When enough is enough (AF quality on g70)

[ChrisRay]

Well it is about time this was addressed. Hopefully the fix will come soon.

It will interesting to see if there is much of a performance hit with the fix.

I think some might be surprised by how relatively minor it will likely be.

 

[Randell]

Oh and if you want the worth shimmering ever, play on a 8500/varients and fire up neverwinter nights.
It was dreadful in neverwinter nights and no driver option reduced it all.
It wasn't nearly as bad in other games, but NWN looked like horrible.
With AF and without btw.

I always said NWN was the only game I played on my 8500 where the angle-dependancy of the 8500 AF was so noticable. So bad that pure tri was better than AF-Bi.

However the cause of the 6xxx/7xxx shimmering is different, it's the way optimisations are applied to mip-maps and the way those gpu's handle LOD bias.

Default 6800 AF was noticably worse to me than my 9700Pro and HQ was worse than 9700Pro quality AF, then you added the CoH and Bf1942 showstoppers in the LOD handling dept and I almost took my 6800GT straight back.

 

[Moloch]

I always said NWN was the only game I played on my 8500 where the angle-dependancy of the 8500 AF was so noticable. So bad that pure tri was better than AF-Bi.

However the cause of the 6xxx/7xxx shimmering is different, it's the way optimisations are applied to mip-maps and the way those gpu's handle LOD bias.

Default 6800 AF was noticably worse to me than my 9700Pro and HQ was worse than 9700Pro quality AF, then you added the CoH and Bf1942 showstoppers in the LOD handling dept and I almost took my 6800GT straight back.

I didnt notice a different in shimmer even if disabled AF

 

[Rys]

If i would thank anybody it would be tridam. The only one who had the guts to put the bad word into his review.

Well, some of us said we were looking in to it. I did in my GTX article and again in my GT preview. At the time I was waiting for a response from NVIDIA (and still am), and wasn't quite sure whether undersampling was what they were doing, compared to other drivers.

I don't quite have Damien's skills when it comes to seeking the raw details of what was going on (until recently when I figured out how to check), but I've definitely been looking into it, and have said as such.

He does deserve the credit, though.

 

[Ailuros]

I don't recall anymore whether I published in June or July 2004 this one:

http://www.mitrax.de/?cont=artikel&aid=24&page=14

Sadly enough out of the batch of games I tested back then, UT2k4 was the only game that uses a negative LOD bias for textures.

Tridam definitely deserves the credit, but some of us weren't exactly "blind" for quite some time now. I still haven't finished building my system to host a G70 (lame excuse of the day ).

 

[Mendel]

to misquote a movie I recently saw...

When someone likes particular ihv's graphics products very much but things don't quite work out as planned, when do you reach the point that enough is enough?
-Never.

 

[Graphics_Krazy]

Can some one with a 7800 try the mipmap tool (from 3dcenter downoads) with 8xAF activated to confirm these results

1) 8xCP OGL is bilinear
2) Changing LOD levels shows weird popping behaviour at low levels - could this explain the shimmering??
3) in HIGH QUALITY the switch to lower mipmaps on the entire screen is immediate

There is also a tool for D3D mipmaps
4) In Control panel 8xAF, D3D stages 1-7 are all bilinear in quality mode
5) Changing the LOD levels shows wierd popping again at low LOD's
6) in App 8xAF, at extremely low LOD levels the entire screen begins to "pop" - its even sooner in HIGH QUALITY mode.

 

[caboosemoose]

Well, some of us said we were looking in to it. I did in my GTX article and again in my GT preview. At the time I was waiting for a response from NVIDIA (and still am), and wasn't quite sure whether undersampling was what they were doing, compared to other drivers.

I don't quite have Damien's skills when it comes to seeking the raw details of what was going on (until recently when I figured out how to check), but I've definitely been looking into it, and have said as such.

He does deserve the credit, though.

Well, with all due respect, unless I've missed something in the Hexus reviews (and i may have done i ahvent read the whole of either piece), it would be impossible for someone who wasn't previously aware of the shimmering effect to glean it from the Hexus reviews. In the GTX review your conclusion states "Only lacklustre video support (especially with my recent acquisition of a Sony PSP and a big desire to playback MPEG-4 AVC video on my PC) spoils it for me somewhat."

And you conclude thusly on the GT:

"There's not really much more to say. It's the same price as ATI's current best, performs as well in games, and does significantly better in almost every other facet that you'd consider. More forward-looking features, less heat, less power consumed. The victory is a sweet one."

As far as I can see we are dealing with cryptic references to talking to NVIDIA about filtering - no mention of poor IQ as such and no specifics. That's not much help to readers who'd like to have a good idea of how the 7800 GT/X currently performs. To be clear, I have done no better myself - an unfortunate oversight. In general, the 7800s have been somewhat overrated IMHO, not least in comparison to the improvement in performance that arrived a little over a year ago. The 7800 GTX is often only marginally quicker than the fastest previous board and remains bandlidth limited and offers relatviely poor IQ in some instances. Still, it's NVIDIA and I guess a lot of people want to keep relationships sweet. I'm not making any specific accusations, so don't get me wrong, but I'm confident the reviews would have been different as a group if no-one cared about keeping relationships well maintained. If you read most of the reviews, you'd think the 7800 was perfect.

 

[Jawed]

I think it should be forum rules that if anyone writes articles for other review sites (don't need the word "competing" here) then their signature or at least their profile should indicate which sites they are.

Which reviews have you written Caboosemoose?
Jawed

 

[caboosemoose]

Sorry, but I can't agree with the idea that people should have to say who they write for if they do. Anyway - it's totally unenforceable, people would wanted to remain anonymous would simply start posting under a new handle, and therefore pointless.

I'd rather remain more or less anonymous in general, there's no benefit having identities known that I can think of, so why create risk?

 

[Jawed]

What risk? You're so pretty that London Boy will start stalking you, or something?

Jawed

 

[Rys]

I'm not sure his points apply, unless he's talking about his own experiences working with NVIDIA, which I'm sure he's not. I'm not sure how you can discuss the relationship or otherwise with an IHV for publications you don't work for He's fishing.

The sites I'd trust to review the hardware properly really should have no qualms about writing NVIDIA up in any way they see fit. That includes the one I write for. The relationship is absolutely not maintained in the way I think you're insinuating it is, I hope you understand that. I certainly don't assume you maintain your own relationship with them in that fashion

Yeah, I could have made specific mention of the filtering issues I wanted to look at, rather than just say "I'm looking at something, check back later", but in a couple of the cases I wasn't even sure what the issue was at the time, and it took further analysis of what was going on to pin it down. It's hard to know what to write in that case. "Something is up but I'm not sure what" isn't something I'm keen to publish. I'd rather "I'll look into filtering later". Which I have done The language to use is hard to decide on.

At any rate, sites are undoubtedly quite close to finishing their investigations and writeups (if any) into the issue, why it happens (if indeed it's understood), how to fix it and what NVIDIA are going to do about it (that's become clear now, or at least 95% clear, at least to me).

 

[trinibwoy]

In general, the 7800s have been somewhat overrated IMHO, not least in comparison to the improvement in performance that arrived a little over a year ago. The 7800 GTX is often only marginally quicker than the fastest previous board and remains bandlidth limited and offers relatviely poor IQ in some instances.

That's a bit harsh and misleading as well. The 7800's not only provide significantly higher performance in more situations than not, but they do so at lower power consumption than the last generation which is impressive in itself. Availability and pricing are also responsible for the high regard for the 7800 series.

Considering I bought a 6800GT earlier this year for $330 and 7800GT's are now available for $400 I find your statement quite ridiculous actually.There's a lot to like about the 7800's in the absence of true competition. We'll see how it measures up once ATi's feature competitive products come to market.

 

[caboosemoose]

One other thing that has crossed my mind and may account to some degree for reviewers' reticence to put the boot in over the the filter thang goes something like this:

1. Reviewer finds IQ/performance issue
2. Reviewer goes to town, making a big deal out of it
3. Chip maker releases new driver that fixes the problem at the same time as or soon after review is published
4. Reviewer looks just a little silly for kicking up such a fuss

I remember when the 6800 was released and CryTek hadn't released a patch that properly supported the 6800 and it was running the path with the b0rked partial precision shader quality a la GeForce FX and many thought, "oh dear, is the 32-bit performance still awful" etc and that turned out to be largely a false concern.

Even if a problem seems straightforward and unambiguous it often pays to take as eve handed a line as possible I guess. All pure specualtion, and probably not often a conscious decision, but just a thought.

 

[caboosemoose]

That's a bit harsh and misleading as well. The 7800's not only provide significantly higher performance in more situations than not, but they do so at lower power consumption than the last generation which is impressive in itself. Availability and pricing are also responsible for the high regard for the 7800 series.

Considering I bought a 6800GT earlier this year for $330 and 7800GT's are now available for $400 I find your statement quite ridiculous actually.There's a lot to like about the 7800's in the absence of true competition. We'll see how it measures up once ATi's feature competitive products come to market.

Don't get me wrong, the 7800 is great, just not Jesus Christ made manifest in pixel rendering silicon as some websites/mags seem to have suggested. One high profile site intro'd a new section called had a section called "NVIDIA kicks ass with the 7800GTX".

Does anyone here think the 7800 raises the bar as much as the 6800/x800 did last summer? I just think that some perspective is required, that's all.

 

[Jawed]

I think the reviewer should observe, describe it to their readers and discuss whether it's likely to impact a purchase decision. Particularly as NVidia has something of a record in foisting substantially below-par IQ on its users.

If a reviewer wants to bitch about NVidia's lame driver release policy, leaked betas and the rest, then feel free too.

Has the HL-2 transparent-AA bug been sorted? Will it be sorted before xmas? Ever?

How many drivers have there been since 7800GTX was released?

I should say at this point that B3D itself seems rather reticent about investigating shimmering problems. I hope Dave is, right at this moment, putting comments on this subject into his 7800GT review.

Jawed

 

[caboosemoose]

I'm not sure his points apply, unless he's talking about his own experiences working with NVIDIA, which I'm sure he's not. I'm not sure how you can discuss the relationship or otherwise with an IHV for publications you don't work for He's fishing.

The sites I'd trust to review the hardware properly really should have no qualms about writing NVIDIA up in any way they see fit. That includes the one I write for. The relationship is absolutely not maintained in the way I think you're insinuating it is, I hope you understand that. I certainly don't assume you maintain your own relationship with them in that fashion

Yeah, I could have made specific mention of the filtering issues I wanted to look at, rather than just say "I'm looking at something, check back later", but in a couple of the cases I wasn't even sure what the issue was at the time, and it took further analysis of what was going on to pin it down. It's hard to know what to write in that case. "Something is up but I'm not sure what" isn't something I'm keen to publish. I'd rather "I'll look into filtering later". Which I have done The language to use is hard to decide on.

At any rate, sites are undoubtedly quite close to finishing their investigations and writeups (if any) into the issue, why it happens (if indeed it's understood), how to fix it and what NVIDIA are going to do about it (that's become clear now, or at least 95% clear, at least to me).

I think you're taking what I am saying a little too literally. I'm not suggesting a bias, which everyone is so petrified of, or being NV or ATI or anyone's lapdog. But, if you are claiming that you never give a second thought to the reaction of NV/ATI/Intel/AMD whoever to a negative comment you are considering putting in your review about their products, well, sorry, I don't believe you.

As for the language you used, for me it's the perfect example. If there was no-one to offend, why not say : "this looks a little shitty, not sure what is going on, might be nothing after all, but currently looking it." But there is someone to offend, so therefore you choose to tell readers that you are currently looking into filtering (without even going so far as to say that you are looking into it because there's a problem - in the bits that I have read it's ambigously put enough to make it almost possible to think you are looking into NV's filtering because it's so damn fine!) and will report back, will allows you to serve your readers to a some degree by flagging something without pissing off NV if the problem turns out to be transient / non-existent. Look, I don't mean to offend, I'm just playing devil's advocate.

The other way to look at it is that if you are trying to be as fair and even handed aas possible, that pushes people towards not taking extreme views, and taking care with making judgements until all the facts are in etc.

I am not fishing.

 

[Randell]

I didnt notice a different in shimmer even if disabled AF

well it was rock solid tri for me, no shimmering, but a loss of detail. This was before the free camera added in a later patch, so maybe I never got the same camera angle. I would say I never actually saw shimering on my 8500 in the same way as on the 6800, just badly messed mip-map transistions in NWN as you moved around, with soem sections of floor blurry and some sharp.

 

[Moloch]

well it was rock solid tri for me, no shimmering, but a loss of detail. This was before the free camera added in a later patch, so maybe I never got the same camera angle. I would say I never actually saw shimering on my 8500 in the same way as on the 6800, just badly messed mip-map transistions in NWN as you moved around, with soem sections of floor blurry and some sharp.

Well it was just way to sharp to allow for good looking textures.

 

[caboosemoose]

What risk? You're so pretty that London Boy will start stalking you, or something?

Jawed

Exactly.

In all seriousness, the risk is the same as any time you open your mouth in public. If there's no need for it to be public (ie because you can remain more or less anonymous and therefore cannot be construed as speaking on behalf of your publication) then it's much more difficult to say something you might later regret. Absolute anonymity isnt really the issue, but I'm sure many appreciate being able to speak for themselves and no-one else, and anonymity certainly helps one do that.