Radeon X850XT PE vs GeForce 7600GT

[cubeless]

get the 7900 gs at outpost for $130 and maybe this thread can die... it whoops them both... and should make the newer games playable...

or we can get into the '256 vs. 512 mb memory' discussion...

 

[apoppin]

Originally posted by: cubeless
get the 7900 gs at outpost for $130 and maybe this thread can die... it whoops them both... and should make the newer games playable...

or we can get into the '256 vs. 512 mb memory' discussion...

BFG GeForce 7800GS AGP $175 after $60 rebate
go for the 512 MB version

 

[Gamingphreek]

Originally posted by: apoppin

Originally posted by: Gamingphreek
I do know what I am talking about as I have demonstrated numerous times.

.e. speaking in generalities using an OLD game to support your rather tenuous position.

I wasn't aware that I was under time constraints here. Last I checked we were discussing games with SM 3 support.

you gave ONE example ... try another

Yes ... discussing GAMES ... not "theoretical performance increases" ... turning on SM3.0 can go either way - depending on what the Devs intended ... from MY experience, it tends to cripple or slow performance for many mid-level cards for many new games [i bought 5 or 6 new games since Oblivion].

SM 3 can only go one way. It can only increase performance as evolucion8 clearly stated. The features that are automatically turned on or that are allowed to be turned on (Parallax and Displacement Mapping) with SM 3 path enabled are what causes the performance decreases. SM itself does not cause those performance decreases and many times those features can be disabled.

We are speaking of "enabling SM3.0" in a GAME ... in some games turning on that "feature" definitely causes FPS drop ... quoting another forum member does not solidify your case

having BOTH a x850xt and a x1950p in the same system ... there isn't really that much difference in the "visuals" with SM3.0 'on' or 'off'.

I agree, seeing as I stated that earlier in the thread. SM 3 decreases the number of calculations required for shaders and increases the maximum number of instructions available. While some IQ featues (As I said before displacement mapping) can be enabled in the SM 3 path, "visuals" were not the focus here.

-Kevin[/quote]my last comment is speaking PURELY of the "visuals" ... "enabling SM 3.0 makes very little difference over the SM 2.0 path ... practically ... in many modern popular titles

EDIT: i don't believe any new games actually require a SM3.0 card to run.

well ... i checked ... evidently there IS a reason for SM 3.0 ... IF you Play SC Double Agent or the direct ports from Xbox360 that use SM3.0 as a minimum
:Q

i am corrected[/quote]

I believe Chaos Theory also requires it unless you want to run in the SM 1.4 (IIRC) path.

you gave ONE example ... try another

1. Far Cry
2. Splinter Cell: Chaos Theory
3. Splinter Cell: Double Agent
4. Chronicles of Riddick (Though enabling HDR absolutely TANKS performance)
5. LOTR Battle for Middle Earth I and II
6. Madden 05 and 06
7. Painkiller
8. FEAR

Need I go on?

We are speaking of "enabling SM3.0" in a GAME ... in some games turning on that "feature" definitely causes FPS drop ... quoting another forum member does not solidify your case

ENABLING SM 3.0 DOES NOT CAUSE A PERFORMANCE DROP. The features that are then unlocked afterwards is where the performance drop comes from. SM 3's only feature that I know of that can cause a performance decrease is Parallax/Displacement Mapping; a feature that is not required to run the SM 3 path.

Again SM 3 does not lower performance. Any performance decrease experienced is from some other feature that is enabled.

I can go more in depth in SM 3 if you need me to spell it out for you.

my last comment is speaking PURELY of the "visuals" ... "enabling SM 3.0 makes very little difference over the SM 2.0 path ... practically ... in many modern popular titles

And I said I agree with you lol. I added on by saying the features that lower performance (Parallax/Displacement Mapping) are the only increases in IQ you will see depending on if you have the "horsepower" to run it. If you do not turn those on, IQ is the same and you receive a performance boost by using the SM 3 path (Dynamic Branching, Use of Interpolators, "early-out")

-Kevin

 

[apoppin]

yes ... go on ... i want to see more than a list of games.

... and Riddick contradicts you 100% ... you get a massive performance hit

i NEVER said --by itself - "enabling SM3.0 causes a performance drop" ....

otOh ... i DID say enabling the "feature set" that is included when you enable SM3.0 generally DOES cause a performance drop -- in many games.

i believe i am speaking from what i observe when i enable it and you are speaking of 'theoretical/real advantages" of SM3.0 over 2.0 ... it isn't usually 'that simple' when you enable SM3.0 ... you may get more eyecandy - at the expense of performance - if the Devs figure you have a newer card.

[btw, Gothic3 - using GameBryo, i think - seems to run better with "post processing" [inc SM 3.o] than without ... but it is buggy as hell and hard to tell *anything*]

ANYway, that argument is moot IF you want to play ALL games you NEED sm3.0 ...
[nevermind that SC-DA is a pretty disappointing port by many accounts]

 

[Gamingphreek]

Originally posted by: apoppin
yes ... go on ... i want to see more than a list of games.

... and Riddick contradicts you 100% ... you get a massive performance hit

i NEVER said --by itself - "enabling SM3.0 causes a performance drop" ....

otOh ... i DID say enabling the "feature set" that is included when you enable SM3.0 generally DOES cause a performance drop -- in many games.

i believe i am speaking from what i observe when i enable it and you are speaking of 'theoretical/real advantages" of SM3.0 over 2.0 ... it isn't usually 'that simple' when you enable SM3.0 ... you may get more eyecandy - at the expense of performance - if the Devs figure you have a newer card.

[btw, Gothic3 - using GameBryo, i think - seems to run better with "post processing" [inc SM 3.o] than without ... but it is buggy as hell and hard to tell *anything*]

ANYway, that argument is moot IF you want to play ALL games you NEED sm3.0 ...
[nevermind that SC-DA is a pretty disappointing port by many accounts]

Haha...well then I agree that some of the features included with SM 3 really tank performance, but more times than not you can disable them and you can take advantage of the very slight performance gains.

Merry Christmas

-Kevin

 

[evolucion8]

Originally posted by: apoppin

Originally posted by: Gamingphreek

Originally posted by: Zstream

Originally posted by: Gamingphreek

After all there's nothing that SM 3.0 can do that SM 2.0 and it's iterations (2.0b, 2.0a) can't.

Uhhhh NO. There is a reason it is 3.0 and the others are 2.0 rev a,b. I'm not going to get into all the differences with you but you are right in that SM 3.0 mainly improved performance and efficiency compared to IQ.

Zstream you have absolutely no idea what you are talking about. Enabling SM 3 has nothing to do with the reasons the games run like crap. The features you enable when using the SM3 path are what kills performance, no the SM3 path itself. Unless you know what you are talking about stop insulting everyone else!!

-Kevin

Emmm I know what SM3.0 is and know pretty well what it does and performance difference between 2.0 and 3.0

Like I stated before using a 7600GT with sm 3.0 is pretty poor performance. No game that I can think of use the features of sm2.0 and just export those to 3.0.


Sm 3.0 increases the shader calculations and therefore more shaders are used in games which drops the performance of a mid range card using 3.0 to a low end card in terms of performance.

WRONG

As evolucion8 stated earlier, SM3 DECREASES the amount of calculations needed.

As was stated just afterwards Far Cry demonstrated performance increases when using the SM 3 path.

Just accept that you don't know what you are talking about and stop spreading misinformation.

-Kevin

neither do you ... and lack of facts never stopped you from posting ... i.e. speaking in generalities using an OLD game to support your rather tenuous position.

turning on SM3.0 can go either way - depending on what the Devs intended ... it tends to cripple performance for many mid-level cards for many games.

having BOTH a x850xt and a x1950p in the same system ... there isn't really that much difference in the "visuals" with SM3.0 'on' or 'off'.

`
SM 3.0 in Far Cry was only to improve performance reducing in 2 passes the lights rendering effects, and in 3 passes on the SM 2.0b. The only new thing that would increase the image quality in SM 3.0 is True Displacement Mapping, and is a Vertex Shader Feature missing on the current X1X00 series. (Can be done on any ATi DX9 card using the Render to Vertex Buffer command) But as far as I know Parallax Mapping or Virtual Displacement Mapping are the same, can be done under SM 2.0, Virtual Displacement Mapping uses a combination of Normal Maps and Bump Mapping to create an ilusion on the texture surface, but theorically the surface still flat. (Like F.E.A.R.), but True Displacement Mapping uses the Certex Shader to really deform the surface, but it may cause some artifacting that looks like colliding polygons and some incorrect surface perspectives. And Dynamic Branching is a dangerous feature that must be used carefully, that's why we will never see games exceeding the 1,536 shader instruction count (512 on each Scalar, Vector, Texture) Cause using Dynamic Branching would increase performance when the shader count exceeds the hardware limits, otherwise is not used but since we are predicting Floating Point Data, a misprediction would require to flush all the pipelines and reloaded them again, incurring in a performance hit, we all know that GPU doesn't like Dynamic Jumps, they tend to be more Parallel than a CPU. So if we move from a SM 2.0 card with it's 160 shader instruction count limit to SM 3.0 card, surely the performance will be increased particularly if the game exceeds the hardware limits of the SM 2.0. But if we move from a SM 2.0b card with 1,536 shader instruction count thanks to static branching to a SM 3.0 card with 1,536 shader instruction count thanks to static branching, nothing will be improved, but if we move it to SM 3.0 with 65,536 shader instruction count thanks to Dynamic Branching, the performance can be improved, or decreased, depending on the case. That's why we saw performance improvements in Far Cry, cause that game pushed the SM 2.0 too far requiring of multiple passes to render effects, now requiring fewer passes on SM 2.0b and SM 3.0 cards. Sorry i wrote too much jejeje. Anything wrong here correct me please.

 

[apoppin]

looks OK ... doesn't contradict anything i wrote.

again "theory"

[correct in every way] BUT there are NO games that have ANYthing like [even] 1,500 shader instructions [nevermind 65K --afaik ... correct me if i am wrong ... but don't the most complex current modern games max somewhere around 100 shader instructions?]

 

[evolucion8]

Originally posted by: apoppin
looks OK ... doesn't contradict anything i wrote.

again "theory"

[correct in every way] BUT there are NO games that have ANYthing like [even] 1,500 shader instructions [nevermind 65K --afaik ... correct me if i am wrong ... but don't the most complex current modern games max somewhere around 100 shader instructions?]

It wasn't to contradict you, is just an addon to what you said The shader count can be sometimes a bit more, sometimes a bit less, only the techdemos like Double Cross Ruby, Toy Shop and Nalu for example uses more than 300 shaders. But games like Far Cry don't exceed the 90 shader instruction threshold in the most intensive pixel shader levels. So in a SM 2.0 card, it has 32 instructions for textures, 64 for scalar and 64 for vector. If for some reason a game exceeds the 32 texture shader count, multipassing must be done incurring in a performance hit. So if a game uses 100 instruction shaders, can be for example 45 vector, 45 scalar and 10 for textures or something (Just an example) So a SM 2.0 card can output a 160 instructions max using the maximum instruction slots on each component, that's something that doesn't happen in games, some uses more vector instructions, some other uses more scalar instructions, blah blah blah.

 

[apoppin]

Originally posted by: evolucion8

Originally posted by: apoppin
looks OK ... doesn't contradict anything i wrote.

again "theory"

[correct in every way] BUT there are NO games that have ANYthing like [even] 1,500 shader instructions [nevermind 65K --afaik ... correct me if i am wrong ... but don't the most complex current modern games max somewhere around 100 shader instructions?]

It wasn't to contradict you, is just an addon to what you said The shader count can be sometimes a bit more, sometimes a bit less, only the techdemos like Double Cross Ruby, Toy Shop and Nalu for example uses more than 300 shaders. But games like Far Cry don't exceed the 90 shader instruction threshold in the most intensive pixel shader levels. So in a SM 2.0 card, it has 32 instructions for textures, 64 for scalar and 64 for vector. If for some reason a game exceeds the 32 texture shader count, multipassing must be done incurring in a performance hit. So if a game uses 100 instruction shaders, can be for example 45 vector, 45 scalar and 10 for textures or something (Just an example) So a SM 2.0 card can output a 160 instructions max using the maximum instruction slots on each component, that's something that doesn't happen in games, some uses more vector instructions, some other uses more scalar instructions, blah blah blah.

thank-you for the clarification
:thumbsup:

 

[cubeless]

Originally posted by: apoppin

Originally posted by: cubeless
get the 7900 gs at outpost for $130 and maybe this thread can die... it whoops them both... and should make the newer games playable...

or we can get into the '256 vs. 512 mb memory' discussion...

BFG GeForce 7800GS AGP $175 after $60 rebate
go for the 512 MB version

but that's only 256mb...

and i guess this is an agp thread...

 

[ROEHUNTER]

Originally posted by: schneiderguy

Originally posted by: rogue1979

I wouldn't worry too much about SM 3.0.

if you want to play Rainbow Six Las Vegas or Splinter Cell Double Agent you need an SM3 card.

Exactly, I would hate to spend money on a new card that already can't run some new games.
A few months ago (before SPCell Chaos Theory came out ) I bought a X800GTO and while it is a decent card (unlocked to 16 pipes) I am now wishing I had bought a Nvidia card for shader model 3.0.

 

[cubeless]

from the at review it looks like rslv totally blows on anything under a hot 7900, and just blows on anything better than that...

 

[schneiderguy]

Originally posted by: apoppin

... and Riddick contradicts you 100% ... you get a massive performance hit

HDR isnt part of SM3

 

[Zstream]

Originally posted by: ROEHUNTER

Originally posted by: schneiderguy

Originally posted by: rogue1979

I wouldn't worry too much about SM 3.0.

if you want to play Rainbow Six Las Vegas or Splinter Cell Double Agent you need an SM3 card.

Exactly, I would hate to spend money on a new card that already can't run some new games.
A few months ago (before SPCell Chaos Theory came out ) I bought a X800GTO and while it is a decent card (unlocked to 16 pipes) I am now wishing I had bought a Nvidia card for shader model 3.0.

Sorry mate but if you expected to play SPCDA or RSLV on a Nvidia card = the price you paid for the GTO you would be playing at 640 * 480 resolution.

The fact is the OP should not even buy any of those cards. Save a little more and buy a midranged setup for 300$.

 

[imported_browsing]

Originally posted by: Captante
OP, you sum up the pros & cons of these cards very well in your first post ... if noise, power consumption & image quality are top priority then get the 7600GT, if raw framerates are the most important thing then go for the used X850XT.

Of course if you can afford the extra $'s for the 7900GS, that is a solid choice as is an X1900GT @ $149.

Sapphire X1900GT @ Newegg

Is this the best 'bang for your buck' deal?

 

[Captante]

Originally posted by: browsing

Originally posted by: Captante
OP, you sum up the pros & cons of these cards very well in your first post ... if noise, power consumption & image quality are top priority then get the 7600GT, if raw framerates are the most important thing then go for the used X850XT.

Of course if you can afford the extra $'s for the 7900GS, that is a solid choice as is an X1900GT @ $149.

Sapphire X1900GT @ Newegg

Is this the best 'bang for your buck' deal?

Again it depends on your budget... if $150 +/- a little is the most you can swing, then yes this card is about the best you can do, ATI in my opinion pretty much has a lock on the mid-range in Grapics cards right now (with the possible exception of the 7900GS) until Nvidia can get some mid-range G80 based parts to market... Nvidia still owns the low end with the 7600GT and of course the high end with the 8800's.

In terms of "Bang for the buck" in ascending order:

Nvidia 7600GT

ATI X1900GT

Nvidia 7900GS** (If you can find a great price ... for $200+ these arn't worth it compared to the X1950 Pro)

ATI X1950 Pro

ATI X1950XT** (Best value in a high perf card)

 

[ScrewFace]

Shader Model 3.0 and 4.0 will be long gone and laughed at compared to the upcoming (2010) DirectX Next (DirectX 11.1) SM 5.0 and 6.0 visuals. And we'll be running SM 6.0 cards when nVidia release the GeForce 10800GTX and ATI releases their Radeon X4800XT.



What movie's this line from (for $1,000,000 Turkish Lira!):"Congratulations, you're still alive. Most people are so ungrateful to be alive...But not you. Not anymore."

 

[apoppin]

Originally posted by: schneiderguy

Originally posted by: apoppin

... and Riddick contradicts you 100% ... you get a massive performance hit

HDR isnt part of SM3

WTF are you talking about?

i am referring to enabling SM 2.0+++ in the Riddick options.

it KILLS performance and it is "shaders"