Gfx Card Test
- Thread starter Doomulation
- Start date
sheik124
Emutalk Member
Reznor, ATI always has the floor in DX9, its like a law of thumb now 
All I am saying is that HomeWorld benchmark is rather intruiging, imagine if they used shadow-mapping for the faster cards? And the reason anand said NV30 would p3wn R300 is because most people at the time thought that NV30's dual vertex shaders or whatever the heck they're called would perform great, but nVidia failed to properly utilize them
All I am saying is that HomeWorld benchmark is rather intruiging, imagine if they used shadow-mapping for the faster cards? And the reason anand said NV30 would p3wn R300 is because most people at the time thought that NV30's dual vertex shaders or whatever the heck they're called would perform great, but nVidia failed to properly utilize them
Wouldn't even bother comparing r420 and NV40 without properly optimised drivers and knowing the prices. They are so equal it's not like there is a clear favourite this time. They perform within 5% of each other most of the time, each beating the other card by a small amount. The NV40 is a new architecture so the drivers are in their infancy stages compared to ATi's relatively mature drivers since the core is modified from the 9800 core.
On one hand you have ATi whose main goal seems to be having the fastest card in terms of speed of DX9 and AA/AF filtering for right now but at the cost of using lower precision for Pixel Shaders than NV40 and no SM3.0 compliancy.
On the other you have the massive nVidia card that has more raw power (no AA/AF) with more transistors but slower filtering and DX9, but has much better performance in OpenGL right now and has support for SM3.0 and uses 32-Bit precison for Pixel Shaders. Lots of people don't like the idea of having two molexes or using two slots for it, but then the GT version does not and has 16 pipes vs X800 Pro's 12. The NV40 also has that video encoder thing(?) which lots of people are interested in. I can't see a *clear* winner here. It's arguable whether SM3.0 will be actually useful, but no one really knows other than the handful of games annonced that will support it.
You could argue either way which of these strengths are more important, but it would be pointless since people all don't play the same games or want the same thing from a video card. Blindly favouring one over the other without taking into account all the facts from both parties = fanboy.
On one hand you have ATi whose main goal seems to be having the fastest card in terms of speed of DX9 and AA/AF filtering for right now but at the cost of using lower precision for Pixel Shaders than NV40 and no SM3.0 compliancy.
On the other you have the massive nVidia card that has more raw power (no AA/AF) with more transistors but slower filtering and DX9, but has much better performance in OpenGL right now and has support for SM3.0 and uses 32-Bit precison for Pixel Shaders. Lots of people don't like the idea of having two molexes or using two slots for it, but then the GT version does not and has 16 pipes vs X800 Pro's 12. The NV40 also has that video encoder thing(?) which lots of people are interested in. I can't see a *clear* winner here. It's arguable whether SM3.0 will be actually useful, but no one really knows other than the handful of games annonced that will support it.
You could argue either way which of these strengths are more important, but it would be pointless since people all don't play the same games or want the same thing from a video card. Blindly favouring one over the other without taking into account all the facts from both parties = fanboy.
Reznor007
New member
X800 series also has hardware video encoding, and tests show it uses less CPU power than 6800 currently. While 6800 supports FP32, the benchmarks don't show it. Basically every shader using program detects nvidia hardware and uses partial precision shaders, which go down to FP16 or as low as FX12, while Radeon cards are strictly limited to FP24(that's all it can do). I seriously doubt anyone could notice the difference in FP24/FP32 since the numbers are per component(96bit VS 128bit precision). The only time you could really see a difference is if the shader is extremely long, so much that both cards would crawl trying to run it.
And Clements, Geforce 6800 is actually derived from NV30. It's more of an upgrade than R360->R420, but it is still a derivitive design.
The biggest difference between the 2 is that people already own X800 Pro cards(Best Buy has already had them), whereas I don't think 6800 cards will be available until late May or June at the earliest. And no, I'm not a fanboy...my 9600 was the first ATI card I've bought...and that was mainly because it was a very fast card for the price($75 new at Best Buy).
And Clements, Geforce 6800 is actually derived from NV30. It's more of an upgrade than R360->R420, but it is still a derivitive design.
The biggest difference between the 2 is that people already own X800 Pro cards(Best Buy has already had them), whereas I don't think 6800 cards will be available until late May or June at the earliest. And no, I'm not a fanboy...my 9600 was the first ATI card I've bought...and that was mainly because it was a very fast card for the price($75 new at Best Buy).
Last edited:
From my personal perspective, I'd go for nVidia for my own reasons. I have no DX9 games at all, other than DX9 plugins (Direct64, Pete's OpenGL2, Dolphin) and stuff, which aren't exactly that GPU intensive compared to say, FarCry (games which I don't really play). I have a few DX8 games that work well enough on NV hardware, but I play more games in OpenGL whenever there is an option to, and I have a *lot* of OpenGL games since I prefer the API over Direct X at the moment, although I don't use Linux. I'm really into id games such as Doom and Quake that I missed the first time around, I play both via Doomsday/glQuake/Tenebrae which are in OpenGL. SM3.0 would be irrelevant unless an N64 plugin utilised them, then it'd be useful to me.
I'm into classic console emulation, particularly N64, so MSAA would not be as good as some super sampling modes to alias those alpha textures, particularly with Jabo's plugin. I don't video encode so that is also a moot point. ATi cards can have issues with PJ64 with the current drivers (not sure how many ATi owners this affects, but there's a fair few) so they can't work in 32-Bit mode.
So, from this I believe that from what I like to do on my computer, an nVidia card would provide the best solution rather than an ATi card. It may not be even worth me upgrading for this generation since the card I have serves all my retrogaming purposes and more, so I'm happy. These 6xxx are likely to be too expensive for me to afford, so I'd even be happy with something like a 5900XT if it is cheap. Me getting an ATi would be simply idiotic since it's OpenGL is inferior to nVidia at the moment, less compatible with the emulators I enjoy playing, and has DX9 performance I'll barely even use with not one proper DX9 game on my PC. So I've weighed up the options and made a decision.
I'm into classic console emulation, particularly N64, so MSAA would not be as good as some super sampling modes to alias those alpha textures, particularly with Jabo's plugin. I don't video encode so that is also a moot point. ATi cards can have issues with PJ64 with the current drivers (not sure how many ATi owners this affects, but there's a fair few) so they can't work in 32-Bit mode.
So, from this I believe that from what I like to do on my computer, an nVidia card would provide the best solution rather than an ATi card. It may not be even worth me upgrading for this generation since the card I have serves all my retrogaming purposes and more, so I'm happy. These 6xxx are likely to be too expensive for me to afford, so I'd even be happy with something like a 5900XT if it is cheap. Me getting an ATi would be simply idiotic since it's OpenGL is inferior to nVidia at the moment, less compatible with the emulators I enjoy playing, and has DX9 performance I'll barely even use with not one proper DX9 game on my PC. So I've weighed up the options and made a decision.
Falcon4ever
Plugin coder / Betatester
my ati 9600 pro...
nice app doom
nice app doom
Reznor007
New member
Clements said:
I know that nvidia's older OpenGL compliance is good, but have you seen their tests on the GLSL compiler test? It fails about 50% even on the newest drivers. You can download the program here http://www.3dlabs.com/support/developer/ogl2/downloads/glslparsetest.exe
Radeons score around 93-94%. This is relevent because Doom3 is going to use GLSL.
As for PJ64 with Jabo's plugin(and Rice's, and Orkin's)...works great here :shrug: Never had a problem with it. I also play Tenebrae(and other Quake builds...I've kind of started to like Darkplaces more than Tenebrae). nVidia's better OpenGL really only comes into play in professional programs though...for the most part ATI works just as good in games(and at the speeds both cards run GL games it doesn't even matter).
Yeah, OpenGL2 is still in it's early stages in the nVidia drivers so I'm not supprised it crashes. Pete's OpenGL2 plugin is the only thing on my computer that uses OpenGL2 (I don't like PSX, but I have the emulator) and by the time Doom 3 is released OpenGL2 support will probably be improved to ATi's current implementation. Older OpenGL "works good" on ATi but that's it. I might as well get nVidia card for the same money and get faster frame rates in my OpenGL games, and that's what I did, and I'm happy. ATi doesn't 'own' nVidia in everything.
cooliscool
Nintendo Zealot
Pete's OGL2 doesn't actually use OGL2, it's just a name for his next gen OGL plugin.
Reznor007
New member
Considering that my 9600(basically the low budget ATI card) is basically CPU limited at 1024x768 in Quake3(with an AthlonXP 3200+) I doubt an nvidia card would help. On a related note, I wish id would do a final patch for Quake3 that would enable SSE on AthlonXP's...the current code in it that checks for SSE will only check for SSE if the CPU ID's itself as an Intel
Tagrineth
Dragony thingy
Clements said:It really does use OpenGL2 now unless it's config is a lie
OpenGL2 isn't required for pixel shader support.
See, the thing is, Pete's OGL2 works, and yet OGL2 isn't even fucking DEFINED all the way yet, let alone released or anything like that. So how can it be using an API version that technically doesn't even exist yet?
Tagrineth
Dragony thingy
Reznor007 said:Considering that my 9600(basically the low budget ATI card) is basically CPU limited at 1024x768 in Quake3(with an AthlonXP 3200+) I doubt an nvidia card would help. On a related note, I wish id would do a final patch for Quake3 that would enable SSE on AthlonXP's...the current code in it that checks for SSE will only check for SSE if the CPU ID's itself as an Intel
400...FPS...NOT...ENOUGH...
MUST...HAVE...450...
I think the reason Carmack hasn't bothered with said patch is, simply, because Q3A's performance is already so ludicrously high.
The game isn't even particularly CPU limited on my 800MHz CuMine P3... let alone a 1.4+ GHz Athlon XP.
Reznor007
New member
My 3200+ maxes out around 230FPS in the default demo that comes in the 1.32 patch. I'm not really asking for more speed...just to use a feature that the CPU has. The current CPU capability detection is pretty crappy.
Once the Q3 source is released I'm sure it will be the first patch done to it.
Once the Q3 source is released I'm sure it will be the first patch done to it.
OP
Doomulation
?????????????????????????
- Thread Starter
- #37
Updating program to version 1.1.
See 1st post for the new version and changes.
I'll also need input on how the test performs on low-end cards such as built in intel crap, as I want to see some things.
See 1st post for the new version and changes.
I'll also need input on how the test performs on low-end cards such as built in intel crap, as I want to see some things.
OP
Doomulation
?????????????????????????
- Thread Starter
- #38
Right, so I was wondering really, what EXACTLY the plugins require to work. All I know is that intel = bad, gf2 mx = okay. And Direct64, no pixel shader = no workie, pixel 1.1 = bad, pixel 1.3 = okay, pixel 2.0 = good
So I know what they need... I could incorporate a feature that tells which plugins it's compitable with.
So I know what they need... I could incorporate a feature that tells which plugins it's compitable with.
MasterPhW
Master of the Emulation Flame
Clements said:
But the pixel shader in Pete's OGL2 only is working with an ATI card, aren't it? I remember that I was reading it in the readme file and tried out with an GF 5600 but doesn't worked...
BTW: the prog is great, I'll post a screen, if I'm back @ home
MasterPhW_DX said:
Worked for me, since I was using drivers that have preliminary OpenGL2 support. Before it didn't work at all, thus the note in the readme, but since there are drivers available that allow the shader effects. I still far prefer the original OpenGL plugin much better since AA works with it so the graphics are at least sharp (like most N64 plugins) and OpenGL2 either looks plagued with jaggies or is really blurry. That's just an opinion of mine.