A rather disapointing review:

[carlmeister]

Fullon3D: http://www.fullon3d.com/hardware/reviews/0899/G400/

Reviews a vanilla G400 on a Celeron, K6-3 and a Athlon against a TNT Ultra and of course shoots it down because he feels it isn't fast enough in Quake II and Unreal. Maybe the Acolytes of Matrox need to straighten him out of his confusion.

[mragsdale]

You must have read it differently than me. I came away with the review feeling that he was saying that the G400 and the TNT2 were the only 2 cards worthy of having in a newer system. He said he would down score the review if the drivers did not improve in the upcoming weeks.

I noticed he didn't think too highly of Tin Yau Lau's 3dfx crud, though.

Rags

[Kruzin]

What review did you read?
The one you linked to gave the card a score of 90%, and had comments like
"I like the board and would not hesitate to recommend it for it's well rounded performance as a 32 bit renderer and all-around work-and-play graphic card for high-end CPU owners. I could well imagine it on the 3D designer desktop."

The review said some very good things, and pointed out what we already knew...it's not the fastest OGL gaming card right now, although "the G400 has more than plenty of 3D power for the sub-500 MHz machines."

Screw Unreal. That's a hacked up pile of code that only really runs well on a couple of cards (the ones that paid Epic to package the game with the cards ). Wait for UT, then (if Epic gets it right this time), we will have something to talk about.

Q3? Thats not even a game yet. It's a pre-beta test. It will change 500 more times before it's a game, so why even worry about it now?

Plus, he was using 5.13 for his Q2 tests. Ant has already tested a newer ICD, and posted those scores a while ago, so we know the OGL scores are only going to get better

Anything said about OGL in any review has to be taken with a grain of salt. We all know Matrox are still in devepment on the ICD. He made that clear in his review, so I don't see any real problems there.

Seemed to be a fair review overall to me, which he noted would be updated when updated drivers are released...maybe he'll bump it to 95% then

[Joel]

I agree with Kruzin. What is so rather disappointing about this review? Also realize that he is reviewing the regular G400 at default setting 125/166 pitting it againist cards at higher setings. I think it does pretty good to hold its on all in all.

Joel

[The PIT2]

Interesting review however it doesn't seem to recommend the card for K6-2 and K6-3 see below

This may just be too low on K6-2s and K6-IIIs because of poor driver performance, so stay away unless you wish to upgrade to an Athlon later this year and want to buy a new 3D accelerator for it already (however much sense that makes).

But gives it thumbs up for 500 plus inc the K7.

[Maggi]

Hi Guys !

Already seen this one ?
http://www.extremehw.com/reviews/3dv...0/index13.html

It compares the regular G400 with a TNTU2 and some numbers are done with pretty high clock rates

The full review starts here:
http://www.extremehw.com/reviews/3dv...400/index.html

------------------
Cheerio,
Maggi

Asus P2B-S @ 112MHz FSB - Bios 1009 final
Celeron300A @ 504Mhz
128MB 7ns SDRAM
G400 DualHead 32MB SGRAM @ 201 MHz memory clock

[SteveC]

Maggi - now that's what I call a review! They seemed to like it....

------------------
Cheers,
Steve

My PC? Not that bad, got all sorts of crap in it, and all sorts of crap around it and my desk is also messy. Now what does that say about me? ;¬)

[carlmeister]

It was my take on his conclusion that was disapointing:

(The Millenium G400 is not the TNT2 killer that it was made out to be by pre-release hyping - not from Matrox, I mind you, but
rather from some overly excitable websites.

It falls short of the performance of the standard TNT2 in the most used real world benchmarks, which I largely attribute to a
lack of driver performance due to a 3 month advantage from NVidia - but who said the world is fair?)

And the fact he used the worst two benchmarks for the G400 (real world benchmarks?). At least he should of tried another D3D benchmark besides Unreal, I agree with Kruizin on that. I know our Quake II benches are improving with every new driver. Also the fact the he throws it up against a higher clocked Ultra and makes no mention of the fact that G400 Max is available.

------------------
Epox MVP3G-M, K6/2-450, 256 meg PC100, G400 SGRAM 32 meg single, Diamond Monster 3DII 8meg SLI, Ensoniq Audio PCI, WD 6.4 HD, Toshiba 32x, Toshiba 12x SCSI, Diamond Fireport 20, Yamaha 4x6 CD writer SCSI, Generic NEC2000 network card, Viewsonic E771 monitor Win95 OSR 2.1.

[franksch3]

A quote from his article:

"To closely analyze the G400’s 32-bit color performance, it is important to look at 1600x1200 scores, because 1600x1200 is the only resolution where the G400 is not fill-rate limited. When comparing the 32-bit performance to the 16-bit performance, we see a significant performance drop. On average, the G400’s performance dropped 25-35% when using 32-bit color, which is okay, but nothing to write home about. Since all the G400’s internal rendering is done in 32-bit color, one would think that the 32-bit color performance would be closer to the 16-bit color performance, but apparently not. "

1) He says that 1600x1200 is the only resolution where the G400 is not fillrate limited. Either this is a typo or he doesn't understand the basics.

2) Since all the G400’s internal rendering is done in 32-bit color, one would think that the 32-bit color performance would be closer to the 16-bit color performance, but apparently not.

He really really doesn't understand that the internal rendering has nothing to do with the speed hit. Wheter you calculate at 16 or even 64 bit doesn't suffer a speed hit.

The only limiting factor is the memory bandwidth.