TurboGL?

[JackYi]

Where are the TurboGL drivers since Matrox's default one stinks? Where can I download them?

[Maggi]

http://www.matrox.com/mga/support/dr...us/prv_drv.cfm

but you could also try the ICD that was just released with their NT drivers ...

Seriouly, there shouldn't be a benefit when going back to TurboGL and it's also not officially supported when using Poweresk 6.x, although some got it working.

[Gurm]

The "default one" meaning the ICD?

It's more complete, and faster, than the TurboGL.

*shrug*

- Gurm

------------------
Listen up, you primitive screwheads! See this? This is my BOOMSTICK! Etc. etc.

[Pace]

What drivers are you using? I'm using 5.33 in Win2k I think, and they don't stink

I'll be trying Win98 soon as well.

P.

[xortam]

<font face="Verdana, Arial" size="2">Originally posted by Gurm:
... It's more complete, and faster, than the TurboGL ...
</font>

In many cases, the TurboGL drivers will be faster. Someday I'll have to post the results of my extensive benchmarking that demonstrates this.

[Liquid Snake]

Well, TGL is only for Win9x. Anyway, AFAIK, the only reason it's faster is because it omits some features, sometimes leading to decreased image quality. I find that it is actually quite fast if you use it with PD 6.x. But it seems that there are some bugs when used with PD 6.x, causing more features to be omitted, leading to more speed. But it's really buggy and I wouldn't recommend using it. It works fine with PD 5.52, but it's about the same speed as the newer ICDs found in PD 6.x.

[Indiana]

Why are most of the people thinking Q-III speed = OGL speed??? I found Matrox' newer OGL ICDs to get slower in most other apps except QIII and this while not really being able to run more OGL software, GLMame still failed like most OGL screenblankers, most OGL scene-demos,....

A test with Tirtanium rates the newer ICD MUCH slower than the old beta one, and the Win2k ones apparently are a joke.
O.K., Tirtanium is just a single application, like Q-III without necessary predictive value for other OGL apps - but this at least shows that while those newer ICDs seem to be well trimmed for QIII they're in fact slower in some other areas.
So I guess it's just up to the OGL programs you use which ICD suits you best.

[Rags]

Indiana,


You have just proven to me and others that know what's going on that you are talking out of your ass.

The ICD for win9x, Win2K, and NT are interchangeable now, that means there are only differences between ICD versions, no difference between OS's.


Please, if you are going to blow smoke, at least make it believable.

Rags

[omegaRED]

Does that mean that now we can use ANY OpenGL ICD ( from PDesk 6.50 Win98/Me, PD 5.52 W2K and forth ) for ANY OS?

------------------
Seth, are you ok ?
I`m peachy, Kate. The world is my oyster.

[Rags]

Bingo. They are all interchangeable. So if one ICD has a known fix that you are seeking, just pop in the ICD from the OS that had the fix.

Rags

[Indiana]

Sigh,



Apparently the newer ICD does perform worse in at least some cases - and, even if it's the same ICD, it does work slower in Win2k: 1024x768x32 in Win98SE is faster than 800x600x32 in Win2k.

So my advice still stands: performance of the different ICD versions is highly dependant on the apps/system config you're using, there is no "generally fastest ICD for all cases".
If the newer ICD doesn't contain a fix you need, you should test the speed of the various ICDs against each other and simply take the one that performs best for you.

[Maggi]

Hi Indiana,

did you benchmark all those drivers on the same revision of Tirtanium ?

Thing is, Tirtanium was chronologically getting slower with each release and thus a comparison between scores done on different versions is not valid.

Besides that is Titanium a bit CPU dependant, ie. accelerating the CPU results in higher score gains than making the vid card faster and thus it is not really a gfx card benchmark.

Try it the old fashioned way, using Q1 & Q2 ...

Cheers,
Maggi

PS: My Mystique G200 is still in that results browser's database, guess why it got so high scores ...

Cheers,
Maggi

[This message has been edited by Maggi (edited 31 March 2001).]

[omegaRED]

Hmm. Indiana, were the 5.41 benchmarks with the TurboGL drivers or the 5.41 ICD? Just to be sure. ~20 FPS difference is quite a big one.

------------------
Seth, are you ok ?
I`m peachy, Kate. The world is my oyster.

[Indiana]

@Maggi:
They were with the same version of Tirtanium (1.9) as well as the same CPU (a Duron800 oc'ed to 1036) with the G400 running at PLL 410, dividers at 2.5/2.0/2.5 core/mem/warp.
I know that Tirtanium is heavily CPU-dependant. this is how I noticed this in the first place: I was actually getting worse results in Tirtanium with the Duron@1036 than I got with my old SlotA Athlon600, and so I retested everything again.
I know that you get somewhat different results with the Quake-engines, but even there you can see in the low-res scores that the newer ICD doesn't scale too well with the CPU - it tends to give better results at the G400-limited benches (1024x768x32) than the old drivers, but worse results at the CPU-limited resolutions (640x480x16). Maybe this is just due to a bad 3DNow! implementation, I currently don't have an Intel to test it.


@omegaRED:
The 5.41 tests were done with the included ICD and with the combination PD5.41/old beta ICD. TurboGL doesn't work with Tirtanium

The main difference is between the 5.41 ICD and the 6.10 one, not between the old Beta and the 5.41. The old beta had some issues (trashed colors in 16Bit) while being only marginally faster. The PD5.52 ICD (which, unfortunately, is missing from this comparison) was even faster IIRC, so it's the one I would recommend, in combination with TurboGL for the supported games.


These results don't say much about OGL-performance in general (the same is true for Quake, btw.) but they at least show that there are some speed-issues with the current ICDs.

EDIT: the answer to omegaRED didn't really fit his question...


[This message has been edited by Indiana (edited 01 April 2001).]

[Admiral]

Hi guys, hope these Quake 3 benchmarks help
test setup: Quake 3 v1.25y, timedemo000, bilinear filtering, high geometric detail, texture detail bar to the right (high), lightmap lighting.
PD 6.50, G400 vanilla@150/200, 1xAGP, 256 aperture


ICD default settings:

800x600x16 60.1 fps
800x600x32 42.2 fps
1024x768x16 42.5 fps
1024x768x32 28.8 fps

ICD optimized for accuracy:

800x600x16 58.3 fps
800x600x32 42.8 fps
1034x768x16 42.3 fps
1024x768x32 28.7 fps

ICD optimized for speed:

800x600x16 59.3 fps
800x600x32 44.2 fps
1024x768x16 42.5 fps
1024x768x32 28.9 fps

TurboGL v1.3:

800x600x16 65.7 fps
800x600x32 49.4 fps
1024x768x16 46.3 fps
1024x768x32 31.5 fps

------------------
P3 650@900, MSI 6163 Pro, 256mb RAM, 45 GB IBM Telesto, 4.3 GB Quantum Fireball LE, 32X TEAC CDROM, 32mb DH Matrox Millennium G400, Vortex 2, 56k Rockwell HCF Modem, hp 970Cxi printer, 17" TM 4296-1, MS Natural Keyboard (PS/2), Genius USB mouse, Windows 98, DX 8.0a