Shack GeForce256 Benchmarks

I just think its sad that this is getting more of a response than the babe hunt.... get your priorities right dammit.

Hury up and post your specs Steve! :-) Meanwhile here\'s what I get with my V2 SLI Config:

Who cares about TNT2 vs. V3 - everyone knows that Canopus V2 SLI ruleZ! ;-)

System Specs:

Retail Malaysian Celeron 366 PPGA, Week 28 - OCÂ’d to 567 @ 2.0 volts

IÂ’m using an Alpha heatsink and GlobalWIn fan on the front of the MSI Slocket with a thin layer of thermal compound between the CPU and the heatsink. I am also using a brushless fan from Radio Shack blowing on the back of the CPU.

64 MB Meall PC-100 SDRAM (Cas-2)
Abit BH-6 Motherboard
Two 12 MB Canopus Pure 3-D II LXs O/C\'d to 100 Mhz with 1 brushless fan blowing over the cards.
8 MB ATI X-pert 2-D Card (GAY but $30)
Western Digital Caviar 5.1 GB EIDE Hard Drive
SoundBlaster AWE 32 Sound Card
Realtek 10-Base T Ethernet Card
USR Sportster 56K Modem
Enlight 7237 Mid-Tower Case w/250 watt power supply and 2 fans. (One fan sucks air in at the bottom of the case, and the other fan blows air out the back up where the power supply is.)

BIOS Settings: I am currently using the “LN” BIOS
CPU Soft Menu II
CPU Operating Speed: User Define
Turbo Frequency: Disabled
External Clock: 103 Mhz
Multiplier Factor: 5.5
Sel 100\\66 Signal: High
Speed Error Hold: Disabled
CPU Power Supply: User Define
Core Voltage: 2.0 volts

Results:

All framerates were measured with V-sync OFF, and utilizing the Autoexec.cfg files specified below:

Quake 1 Autoexec.cfg
gl_keepjunctions \"0\"
r_wateralpha 0.5
sv_aim 2
gl_triplebuffer 1
+mlook
gl_playermip \"0\"
gl_polyblend \"0\"
exec proxyautoexec.cfg
m_filter \"5\"
scr_conspeed \"4000\"

Quake2 Autoexec.cfg Settings
set cl_particles \"1\"
set gl_dynamic \"0\"
set gl_polyblend \"0\"
set gl_playermip \"2\"
set gl_flashblend \"0\"
set gl_ext_swapinterval \"1\"
set gl_swapinterval \"1\"
set gl_ztrick \"1\"
set cd_nocd \"1\"
set in_mouse \"1\"
set m_filter \"1\"
set in_joystick \"0\"
set gl_modulate \"2.1\"
brighter. default is 1
set vid_fullscreen 1
scr_conspeed \"4000\"

Quake 3 Settings
High Quality Sound
16-bit color
16-bit textures
Texture detail – Default
(4th black line from far right.)
Low Geometric Detail
Lightmap lighting
Bilinear filtering
Marks on walls: Off
Ejecting brass: Off
Dynamic Lights: Off
I.D. Target: On
High Quality Sky: Off
Sync Every Frame: Off

CPU Settings: 5.5 x 103 (567 Mhz)
Frames Per Second
Benchmark 800 x 600 1024 x 768
Q1 – Timedemo demo2 165.8 104.6

Q1 – Timedemo bigass1 123.6 99.6

Q2 - Demo1.dm2 139.0 100.8

Q2 - Crusher.dm2 71.9 54.6

Q3A – q3demo1 81.3 73.9

Q3A – q3demo2 62.7 57.3


Conclusions:
Now that IÂ’m running my Celeron at 567 Mhz life is better, but not by much. As you can see, at a resolution of 1024 x 768, I am still maxing out the fill-rate of my SLI configuration. At 800 x 600, performance is increased substantially, but not enough for me to drop my resolution down. So, until a really awesome next-generation graphics card comes out (meaning either Rampage or NV-15) IÂ’ll still be cruising with my SLI rig.

~KK

#267

Exactly.

FrO

#264, 68 fps @ 800x600x32 isn\'t good enough for you? Ha! I have to deal with an average of 25 fps @ 640x480x16 on an empty map!! Get a good FFA match going in Q3tourney with a 4-5 people and I\'m at about 10 fps. 68 fps @ 800x600x32 is heaven for me. You people have no idea how lucky you really are.

no its not

First post divisible by ((2^2)^2)^2!

#240 God dammit, GEFORCE SUPPORTS DIRECTX TEXTURE COMPRESSION IN HARDWARE. THIS HAS BEEN CONVIRMED BY NVIDIA EMPLOYEES. It\'s the rumor that just won\'t die.

#250 Based on what you posted, you are dangerously close to having no idea what you\'re talking about.. visibilty determination has zip, zilch, nada to do with rendering. If you are talking about depth complexity and overdraw, that\'s an architecture issue that all current 3D cards have to deal with -- except for the funky tile-based rendering architecture in the PowerVR and dreamcast. Who knows, maybe that is the future.. but the IMMEDIATE problem now is not fill rate (which is how you solve overdraw, just draw faster) but memory bandwidth which is gating current fill rate throughput anyway.

That\'s just great chud, now he\'s gonna start talking again. :(

Funny that people seemed to start making intelligent conversation around the same time creed stopped saying anything.

#247, well, I\'m only talking about the Virge cause I had one and did actually lose sleep over that damn thing. Never owned a Savage4 though, and from what you\'re saying I\'m glad I didn\'t. hehe.

CHUD, you think the S3 Virge gave S3 a bad name-- how about the current Savage4? They have user PETITIONS on fan sites for better drivers! I have a Savage4 for my wife\'s computer and it\'s pretty much a crapshoot what will run properly on it, even with all the latest drivers.

Why anyone listens to S3 at all, I have no idea. Their credibility is just about nil, yet all the news pages were eating up their \"800mpix\" Savage 2000 as if it was a shipping, tested product with stable drivers.

#240, fill rate is not the issue any more. It\'s memory bandwidth.

You have to read and write pixels FROM MEMORY. The data isn\'t just coming magically from inside the 3d chip somehow, it\'s textures in memory. The reason we haven\'t run into this brick wall so hard until now is because nobody has (significantly) broken the ~300mpix barrier until the GeForce-- *and* the memory bandwidth requirements are doubled at 32-bit!

Why do you think all these 3d company freaks are doing embedded memory, like the L1 cache on your celeron and P2 chips? It\'s the only fucking way they can get memory fast enough to *SUPPORT* those high fill rates.

In summary, you can quote super-high fill rates. But it won\'t mean dick in the real world unless you have MEMORY that can PROVIDE DATA fast enough for the chip to fill the polygons.

End of lesson.

Crap, didn\'t get to finish my last post, little keyboard mishap.

OEM 16mb TNT2 cards can be had for $80.. and every single TNT2 I\'ve ever used (about 6 total) will overclock to Ultra levels no problem (that is, from 125/150 to 150/175 or beyond). It\'ll generate more heat, but then again even 3dfx only put a heatsink on the V3-2000 and V3-3000.

Just be careful what kind of memory you get, because some of these guys put 7ns memory on their cards which is only rated to 1000/7 = 143mhz.

>>I just want to know why everybody thinks 3dfx is so evil when I can build a solid new system with a $80 V3 versus one with a $150+ TNTUltra that still won\'t work right half the time??? <<

Dude, you\'re on crack. Get an OEM 16mb TNT2 for $80 and overclock it.

#240, it\'ll take quite a bit to get me to support S3. It took me a long time to get over those nightmares I was having about my old 2mb S3 Virge.

#237, they\'re good cards, but 3dfx just doesn\'t ever seem to add very new features in their cards. Sure they keep getting faster, but the visual quality hasn\'t improved much in the last few years. I just think it\'s time for 3dfx to put an end to the Voodoo cards, and move on to something else. Just like the way Microsoft doesn\'t seem to want to give up with the Win9x kernel. It\'s like they make money off of a product, so they milk it to death, and it doesn\'t ever seem to improve much.

Why do people always get into 3dfx vs. nVidia fights? straight up, this is what I think of the entire thing:

instead of telling somebody that their TNT2U/V3 is a piece of shit and they should get the one that you have, you don\'t you think about their needs first. their needs could be different than yours, and this would call for a different video card. I built a system for a girl who was going to do NO gaming (or not much). I put a G400-16 meg in her machine. while another friend wanted hardcore gaming capabilities. I put a TNT2U/V2-SLI in his. it\'s all about what they need, and which card fits that need best. I\'m going to be in the market for a new Video card pretty soon, and the GeForce looks good to me (DDR ram is too much though).

essentially:
quit the bitching and fighting. everyone\'s needs are different.

FrO

I just want to know why everybody thinks 3dfx is so evil when I can build a solid new system with a $80 V3 versus one with a $150+ TNTUltra that still won\'t work right half the time???

Enquiring minds demand the truth!

And to anybody that wants to give me shit... blow me.

I still think the Geforce has some pretty lack luster fill rate. 20fps faster then our current TNT2u\'s? Come on $200 for 20fps. NVIDIA your going to have to offer me more then that if you want my money.

prive=price