9500 hack + FSAA strangeness [Archive]

View Full Version : 9500 hack + FSAA strangeness

Ante P

13-Jan-2003, 02:10

Well no one has missed all the 9500 hacking.
Today a software patch was released that enables the 4 additional pipelines.

I read a very strange thing in the forums at [H] though.
As you probably know some people get artifacts (checkboard effect all over the screen) after modding/hacking.
Apparently turning on 2x FSAA fixes this for many people.

Any ideas why turning on FSAA would fix it? (I assume it doesn't disable any pipelines or something crazy like that.)

OpenGL guy

13-Jan-2003, 02:36

Doomtrooper

13-Jan-2003, 02:47

We have ways to make you talk :wink:

Ante P

13-Jan-2003, 02:54

P.S. If you do your hack and enable 2xAA, is performance improved over the original 9500?

ie perhaps it does indeed disable those pipes for whatever reason I assume..

hmm I don't have the card myself and I can't be bothered to register at [H] just to ask them ;)

but it only works for soem of them, for most people it still didn't work

http://w1.461.telia.com/~u46115957/whoa.jpg
whoa dude this is like a total mystery or something :lol:

Nagorak

13-Jan-2003, 04:49

If this is true, I wonder if that means 2X FSAA also sees through the hack even on a working 9500 hack... Then again a 9700 Bios flash probably would solve that problem (which might be why some people's problems are solved this way).

Also, I wouldn't be surprised if ATI starts trying to put more "blocks" in their future drivers against this. (Or maybe they'll just take a clue from the failure of the MPAA and RIAA to stop hackers and not bother). Either way, this still isn't as convenient as owning your own R9700, since you have to keep downloading new hacks. Then again, for the price...

Dave Baumann

13-Jan-2003, 08:23

Could the artefacts be an indication that those enabled pipes were actually broken?

T2k

13-Jan-2003, 08:29

On the site, the guy said about his patch: This does not enable the 256bit memory bus. On all 9500 cards using 9700 PCBs the memory bus is already 256 bits. - isn't it BS? :shock:

Edit: typo

Dave Baumann

13-Jan-2003, 10:16

On all 9500 cards using 9700 PCBs the memory bus is already 256 bits.[/i] - isn't it BS? :shock:

How many times must we go over the same thing :?: :?: :?:

http://www.beyond3d.com/forum/viewtopic.php?t=3532

Dave Baumann

13-Jan-2003, 13:46

Finally - XBit reaches the conclusion that we've been saying since the introduction of 9500:

http://www.xbitlabs.com/news/story.html?id=1042408060

Version 2. To tell the truth, it is a very unpleasant one. Some RADEON 9500 chips are none other but the “defective” RADEON 9700 ones. During the chips testing, they check whether all its blocks are functioning properly and if some of them do not work, the chip is marked as defective. However, if the defective blocks are inside the pixel pipelines and there are 4 pipelines at the most that do not work well, then these 4 non-functional pipelines can be simply disabled and the chip can become a RADEON 9500.

Why this is 'not very pleasant' is beyond me since the card is working as advertised, regardless of whether it may be an R300 with some pipelines disabled.

Here's some images with the defects:

http://www.xbitlabs.com/news/images/2003-01/9500pol-2-b.jpg

http://www.xbitlabs.com/news/images/2003-01/9500pol-3-b.jpg

Note, in those two instances the artifacts appear mostly in the areas where PS is used.

Ante P

13-Jan-2003, 14:30

Was anyone expecting anything else?
And I gotta agree with Dave, what the hell is dissapointing in having card that deliver what they're supposed to deliver.
It's a great way to optimize their production which in the end leads to better prices for us consumers.
What's not pleasant with that? ;)

Yeah the artifacts seems to happen mostly when there's heavy multi texturing/pixel shading in the works.

BTW I spoke to Wizzard who developed the software patch and he's working on disabling specific pipelines, so perhaps if only one is broken you can still use 7.

I did some predictions based on the results I already have:
http://www.nordichardware.se/Undacbilder/hack.gif
(UT2003 BotMatch Antalus 1024x768 with 4x FSAA and 8x Quality Aniso)

GraphixViolence

13-Jan-2003, 15:01

I guess the "unpleasant" part was finding out that you can't get something for nothing after all. :roll:

Ante P

13-Jan-2003, 15:25

I guess the "unpleasant" part was finding out that you can't get something for nothing after all. :roll:

Most still do, the success rate is higher that the failure rate from what I can tell.
And if wizzard really finds out how to enable/disable spcific pipes all is good. :)

2B-Maverick

14-Jan-2003, 09:54

Was anyone expecting anything else?
I did some predictions based on the results I already have:
http://www.nordichardware.se/Undacbilder/hack.gif
(UT2003 BotMatch Antalus 1024x768 with 4x FSAA and 8x Quality Aniso)

shouldnt the supposed 9500 7 to 5 pipes scale from the 9700 down to the normal 9500 (and not only down to the 9500Pro) ?

And: i dont know about those 7 pipes, but i guess the drivers are optimized for 4 or 8 pipes, and not for 7.

So i doubt you will get good performance out of 7 pipes..... but you never know of course... untill you try.

RussSchultz

14-Jan-2003, 12:30

Interesting, my errors do NOT look like that. Mine are large black squares that run in several lines across the screen.

[-----------------------------------------------------]
[ # ]
[ # ]
[ # ]
[ # ]
[ # ]
[ # ]
[ # ]
[ # ]
[ # ]
[################################################# ####]
[ # ]
[-----------------------------------------------------]

Everything else but these is rendered correctly.

Mr.huang

14-Jan-2003, 13:45

I Think it's the Voltage Problem,the R300 Chip in R9500 64MB or R9500 128MB are good die, just ATI use the voltage trick.

kmolazz

14-Jan-2003, 14:04

can you explain better?

Colourless

14-Jan-2003, 14:21

I would be quite amazed if the chip would work with 5, 6 or 7 pipelines. It looks to me like it's configure to use sets of 4 pipes.

From what I can tell from the screenshots is the screen is divided into 16x16 tiles. Each tile is then rendered using a block of 4 pipelines. I would imagine that that the 8 pipe card render 2 tiles at a time, while the 4 pipe cards only renders 1.

Most of the problems seem to be related to blending and/or pixel shaders. It would seem in most cases with artifact, the pipelines are only partially broken. However, not always though. Here are some shots that I found posted over at Rage3D. Pretty obvious whats going on here. 2 of the pipelines are completely busted.

http://www.steellam.pwp.blueyonder.co.uk/Car.JPG
http://www.steellam.pwp.blueyonder.co.uk/Lobby.JPG
http://www.steellam.pwp.blueyonder.co.uk/Dragothic.JPG

Mr.huang

14-Jan-2003, 14:26

I meant the problem is due to the voltage ic, there are 2 kinds of voltage ic in each R9700/R9500 series which control GPU and memory 's voltage.

Doomtrooper

14-Jan-2003, 15:27

I think its a pipe issue myself, not sure if its binning or possibly something to do with 256-bit interface (noise)..

If you look at the cores...

R9700 Pro

215R8CBGA13F

vs 9500 NP

215R8ABGA13F

They are etched with different part number, the success rate is dropping like flies though on this but I'm willing to give it a whirl.

What makes things confusing if a working 9500 Pro uses the soft-mod they get chessboard artifacts http://64.207.13.28/mysmilies/contrib/ruinkai/dazeda.gif

Now some people have made the FireGL mod work and claimed it was working by moving the resistor to 6 and 5.

http://www.darkcrow.co.kr/image/news/0301/95009700pro2.jpg

Oh this is so much fun..really.

RussSchultz

14-Jan-2003, 16:45

My "checkerboard" artifacts look very much like those in the picture--except they follow the pattern I drew out (only the areas of the screen marked with a # are affected).

Wierder yet, these things were obviously casting "shadows" as in one of the pixel shader tests (the one with the fish) I had white squares/checkerboards that were "reflecting" off of the water.

Colourless

14-Jan-2003, 19:45

RussSchultz that's probably because the test is doing a render to texture.

Also, why don't you take a screenshot

Nagorak

14-Jan-2003, 23:24

Partial repost from the other thread:

Ok, I got 2 9500s today and I have to say that this hack is just BS. I don't believe any of the 9500s are working. Of the two I received, one is totally jacked, meanwhile the other has VERY SMALL errors. You have to look very closely in order to find these errors. As far as I'm concerned it's likely those with a "working" 9500 actually have one that has very small errors that they haven't noticed yet.

Let me explain my reasoning. The theory of this "hack" was ATI was releasing fully functional cards as 9500s because the demand was so high. However, my feeling is ATI would *NOT* release the cards as 9500s if the pipes are all working, instead they would set them to be 9500 Pros. The difference in cost between a 9500 and a 9500 Pro is negligible ($20). The difference in sales between them should be more or less negligible also, IMO, but on the one hand ATI gets more money. I really feel the only reason for the 9500 non-pro's existence is to dump bad cores. There's no other way it makes sense because with a fully functional core its selling price is so much lower than the 9700, yet it would cost the same amount to make (even the same PCB).

That's my gut feeling anyway, after receiving two bum cards (so much for a 30% failure rate... :roll: ). Also, even assuming that there are some actual working cards, my 0% success rate is a good indication that we're looking at a very poor likelihood of getting one. It's not a "30% failure rate", at best it's a 30% success rate. Either that or I have really bad luck...

Anyway, I'm just going to fork over the extra money for a working card. It's just not worth the hassle buying 500 9500s just to save $100.

Doomtrooper

14-Jan-2003, 23:32

I've seen a working card, it is not BS but the success rate is low.

Nagorak

14-Jan-2003, 23:43

I've seen a working card, it is not BS but the success rate is low.

Fair enough.

Doomtrooper

15-Jan-2003, 00:39

If option 1 fails then there is always option 2:

http://www.rage3d.com/board/showthread.php?s=&threadid=33658209

Doomtrooper

18-Jan-2003, 00:37

Ok after unsucessful software mod on my Saphire 9500 NP 128, I ventured into the 9700 hardware mod..which of course failed :(

I then tried the Hardware mod (using both resistors) for a FireGL X1...which of course failed
:(

I then tried removing the X1 bios and tried the Z1 bios still modded for FireGL mode using conductive grease (same stuff used for unlocking XP's).

To my surprise it worked :D

Although nothing to brag about (its still a 4 pipeline card) ...B4 1942 plays very well with 4xAA and scored 7500 on 3Dmark with 4X FSAA Not bad.