Frag me --- that download I linked for the codec pack is 983 MB of bloatware + cruft + corporationary non-sense.

It wasn't like that a couple of years back.

Going to link a codec pack is **just the damn codec ***. 

input does lag somewhat when the stuttering occurs, but that isn't the main issue. visually, it looks like frame loss (like down to 10-20fps or so) but the fps counter doesn't budge. I've also experienced total loss of sound on one occasion, I don't know if it's related. 

And on this Windows 10 laptop I'm testing this out on, it doesn't even make the codec available as far as I can tell.

What nonsense. 

What does the Open GL version do? If the Open GL version doesn't have this kind of issue and the WinQuake one does, that tells me something.

Huge difference on how they update the screen.

On the Mac version of Mark V, it actually uses the Open GL method of drawing in the software renderer. 

no issues withe the gl version. Also, the stuttering occurs when viewing demos as well, this might rule out input being the culprit. 

Sounds like the buffer transfer of WinQuake is the culprit. What's your display resolution? 

1920x1080

I didn't mention this earlier because I have to do more testing, but the screen stretching may have an impact... or maybe it's more obvious when enabled, I'm not sure yet. 

http://i.imgur.com/mZrY3ws.png
http://i.imgur.com/1hujKqG.png

Looks like some of the flame polygons aren't being removed when the particle flame is enabled. 

You noticed ;-)

It's a temporarily measure that is supposed to be hard to notice.

QMB needs a flame.mdl without a flame. I can't roll up flame.mdl and include in the binary because flame.mdl belongs to id Software and is not licensed as open source.

My temp measure was to shave off verts above a certain point until I determine best way to have Mark V conditionally render flame.mdl in 2 different manners.

I will probably do this at some point in time by detecting unwanted verts via texture coordinates by haven't done this yet. 

If you are 100% sure GL that the GL version never has the issue, this is likely solved if the issue is buffer transfer.

I wouldn't overthink it.

I don't know when I'll have time to rework it, because it is a bit of a time consumer (might take 20-25 hours).

It's always been on my list do because then WinQuake gets Open GL's vid_vsync capability. 

That being said, you might try a lower resolution far closer to the original and it might work just fine right now.

1920 x 1080 x 24 bits per pixels is a lot of buffer transfer going on every second. Open GL was built for that. Windows API? Don't know.

Lower resolution might help quite a bit. Or might not. 

*Correction: Not every second --- 72 times a seconds or more. 

I've spent far less time in the gl version than the software version. I'll play the fuck out of the gl version tonight then get back to you. I don't want to give you false info thus sending you on a wild goose chase. I appreciate your time and this engine. I played both Beyond Belief and Dimensions of the Past with it and it was a real treat! 

I appreciate that. Wild goose chases suck! 

FYI, build 1009 does not understand the "install" command. 

Thanks for the headsup, Johnny. In this particular instance, I noticed the oversight after I uploaded it but since this is a temp build not linked as main download anyway, I figured I'd wait until I had the "non-temp build" done. ;-) 

If Mark V supports shaders, you may want to try Seven's new_torch model found here: http://quakeone.com/forums/quake-mod-releases/finished-works/11595-a.html
Looks great and works like a charm (at least in DP). 

Uh --- Mark V certainly does NOT support shaders. And certainly never will support them either.

Do you have the slightest clue what dwere was posting about? No?

Ok -- now get lost.

/You are nice guy, but my tolerance for incompetence in a technical thread is zero. Don't do it again. 

Sorry if I upset you Baker, I didn't mean to. *tiptoes out* 

I'm sorry I posted too intensely, I try not to do that, but there's no edit.

You are a good guy. 

I ran through an episode with the gl version and it performed flawlessly. It may be safe to say that this issue is winquake specific.

I'd like to draw your attention to another curiosity, not because I'd like it changed necessarily, just interested to know why it is:

I like to use a faithful q1 viewwmodel fix that tweaks some verts and alignments. As you can see in this comparison, the texture alignment appears to be off even though we know it isn't (in winquake). 

Yeah I doubted the entire Divx suite needed to be installed hah.

Anything you want me to test as far as my black capturedemo issue?

As far as the sound issue where there was noise...I think 11k has noise and 41k does not. But at 41k the sound is "airy" for lack of a better word. QS does not exhibit this behaviour but it looks like it uses different method for sound processing?

Unrelated but using your advanced demo playback features to watch retrojam5 demos is a godsend. Changes everything! 

As you can see in this comparison, the texture alignment appears to be off even though we know it isn't (in winquake).

Software Quake and GLQuake treat MDL texture coordinates slightly differently. This is a giant PITA for me, because a model tweaked for one set of engines will have broken alignments in another.

I rarely miss an opportunity to rant about it.

BTW, it's funny how even a stock id model works better in a GLQuake-like engine. 

software rendering uses affine interpolation when drawing mdls, because its simpler/faster.
whereas brush surfaces use proper perspective correction by recalculating the texture coords every 16 or 8 pixels (d_subdiv16).
(quake2 added perspective correction on its md2s. I believe this to be the reason why quake used bsp ammo boxes while quake2 was free to use md2s instead)

glquake has an 'gl_affinemodels 1' cvar setting that asks the driver to replicate software rendering's crappy interpolation. however its one of GL's 'hints', which means modern drivers just ignore it and always use proper perspective.

if you have glsl 130 (gl3), there is a 'noperspective' interpolation modifier which ensures the same appearance. 

I should probably add that I was talking about a different problem. So there are at least two ways for a model tweaked for GL to be screwed by software rendering.