Announcement

**Genom** · 8 December 1999, 15:00

128? kinda overkill me thinks. I know it's easier to borrow a million bucks than 100, but lets not go crazy here

BTW Himself, 5.30 is all ahm gonna be using fer the benchmarks as I said before, I am not breaking System Shock 2 agian for a benchmark that is obviously been optimized to be very biased towards GeForce cards. If there where more than 1 Bump mapping test and no T&L stuff I am sure the G400 would show higher scores than the GeForce. Just that these guys have laid more emphasis on a feature on one card than an another. Notice non of the tests do dualhead for example? something that I personally think is among the best G400 features. And also something that already some stuff has used (like Flight Sim). Yet it gets ignored. and T&L that is not used is put in. Kinda makes ya wonder no?

Just my 2 dimes, 3 pennies and a nickel.

------------------
------------------------
More hardware than any sane person should have.
-------------------------

**box** · 8 December 1999, 17:20

And an AMD result

Athlon 700@770, G400MAX, Default test: 3086 3DMarks

(not done pushing it yet)

**Scytale** · 8 December 1999, 17:35

No doubt all next gen cards will do T&L. It's erm... part of the pattern.

Now look at 3DMark2000 and think about how things will change in the coming 5 years or so...

**Himself** · 8 December 1999, 19:46

Genom,

Whatever you like, I just think that with a busted helicopter run through and a busted single textured run through your numbers are meaningless for comparison, anyway I was refering to the guy who posted before me.

How would you benchmark dual head? And wouldn't a EMBM benchmark be really just a G400 benchmark right now? Benchmarks like Tirtanium have had lots of detail in them for over a year now, they are no more biased than this one is, that we know of so far. It fills my requirement of being sensitive to overclocking changes, that means it can actually measure fillrate performance between cards unlike 3dmark99max.

BTW, you really should run through the time once with working drivers, that helicopter thing looks horrid otherwise.

On that topic, Anyone else notice single buffering looking like crap?

**Agent31** · 8 December 1999, 20:39

I did post the PIII enhanced score using a GeForce as well you may notice.

**alessandro** · 8 December 1999, 21:56

In 1024*768-16bit-z16 i'm getting:3DMark Result

--->1820 3D marks

CPU Speed
97 CPU 3D marks

Game 1 - Helicopter - Low Detail
40.2 FPS

Game 1 - Helicopter - Medium Detail
24.6 FPS

Game 1 - Helicopter - High Detail
9.5 FPS

Game 2 - Adventure - Low Detail
43.4 FPS

Game 2 - Adventure - Medium Detail
21.8 FPS

Game 2 - Adventure - High Detail
12.1 FPS

Fill Rate (Single-Texturing)
270.7 MTexels/s

Fill Rate (Multi-Texturing)
269.1 MTexels/s

High Polygon Count (1 Light)
1318 KTriangles/s

High Polygon Count (4 Lights)
899 KTriangles/s

High Polygon Count (8 Lights)
645 KTriangles/s

8MB Texture Rendering Speed
213.1 FPS

16MB Texture Rendering Speed
196.0 FPS

32MB Texture Rendering Speed
118.7 FPS

64MB Texture Rendering Speed
56.3 FPS

Bump Mapping (Emboss, 3-pass)
99.2 FPS

Bump Mapping (Emboss, 2-pass)
127.3 FPS

Bump Mapping (Emboss, 1-pass)
227.1 FPS

Bump Mapping (Environment)
89.5 FPS

------------------
Pentium II 450Mhz@504,Asus P2B rev.1011,17" Sony Multiscan 200PST,128MB PC100 ram,Matrox Millenium G400 MAX 32MB 5ns SGRAM,Quantum Firebal EL 10.2Gb,Epson Stylus Color 740,Sound blaster Live!,Cambridge Soundworks 5.1,Creative PC-DVD 5X,Ricoh MP7040S,Adaptec SCSI 2920C,Diamond SupraExpress 56e PRO,Iomega Zip Drive.

**Himself** · 8 December 1999, 22:30

Anid,

<h3>Here are some results at 1280x960x16,
note how AMD 3D-Now! mode is slower than DX7 software T&L mode.</h3>

SYSTEM
Software Versions:
Windows Version Windows 4 B, Build 1212 Win95b
DirectX Version 4.07.00.0700
Bios Version Award Modular BIOS v4.51PG
Bios Date 08/03/99

Memory:
Total Physical Memory 128MB
Free Physical Memory 56MB

PROCESSOR
CPU Information:
Processor Type AMD-K6(tm) 3D+ Processor
Processor Speed 448MHz
Processor Caps 3DNow!, MMX

Display:
Type: Internal
Width: 1280
Height: 960
Depth: 16-Bit
Buffering: Triple
Z-Buffering: 16-Bit
Refresh Rate: VSync Off
Texture Format: 16-Bit

<h3>CPU Optimization: D3D Software T&L</H3>

RESULTS
Platform: Internal
3DMark Result: 1455 3D marks
CPU Speed: 78 CPU 3D marks
Game 1 - Helicopter - Low Detail: 32.8 FPS
Game 1 - Helicopter - Medium Detail: 20.1 FPS
Game 1 - Helicopter - High Detail: 7.1 FPS
Game 2 - Adventure - Low Detail: 33.9 FPS
Game 2 - Adventure - Medium Detail: 17.4 FPS
Game 2 - Adventure - High Detail: 9.8 FPS
Fill Rate (Single-Texturing): 265.0 MTexels/s
Fill Rate (Multi-Texturing): 262.3 MTexels/s
High Polygon Count (1 Light): 975 KTriangles/s
High Polygon Count (4 Lights): 711 KTriangles/s
High Polygon Count (8 Lights): 542 KTriangles/s
8MB Texture Rendering Speed: 150.1 FPS
16MB Texture Rendering Speed: 142.8 FPS
32MB Texture Rendering Speed: 90.3 FPS
64MB Texture Rendering Speed: 47.9 FPS
Bump Mapping (Emboss, 3-pass): 73.9 FPS
Bump Mapping (Emboss, 2-pass): 92.5 FPS
Bump Mapping (Emboss, 1-pass): 158.3 FPS
Bump Mapping (Environment): 60.1 FPS

<H3>CPU Optimization: AMD 3DNow!(tm)</H3>

RESULTS
Platform: Internal
3DMark Result: 1367 3D marks
CPU Speed: 70 CPU 3D marks
Game 1 - Helicopter - Low Detail: 28.8 FPS
Game 1 - Helicopter - Medium Detail: 16.6 FPS
Game 1 - Helicopter - High Detail: 4.6 FPS
Game 2 - Adventure - Low Detail: 34.4 FPS
Game 2 - Adventure - Medium Detail: 18.9 FPS
Game 2 - Adventure - High Detail: 10.6 FPS
Fill Rate (Single-Texturing): 262.5 MTexels/s
Fill Rate (Multi-Texturing): 260.8 MTexels/s
High Polygon Count (1 Light): 892 KTriangles/s
High Polygon Count (4 Lights): 810 KTriangles/s
High Polygon Count (8 Lights): 725 KTriangles/s
8MB Texture Rendering Speed: 148.2 FPS
16MB Texture Rendering Speed: 142.1 FPS
32MB Texture Rendering Speed: 99.3 FPS
64MB Texture Rendering Speed: 51.8 FPS
Bump Mapping (Emboss, 3-pass): 70.1 FPS
Bump Mapping (Emboss, 2-pass): 88.9 FPS
Bump Mapping (Emboss, 1-pass): 155.4 FPS
Bump Mapping (Environment): 59.4 FPS

**[SITH}PELADO** · 8 December 1999, 22:39

1024 default settings
G400 DH (118.8 matrox tweak utility)
P3-450@630
2771 3dmarks

**alessandro** · 8 December 1999, 23:53

THERE IS NOT A T&L GAME YET!!!!!!

So any numbers getting from 3dMark 2000 it has not real valiue.
For the time been,3dMark 99 MAX is the bencmark to measure any graphic card.
I,m telling you this because 3dMark2000 is HEAVILY hardware T&L oriented ,thus 100% G-spot bias.

**alessandro** · 8 December 1999, 23:55

THERE IS NOT A T&L GAME YET!!!!!!

So any numbers getting from 3dMark 2000 it has not real value.
For the time been,3dMark 99 MAX is the bencmark to measure any graphic card.
I,m telling you this because 3dMark2000 is HEAVILY hardware T&L oriented ,thus 100% G-spot biased.

**Himself** · 9 December 1999, 02:22

<H3>Here are some results at 1024x768x16,
again, note how AMD 3D-Now! mode is slower than DX7 software T&L mode.</H3>

SYSTEM
Software Versions:
Windows Version Windows 4 B, Build 1212
DirectX Version 4.07.00.0700
Bios Version Award Modular BIOS v4.51PG
Bios Date 08/03/99

Memory:
Total Physical Memory 128MB
Free Physical Memory 52MB

PROCESSOR
CPU Information:
Processor Type AMD-K6(tm) 3D+ Processor
Processor Speed 448MHz
Processor Caps 3DNow!, MMX

Display:
Type: Internal
Width: 1024
Height: 768
Depth: 16-Bit
Buffering: Triple
Z-Buffering: 16-Bit
Refresh Rate: VSync Off
Texture Format: 16-Bit

<H3>CPU Optimization: D3D Software T&L</H3>

RESULTS
Platform: Internal
3DMark Result: 1575 3D marks
CPU Speed: 78 CPU 3D marks
Game 1 - Helicopter - Low Detail: 34.9 FPS
Game 1 - Helicopter - Medium Detail: 21.5 FPS
Game 1 - Helicopter - High Detail: 7.5 FPS
Game 2 - Adventure - Low Detail: 39.1 FPS
Game 2 - Adventure - Medium Detail: 18.1 FPS
Game 2 - Adventure - High Detail: 10.1 FPS
Fill Rate (Single-Texturing): 267.4 MTexels/s
Fill Rate (Multi-Texturing): 264.8 MTexels/s
High Polygon Count (1 Light): 1057 KTriangles/s
High Polygon Count (4 Lights): 749 KTriangles/s
High Polygon Count (8 Lights): 563 KTriangles/s
8MB Texture Rendering Speed: 215.8 FPS
16MB Texture Rendering Speed: 182.6 FPS
32MB Texture Rendering Speed: 103.9 FPS
64MB Texture Rendering Speed: 49.3 FPS
Bump Mapping (Emboss, 3-pass): 83.7 FPS
Bump Mapping (Emboss, 2-pass): 124.8 FPS
Bump Mapping (Emboss, 1-pass): 235.4 FPS
Bump Mapping (Environment): 94.0 FPS

<H3>CPU Optimization: AMD 3DNow!(tm)</H3>

RESULTS
Platform: Internal
3DMark Result: 1503 3D marks
CPU Speed: 72 CPU 3D marks
Game 1 - Helicopter - Low Detail: 30.3 FPS
Game 1 - Helicopter - Medium Detail: 17.3 FPS
Game 1 - Helicopter - High Detail: 4.7 FPS
Game 2 - Adventure - Low Detail: 41.8 FPS
Game 2 - Adventure - Medium Detail: 20.0 FPS
Game 2 - Adventure - High Detail: 11.1 FPS
Fill Rate (Single-Texturing): 267.7 MTexels/s
Fill Rate (Multi-Texturing): 265.2 MTexels/s
High Polygon Count (1 Light): 915 KTriangles/s
High Polygon Count (4 Lights): 848 KTriangles/s
High Polygon Count (8 Lights): 752 KTriangles/s
8MB Texture Rendering Speed: 230.2 FPS
16MB Texture Rendering Speed: 198.5 FPS
32MB Texture Rendering Speed: 119.9 FPS
64MB Texture Rendering Speed: 56.9 FPS
Bump Mapping (Emboss, 3-pass): 75.1 FPS
Bump Mapping (Emboss, 2-pass): 112.4 FPS
Bump Mapping (Emboss, 1-pass): 216.6 FPS
Bump Mapping (Environment): 94.7 FPS

PS. Tommy Two Times, 3DMark99Max sucks.

**Genom** · 9 December 1999, 13:47

Himself, please read my post carefully. I am saying that I think 3dmark 2000 is useless as it is obviously biased towards the GeForce. Why? well, it's the only card with T&L! DO you understand my bump mapping example now? The dual head reference was in the same vein. I will not place ANY importance on a benchmark that is designed to make one specific card shine. As it stands right now, 3dmark2000 is to me like running the geforce tech demos. It will highly inflate the numbers for that card, using stuff nothing else will. If I tried to run a becnhmark that relied heavely on EMBM and no T&L at all, I am pretty sure the results would be very skewed towards the G400. Is this a usefull benchmark? Not at all. See the point I am trying to get at? 3dmark99 is in my opinion (as far as I have one for becnhmarks anymore) more reliable, and even so, I find the exageration for SSE instructions to be a bit on the extreme side. However, TurboGL has shown how these instructions CAN increase performance grwatly in some aplications, so I am willing to let it slide.

This i not meant to be a flame. Just trying to get a point across as best I can.

**Himself** · 9 December 1999, 14:20

Genom,

You have have whatever opinion you like about the benchmak, who cares really? I am just pointing out that posting benchmarks with broken drivers is a waste of time and you are missing on on seeing the demo as it's meant to be seen.

**Jorden** · 9 December 1999, 14:33

That's the pool with EMBM, Himself... now show us yours (press F12 to make a nice piccy)

Jorden.

**DentyCracker** · 9 December 1999, 17:13

Me needs to get G400 of any sort soon

Announcement

3DMark 2000 has been released...

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment