G450 and 3DMark and scores

[chip]

installed new via.my system specs-
kt7 raid-mode 0 stripping.
t-bird 1 gig.
384 megs pc 133 ram.
radeon 64 vivo-oc 200-200
dvd,50x cd,8x4x32 burner.
tv card.
sb live.
fm card.
10/100 lan.
win me.
all hooked up to LAN on home network with 4 other pc's-i own them all!!it's my gaming system!!friends come over-i frag them all-hehe.
1024x768 32 bit-5871 3dmarks.
THIS THING ROCKS!!!!!

------------------
IT'S NEVER FAST ENOUGH!

[CannyOne]

The one other small(?) fact that Himself fails to mention is the "auto-dithering" effect that occurs at any resolution above 1024x768! (The rest of us would call this 'blurry', but I'm really tryin to be nice.)

As for Chip, I'd post my GF2 scores, but I refuse to debase myself by posting non-Matrox statistics on this board. The one suggestion I do have for you is to install DX8 and the latest "special purpose" drivers. You should be quite happy with the results. Though your 3D Mark 2K scores will drop a little! A friend found the anti-alaising to be much improved amoung other things.

[Indiana]

Originally posted by chip:

1024x768 32 bit-5871 3dmarks.
THIS THING ROCKS!!!!!

I didn't post my benchmark results for the overall score (which is very low nowadays, when you consider that this is on a 1Gig CPU), but for the AGP texturig results.
Here, I think the G400 does exceptionally well (remember: the card has only 16MB and it's running "only" in AGP2x mode), can you beat the 136fps in the 64MB AGP-test with your Radeon/Geforces?

[Himself]

Textures, my scores are around 160, 150, 90, and 10, if you are interested, they never change, it's too boring to keep testing it. BTW, it doesn't factor into the final score, there is no point waiting for them to complete, press escape after the cpu tests, if you want those marks included. I barely have the patience to wait for the thing to complete as it is.

My scores from my G400 32MB on my old system were:

8MB Texture Rendering Speed: 228.2 FPS
16MB Texture Rendering Speed: 204.9 FPS
32MB Texture Rendering Speed: 127.3 FPS
64MB Texture Rendering Speed: 63.5 FPS

Bottom line, all the spec sheet BS aside, and the my finger is longer than yours crap, there are no games out there that a V3 can't handle. If there were I'd upgrade. It's pointless to chase shec sheets and benchmark numbers, what does it get you? Until there is software worth a damn, it doesn't matter if my video card isn't worth a damn. Like I said before, it's the same speed as a G400, when you factor out features you can't actually use, it's no better than a V3 anyway.

I never used 32 bit when I had my G400, too slow, don't miss it. It is one possible reason for me to upgrade, or fillrate in general. Then again, I wasn't impressed by the difference in quality in the games out there between 16 and 32 bit. Played the Sacrifice demo lately, didn't want to play the game much, the combats are too automatic, but the game looked good. Speed at 1024 wasn't an issue, no jerkiness, smooth and fast.

If you can see dithering at resolutions over 1024x768 on a 17" monitor you are obsessed with it. Grabbing that Alice demo now, screenshots for that look to be standard fare, corners everywhere, even on a GTS Ultra. BTW, 1600x1200 FSAA 4X was enabled in those screenshots, LOL! Talk about auto non features at higher resolutions. (and stupid webmasters).

Other than that, the drivers for the V3 are great, I'm rather picky with drivers, but the ones for the V3 just work, no screwing around. Hours of aggravation avoided is worth a dozen gee whiz features I can't use.

[Novdid]

The V3 and G400 are at same speed (not the MAX), BUT!!!! the G400 has a image qualty that beats the crap out of the V3, that´s what makes a good card. When I play games I want sharp images with colors that look great even in 16 bit color, considering that a V3 sucks bigtime compared to the G400.

[rubank]

Answer to Indiana:

Yes, even my MX can.
But it´s really not an interesting issue, since this benchmark doesn´t relate to actual gameplay performance. Try running MBTR demo benchmark and you´ll see what I mean: it uses textures heavily, and in spite of the 3Dmark AGP numbers, the G400 is practically unable to run MBTR at 10x7x32, even without enabling 24 bit textures. (Would be interesting to know about the Radeon here).

I´ve said it several times before: 3Dmark is a shitty benchmark whos grossly inflated numbers mean dick. But it looks nice (if you disregard the supposed-to-be people) and those 5000+ scores makes you happier than 5.000+ would, don´t they?

rubank

[Himself]

I'm tempted to put up an old G400 screenshot and say it's a V3 screenshot, just to see if y'all would rag on it.

Like I said, I don't miss anything about the G400. I thought I would, I expect to only have the V3 a few months, I bought it as a basic 2D card to hold me over until I upgraded. I was surprised at the lack of difference between the two.

[Casey Jones]

You are so clever. Your mom must be proud.


--------------------------------
Drivin' That Train To see Uncle John's Band, Smelling the Sugar Magnolias, With a Touch Of Grey In The Skies. Maybe I Should Watch My Speed.

[Ashley]

After 1000000s of hours of testing
here are some actual 3dmark2K scores at stock speeds:

G400 16MB SDR SGRAM @ 168MHz = 2859
G450 16MB DDR SDRAM @ 166MHz = 2551
G450 16MB SDR SDRAM @ 166MHz = 1702
(and for Himself )
V3K 16MB SDR SDRAM @ 166MHz = 3090

[yepitsme]

Okey, thank you very much everybody, but some of you (=all) missed the point here,
i think i know what i was bying at, i knew the scores i was about to get, all i asked for if someone of you would please send 3dMark score to compare with ( another G450 with SDR 16Mb? )

This card is running very well with 2D and
all the games i have played with ( toca 2 ,
H&D , CFS 2 etc. ) are running OK.

i just wanted to check it out

Thank You very much, Elvis has left the building.

yepitsme again

[Himself]

Ashley,

I guess those scores indicate the SDR version of the G450 is 64 bit or the DDR version is 128 bit. Odd. SDR at 128 bit should be very close to the DDR 64 bit scores.

[Greebe]

Himself, your getting warmer!

[Ashley]

These were store bought boards, and in the case of DDR I was giving the actual speed not the "effective speed":

G400 16MB SDR 128-bit SGRAM @ 168MHz = 2859
G450 16MB DDR 64-bit SDRAM @ 166MHz = 2551
G450 16MB SDR 64-bit SDRAM @ 166MHz = 1702

Overclocked LE version:
G450 16MB SDR 64-bit SDRAM @ 200MHz = 2064
(up from 1702)

The performance hit for DualHead:
G450 16MB DDR 64-bit SDRAM DH enabled = 2097
(down from 2551)

I don't usually spend time benchmarking, but with these 3 cards eveything, right down to the driver, is held constant. The variance here can be attributed directly to 64-/128-bit and DDR/SDR differences.

[Indiana]

Originally posted by rubank:
Answer to Indiana:

Yes, even my MX can.
But it´s really not an interesting issue, since this benchmark doesn´t relate to actual gameplay performance.

Of course 3DMark2000 scores have nearly nothing to do with actual game performance. They're useful to test the speed of selected operations, though.
It was claimed that the low AGP-texturing fps were due to a card having only 16MB onboard Mem and this is simply not true. And the score shows what performance you can get of a G400/16MB DH - which BTW is cheaper than a G450/16MB if you can still get it.

Besides (I have just rechecked on the MadOnion performance Orb) I haven't found a score of more than ~120fps for the 64MB texturing test even on the Geforce2GTS. Not so good values considering that those cards mostly had 64MB onboard RAM and thus don't really have to transfer large amounts over the AGP bus. On the Geforces you can always notice a large drop between the 32MB and the 64 MB test (this is the point where the card actually has to use AGP) - this is not the case on the G400, at least not here.

I'm this anal about this because it was one of the major annoying points I had with my old Asus V3400TNT, the "hickups" when textures were loaded (e.g. Unreal). Now, a friend of mine has a Geforce256DDR and I saw the same behaviour there (not so strong, but it was noticeable). Seems NVidia still has the same problems with the handling of large textures they had in the past.
It's not so nice having a framerate of 120fps going down to 15 fps for a fracture of a second...
I really would like to know if ATI does this better.

[rubank]

Well Indiana,

I won´t argue your point about AGP vs onboard RAM; I made that very comment myself earlier in this thread.

But I still think 3Dmark is a lousy benchmark, and I don´t even regard the texture rendering speed test as valid, since I always get considerably higher values in the 64meg test in 32-bit colour than in 16-bit colour (with my MX card). Whether this is true also with the G400 I don´t remember - I didn´t save those scores, other than in my inadequate memory, but maybe you can tell me if this is so?
This may also explain the numbers you relate to from MadOnions database, the default test is in 16-bit colour.

This behaviour could very well be a "feature" of the drivers of course, but then again - real games/apps speak another language. So I stand by my opinion regarding 3Dmark.
Prove me wrong and I will stand corrected.

And by all means, try the MBTR demo. It looks better than any other game, even as a still-image show, and you should feel right at home

rubank