Tigre Marino

If being the 85% of the Windows ecosystem like XPSP2 is now isn't relevant... hello!!! I don't know what that is. Face it: Windows Vista is a mediocre, bug-ridden, slow, slouch and DRM-infested operating system. There's nothing you can do on Vista that SP2 can't (except playing games with DX10 which has been overhyped both in IQ and performance). Vista is like one fuel-thirsty SUV that rolls over without apparent cause (like the Ford Explorer). Now we now why Japanese, Indian and Korean auto industries are crushing American ones, if they build cars like they now program operating systems like Vista. If Windows 7 isn't as good as Windows XP, I'm moving to MacOS or Linux, because Windows Vista is like Windows Millenium Edition 2.

arfster

One key fact explains 99% of the differences observed in the article: by default, ATI 2xxx/3xxx series cards expand the colour space to PC levels (ie 0-255), presumably because they think most people will be viewing on a PC monitor. Note this doesn't apply to SD - ATI bizarrely expand HD but not SD, and you have to apply registry tweaks to get them to behave consistently (I've been complaining to their tech support for 9 months about this, but no resolution).

For a fairer comparison with Nvidia, I'd suggest reversing the expansion via brightness16/contrast86 (this uses the same bt709 method as the original expansion, so it's near lossless).

arfster

ATI's denoise only works for interlaced material, and only with the more advanced modes of deinterlacing selected(motion-adaptive & vector-adaptive). It's also disabled for all HD in recent drivers, along with sharpening - whether by design or not I've no idea. Their drivers are very buggy, things regularly get broken/fixed/rebroken every few driver cycles.

osho_gg

Thanks for a very thorough review.

I noticed in the Nvidia Control Panel's screenshots that inverse telecine was unchecked. I understand that the default is to enable it. I was wondering did you notice any difference in your tests with this setting enabled/disabled? and if so, what?

Osho

Eeastcoasthandle

I completely disagree with the conclusion. IMO based on reading the article, the IQ & performance wasn't properly weighed out. When attempts were made to adjust video or color settings to the nvidia card to make it as good as ATI. Should have been the first indication that ATI IQ is better. The second is when the tweaked settings for the nvidia card had negative side effects should have also weighed heavily into the conclusion. When most people haven't the slightest clue on how to tweak those settings the nudge should have went to ATI. So what that Nvidia offers playback with Aero, I have no use for it. The direction of this review is about IQ and performance not which offer more features. If ATI clearly demonstrates that the IQ and performance is better without playback using Aero it should have won with a cliff not stating that in your opinion it's a feature worth adding.

Zardon

And?


This is the last time I will dignify these questions with a response. This has already been answered in the thread (and in the article itself) and it was stated that these are the settings ATI stated were the best. Obviously as they design the software and hardware then we would assume they would know, not yourself. Also if you feel this article wasn't "properly weighed out" then I suggest you perhaps test them yourself, publish it and post in here so I can see how you can improve on capturing raw data from the cards then analysing the raw pixel breakdown.

Rather a sweeping statement isnt it? We mentioned that certain aspects of the IQ are better, but not all. I can therefore already assume you have either a: glanced over it or b: not bothered to correctly read the whole editorial.

Why? they can be turned off and aren't the default out of the box settings.

Well I guess you just answered your own question. If people don't know how to tweak the settings, then the enhanced Nvidia settings which AREN'T default wouldnt be used.

Well as difficult a concept as it might seem, you are just one person. Are we to start reviewing products and omitting bugs and issues because some people might not care? If this is a condemning point for you then I suggest you perhaps start frequenting another site which omits issues based on loyalties to specific manufacturers because DH always tells it like it is, regardless of the company involved.

Really? I wasn't aware you were in charge of handling editorials and the direction of the content on DriverHeaven, I thought that was my job.

This article was ascertaining IQ, however during this process if we find issues relating to the playback of this content we will mention it. It seems to me that quite honestly you have no interest in an unbiased article but are more concerned with the fact we should be ignoring issues, if they are relating to ATI or their products.

Well thats the point, it didn't. Quite what a cliff has to do with anything however .......

Incidentally if I see these same questions repeated again and again in this thread they will be removed. We have no time to be readdressing the same mindless questions which have already been answered in both the editorial and this thread.

Zardon

We are actually aware of this, however this is a real world article and we are detailing the relevant image quality with basic out of the box settings, or settings that either ATI or Nvidia have recommended we use. (there is a reason you have no resolution from your reports to ATI, they dont feel its a change people should be making).

Like everything, whether it is hardware or software there are always tweaks and (possibly subjective) improvements an educated enthusiast can make to have something run better (bios, registry etc). However for the majority of end users this will be irrelevant as they will never know a tweak like this, therefore it is pointless testing. The fact that ATI told us not to alter registry settings or panel settings beyond what we used is the only feasible way we can analyse the products. I mean we know some tweaks for Nvidia also, but we didn't use them either for exactly the same reasons.

If ATI feel the settings they provide are the ones that people should be using then this is the way we test it, the same with Nvidia. If we were to analyse something we had tweaked significantly to the point that only 1% or less of the populace would change then this becomes a totally irrelevant article.

Alex

Yeah their drivers are very buggy, I noticed things working or breaking on a regular basis between different driver revisions. Must be a nightmare for reviewers to get results when the drivers are really so poor at the end of the day.

Veridian3

On our test system the default was unchecked. As for the differences when used, for the tests we were doing the option has no impact on the results.

pouyoux

I've find your article very very relevant to what lots of people are wanting to know about video cards.

I'am totally stunished about the lack of transparency in the process of decompressing HD material ..

I've got 2 questions :
- why not to include in this article a third concurrent with the picture you get in 100% software rendering mode ?
- do you plan to do the same thing with other video formats (xvid, mpeg etc etc), to verify if ati and nvidia have the same footprint in decoding those files ?

Veridian3

Thanks for the feedback.

With regard to the software rendering, we will consider that for the next article. I can see it making some interesting comparisons.

At this time we don't have any plans for other video formats, if anything i would be thinking about .ts or .mkv performance rather than xvid/mpeg etc. But once again we will keep it in mind for the next video article we do.

pouyoux

as almost all mkv are using x264 stream, I think you already have results in the "casino royal" part of your article.
But comportments of those cards with SD (small definition) movies like DVD, xvid files would be interesting.
Another interesting thing would be to do the same article under Linux, as Linux is a valuable platform to make Home Theater PC.

Veridian3

At this point with ATI/NV i would take nothing for granted about where results would be, performance of the system would also be interesting in my opinion.

pouyoux

as Casino royale is h264 encoded I don't see what could makes a difference with a h264 stream in a mkv. But you can show me I'm wrong

as you said in your conclusion that none of the nv/ati decoders are perfect, I am on hurry to see/validate that software rendering just do it.

temeteus82

btw. As far as I know there is only DivX acceleration on ATI cards. And mostly Xvid decoding is done by the cpu so there aren't much to test in my opinion...

__________________

Judas

that's irrelavent due to the fact that when xp was out for just over a year, the mass amount of users still using windows 98 was roughly the same. and the same claims were made (sp1 would have been either just recently available as well, or becoming available soon)

Everthing you've stated about vista is in FACT.... completely false/inaccurate/laughable.

There's simply nothing to be discussed, Vista is the OS to use... there isn't any need for XP any more.

__________________

Vikingod

Lets keep the discussion to the articel at hand, please. Irrelavent replies should be ignored. No need to justify them, their merit is obvious to everyone reading the thread.

Great article Allan and Stu, it was very informative.

skibum5000

I thank you for the report and there is much valuable and useful information, however, I fear that a very major and fairly basic error was made in the IQ comparisons.

Doesn't nvidia currently auto expand levels while ATI does not?
edit: ati apparently does too with recent drivers for HD, maybe your special method used older drivers scren captures, or more likely looking at color values it seems like you black crushed nvidia by doing something to make it carry out levels expansion twice.

below still applies only it is now likely that everything I said needs to be scaled by another 5%, with nvidia getting black crushed and ATI arriving at proper expansion levels (makes sense since it didnt look that faded to me as 15-235 would've).
So perhaps below is what happend, but it seemsmore like you had ATI 15-235 expanded to 0-255 and nvidia already expanded and then expanded again.
Did you set some something weird somewhere? I can watch that bluray on my nvidia and not get double expansion (i have all sliders neutral default). Fine on my system.

No wonder you found the mysteriously 5% deeper blacks for the nvidia because guess what 5% off 255 is... 13 and guess what the offset is for levels expansion 15! That is all it is. NVidia, as of the last few months of drivers automatically stretches 15-235 to 0-255 so you should view the images with TV set to handle 0 as black and 255 as white but with ATI you should set the TV to expect 15 as black and 235 as white. The ATI way does give you a little leeway to adjust things yourself and allow for a little blacker than black and whiter than white signal, although technically beyond spec there can sometimes be some info there and it is up to you if you a little lost contrast and more noise or a little more detail into black and white, but strictly speaking all that data was meant to be cut off if you exactly follow the standard. If you watch ATI as if 0 is black though you are getting a great loss in contrast, saturation, image pop and muted black levels. If you watch NVIDIA with 15 as black you are getting lots of black and white shades clipped off. If do it in reverse, then both look as they should.

If you do the test properly (by, for instance, setting normal when viewing the nvidia images and low when viewing the ATI ones for black level on HDMI input or by stretching 15-235 to 0-255 in CS3 for the ATI images) the luminance are pretty much the same on both cards! There is no detail loss on Nvidia, they look pretty much the same! And the ATI is not mysteriously missing pop and saturation, again it's pretty much the same!

The color balance is different though and that is weird and an important finding (note the level balancing enhances the apparent difference and increases apparent red strength for the expanded versions but it still very much there).

skibum5000

but what if nvidia happened to have expanded 15-235 to 0 to 255 before your capture while ATI had not, this appear to be the case. Then everything regarding saturation, rightness, black and white details will be wrong. Maybe you accidentally had things set so as to trigger double levels expansion for NVidia and have ATI correct (EDIT: this is what happened for sure, you are viewing ATI at proper levels expansion and nvidia with a DOUBLE expansion applied due to who knows what, but it does not do that on my machine using any of three different drives, inlcuding latest beta and certified and an older one). Anyway, something was messed up. The two cards, in fact, put out pretty much the same levels, if not color balance.

arfster

Yup - downloading and looking at the screenshots, this is exactly what's happened. The ATI images have been expanded to 0-255, as their drivers always do with HD, while the Nvidia ones are way beyond that. Most likely explanation is a double expansion by the Nvidia drivers, which I have heard of happening occasionally.

IMO the Nvidia ones are unwatchable, even on a PC display. All dark detail is totally wiped out, like watching a badly encoded xvid.

Tigre Marino

Only if XVID codec is installed. DiVX can takeover on any generic MPEG-4 decoding if there it is not a competing codec installed in the system, so if the video card supports DXVA (DirectX Video Acceleration) and the driver supports it, DiVX can use this support to enhance the experience.

Tigre Marino

The same happened with Windows Millenium Edition... and I've seen tons of people "degrading" their new Vista laptops/desktops to Windows XP, which is more of an upgrade, btw. Even I did with my new Gateway MX6947M notebook using the license from my previous dead laptop.

I rather use Ubuntu or MacOSX if there it is no choice to use XP.

arfster

Vddobrev over at avsforum has a suggestion. Apparently Nvidia's 174.74 expand just like ATI normally, but .....

"Edit: I think I know how they triggered double expansion. In the nVidia control panel, there is a setting "Select HDMI color format" - RGB or YCbCr444. If they selected YCbCr444, then this is how they got it."

Pretty nasty bug.

Veridian3

I'm going to ignore most of the recent comments above, we have explained the answers to these several times now.

As for the YCrCb comment, when i was starting the testing i compared the output of that option compared to RGB and don't remember seeing any real difference in IQ on the scenes I tried. I will go back and check again over the next day or two and confirm the results using the two settings. However, if there is a bug in the NV driver it doesn't invalidate the results/testing. The images still reflect the output of the Nvidia card and if all NV need to do to improve things is fix that bug, then great.

Once again, please remember... the point was not to get perfect IQ through our own configuration... it was to test what each card is currently outputting... bugs or no bugs... these images are what consumers will see.

Zardon

As stuart said if there are nvidia or ATI driver bugs, it will be exactly what the end user will see - we mentioned everything we noticed. So it is relevant. We have answered many of these questions earlier in the thread relating to driver settings, why they were selected and the input we have received from Nvidia. Changes are being made by them already for future revisions.

Perhaps we could follow the article up with our own "optimised" conditions - making registry changes, changes to the driver etc, to get multiple results from both sets of hardware, however as it stands our testing basically shows what the end user will see with settings out of the box as well as settings that nvidia asked us to use (ATi said default was their best). People in this thread who are attempting to ascertain quality comparisons via possibly uncalibrated or incorrectly calibrated screens are wasting their time, we could have done that eons ago, but we have analysed colour breakdown from raw captures directly taken from the hardware without anything else in the mix (panel, calibration, ambient light etc) to taint results.

While we appreciate the interest in the article, much of this appears to be going in circles with many people debating the supposed merits of outdated operating systems and other such irrelevancies. If you have any questions which weren't detailed in the article or in this thread then drop us an email, however if they have already been answered then we really won't be making the effort to reply to them. So please make sure to read everything beforehand.