G400Max/Voodoo3 3D WinBench99 results

[DrFizix]

I have compared a G400Max (32MB Dualhead) on my home machine (Asus P2B-F, 192 MB PC100) to a Voodoo3 3000 on my work machine (unknown 440BX chipset in a Gateway), with Pentium III 450 MHz CPUs on both systems. Both systems are running at factory spec (i.e., no overclocking). I realize this isn't a perfect benchmark because the systems aren't identical, but it will give you some information to go on.

G400Max (32MB), P3-450, 192MB RAM, 16-bit: 1120 3D Winmarks

G400Max (32MB), P3-450, 192MB RAM, 32-bit: 878 3D Winmarks

Voodoo3 3000 (16MB), P3-450, 256MB RAM: 16-bit: 847 Winmarks

There is of course no 32-bit score for the Voodoo3.

Hope this helps someone!

[Maggi]

mmmmhhhh ... sweeeet !!!

Maybe Matrox could use those to post 'em up to their 'Competitive Analysis' page ...

...

What settings did you use ?
(resolution, color depth, refresh rate, buffering)



------------------
Cheers,
Maggi
________________________

Working Rig:
Asus P2B-DS @ 103MHz FSB
Double Pentium III-450 @ 464 MHz
4 x 128MB CAS2 SDRAM
Matrox Millennium G400 32MB DualHead
Nokia 445Xi (21")
Nokia 447Xpro (17")

Home Rig:
Asus P2B-S Bios 1010 @ 112MHz FSB
Celeron 300A @ 504MHz
2 x 128MB CAS2 SDRAM
Matrox Millennium G400 32MB DualHead @ 150/200MHz
CTX VL710T (17")

[DrFizix]

Maggi,

The most interesting thing for me in this exercise was that, prior to buying the G400MAX, I saw a full-page advert for it in the Nov. issue of MaximumPC, in which Matrox claimed a 3D WinBench score of 880 at 32-bit (IIRC, I'm sure it was 88x-something). I ran the same test on my homebuilt, stock configuration without having to resort to heroic measures (i.e. disable this, clean-boot that, turn off wallpaper and sound, etc.) and got 878 in 32-bit. After all the hype and lies from Nvidia and 3dfx, that's a very welcome bit of truth in advertising. Way to go, Matrox!

On both systems I ran the benchmark at 1024x768 desktop; 3D WinBench runs most of the fps benchmarks at the desktop setting, i.e. 1024x768, although it switches to 640x480 for some of the primitive rendering tests.

I ran the test on the G400Max at both 16-bit and 32-bit depth at 85 Hz on my Sony 17" at home (thought the poor thing was going to die from all the resolution switching -- click, click, click), and I ran the Voodoo3 at 16-bit only (32-bit 3d rendering not available on that card) at a 75 Hz rate (had to run that rate because that machine has an LCD flat panel) at work. The buffering was also varied by the benchmark program between double and triple.

I know it would be a more clear comparison if I went back and installed my old Voodoo3 on my home system so it was running on the same motherboard as my new G400, but there's no way I'm going back to that fuzzy Voodoo desktop even long enough to run a benchmark.

The 3D WinBench test is a very comprehensive test of Direct3D capability (it better be comprehensive at a 75 MB download!), and the index it spits out is a weighted suite of approx. 30 speed tests in 3d scenes with varying buffering, mipmap settings, and texture sizes. It also generates a full table of framerates and Direct3D capabilities that can be ported to Excel.

I have an Excel spreadsheet of the comparison that I can e-mail you if you want to see all of the guts and eyeballs; it is approx 100KB. I just didn't want to post the whole damn thing here and incur the wrath of the MURCers (I'm a one-week newbie, only had the G400MAX that long, and I don't want to abuse the great discussion board here).

[xortam]

I was wondering if anyone else out there ever uses the ZD 3D benchmark: Finally someone to compare with. I'm a bit dissapointed to see that my nmbers are a little lower than your's on the same CPU (see specs in signature, all bios and drivers up to date (including G400 Max)). Also reported some O/C'ed #s (Matrox tweak utility).

100% clock 1024x768x32 @ 85 Hz -- 861
100% clock 1024x768x16 @ 85 Hz -- 1030
100% clock 1280x1024x32 @ 74 Hz -- 545 (refresh is limited by monitor)
100% clock 1280x1024x16 @ 74 Hz -- 746

110% clock 1024x768x32 @ 85 Hz -- 918
113% clock 1024x768x32 @ 85 Hz -- 934 (not quite stable)
110% clock 1280x1024x16 @ 74 Hz -- 805

Have seen 1280x1024x32 run at 545 and 573 on my first defective Max. I'll have to test some more on the new board to see if this number is again reached on the new card.
Note that lower resolutions are bottlenecked by the 85 Hz refresh rate restriction (unlike you, the highest I managed was 1110).
I would be interested in seeing you're 1280 resolution numbers (at my refresh rate would be really helpfull).


------------------

• ASUS P2B-S, PIII 450MHz, Award ACPI BIOS v1010, 128 MB RAM
• MYLEX FlashPoint RAID+ (BIOS v2.02N) running RAID 0 on two 9 GB IBM DDRS 39130D Disks
• Diamond MX300 sound card
• Matrox Millennium G400 Max Dual Head - English
• NEC 5FG monitor
• YAMAHA CRW4416S and NEC Multispin 3x CDs
• 3Com Fast EtherLink XL 10/100Mb TX NIC
• US Robotics 56K Voice FaxModem Pro
• Pioneer DVD-303S SCSI
• Note--All SCSI devices (except disk drives on RAID) are connected to onboard AIC7890 U2W SCSI
• Mainly running Win98 v4.10.1998

[DrFizix]

xortam,

Sorry, my monitor (Sony Multiscan 200sx) tops out at 85 Hz at 1024x768 and 60 Hz at 1280x1024. I'll be glad to run it at 1280 at that refresh if you want.

[DrFizix]

xortam,

I noticed in your spec list that you're running a Diamond MX300 card (Aureal Vortex 2). Some months ago, I tried a Vortex 1 (i.e., cheap) card and found that it was choking my framerates in Quake 2 and slowing down CPU speed testing in Wintune 98 (I know, sorry benchmark, but it's quick and dirty). I don't recall the specific number or percentage of frame rate drop in Q2, but I do recall that it was causing my P3-450 to report as a P3-415(???) in Wintune 98, where the "415" speed was based on a looped algorithm that measures CPU speed. I replaced it with an SB Live Value and everything was back up to speed.

I'm the last person around that would shill for Creative Labs, and I know the MX300 is the better card, but there may be something about it (esp. in combination with your ASUS P2B) that is slowing your AGP down. I don't have a clue as to what it could be specifically, but it's just a thought from my experience.

[xortam]

DrFizix, I don't think the sound card is at issue here. The ZD benchmark doesn't use any audio. I believe the benchmarks are designed to test only 3D video performance which in this case includes only CPU, memory, and the display subsytem. This would eliminate things like audio performance, larger available memory, faster disk access, etc.

I always set the resolution of the desktop to the same resolution that I want to run the benchmarks. I noticed you mentioned that you turned off you're wallpaper; I don't. I wouldn't expect that to matter but I can try a run in that mode. If you want to do some runs at 1280, I'll set my refresh to 60 and we can compare those numbers.

[carlmeister]

I have to agree with Dr Fizix, the the Vortex 1 has to much overhead on the processor. I installed a Montego on mine about 6 months and I was amazed to find it lowered my Norton Sys Info score (this dosen't use sound either) and dropped my Quake II framerates by 5 fps. So I yanked it out and gave it to my girlfriend and poped back in my trusty Esoniq Audio Pci card. I didn't hear that much of a difference to accept the performance drop. The Vortex 2 is supposed to be less cpu intensive, but I haven't seen any real proof to that.

[xortam]

I'm welling to try disabling my sound card and seeing if it makes any difference (but I'm not holding my breath). I'm not sure how quickly I'll get to some of these tests. I got pretty burnt out on trying different system setups and running benchmarks while I was trying to prove that my first Max was bad (it was).