HD 5870 Discussion thread.

[Bo_Fox]

Yeah, I do not know if it's just the 57xx cards. Heck, it could just be that 9.10 driver set in combination with a 5770 as tested by BFG10K.

http://alienbabeltech.com/main/?p=12648&page=3

I really do hope that you are right about the 58xx cards not having any of the brilinear problems as shown in the link. It just would not make sense for ATI to unveil a new perfectly circular AF method, and then cheat by using brilinear AF (unless they wanted their 5770 cards to look good in benchmarks, hoping that nobody would check the IQ once again after the 5870 was already released a few weeks beforehand).

Thanks for the heads-up, erocker. Nobody else seems to be complaining with this particular AF problem, but the problem for me is that nobody else has proven BFG10K to be wrong about the HD5k series.

I wish I tested the IQ on the 5750 that I bought for my brother. I only put it in my PC to see if it was DOA or not, and forgot to test the AF quality. Then I returned it after finding out that it was not compatible with older NForce chipsets (NF4). If I bought it for myself, I would've returned the card anyways if there were the same AF problem in the above link without being able to fix it.

 

[Mussels]

from what i've seen, its ONLY 5770 cards (not 58x0), and its definately fixable in drivers

 

[Steevo]

Yeah, I do not know if it's just the 57xx cards. Heck, it could just be that 9.10 driver set in combination with a 5770 as tested by BFG10K.

http://alienbabeltech.com/main/?p=12648&page=3

I really do hope that you are right about the 58xx cards not having any of the brilinear problems as shown in the link. It just would not make sense for ATI to unveil a new perfectly circular AF method, and then cheat by using brilinear AF (unless they wanted their 5770 cards to look good in benchmarks, hoping that nobody would check the IQ once again after the 5870 was already released a few weeks beforehand).

Thanks for the heads-up, erocker. Nobody else seems to be complaining with this particular AF problem, but the problem for me is that nobody else has proven BFG10K to be wrong about the HD5k series.

I wish I tested the IQ on the 5750 that I bought for my brother. I only put it in my PC to see if it was DOA or not, and forgot to test the AF quality. Then I returned it after finding out that it was not compatible with older NForce chipsets (NF4). If I bought it for myself, I would've returned the card anyways if there were the same AF problem in the above link without being able to fix it.

Its BIlinear, not brilinear. If you are going to be condescending to others at least do it with correct spelling.

A quick search turnd up lots of people who are running a NF4 chipset and 57XX series cards.

Mabey it isn't that they have not proven him wrong, but that no one else notices, cares, or they understand the obscure article that was written about a problem with one test system, with one driver didn't match up to the results posted by many others.

And your link is dead.

http://www.hardocp.com/article/2009/09/22/amds_ati_radeon_hd_5870_video_card_review/6

A quick lok at this page shows the interference pattern on the 5800 series card was the same as the 285, but the filter quality is higher, fuzzier maybe yes but overall users would prefer angle independant so that some edges don't appear warped.


So IQ, the 5XXX series wins, no one else has issues with their NF4 and a 57XX card that I found, besides the bottleneck a NF4 board/CPU would put on a newer card.

Your entire post, and this thread is about something that doesn't exist, and rumours that you continue to propogate. Please stop.

 

[Bo_Fox]

from what i've seen, its ONLY 5770 cards (not 58x0), and its definately fixable in drivers

Hopefully, you are right.

 

[Mussels]

Steevo: brilinear was a term coined back when Nv was cheating in the FX series cards, i believe.

When they were supposed to do trilinear, they did half tri, half bi - hence the name (and their godawful image quality)

 

[Bo_Fox]

Its BIlinear, not brilinear. If you are going to be condescending to others at least do it with correct spelling.

A quick search turnd up lots of people who are running a NF4 chipset and 57XX series cards.

Mabey it isn't that they have not proven him wrong, but that no one else notices, cares, or they understand the obscure article that was written about a problem with one test system, with one driver didn't match up to the results posted by many others.

And your link is dead.

http://www.hardocp.com/article/2009/09/22/amds_ati_radeon_hd_5870_video_card_review/6

A quick lok at this page shows the interference pattern on the 5800 series card was the same as the 285, but the filter quality is higher, fuzzier maybe yes but overall users would prefer angle independant so that some edges don't appear warped.


So IQ, the 5XXX series wins, no one else has issues with their NF4 and a 57XX card that I found, besides the bottleneck a NF4 board/CPU would put on a newer card.

Your entire post, and this thread is about something that doesn't exist, and rumours that you continue to propogate. Please stop.

I'm only educating you about stuff--it's not to be condescending. You certainly could learn a lot about "brilinear". Google is your friend. That is why I wrapped it with quotes, to imply how it is a combination of bilinear and trilinear.

Usually, it can be fixed by disabling the Catalyst A.I. optimization, but BFG10K seemed to imply that he has tried everything to fix the problem (and he certainly did show how much knowledge he had--which is pretty much required to write the article).

Yes, the link is now dead, as of 1-2 hours ago. Just a couple hours ago, it was up and running on the alienbabeltech site. Perhaps it was because of the recent comment that I posted there that prompted him/them to take it down? Oh well, if I ever decide to buy a 5770 or a 5870 (which is unlikely anyways), I'll be sure to test for this problem and perhaps do some more extensive testing on how bottlenecking the memory bandwidth is.

About the 5750 incompatibility with Gigabyte and Abit Nforce4 chipset motherboards, here's a link for you, Steevo: http://www.eggxpert.com/forums/permalink/585964/585838/ShowThread.aspx#585838 If you want more, I can provide some more.

No need to be so belligerent or cynical, Steevo. Take it easy, bud. It's only a video card and if you have more questions, I'll keep on answering them for you.

 

[Steevo]

Steevo: brilinear was a term coined back when Nv was cheating in the FX series cards, i believe.

When they were supposed to do trilinear, they did half tri, half bi - hence the name (and their godawful image quality)

I remember it. Just don't have any idea why it is still being used, that was a blatent attempt to cover up a poor design, and to try and achieve the performance a competitor was getting. However the only compnay that has ever done that time and again is NV.


If we are going to have a discussion about a card, hardware, fine. If this is just going to be a bitch session from haters then move it to GN where it belongs.


It really detracts from the forums overall good name and reliability, we have approved this card and lots of users are loving it, however you get some to make a bad purchasing decision based off one odd thread here that has been disproven, time and again. It's a junk thread. Bo you make some great threads elsewhere, but this one is a stinker.

I'm only educating you about stuff--it's not to be condescending. You certainly could learn a lot about "brilinear". Google is your friend. That is why I wrapped it with quotes, to imply how it is a combination of bilinear and trilinear.

Usually, it can be fixed by disabling the Catalyst A.I. optimization, but BFG10K seemed to imply that he has tried everything to fix the problem (and he certainly did show how much knowledge he had--which is pretty much required to write the article).

Yes, the link is now dead, as of 1-2 hours ago. Just a couple hours ago, it was up and running on the alienbabeltech site. Perhaps it was because of the recent comment that I posted there that prompted him/them to take it down? Oh well, if I ever decide to buy a 5770 or a 5870 (which is unlikely anyways), I'll be sure to test for this problem and perhaps do some more extensive testing on how bottlenecking the memory bandwidth is.

About the 5750 incompatibility with Gigabyte and Abit Nforce4 chipset motherboards, here's a link for you, Steevo: http://www.eggxpert.com/forums/permalink/585964/585838/ShowThread.aspx#585838 If you want more, I can provide some more.

No need to be so belligerent or cynical, Steevo. Take it easy, bud. It's only a video card and if you have more questions, I'll keep on answering them for you.

So just due to a BIOS problem, not a NF4 problem you label a card as having a problem? Sounds backwards to me.


I'm not being anything. I just like facts.


Does the 5XXX series need more bandwidth? Nope. Proven it time and again. Its at the point of minimal gains on that front.
Does the 5XXX series have a IQ problem? Not according to most sites, people, and users. Some that have issues are possibly the game, too high of overclocks, bad drivers/installation, and or other unknown compatability issues.
Does the 5XXX series have below expectations problems? Only to those not using them apparently, or those foolish enough to get wrapped up in hype, marketing, and spin.


Does this thread serve a useful purpose other than a rant thread for you?

 

[Bo_Fox]

I remember it. Just don't have any idea why it is still being used, that was a blatent attempt to cover up a poor design, and to try and achieve the performance a competitor was getting. However the only compnay that has ever done that time and again is NV.


If we are going to have a discussion about a card, hardware, fine. If this is just going to be a bitch session from haters then move it to GN where it belongs.


It really detracts from the forums overall good name and reliability, we have approved this card and lots of users are loving it, however you get some to make a bad purchasing decision based off one odd thread here that has been disproven, time and again. It's a junk thread. Bo you make some great threads elsewhere, but this one is a stinker.



So just due to a BIOS problem, not a NF4 problem you label a card as having a problem? Sounds backwards to me.


I'm not being anything. I just like facts.


Does the 5XXX series need more bandwidth? Nope. Proven it time and again. Its at the point of minimal gains on that front.
Does the 5XXX series have a IQ problem? Not according to most sites, people, and users. Some that have issues are possibly the game, too high of overclocks, bad drivers/installation, and or other unknown compatability issues.
Does the 5XXX series have below expectations problems? Only to those not using them apparently, or those foolish enough to get wrapped up in hype, marketing, and spin.


Does this thread serve a useful purpose other than a rant thread for you?

That's your rant. Let's please keep it a level-headed discussion about why a 5870 is "below expectations"? I do not agree with your bias, and most of the reasonable people here reading this thread would see how a 5870 could use more bandwidth.

I might be talking to BFG10K more about the AF quality (especially if I think about buying one), and I do believe erocker and mussels that the 58xx cards do not have this brilinear problem.

Also, it is rumored that AMD is going to release a better version of 5870 (perhaps 5890) in a month or two.

For the record, I did not make this thread. Peace out, Steevo.

(Edit: many older NF4 motherboards do not have the bios option that supports the power-savings tech of the HD5xxx series, and it was reported that one of the companies other than XFX released an updated bios for their 5750 card. )

 

[Benetanegia]

Its BIlinear, not brilinear. If you are going to be condescending to others at least do it with correct spelling.

A quick search turnd up lots of people who are running a NF4 chipset and 57XX series cards.

Mabey it isn't that they have not proven him wrong, but that no one else notices, cares, or they understand the obscure article that was written about a problem with one test system, with one driver didn't match up to the results posted by many others.

And your link is dead.

http://www.hardocp.com/article/2009/09/22/amds_ati_radeon_hd_5870_video_card_review/6

A quick lok at this page shows the interference pattern on the 5800 series card was the same as the 285, but the filter quality is higher, fuzzier maybe yes but overall users would prefer angle independant so that some edges don't appear warped.


So IQ, the 5XXX series wins, no one else has issues with their NF4 and a 57XX card that I found, besides the bottleneck a NF4 board/CPU would put on a newer card.

Your entire post, and this thread is about something that doesn't exist, and rumours that you continue to propogate. Please stop.

That hardopc link is showing the same IQ problem, so there is/was a problem. The comparison pictures seem to confirm it's a driver issue, because the HD4890 is showing a similar (although still different and less dramatic) problem. You can deny the problem until the end of times, but it will still be there, or at least it was there and it's not rumor. I've seen the same problem in many games from different sources and most of them were user uploaded images that were just trying to show DX11 features. What I mean is, no manipulation there. It might be a driver issue and it might have been fixed with the latest drivers, but until I see any evidence showing the contrary, I'll conclude the issue is still there.

PD. What users think is irrelevant, 99% of people is always happy with their new adquisitions and will never admit any defect.

 

[Bo_Fox]

Thanks, Bene for pointing this out!! I never actually looked at the HardOCP image myself, other than the one provided by BFG10K! Here's the link to the brilinear image for the 5870: http://www.hardocp.com/image.html?image=MTI1MzU4OTM1NVlDbXBla3ZKZm5fNl8yX2wucG5n

The image for the GTX 285 does also seem a bit bilinear, since lots of the moire pattern is not showing, but I think that it's because [H] did not completely disable AF optimizations in the control panel.


Edit: I think the alienbabeltech site is just temporarily down: http://alienbabeltech.com

It shows some real-world applications of how this brilinear filtering makes things more blurry than truly trilinear 16x AF.

 

[Steevo]

That hardopc link is showing the same IQ problem

Let’s start right off the bat looking at filtering quality since ATI has put a lot of effort into making it "perfect" with the 5800 series.






The first screenshot above shows all three video cards being tested here today with Trilinear filtering, i.e. no Anisotropic filtering, their default settings. What will catch your eyes immediately is the fact that the Radeon HD 5870 exhibits a perfect circle on all filtering levels as we look down the tunnel. The GeForce GTX 285 seems to be better than the Radeon HD 4890, so the HD 5870 is a big improvement over AMD’s previous generation.


In the second screenshot we have enabled 16X AF. Once again we find the Radeon HD 5870 produces the best 16X AF out of the bunch. It is a perfect circle, with no dependencies on angles like the Radeon HD 4890. I would say ATI has succeeded in providing the absolute best filtering quality in a gaming graphics card

Yeah.Sounds like a problem there. The second image shown they decide not to fuck with the few remaining pixels beyond a certain point, so the might amount to a 1% loss of random white or truly black pixels? Who cares. If you are walking around in a game looking for those 1% in the far distance you need help.


AA with supersampling performs aa on EVERYTHING so textures look blurred. Do don't use it. Simple fix huh, specially considering that a large screen doesn't need 32X AA to look good.


Brilinear from 2004, not 2008, or 2009. http://www.tomshardware.com/reviews/ati,819-10.html

They used a intelligent filter method to make the whole thing go, and actualy spanked the competition on IQ with the next up X1800XT in all departments.


I guess I haven't seen the proof in the pudding in a game screenshot, from multiple sources where tehy tried to isolate the problem.

I really hope the Fermi comes out soon so there is something for you guys to jerk off to instead of bashing the red team.

 

[Bo_Fox]

Yeah.Sounds like a problem there. The second image shown they decide not to fuck with the few remaining pixels beyond a certain point, so the might amount to a 1% loss of random white or truly black pixels? Who cares. If you are walking around in a game looking for those 1% in the far distance you need help.


AA with supersampling performs aa on EVERYTHING so textures look blurred. Do don't use it. Simple fix huh, specially considering that a large screen doesn't need 32X AA to look good.


Brilinear from 2004, not 2008, or 2009. http://www.tomshardware.com/reviews/ati,819-10.html

They used a intelligent filter method to make the whole thing go, and actualy spanked the competition on IQ with the next up X1800XT in all departments.


I guess I haven't seen the proof in the pudding in a game screenshot, from multiple sources where tehy tried to isolate the problem.

I really hope the Fermi comes out soon so there is something for you guys to jerk off to instead of bashing the red team.

No, they did not use Supersampling AA. That is not why the brilinear filtering problem exists.

Over at alienbabeltech.com (which is temporarily down, not dead), BFG10K also included a screenshot of supersampling with 16x AF which made the colored bull's-eye pattern twice as small.

 

[Benetanegia]

Yeah.Sounds like a problem there.


AA with supersampling performs aa on EVERYTHING so textures look blurred. Do don't use it. Simple fix huh, specially considering that a large screen doesn't need 32X AA to look good.


Brilinear from 2004, not 2008, or 2009. http://www.tomshardware.com/reviews/ati,819-10.html

They used a intelligent filter method to make the whole thing go, and actualy spanked the competition on IQ with the next up X1800XT in all departments.


I guess I haven't seen the proof in the pudding in a game screenshot, from multiple sources where tehy tried to isolate the problem.

Anisotropic filtering is not only about angle dependecy which is the only thing they are mentioning there. Suuuuure, the HD5xxx is producing a perfect circle so it has perfect AF. Yeah great conclusion. Quality matters too and as we can see in the Stalker screenshots I posted above and in the Doom3 SS in alienbabeltech, that's more than obvious, because of the simple fact that the HD4890 is producing much better picture quality than the HD5770, even if it has a much worse angle dependency and well, it's not producing a line between the different LOD textures, something that I have never seen since year 1997 I believe. It's like... come on...

And please don't talk about previous generations because it's offtopic and has nothing to do with the current one. Everybody knows that Ati did a superb job with AF in the X1xxx series, they really destroyed Nvidia in that department. Later Nvidia surpassed them with G80, while AMD's didn't change. Now AMD has greatly improved angle dependency, but at the cost of destrying the quality. Maybe it's a bug and it will be fixed, but as it stands now, for those of us who appreciate picture quality, it's garbage quality.

 

[Goodman]

Thanks, Bene for pointing this out!! I never actually looked at the HardOCP image myself, other than the one provided by BFG10K! Here's the link to the brilinear image for the 5870: http://www.hardocp.com/image.html?image=MTI1MzU4OTM1NVlDbXBla3ZKZm5fNl8yX2wucG5n

The image for the GTX 285 does also seem a bit bilinear, since lots of the moire pattern is not showing, but I think that it's because [H] did not completely disable AF optimizations in the control panel.


Edit: I think the alienbabeltech site is just temporarily down: http://alienbabeltech.com

It shows some real-world applications of how this brilinear filtering makes things more blurry than truly trilinear 16x AF.

Let see if i get this right you all b** about an image that looks a little more blurry then the other 2 pictures?

All i see there is that the 5870 got a better image almost perfect circle & beside all the other circles (red , green) are smaller so of corse it's look more blurry

You guys should get a life , looking for problems where there is none

 

[Benetanegia]

The second image shown they decide not to fuck with the few remaining pixels beyond a certain point, so the might amount to a 1% loss of random white or truly black pixels? Who cares. If you are walking around in a game looking for those 1% in the far distance you need help.

I think you have to take a look again:

It's not any pixel or 1% of anything, I think you missed the point or you are showing a limited understanding. It's not any blurred pictures, it's the abrupt change between the textures. The artificial line you are seing in the ground in the DX11 IS NOT different lighting, it's not that there is a shadow on the front and the sun is hitting hrd on the distance. Both DX10 and DX11 SS's have the exact same lighting, but only the DX10 one is doing trilinear filtering and thus producing a good looking image. Don't know about the rest of you, but I don't want those lines in my games, I have never wanted them since trilinear filtering made it's appearance soooo many years ago. It's really unacceptable.

EDIT: Now that I look at it, it's not only that. It has flat shading across the entire texture, while the DX10 one is showing a perfect gradient that nicely (unnoticeably) extends to the next LOD texture...

 

[Wile E]

Are we sure this isn't a DX11 bug, or a game bug? How can one be so sure ATI is to blame here? It's not like we have NV DX11 hardware to confirm the difference.

 

[Steevo]

The uh, darker one is the DX11. If you look at the whole picture from the site you will see that the plants when up close have tesselation applied or someting, and tehy new shadows..... and yes ATI finally did the object shodows they were missing for so long.

Notice how the screenshots weren't in order on that site?

I guess what you see is dependant on your preconceptions of how bad you want it to look.



Looks good to me, and these were from September, so....... 3 new drivers.

 

[Mussels]

to me, that line is clearly from the shadow of the building.

You cant say "they used the same lighting between DX10 and 11" or whatever, because its an entirely different rendering path. if you set it to DX8.1, 9, 10, 10.1 and 11 and compared the screenshots, i bet you'd find that shadow/line varies between all of them.

 

[Benetanegia]

The same line appears in the alienbabeltech Doom3 screens and also in the other screenshot I posted earlier. Plus many other user uploaded images I've seen in other forums and also videos on youtube.

@Steevo the DX11 one is the one with the glitch, the one on the top on your post. It has SSAO and better shadows.

 

[Bo_Fox]

Let see if i get this right you all b** about an image that looks a little more blurry then the other 2 pictures?

All i see there is that the 5870 got a better image almost perfect circle & beside all the other circles (red , green) are smaller so of corse it's look more blurry

You guys should get a life , looking for problems where there is none

This is un-called for. If you spend $500 on a video card just so you can play with 4x AA instead of 2x AA, you are basically telling yourself to get a life.

Also, you do not understand what we mean by brilinear (actually, strongly bilinear) filtering. Just looking at how it's perfectly circular does not give you the answer.

 

[Bo_Fox]

The same line appears in the alienbabeltech Doom3 screens and also in the other screenshot I posted earlier. Plus many other user uploaded images I've seen in other forums and also videos on youtube.

@Steevo the DX11 one is the one with the glitch, the one on the top on your post. It has SSAO and better shadows.

Thanks, Bene, for helping to keep this a sincere discussion about the IQ. I do hope that there is not such a brilinear filtering problem with the latest drivers, but if there is, at least this thread will be one of the legendary threads on TPU for exploiting this issue.

 

[Goodman]

This is un-called for. If you spend $500 on a video card just so you can play with 4x AA instead of 2x AA, you are basically telling yourself to get a life.

Also, you do not understand what we mean by brilinear (actually, strongly bilinear) filtering. Just looking at how it's perfectly circular does not give you the answer.

I'll never be nuts enough to spend that much money on a card when an -+$200 card will still be able to play the games just fine.

It's like people complaining about a movie not been real & what not ...it's just a f*! an movie...

Same for a games it's just a game doesn't have to be perfect just fun to play with & time killer.

So i'll say again , get a life..

 

[Bo_Fox]

I'll never be nuts enough to spend that much money on a card when an -+$200 card will still be able to play the games just fine.

It's like people complaining about a movie not been real & what not ...it's just a f*! an movie...

Same for a game it's just a game doesn't have to be perfect just fun to play & time killer.

So i'll say again , get a life..

I'll say this to you.. what are you doing crapping on this thread here? Please kindly get out of this thread and get a life yourself.

Some of us just want good image quality in the games that we love. You might be happy with a 6-bit color TN panel, but many of us here would not.

 

[Bo_Fox]

The detailed moire pattern of the background is shown in the 4890 image, but not in the 5870 image. This is the difference between bilinear and trilinear filtering. It is bad enough that none of the details show up in the 5870 image within the invisible line. The colors of the inner circles still blend nicely, like as if it's still trilinear, but the moire background shows bilinear (hence "brilinear").

 

[Goodman]

Some of us just want good image quality in the games that we love.

Wanted an good image quality is fine but what you want is extreme & you probably wont see any of it while playing the game because everything goes so fast ... unlike a still picture (screen shot)