cxm717 wrote:

It's a 256MB Parhelia. Its clocked at 250MHz. The 128MB Parhelia is clocked at 207MHz. I ran some benchmarks on the 2 Parhelias and the 256MB one is much faster, sometimes as much as 40%. I'm not sure why as its only clocked 20% higher.

Possibly a respin of the GPU silicon? Parhelia was rather buggy initially. Are both AGP 8x?

If you're really curious you can find everything about any Matrox card on the old Matrox Users Resource Center forum. You'll probably get a browser warning about it. They can't figure out why or how to remedy it. I used to hang out on there back in the G400 days.
http://www.murc.ws

I ran some more benchmarks on the G400 with high textures in quake3. To see if the slowdown bug is caused by the OS or hardware. All the tests were done with the highest textures (r_picmip=0), 32bit colour, 32bit textures. I ran them all on the same P3 I did the other windows 98 tests on. P3@1GHz in an intel d815eea board, although I added another 256MB stick of PC133 for the XP and 2k tests. https://docs.google.com/spreadsheets/d/1QO3Xd … dit?usp=sharing. I think it's a driver bug. Running the timedemo on the drivers that have the bug, the demo runs fine at first but it gets to a certain point and just slows right down. After the timedemo finishes even the game menu is very slow.

swaaye wrote:

Possibly a respin of the GPU silicon? Parhelia was rather buggy initially. Are both AGP 8x? […]
Show full quote

cxm717 wrote:

It's a 256MB Parhelia. Its clocked at 250MHz. The 128MB Parhelia is clocked at 207MHz. I ran some benchmarks on the 2 Parhelias and the 256MB one is much faster, sometimes as much as 40%. I'm not sure why as its only clocked 20% higher.

Possibly a respin of the GPU silicon? Parhelia was rather buggy initially. Are both AGP 8x?

If you're really curious you can find everything about any Matrox card on the old Matrox Users Resource Center forum. You'll probably get a browser warning about it. They can't figure out why or how to remedy it. I used to hang out on there back in the G400 days.
http://www.murc.ws

The 256MB Parhelia is an x8 card. The 128MB one is 4x. I pulled the heatsinks to have a look and yeah, they look different. https://imgur.com/a/gJDKATp. Honestly the only reason I bought the Parhelia cards is because I got a good deal on 3 identical old LCD monitors and Id always wanted to try out a Parhelia. Back when the Parhelia was released I already had a geforce3 or 4 (cant really remember tbh) in my gaming system. I'll check out that forum though. Thanks for the link

cxm717 wrote:

I ran some more benchmarks on the G400 with high textures in quake3. To see if the slowdown bug is caused by the OS or hardware. All the tests were done with the highest textures (r_picmip=0), 32bit colour, 32bit textures. I ran them all on the same P3 I did the other windows 98 tests on. P3@1GHz in an intel d815eea board, although I added another 256MB stick of PC133 for the XP and 2k tests. https://docs.google.com/spreadsheets/d/1QO3Xd … dit?usp=sharing. I think it's a driver bug. Running the timedemo on the drivers that have the bug, the demo runs fine at first but it gets to a certain point and just slows right down. After the timedemo finishes even the game menu is very slow.

OK so on 2000/XP 5.83 is as far as one should go for gaming performance, beyond that is no-no. So strange that such a bug would persist in the drivers for so long. At least it doesn't exist on Win98.

I was doing some testing with my G400 under windows NT4 yesterday and I found that the latest NT4 driver has the same bug. I'm going to see if I can find some older drivers and figure out which one is the last good driver. I tried the NT4 drivers off the G450 disk (got it from the vogons driver library) and it doesn't have the bug.

RichB93 wrote:

I think that over time Matrox just focused purely on driver stability as opposed to performance, as they sold (and still do sell!) these cards for basic display systems, such as signage and basic dual display output. Under Windows XP, I found that the whilst the 5.93.009 drivers are the last to offer Direct3D support, they cannot render into textures, causing 3Dmark 2001 to refuse to run! The previous 5.92.006 drivers do work, but I would be very interested to see how much better even earlier drivers perform.

Actually, the last driver under XP that supported D3D for the G4x0 series is 5.96.004. I keep a copy of it on my storage drive for this very reason. It's the last version viable for gaming and since I own a number of G4x0 cards as well as a G200, I keep it around.

edit... I just looked at Matrox's website, there's three different versions of the 5.96.004 driver. One, which I've never seen until today, explicitly states that it doesn't support D3D. It's dated Jun. 29, 2006. Of the other two, one is the WHQL version also dated June 29th, 2006 and the third one is dated May 8, 2006. I have no idea which one I have, but I've been using this exact same file since sometime in 2006 (I have a CD backup with it on there as well). I've run the version I have with plenty of games, from Unreal and UT99, to all manner of old games, demos of games, etc all of which use D3D.

This thread has me ready to yank the Radeon 7000 out of my P3 just to install my PCI G450 Dual Head card just to run Quake 3... And I'll do it tonight if anyone can point me to a how-to to get timedemos back. (I have the Linux Q3A, but using ID's instructions to get it to run under Windows. Just install the point release for Windows and copy PAK0.pak to base3A). I'll also install Unreal on it and take screenshots just to show it running D3D.

If you're running the newest version of Q3 demo4.dm_66 should be in pak6.pk3. I added the demo as an attachment though. I'm curious how the PCI version performs vs the AGP version.

Edit: It looks like the PCI version has SGRAM while the AGP has (DDR) SDRAM. Is this correct?

cxm717 wrote:

If you're running the newest version of Q3 demo4.dm_66 should be in pak6.pk3. I added the demo as an attachment though. I'm curious how the PCI version performs vs the AGP version.

You'll find out in a bit...

It's installed and the machine is up. It's a fickle card though, and I almost always forget about that aspect of it. It refused to boot to the card until I moved the soundcard one slot further away. The machine had XP SP1 installed, hadn't gotten around to updating it because I wasn't sure if I was going to keep it installed or switch to 98SE. SP3 is installing right now (forgot how long it takes on a P3). Once that's done I
ll install the driver for the G450 and for the soundcard, then install Unreal and Quake 3.

cxm717 wrote:

Edit: It looks like the PCI version has SGRAM while the AGP has (DDR) SDRAM. Is this correct?

Just saw this. I have it updated and the driver for the G450 and the AW744 are in place. This is the card's properties:

It's got DDR RAM on it.

Almost forgot:

As it shows here as well as in the card properties screenshot, D3D is supported and enabled by the 5.96.004 driver.

Working on Quake 3 now but I might not get to posting it until the morning.

swaaye wrote:

Dirk Daring wrote:

Thanks... I'm gonna give it a try later tonight...... not sure how much good it would do with my CPU bottleneck though, 🤣 I guess I could throw it into my Athlon 1800+ system but the TurboGL driver is Win98 only and that machine is on XP.

Having gobs of CPU power might bring more speed at low resolutions.

The forum for all Matrox knowledge used to be MURC. Matrox Users Resource Center. http://www.murc.ws. That place is a treasure trove of info for all Matrox products back to G100/G200.

sheesh... I tend to forget about murc. I was a regular member for years and whatever the Matrox Users site was before that in the late 90's as well. Login still works over there too.

and here are the results for Q3A on a G450 PCI card:

Pretty not good. 32fps at 800x600x32 with just about everything turned up. I think the measly 16MB of RAM on the card may have contributed to it slowing down on several sections, but I don't have a machine with an AGP slot open that I could use for an a/b comparison.

... Well... technically I do. I do have a Gateway E3200 SFF NLX desktop and I do have a pair of NLX form factor G400 cards (and the 8MB G200 I have is NLX as well), but they're both 16MB as well but AGP is supposed to alleviate that (ha!). Plus I'd have to put the machine back together which means taking the Slot I P3-750 out of the drawer it's in as well as the 768MB of RAM, then I need to slap a hard drive in it and install an OS... And I'm lazy right now.

I think what would help it significantly would be to install Win98SE on this machine and re-run the test using the 6.83.017 driver.

misterjones wrote:

This thread has me ready to yank the Radeon 7000 out of my P3 just to install my PCI G450 Dual Head card just to run Quake 3... […]
Show full quote

This thread has me ready to yank the Radeon 7000 out of my P3 just to install my PCI G450 Dual Head card just to run Quake 3... And I'll do it tonight if anyone can point me to a how-to to get timedemos back. (I have the Linux Q3A, but using ID's instructions to get it to run under Windows. Just install the point release for Windows and copy PAK0.pak to base3A). I'll also install Unreal on it and take screenshots just to show it running D3D.

You are my hero! Please do this, a direct comparison between the Radeon 7000 PCI vs G450 PCI using different drivers would be absolutely awesome!

I went back and re-read the Anandtech article. The one thing I wish they had done was state the settings used for Q3Test. If they had geometry set lower or lower quality textures it would definitely increase speed on this thing. But as I said before, this is XP and they were using Win98SE. The TurboGL client is built into the 98SE driver and is likely NOT included in the 2000/XP drivers since it wasn't available for NT4. The G-series likely works best for gaming under 98SE and best for professional 3D programs under NT-based versions of Windows.

I'll put the Gateway back together throughout the day and install Win98 on it. I'll probably be able to test sometime this evening. I'll even throw in the G200 to see what it's capable of under the Win9x driver.

misterjones wrote:

As it shows here as well as in the card properties screenshot, D3D is supported and enabled by the 5.96.004 driver.

Are you sure about the version of the driver?

derSammler wrote:

misterjones wrote:

As it shows here as well as in the card properties screenshot, D3D is supported and enabled by the 5.96.004 driver.

Are you sure about the version of the driver?

Yes. If you'd like I'll record a video of Unreal running under D3D on it with my phone.

Re-read what I wrote earlier: There's three versions of the driver on that page, but if you go to the Legacy section, you'll find the same driver that I use for the G4x0 cards and G200 for 2000/XP. Same version numbe, D3D is enabled and all.

https://www.matrox.com/graphics/en/support/drivers/latest/

The driver is found here: https://www.matrox.com/graphics/en/support/dr … p2k_596_004.php

It's the May 8th, 2006 driver that has D3D enabled. The release notes make no mention of D3D being disabled.

ftp://ftp.matrox.com/pub/mga/archive/2kxp/200 … tes_596_004.txt

misterjones wrote:

Actually, the last driver under XP that supported D3D for the G4x0 series is 5.96.004. I keep a copy of it on my storage drive f […]
Show full quote

RichB93 wrote:

I think that over time Matrox just focused purely on driver stability as opposed to performance, as they sold (and still do sell!) these cards for basic display systems, such as signage and basic dual display output. Under Windows XP, I found that the whilst the 5.93.009 drivers are the last to offer Direct3D support, they cannot render into textures, causing 3Dmark 2001 to refuse to run! The previous 5.92.006 drivers do work, but I would be very interested to see how much better even earlier drivers perform.

Actually, the last driver under XP that supported D3D for the G4x0 series is 5.96.004. I keep a copy of it on my storage drive for this very reason. It's the last version viable for gaming and since I own a number of G4x0 cards as well as a G200, I keep it around.

edit... I just looked at Matrox's website, there's three different versions of the 5.96.004 driver. One, which I've never seen until today, explicitly states that it doesn't support D3D. It's dated Jun. 29, 2006. Of the other two, one is the WHQL version also dated June 29th, 2006 and the third one is dated May 8, 2006. I have no idea which one I have, but I've been using this exact same file since sometime in 2006 (I have a CD backup with it on there as well). I've run the version I have with plenty of games, from Unreal and UT99, to all manner of old games, demos of games, etc all of which use D3D.

This thread has me ready to yank the Radeon 7000 out of my P3 just to install my PCI G450 Dual Head card just to run Quake 3... And I'll do it tonight if anyone can point me to a how-to to get timedemos back. (I have the Linux Q3A, but using ID's instructions to get it to run under Windows. Just install the point release for Windows and copy PAK0.pak to base3A). I'll also install Unreal on it and take screenshots just to show it running D3D.

That is a very peculiar card, I have never seen a G450 PCI without DVI out in the wild, and certainly no 16MB G450s.. Here's my card, 32MB with DVI out.

Is yours an OEM card?

misterjones wrote:

and here are the results for Q3A on a G450 PCI card: […]
Show full quote

and here are the results for Q3A on a G450 PCI card:

Pretty not good. 32fps at 800x600x32 with just about everything turned up. I think the measly 16MB of RAM on the card may have contributed to it slowing down on several sections, but I don't have a machine with an AGP slot open that I could use for an a/b comparison.

... Well... technically I do. I do have a Gateway E3200 SFF NLX desktop and I do have a pair of NLX form factor G400 cards (and the 8MB G200 I have is NLX as well), but they're both 16MB as well but AGP is supposed to alleviate that (ha!). Plus I'd have to put the machine back together which means taking the Slot I P3-750 out of the drawer it's in as well as the 768MB of RAM, then I need to slap a hard drive in it and install an OS... And I'm lazy right now.

I think what would help it significantly would be to install Win98SE on this machine and re-run the test using the 6.83.017 driver.

That's not that bad, actually it is about what I expected. 32MB version should be able to reach around 36FPS with those settings, and probably 40fps at 1024x768x16 which is roughly Voodoo 3 2000 levels of performance - that is a solid in my book. I have a 32MB version of this same card and look forward to testing it in my GX110 with a 1GHz P3 CPU myself.

To indulge my curiosity, could you test it in Q3 with the Radeon 7000 and post the results as well? 32FPS at 800x600x32 , more than I thought it would do.

derSammler wrote:

misterjones wrote:

As it shows here as well as in the card properties screenshot, D3D is supported and enabled by the 5.96.004 driver.

Are you sure about the version of the driver?

There is a WHQL certified driver with the same version that has D3D support.

That is a very peculiar card, I have never seen a G450 PCI without DVI out in the wild, and certainly no 16MB G450s..

The one with two vga connectors is actually more common (see http://www.matrox.com/graphics/de/products/le … /g_series/g450/). The one with DVI was called "G450 Dual Head DVI". I own one of each, but both AGP, not PCI. They also have 16 MB.

derSammler wrote:

That is a very peculiar card, I have never seen a G450 PCI without DVI out in the wild, and certainly no 16MB G450s..

The one with two vga connectors is actually more common (see http://www.matrox.com/graphics/de/products/le … /g_series/g450/). The one with DVI was called "G450 Dual Head DVI". I own one of each, but both AGP, not PCI. They also have 16 MB.

I know that there is a dual VGA AGP version, but I never saw a dual VGA PCI version, that's all. As for 16MB, you learn something every day? Never saw a 16MB G450 PCI before. Even Matrox's site lists the card as having only a 32MB option. It's crazy how availability of certain models must have varied by region.