Tested all Win9x drivers - new drivers major flop!

[Harri Hirvonen]

Yesterday I tested all Win9x drivers to G400 starting from 5.52...
The main reason was bad framerate on CMR2.
Test programs were 3DMark99 and 2000. Everything looked rather good from 5.52 to 6.10. Just some minor differences. 6.10 is the fastest and most mature one - and from now on the one I use.
But the new ones, which should be good for DX 8.0. were very very bad. Framerates went down 10 to 30%, mainly 15%. This ment going down from 60 fps to 47fps or from 59fps to 41fps (3DMark99).
Is there something wrong because the new drivers support G450 - is there something wrong concerning different memory pipelines?
Are these drivers crippling G400 128bit memory access??

Take my recommendation seriously, change back to 6.10!

Habba

[DosFreak]

Hang on tight Harri! Your about to be flamed!

[rylan]

Flame away. But I noticed the _same_ performance drop. So something is definately getting screwed up.
6.10 seems noticably faster than any of the newer drivers.

[xortam]

I found PD 5.52 to be the best for me when running Q3A though I didn't try 6.21 or 6.23 (due to others complaints). It really depends on your system, app, and resolutions/color-depth. Your mileage may vary.

[Kruzin]

I think it's pretty common knowledge around here by now that the last couple of releases of drivers have been slower for many people. If you have no problem with older drivers, and they run faster for you, then run them. Nobody is going to hold it against you. If you preffer the added features and fixes in newer drivers, run them.

You need to understand the path Matrox takes with their drivers. Speed optimisations are the last thing they do. More important than counting fps in games is fixing bugs. Newer drivers fix bugs that where there in older drivers. Often fixing a bug will slow something down. Once the bugs are ironed out, they then spend more time on speed, and adding new features.

5.52.015 was pretty late in the PD5 series. By that time, they had ironed out many problems, and spent time tuning the drivers more. PD6 is a whole new beast. They are still working out bugs and getting the features they want in it running. We are only on 6.23.005 in the PD6 series. I'd bet that by the time PD6 gets into the 6.5x.xx range they will have matured quite a bit (just like 5.5x did over 5.2x).

[Rags]

DosFreak,

You seem like you have a relatively good amount of intelligence judging from your posts here (besides the out-right flame bait), and the posts at Ars. Why is it you have a hard time to see why certain posts will get get negative reactions and some will get positive?

Harri's post was very clearly not flame bait, and I would think will not get flames headed towards it. Why? Well, because he offered some real information without the sarcastic remarks that seem to saturate your posts lately. With his post, we have a good foundation for an intelligent discussion about the subject at hand. The ones that leave little room for discussion, and a lot for bantering are the ones that people will flame for. I take it you must not get out much, because if you did, you would see that when dealing with people in real life, you have to talk a certain way so you won't come off the wrong way. Or maybe you do act this way, if this is the case then this is the reason why people shy away from you, because you come off as an arrogant fool. Discussions good! Proclomations bad!

Rags



[This message has been edited by Rags (edited 27 January 2001).]

[rylan]

I'm not complaining about M fixing bugs in the new drivers. And to be honest, I don't count fps in games either. I just go by how it looks, and if its smooth enough to play without getting annoying. I guess whats bothering me about the new driver is that we don't even know what the bug fixes are (are there releases notes someplace yet? ), and performance still dropped. I can definately understand that fixing a bug can lower performance, but it does get frustrating when the drivers get progressively slower over several releases.

[DosFreak]

"I think it's pretty common knowledge around here by now that the last couple of releases of drivers have been slower for many people."

Rags,

This is what I was referring to. I've noticed through the past couple of driver release older members telling people to stop running benchmarks and start playing games. The comments were WAY less subtle than Kruzin's tho and far less kind. I don't think the G400 can afford to lose MORE of the speed since it has so little already! I know that there are MANY things a driver team could do to speed up these drivers so why aren't they? That's what I mean when I said he was about to get flamed. Flamed by Murcers for daring to worry about games/benchmarks running slower!

*No, I don't wanna hear any BS about how hard it is to make a driver. From the ground up? YES. But to optimize a driver it shouldn't be THIS hard. I've always been impressed with Matrox's Window 2000 drivers but I HIGHLY doubt these drivers are as optimized for speed as they could be.

[Rags]

Like I said DosFreak, it's all in how the post is worded. If there is no room for discussion, then what do you excpect?

Rags

[xortam]

SW, in general, will go through cycles where one or more releases are put out to address bug fixes and then there will come a time where a release will need to concentrate on performance improvements. Many times, bug fixes will degrade performance because you're adding code to catch these exception conditions. These bugs can be due to simply an oversight, implementation error, or a defect in the architecture. It difficult to schedule time to rearchitect a product or feature. If Matrox now has a reduced engineering force, it may be difficult for them to expend a lot of resources on sustaining the current HW while developing drivers for the new HW. I generally stick with drivers that provide the fastest Q3A framerates because I don't see most of these bugs that have reportedly been fixed.

P.S. Looks like I reiterated a bit what Kruzin said but I expect it helps drive home the point.

[This message has been edited by xortam (edited 29 January 2001).]

[DosFreak]

"If there is no room for discussion, then what do you excpect?"

Bickering and angry replies!

[TdB]

one of the fixes in the new drivers pd 6.23
is that mirrors and teleporters in quake3 is now shown correctly. take a look at q3dm0

[antdude]

<font face="Verdana, Arial" size="2">Originally posted by TDB:
one of the fixes in the new drivers pd 6.23
is that mirrors and teleporters in quake3 is now shown correctly. take a look at q3dm0 </font>

What was the problem before? Reverse?

[prkl]

<font face="Verdana, Arial" size="2">Originally posted by Harri Hirvonen:
Take my recommendation seriously, change back to 6.10!</font>

I was already wondering how on earth my current system performs worse after hw & pd update... Ath500@700 (7x100) vs. T-bird 800@945 (7x135) result: 3DMark2000 scores went 3% down

[Oleg V. Kuzmin]

It is a little the information:
---------------------------------------
CPU Celeron 458 MHz (83x5.5)
Mainboard Abit BH-6
System memory 128 Mb
Matrox Millennium G400 32 Mb SGRAM drv. 6.04.029 + TurboGL 1.30.002
Q3 HQ
16 bit:
640 - 53fps
800 - 51.4
1024 - 38.1
------------------
CPU Celeron 458 MHz (83x5.5)
Mainboard Abit BH-6
System memory 128 Mb
Matrox Millennium G400 32 Mb SGRAM drv. 5.50.005
TurboGL 5.41.006.001
16 bit
640 - 51
800 - 49
1024 - 41
-----------------------------------------
CPU Celeron 458 MHz (83x5.5)
Mainboard Abit BH-6
System memory 128 Mb
Matrox Millennium G400 32 Mb SGRAM drv. 6.10.013
16 bit
640 - 55
800 - 49
1024 - 43




------------------
P!!!-900/128Mb PC133
WD200BB
DVD Pioneer DVD-115
SB PCI128
CTX - PR500F
MB GA-6OXM7E
MGA G400 SH 32MB