• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Difference between PS3.0 and DXT5?

M

Maulenstein

Weaksauce
Joined
Jun 22, 2004
Messages
94
Alright, I want to figure out something.

I want to get the 6800GT, but ATI's 3Dc is still pretty bitching and actually have a few games supporting it. So it's keeping me a bit back.

But then there's the DXT5 people say, what exactly is it? Will the same games use it?

Thanks.
 
PS3.0 is Pixel Shader 3.0
Text
DXT is DirectX Texture Compression
Text

cant really compare the two because they are 2 differant things
 
I know 3Dc only applies to normal maps. But does DXT5 apply to normal maps as well ?
 
This should explain all your questions. There is even pics for visual comparison.

http://www.nvnews.net/vbulletin/showthread.php?t=30772&page=4&pp=15&highlight=3Dc

As you can see, there is very little benefit of 3Dc over DXT5.

Then check out some of the kickass games ATI is going to get 3Dc support in.

http://www.hardocp.com/image.html?image=MTA4MzAzNjk0NGJ3UHFVM2NZQWJfNl80X2wuanBn

Half Life 2 is the only game in that list worth mentioning lol. Far Cry is now added to that list of course.

3Dc is NOTHING compared to SM 3.0 support.
 
By compressing the textures at slightly higher quality, but also smaller size there should be a noticable performance increase in addition to the slight visual quality improvement (due to less bits needing to be transferred in overhead from videomemory to the GPU)

It should allow for example, a 128MB videocard to store maybe an extra 20MB or so of compressed textures beyond DXTC5. Or at least thats what I think its supposed to do, sort of like Z-compression it can dramatically speed things up.
 
its important to note 3dc is designed specifically for normal maps, which DXTC5 cannot compress without quality loss
 
There was even a rumor somewhere that 3Dc internal calculations were floating-point based, which would explain a heck of a lot as to why it looks "different" than DXTC 1-5. Of course videoprogrammers will probably never know what it uses, its just basically feed it a texture, and whatever pops out the other end (and at whatever quality) is usually what is used.

Getting really close to the texture. Thats just it though, when you die in UT2004 and your head is lying against the floor, the floor textures from that viewpoint are so large you can start to see the pixels. I would think/hope with a better compression scheme it wouldn't look quite so blocky.
 
NV40 said:
take a look a the real life diferences between Dxt5 and 3Dc...

what you see there is the *worst case scenario for non-3dc cards* using vanilla Dxt5. using very high resolutions normal maps.. This is the real "IQ advantage" of ATi vs Nvidia.notice that while there are slighly precision diference with shadows ..3Dc make thing slighty blurier and Dxt5 slightly sharp.

as you see this is not any way similar to ATi marketing (apples vs orange comparisons) that manipulates their screenshots to make it looks like 3dc have significant advantages over very old compression techniques which is not the case.

as i said in the past.. all the real benefits of 3Dc can be done with Dxt5 ,you can use 4x times higher resolutions textures or save performance only diferences is that Dxt5 have slightly less precision ,that not necessarily means that will make things always uglier ,but in most cases slighty diferent. what should be noticed is that in a real game those diferences will be virtually invisible since in games you dont play with your character face atached to the floor. 3Dc is nice..an imporved version of Dxt5 (which ATI tried to push sometime ago before 3Dc), and its not anywhere what ATi is trying to sell. and like everything is have its limitations of where can be used or not.

non 3Dc hardware have the option to..

-use vanilla DXt5 with slighty less precision.. (or improved tweaks to to Dxt5 that will look identical to the human eye .. )

-no compression at all.. here you can get higher precision than 3Dc (since its using the original texture) but it might comes at the expense of more memory use.

-software emulation of 3Dc (which also was said to be possible)

Hl2 use of 3Dc is merely a PR marketing $$ move by ATI since that game barely use normalmaps and mostly is based on precomputed lightmaps .. FArcry and Doom3 will benefit more performance wise since those games really use normal mapps heavily. the point is that you will hardly see any slight diference in a [real game] if Dxt5 is used too propertly in non ATi cards.
Click to expand...

I stuck the parts that should be well noted in bold.

slainfoe said:
not gonna get into a ati vs. nvidia thing here but am going to state that i ran the demo on my 6800 ultra and was hard to notice any problems. when your playing a game your not going to notice little tiny dot out of place or something. same reason why do you need max AA at 1600x1200 when pic looks the same with no AA or low AA.
Click to expand...

DSC said:
SM3.0 is part of DirectX9.0, 3Dc isn't. If other 3d companies doesn't support 3Dc, then it's a dead duck. Remember Truform? 3Dc isn't even invented by ATI, it's DXT5 with their modifications. It's S3's invention(S3TC) and licensed to MS(DXTC).
Click to expand...

Brent said:
its important to note 3dc is designed specifically for normal maps, which DXTC5 cannot compress without quality loss
Click to expand...

Its not really a quality loss. Its just a different type of IQ. DXT5 gives it a sharper look while 3Dc makes it more blurry.

And ATI users think SM 3.0 is worthless. :rolleyes:

Here are some of the differences between SM 3.0 and SM 2.0b.

Dependent Texture Limit
2.0b = 4
3.0 = No Limit

Position Register
2.0b = none
3.0 = Yes

Executed Instructions
2.0b = 512
3.0 = 65536

Interpolated Registers
2.0b = 2+8
3.0 = 10

Intstruction Predication
2.0b = none
3.0 = Yes

Indexed Input Registers
2.0b = none
3.0 = yes

Constant Registers
2.0b = 32
3.0 = 224

Arbitrary Swizzling
2.0b = none
3.0 = yes

Gradient Instructions
2.0b = none
3.0 = yes

Loop Count Register
2.0b = none
3.0 = yes

Face Register (2-sided lighting)
2.0b = none
3.0 = yes

Dynamic Flow Control Depth
2.0b = none
3.0 = 24

Minimum Required Full Precision Shader Capabilities
2.0b = FP24 (96-bit)
3.0 = FP32 (128-bit)
 
Here are some of the differences between SM 3.0 and SM 2.0b.
Click to expand...

Those are just the differences between PS2.0B and PS3.0 I posted btw. There are likely more between VS2.0B and VS3.0 but I haven't seen a solid spec sheet on vs2.0b yet.
 
Brent_Justice said:
Oh that is comedy gold right there.

Mind If I use it in my sig? heh
Click to expand...

You didn't quote the rest of what i said. 3Dc makes the image look more blurry while DXT5 makes the image more sharper. That doesn't mean DXT5 is ugly compared to 3Dc and i dont think there is a chance in hell you can see any difference AT ALL in-game.

NV40 said:
what you see there is the *worst case scenario for non-3dc cards* using vanilla Dxt5. using very high resolutions normal maps.. This is the real "IQ advantage" of ATi vs Nvidia.notice that while there are slighly precision diference with shadows ..3Dc make thing slighty blurier and Dxt5 slightly sharp.
Click to expand...

Those pics were a WORST case scenario using vanilla DXT5 at VERY high resolutions on NORMAL maps lol.

NV40 said:
Hl2 use of 3Dc is merely a PR marketing $$ move by ATI since that game barely use normalmaps and mostly is based on precomputed lightmaps.
Click to expand...

And there will be little if any performance increase in it over DXT5. Especially in Half Life 2.

Noone wanted to use DXT5 and if 3Dc is only marginally better, why is it such a big deal now?

SM 3.0 at least has alot of REAL IQ and performance abilities. 3Dc has very little over the existing DXT5.

Using 3Dc also has lower precision then using no texture compression at all.
 
In a worst case scenario in an ATI employee's graphic demo creation, dxt5 has minor artifacting while 3dc doesn't. Difference is slight, but it is there - will probably be very subtle in most games, similar to the differences between ati's brilinear and nvidias full trilinear filters. Definitely not worth giving up Nvidia's FP16 blend HDR Lighting capabilities for though, that will be very noticable. In a real life game, if DXT5 caused the minor artifacting like you saw in the ATI/Humus 3dc/dxt5 demo, the dev would simply not compress that particular texture.
 
DXTC and 3Dc have nothing to do with pixel and vertex shading (and should not interfere). Texture compression is a fixed function add-on introduced with the good old (but slow) S3 Savage, long before DX8 and 9 even showed up.

Uncompressed most of the time is just plain ugly, as you just can't put the detail in that you can get with a similarily sized compressed image. Even the original S3TC/DXTC1 was superior according to many people, and it took up 3-4 less space than an equivalent 16-bit uncompressed. I can only imagine what level 3Dc is at now.

http://www.digit-life.com/articles/reviews3tcfxt1/

Its a highly, highly underrated capability. One again, ATi has does a stupendous job of not mentioning this quite important technology leap to anyone.
 
ZenOps said:
Uncompressed most of the time is just plain ugly, as you just can't put the detail in that you can get with a similarily sized compressed image. Even the original S3TC/DXTC1 was superior according to many people, and it took up 3-4 less space than an equivalent 16-bit uncompressed. I can only imagine what level 3Dc is at now.
Click to expand...

Not compressing a few particularly troublesome textures while the majority are compressed will do nothing against performance or IQ.
 
True but with 3Dc, supposedly there is nothing left that will not compress. The example ATi gives with DXTC1-5 was with textures on globe type surfaces (curved along two axis) it would tend to "block up". It doesn't anymore, which means the entire scene can be rendered with 3Dc without having to worry about and check for uncompressed textures mixed in (which should give another healthy performance boost beyond just the better compression.)
 
ZenOps said:
True but with 3Dc, supposedly there is nothing left that will not compress. The example ATi gives with DXTC1-5 was with textures on globe type surfaces (curved along two axis) it would tend to "block up". It doesn't anymore, which means the entire scene can be rendered with 3Dc without having to worry about and check for uncompressed textures mixed in (which should give another healthy performance boost beyond just the better compression.)
Click to expand...

That may be true, but I don't see developers dumping DXT5 anytime soon when it is the primary format used by the #1 video card chipset maker. So x800 owners will mainly just get a minor performance benefit out of 3dc. I wouldn't be surprised if Nvidia introduces their own texture compression standard in NV50 to attempt to prevent 3dc from becoming a standard.
 
ZenOps said:
Its a highly, highly underrated capability. One again, ATi has does a stupendous job of not mentioning this quite important technology leap to anyone.
Click to expand...

WTH are you talking about? lol

3Dc has been mentioned in nearly every X800 review and ATI even has slides made up to show the upcoming games that will support it lol.

http://www.hardocp.com/image.html?image=MTA4MzAzNjk0NGJ3UHFVM2NZQWJfNl80X2wuanBn

ATI might not beat their drums as hard about 3Dc as nVidia does SM 3.0 but its definitely been mentioned. They even talked about it in the last Crytek interview.

Maybe ATI doesn't make such a big deal about it because of the fact it has very little benefit over DXT5, and in games like HL2 there will be little to no performance increase.

Or maybe because they tried to push DXT5 and noone wanted to use it.

Or maybe because ATI is just using something that they didn't even really come up with.

3Dc isn't even invented by ATI, it's DXT5 with their modifications. It's S3's invention(S3TC) and licensed to MS(DXTC).
Click to expand...
 
I don't think people realize how DXTC/3Dc works. You don't need 6 different texture sets to support DXTC 1 through 5 and 3Dc, you just need the one original uncompressed texture set at maybe twice the resolution of what you intend to finally have. And change one line of code that specifies the textures to go to the 3Dc input instead of the DXTC1-5 of the older cards.

Whatever the videocard supports, it will compress and then decompress in realtime and use. Which means that you if you can squeeze 300MB of uncompressed textures into at 128MB card using DXTC5, you might be able to squeeze closer to 350MB using 3Dc, have better performance and have it look better too.

What tended to happen with some early games, is that they precompressed the textures into DXTC formats (the second disk of the original UT is a good example)... Its not necessary to do that anymore, it was never necessary to do that actually.

PS: I think the reason Nvidia never mentions DXTC(1) is because of the "Quake ugly sky banding" issue that really never got resolved until about the GF4 era. Its completely invisible now, but is an extremely important part technology. Anyone remember this driver settings page:

http://s89930274.onlinehome.us/dds.jpg
 
ZenOps said:
Whatever the videocard supports, it will compress and then decompress in realtime and use. Which means that you if you can squeeze 300MB of uncompressed textures into at 128MB card using DXTC5, you might be able to squeeze closer to 350MB using 3Dc, have better performance and have it look better too.
Click to expand...

Yea but were not running 128MB cards any more. Were using 256MB cards and soon to be 512MB cards.
 
Well, I'm sure it will.

End-users never see it, but its there, programmers know that better compression is the holy grail, and longer shaders are just a pain in the ass. Just like in 3Dmark2001, when people install it, don't you ever wonder why the textures have to "be generated" and not simply "copied over"? That is the proper way, so that when hopefully NVidia gets a better compression scheme than ATi, it will work with all the earlier games too, be faster, look better and take up less memory.
 
ZenOps said:
Well, I'm sure it will.

End-users never see it, but its there, programmers know that better compression is the holy grail, and longer shaders are just a pain in the ass. Just like in 3Dmark2001, when people install it, don't you ever wonder why the textures have to "be generated" and not simply "copied over"? That is the proper way, so that when hopefully NVidia gets a better compression scheme than ATi, it will work with all the earlier games too, be faster, look better and take up less memory.
Click to expand...

Generating textures (procedural) at load still takes up the same amount of vram as precompressed textures. Just procedural textures save hard disk space. Even procedural textures have to be stored somewhere you don't want to generate them on the fly when the program is running, really slow things down, Main use of procedural textures are for fx and control of that texture at the pixel level is alot easier.
 
JKA Represent said:
I dont think 3dc is really going to matter that much. It will help, but not that much
Click to expand...

There is a benefit for it but the problem is other people have to adopt it into thier cards :/ which for a proproitary method like this is rare. If it was a part of direct x or ogl then it would be.
 
What I'd really like to see is a programmable texture compression engine, now that would just kick ass on everything.

If a game developer then had a game on one DVD with quadruple sized uncompressed textures instead of say one CD with regular sized textures:

You could set up a software (CPU based) compression engine that would continually compress the textures with ever increasing levels of quality and speed based on how much time you had to spare and the GPU and CPU capability you have... Sort of like running a multipass through a Divx encoder to get the best quality. Leave the compression stage on overnight for 8 hours and the game will look better and play faster than it did the night before.

The game would still be playable in uncompressed, IE: just take the right top pixel out of every 4x4 quadruple sized pixel and it will be "regular" again.
 
You must log in or register to reply here.
Back
Top