PDA

View Full Version : What is this?


Pages : [1] 2

Bopple
09-15-03, 09:41 AM
What is this madness?
I have never heard anything remotely similar to this before.
Have I missed anything?

I copied this from THG. The name was Balderdash.

ATI utilizes a 8x1 pixel shader path, with one path at 8 bits. Nvidia, on the other hand, uses a 4x2 path with two paths each 4 bits wide. Currently, any game using PS 2.0 with the FX cards is only accessing shaders at 4x1, due to driver and DX9b limitations (we will see DX9c soon, mark my words) and so, the DX9 games and 45:23 driver is effectively ignoring the second PS 2.0 path.

The preview 51:75 driver alleviates this problem, enabling the full second path for use in the game as much as possible before any update to DX9 is implemented to allow true dual channels as intended by its design.

We see these HL2 benchmark results now because HL2 is seriously dependant on pixel shaders in their current form and it is singly responsible for the framerate discrepancies.

The fix coming with the Det.50 should bring the numbers in line with ATI's, and additionally, the updated DX9c from Microsoft will likely make the FX cards the winner once true dual channel shaders are implemented and dual channel benefits can be accessed.

The next incarnation of DX9 should include the ability to use simultaneous wait states for PS 2.0 textures in DX9 applications. This will greatly reduce the 'problem' shown in these 'benchmarks.' The DX9 SDK was built (without any hardware available mind you) to favor one long pipe (and thus currently favor the ATI 8x1 version) since each texture has to go through a myriad of call back and wait/check states and has a definite FIFO for all textures in the pipe the nV (4x2) pipe is crippled during these operations. With the next version of DX9 you'll see the included paired texture waits in the shader process allowing the nV to actually utilize the 4x2 pipe simultaneously instead of a defined FIFO for each.

EDITED for spelling and clarity...

Behemoth
09-15-03, 09:52 AM
lets hope its not another PR lie to sell their cards.....

Sazar
09-15-03, 09:55 AM
:rolleyes:

The Baron
09-15-03, 09:58 AM
Wow.

'Wait for DX9 games!"
"Wait for Det50!"
"Wait for the next version of DirectX!"

Uhhuh.

NickSpolec
09-15-03, 09:59 AM
If this is the case, then why is this the first time ANYONE has heard about this, or why has Nvidia NOT said anything concerning this?

If this is true, then why has Nvidia lowered percision/detail in games instead of saying "It's a DX thing... wait for an update along with new drivers"?

I really don't buy it. It seems too logical an explaination --- and we all know logic is one thing not involved with anything Nvidia lately (:D)

digitalwanderer
09-15-03, 10:01 AM
Originally posted by Behemoth
lets hope its not another PR lie to sell their cards.....
"Pssssst! It's another PR lie to sell their cards." ;)

Sazar
09-15-03, 10:03 AM
Originally posted by The Baron
Wow.

'Wait for DX9 games!"
"Wait for Det50!"
"Wait for the next version of DirectX!"

Uhhuh.

then m$ said to nvidia " I am your father"

/me gets caught up in that star wars thinger...

ragejg
09-15-03, 10:03 AM
meh...

but, maybe not meh...




...



well, meh for now... :p

:afro:

Who is Balderdash, and how's that thread coming along? Is he pwned? Does this d00d have cred??

Sazar
09-15-03, 10:05 AM
Originally posted by ragejg
meh...

but, maybe not meh...




...



well, meh for now... :p

:afro:

Who is Balderdash, and how's that thread coming along? Is he pwned? Does this d00d have cred??

he sounds like he is related to Baldrick... though I have not conducted any dna testing as of yet...

The Baron
09-15-03, 10:07 AM
Originally posted by Sazar
then m$ said to nvidia " I am your father"
AHHA! So... wait, where does ATI fit in then? As Boba Fett?

MS cut off NV's hand with DX9..

ahha.

got it.

3dfx = Yoda.
NVIDIA = Luke.
MS = Vader.
ATI = Emperor.
DX = Leia
OGL = Han Solo

God, this is the dumbest post ever.

Sazar
09-15-03, 10:11 AM
Originally posted by The Baron
AHHA! So... wait, where does ATI fit in then? As Boba Fett?

MS cut off NV's hand with DX9..

ahha.

got it.

3dfx = Yoda.
NVIDIA = Luke.
MS = Vader.
ATI = Emperor.
DX = Leia
OGL = Han Solo

God, this is the dumbest post ever.

:eek:

and I thought I was the one on meds...

/me runs away from baron...

Johnmcl7
09-15-03, 10:31 AM
As someone pointed out on Rage3d, how is the 9600 pro able to perform so well with its 4x1 architecture?

I'm assuming it's the same guy, he posted on [H]ardocp first with this, http://hardforum.com/showthread.php?s=&postid=1025268849#post1025268849


I'm not going to defend nvidia's decision to implement 4x2 archetecture over a more standard single channel mode, it seems reasonable to think that their mobo chipset driver with dual channels was probably their motivation in trying it. But with that in mind, wouldn't it be a sad state of affairs if nobody ever tried to push the envelope with technilogical advances and only stuck with the tried and true? We would still be using 486's and EGA graphics cards if that were the case.

To call the 4x2 architecture of the Geforce FX innovative/pushing the envelope is laughable to say the least, but to compare pipeline architecture of a graphics card to the ram configuration on a motherboard chipset is just downright odd!

I don't think he has a clue what he's speaking about.

John

creedamd
09-15-03, 10:41 AM
He is an avid known fanboi of Nvidia. If nvidia put a pile of dong in his hand, he would say "well it doesn't smell too bad if you don't breathe"

Sazar
09-15-03, 10:47 AM
Originally posted by creedamd
He is an avid known fanboi of Nvidia. If nvidia put a pile of dong in his hand, he would say "well it doesn't smell too bad if you don't breathe"

:rofl

digitalwanderer
09-15-03, 10:57 AM
Originally posted by The Baron
AHHA! So... wait, where does ATI fit in then? As Boba Fett?

MS cut off NV's hand with DX9..

ahha.

got it.

3dfx = Yoda.
NVIDIA = Luke.
MS = Vader.
ATI = Emperor.
DX = Leia
OGL = Han Solo

God, this is the dumbest post ever.
You've been up all night banning fanboys, haven't you? :eek2:

Johnmcl7
09-15-03, 10:58 AM
Originally posted by The Baron
AHHA! So... wait, where does ATI fit in then? As Boba Fett?

MS cut off NV's hand with DX9..

ahha.

got it.

3dfx = Yoda.
NVIDIA = Luke.
MS = Vader.
ATI = Emperor.
DX = Leia
OGL = Han Solo

God, this is the dumbest post ever.

I think MS should be the emperor (with control over everything) and ATi being vader...

John

ChrisRay
09-15-03, 11:06 AM
Originally posted by Johnmcl7
As someone pointed out on Rage3d, how is the 9600 pro able to perform so well with its 4x1 architecture?

I'm assuming it's the same guy, he posted on [H]ardocp first with this, http://hardforum.com/showthread.php?s=&postid=1025268849#post1025268849




To call the 4x2 architecture of the Geforce FX innovative/pushing the envelope is laughable to say the least, but to compare pipeline architecture of a graphics card to the ram configuration on a motherboard chipset is just downright odd!

I don't think he has a clue what he's speaking about.

John


The 9600 pro has Dedicated FLoating Point Units. I've read and pretty much seen that the FX FP units are not dedicated and act as texel units as well. So Sometimes I believe the Nv card could actually be a 2x2 card in Shader scenerios if those are acting as texel units.

Behemoth
09-15-03, 11:26 AM
Originally posted by ChrisRay
The 9600 pro has Dedicated FLoating Point Units. I've read and pretty much seen that the FX FP units are not dedicated and act as texel units as well. So Sometimes I believe the Nv card could actually be a 2x2 card in Shader scenerios if those are acting as texel units.
yeah that would pretty much explain why fx loses so much to radeon even in ut2003's dm-ice-tomb and rustatorium maps, and bf1942 really because nv3x simply has LESS units to do all things at once. they are not even dx9 problems. i bet.

Johnmcl7
09-15-03, 11:38 AM
Originally posted by ChrisRay
The 9600 pro has Dedicated FLoating Point Units. I've read and pretty much seen that the FX FP units are not dedicated and act as texel units as well. So Sometimes I believe the Nv card could actually be a 2x2 card in Shader scenerios if those are acting as texel units.

I realise there are further implications, my point was more that 'balderdash' seems to have convinced himself that the FX is being run as a 4x1 card which cripples it. Given some of his other statements I don't thin it's worth going any deeper...

John

Slappi
09-15-03, 12:16 PM
Originally posted by digitalwanderer
You've been up all night banning fanboys, haven't you? :eek2:

Looks like he missed one..... :rofl

J/k

YOU KNOW YOU'RE MY BOY DW.....YOU KNOW THIS!

digitalwanderer
09-15-03, 12:21 PM
Originally posted by Slappi
Looks like he missed one..... :rofl

J/k

YOU KNOW YOU'RE MY BOY DW.....YOU KNOW THIS!
I'll keep setting 'em up and you keep knocking 'em home. ;)

aapo
09-15-03, 03:03 PM
This kind of stupidity makes me mad. I hope any of you weren't falling into this...

Originally posted by Balderdash
ATI utilizes a 8x1 pixel shader path, with one path at 8 bits.


WTF? One path at 8 bits? R3X0 can do 8 pixels with one texture fetch per cycle, not freaking bits. The pixels that go to the framebuffer are integer pixels at 4 x 8 bits = 32 bits! The first number in 8x1 tells how many pixels run through the pipeline simultaneously, and the 1 tells how many texture fetches can be done to each one of those pixels. PS2.0 pixels are (3 or 4) x 24 bits with ATi hardware, BTW, so it consumes even much more bits with all the pipelines.


Nvidia, on the other hand, uses a 4x2 path with two paths each 4 bits wide. Currently, any game using PS 2.0 with the FX cards is only accessing shaders at 4x1, due to driver and DX9b limitations (we will see DX9c soon, mark my words) and so, the DX9 games and 45:23 driver is effectively ignoring the second PS 2.0 path.


WTF? The 2 in 4x2 is the number of tecture fetches per pixel (4 pixels) per clock. The '2' doesn't play a frigging role with pixel shaders. NV35 can output max 4 pixels per cycle no matter what.

There is a semen of truth there between the lies, though. Due to driver limitations and design flaws the NV35 (not other NV chips, btw) is forced to calculate max 4 FP ops per clock. It's possible there are free FP units to be used in the architecture, but something rather complicated was wrong with the loopback operations in the NV35 pipeline. Frankly, I didn't understand it at all, so I'm not gonna try to explain it, but I think Uttar or Dragunov do know what's wrong there. :D


The preview 51:75 driver alleviates this problem, enabling the full second path for use in the game as much as possible before any update to DX9 is implemented to allow true dual channels as intended by its design.

We see these HL2 benchmark results now because HL2 is seriously dependant on pixel shaders in their current form and it is singly responsible for the framerate discrepancies.

The fix coming with the Det.50 should bring the numbers in line with ATI's, and additionally, the updated DX9c from Microsoft will likely make the FX cards the winner once true dual channel shaders are implemented and dual channel benefits can be accessed.


These 'dual channels' are nothing but a hallucination.


The next incarnation of DX9 should include the ability to use simultaneous wait states for PS 2.0 textures in DX9 applications. This will greatly reduce the 'problem' shown in these 'benchmarks.' The DX9 SDK was built (without any hardware available mind you) to favor one long pipe (and thus currently favor the ATI 8x1 version) since each texture has to go through a myriad of call back and wait/check states and has a definite FIFO for all textures in the pipe the nV (4x2) pipe is crippled during these operations. With the next version of DX9 you'll see the included paired texture waits in the shader process allowing the nV to actually utilize the 4x2 pipe simultaneously instead of a defined FIFO for each.


:mad: :mad: :mad:

*sigh*

Maybe I just say that NV35 has only one pipeline with some primitive ILDP stuff allowing it to simultaneously do stuff with 4-8 (p/z/t)ixels. R3X0 has 8 traditional pipes. That guy has reveresed the situation to a perverse mess with no useful facts left.

ChrisRay
09-15-03, 04:10 PM
Originally posted by Johnmcl7
I realise there are further implications, my point was more that 'balderdash' seems to have convinced himself that the FX is being run as a 4x1 card which cripples it. Given some of his other statements I don't thin it's worth going any deeper...

John

A Second TMU wouldnt benefit Shader aplications anyway. I dunno why he'd think that.

Johnmcl7
09-15-03, 07:14 PM
Originally posted by ChrisRay
A Second TMU wouldnt benefit Shader aplications anyway. I dunno why he'd think that.

Well some of the die hard Nvidia fans still seem to be hoping that the Detonator 50s are going to magically fix the GeForceFX performance problems and has somehow decided on MS not supporting the 4x2 architecture being the problem - I'm also baffled as to why he'd think that.

John

Sazar
09-15-03, 10:04 PM
Originally posted by Johnmcl7
Well some of the die hard Nvidia fans still seem to be hoping that the Detonator 50s are going to magically fix the GeForceFX performance problems and has somehow decided on MS not supporting the 4x2 architecture being the problem - I'm also baffled as to why he'd think that.

John

I just read the thread he started @ [H] where he is apparently the resident nvidia cheerleader... :)

about 18 pages of odd information..

makes me cringe @ how little some people know about the things they discuss...

I personally don't know as much as some of you others but damn... there is some ignorance out there...