new tools are a far better judge of the level
performance than r_speeds would be if they were available.

The point is that r_speeds are no longer used, and performance isn't
only dictated by polycount. As source uses other technologies such as
normal mapping and shaders, you can't tell how well a map will run only
by the polycount. These new tools are a far better judge of the level
performance than r_speeds would be if they were available.

Sounds really smart... I cant wait to lay my hands on a half-life2 copy.

I would really love to buy hl2 via steam but i dont have a VISA of MASTERCARD card..

So i cant play with the editor until the game is relseased =(.

And the most s**tty thing of all is that the game is released the same
day as the huge big important NATIONAL english test. Man, that really
sucks.

I appreciate the fact that we need to start relying on the 'budget' to analyse performance, but I can't help thinking that a well constructed map at the r_speeds 1/gl_wireframe 2 level will provide a better start point than a less well ordered one.

Having mucked around with the console, I think mat_wireframe and/or r_overlaywireframe equate to gl_wireframe 2, but no specif link to r_speeds (which sounds like it may now be irrelevant anyway),

CJ

Not really. Thing is, as texture resolution gets higher and higher, as there's more and more passes on any one texture (alpha channels, normal maps, spec maps, glow passes, etc.), and as other effects get more and more expensive, polys become cheaper and cheaper to the point where they're almost irrelevant. Unreal's current method of showing scene complexity by displaying a range of green to red based on overdraw and Doom3's r_showlightcount option is the type of thing that will be more and more indicative of performance than any raw number of polys.

I think I remember Valve saying they relied more on general fps on target systems, rather than something like r_speeds.

the taboo version still uses "r_speeds 1" gives a massive amount of info i do not understand.. no of it is familiar.. perhaps i could sneak in a screenshot and compare it to what they released with the budget command???

in the interest of science?? lep? gwil?

No, I don't care what the taboo version uses tbh, it has no consequence on anything...

fine then, no prob..

i was kinda thinkin the same thing, but also was thinking, "what did they change since"

i still think the stolen version was a ploy to extend their time limit for release, you know.. they claimed a release date they couldn't meet and needed to do something..

i don't know much about coding, but it seems to me that to use a completely different set of commands would be quite a feat.. and since the taboo version still uses HL1 commands and this new release does not.. shrugs

whether it was a ploy or not... it worked, they got their extra year :sad:

I doubt that Gabe et al are capable of tricking the FBI!

Leperous said:
I doubt that Gabe et al are capable of tricking the FBI!

yeah well, guess you're right, a simple games design team wouldn't have the skills i suppose. :/

BTW, wasn't it the FBI that determined that mouth sex is not sex if its the president on the receiving end??

not capable of deception, thats funny lep

i moved my post to a new thread

While using the +showbudget command the only thing that spikes signifigantly is the VGUI graph. It stays below 5ms but does approach it. What does this mean, is it bad?

Posts 2 and 6 in this thread answer that ...

Read the thread man! :razz:

Ha! edited my post so I dont look like dumbass. Now you guys just appear abusive. :razz:

<DIV class=quote>
<DIV class=quotetitle>? quoting nickster</DIV>
<DIV class=quotetext>Ha! edited my post so I dont look like dumbass. Now you guys just appear abusive. :razz: </DIV></DIV>
thats ok, we trust them more than you so it all works out :biggrin: