What's that plopping sound? Oh, it's ATI loading their drawers!

[BFG10K]

I'm a little confused as to the current state of ATi's newest drivers. On one hand I see people praising them almost everywhere and on the other hand I see extremely poor performance in the review sites in the current GF4 reviews.

Will ATi ever be able to write decent drivers?
Why does ATi take two steps forward and then one step back?
Why can't ATi produce drivers that run fast and look good? Why do they always have to trade one for the other?

 

[jfunk]

<< that still doesn't answer my question. let me try again. the geforce4 and the radeon8500 both have very similar features. the geforce3 lacks these features. geforce3 is about as fast as radeon8500. geforce4 with SAME features as radeon8500 is much faster than said radeon8500. why don't the advanced features (dual vertex shader, HSR, efficient memory architecture, etc) translate into incredible performance for the radeon8500 like they do for the geforce4? yeah, ATi's drivers aren't great, but i don't think they are responsible for this (at least not totally). is there an architectural flaw in the radeon8500? or maybe the architecture is fine, but there is a manufacturing process has a problem (like the savage2000's TnL not working due to hardware flaws). maybe it is the drivers after all.

[edit: i knew about the data fetch size difference, but i don't think that explains it all. my point is that people claim that the radeon8500 is so much more advanced than the geforce3, and that the geforce4 is only now catching up but the radeon8500 doesn't seem to take advantage of its features.]

--jacob >>



I think the key here is that those "extra features" of the Radeon are not particularly performance enhancing as much as they are just feature support. When newer games come out that support all those features, the Radeon will be able to handle it, where as the GF3 will not, since it doesn't have the proper support. So, the card is not necessarily faster, just supports more features. NVIDIA has is now releasing a card that supports all those features the Radeon does, but is faster, which it should be since it is comming out half a year later.
j

 

[giocopiano]

I'm with DefRef and Wag. Cards not working properly is no fun and nvidia do the better job of making their cards work. Experience is not just some word you know. If processors have errata, then so too will video chips and ATi must have a lot of them. Their drivers are like having too many people in a bed. One persons pulls the sheet over them and exposes someone else and there's no ultimate solution.
I'd think it would be a good bet that GF4 has less problems than the 8500 already.

 

[AA0]

DefRef...

Those drivers were beta drivers from ATI. They fixed bugs, but it killed the speed, the next official release is supposed to fix the speed problems with the betas.

I suggest you learn how to read if you can't find any major problems with the GF4 in those 4 reviews... as more than a couple are mentioned.

 

[Killrose]

Actually the sound you hear is from all the people who pay'd a premium for their Ti500.

 

[Rand]

<< ATI HAS gotten a lot of stuff fixed, but there are still problems, but I guess it's better to have buggy, broken drivers and look forward to vaporhardware than it is to accept that Nvidia executes damn near flawlessly (unless you count their pathetic Personal Cinema) >>



Right, the NMI parity error that's consistently bothering those with dually processors is flawless, the nv4_disp infinite loop error that's been around for god knowns how long is flawless. And you yourself mention the 'wonderful' Personal Cinema, the poor TV_Out unless you use TVTool is flawless.

nVidia is far from flawless, and nVidia has very severe issues with their drivers. Are they better then ATi in tha respect? Yes, far better.
But to call them flawless is very wrong as they quite obviously have a few very severe and widespread compatibility issues.

IMHO, nVidia has been slacking of in driver support. Once it would have been viable to say they were damn near flawless, but that's no longer the case and they most definitely have issues that have been going on all to long now.

And for the record, I personally think ATi has been making dramatic strides in driver support. Performance has clearly improved quite a bit from the early drivers, and game compatibility issues are at a minimum at this point. They still have a number of small issues with a couple games, and HydraVision still isnt 100% on the R8500... but at least their progressing.
They are at least good enough that I would no longer pass on purchasing an ATi product solely because of driver support.
The vast majority of people with an R8500 has been quite happy with it, the people with drastic problems are in the minority at this point.

I have an nVidia graphics board in my PC at the moment, and I have no regrets about choosing nVidia over ATi. It was the better undoubtedly option for me.
But ATi is not the satan people make them out to be, nor is nVidia the angel some would label them.

 

[AGodspeed]

As for the actual chip, the GF4 looked really sweet, until I looked closer. When you notice that the MX cards are the only ones that have all the features on them. The Ti cards are still just speed only, the dual display and dvd is subpar.

AA0, you need to read reviews before making such a ludicrous statement. Not one of them mentions subpar DVD or dual display capabilities.

If you read the reviews around the web of the GF4 you'll see how bad the drivers are.

Utter baloney. Tomshardware was the only review that I've read that mentioned anything about a possible driver conflict, and it was with the GeF4 MX and nView. However, Anand, NVNews, and HardOCP experienced zero problems running nView with the GeF4 Ti/MX. And they all used Athlon systems just like Tom did. In fact, Anand used the exact same motherboard as Tom, and Anand didn't experience any problems.

Tomshardware made a mistake most likely, but only time will tell.

 

[Rand]

<< As for the actual chip, the GF4 looked really sweet, until I looked closer. When you notice that the MX cards are the only ones that have all the features on them. The Ti cards are still just speed only, the dual display and dvd is subpar.

AA0, you need to read reviews before making such a ludicrous statement. Not one of them mentions subpar DVD or dual display capabilities.
>>


I think what AA- meant was that the GF4 Ti lacks the improved DVD that the GF4 MX has, which indicates that it is almost assuredly going to be lacking compared to ATi in this respect still.
I'm not sure what he meant in reference to the dual display capabilities though, they havent really been covered very in depth.
The nView/TwinView dual display capabilities of the hardware and driver feature set is still slightly behind ATi/Matrox's from what I've seen, but I havent heard of any problems... nor am I aware of any major issues with the the dual display functionality in the drivers.

 

[duragezic]

I don't know about the 8500 but I know for sure that the original Radeon's drivers still are *slow*. By that I mean the performance has hardly improved. In some cases in win2k it did with newer revisions (the much needed perf boost in UT) but I've seen driver comparisons from the oldest to the newest and they're all within a few fps of each other. I doubt that the drivers are good and that is as fast as the Radeon will perform. When I seen specs I thought the Radeon would be better than the GF2 and when I seen 8500 specs I also thought it'd be much better than a GF3, because technically it should but it just doesn't. Pretty disappointing.

And in Anand's article, kinda funny how "well this is fixed in X program but the performance dropped a lot in two other ones".

I'm satisfied with my Radeon, and the 8500 is looking a bit better as time goes on, but how long will it take and will it ever get really good? I've only had a few game compatibility problems so I'm more concerned with drivers that improve performance, but I'm probably just playing games that work fine.

This slow fog rendering stuff is still one of the worst, most disappointing parts about the Radeon. I don't even know whether it was determined to actually be fixable by drivers or if it was due to the hardware. Is it even fixed in the 8500?

 

[AA0]

yes, the Ti series doesn't have the DVD or nview capabilities of the MX. Looks like nvidia is screwing with the chips again.....

Tvout implementation according to sites still sucks.

One issue that I've had on a current dual monitor setup is not having the ability to play a DVD movie on my second display was from one site

The Tom's hardware article is definitely showing a bug, and like most bugs doesn't occur to everyone. Two people attempting the same thing, on 2 systems...

2 of the 3 I read had problems, not good. I didn't read the other sites since most of the time they just list what the feature is without testing them.

 

[spanky]

there is only one way to settle this issue about whether gf4 is better then the 8500. DefRef: paypal me $1000 and i will buy both cards and thoroughly test them and personally post u my findings. then i will give u the superior card, and find a way to go on living with the loser. :frown:

 

[FlamingPanther]

phooie, and I just bought an 8500, when the GF4 comes out and the R300 on the way.
eck, oh well, Time To Upgrade!

*Begin Rant*

I will never give up the superior 2D and DVD quality of my Radeon
And I don't give a crap if the GF4 can run at 1600x1200 with Quincunx AA,
because my flat panel's can only do 1024x768.
In my opinion the Radeon's AA image quality is far superior, even if i do go from 200 FPS to 90 FPS in Q3
my refresh rate is 75Hz and V-sync is always on.

So as I see it as nvidia's flagship card now supports a little over HALF of what the radeon 8500 supports,
and it's the half no games/software supports anyway!
(Specifically the second texture unit)

If Nvidia had improved Tv-out (even just a little)
or put HARDWARE DVD decoding (or even just Bob+Weave adaptive de-interlacing)
on the GF4, than I would've pre-ordered one right now.

Also, the Radeon's driver "issues" don't freeze my computer, they just look bad (like Q3A texture "bug")

The GF3 I had before either froze or ran perfect, not much in-between, I don't know about anyone else but
I can't stand to get to the next mission in AVP2 just to have it freeze again.
(Not to say the 8500 doesn't have problems in AVP2, they just don't freeze my machine)

In reguards to TRUFORM, the GF3 always has supported it though N-patches, just its never been implemented in drivers....
And what about the 20% performance hit the GF4 takes with Anisotropic Filtering turned on?

And that Nv4disp.dll thing? no less than three friends STILL suffer from that,
they have either bought a Radeon 8500DV, or gone back to a V3!(except one nvidia fanboy)


Wait, who was it that had crappy drivers? ATI? naaaawwwww

* end rant *

I agree with NicColt, All this is meaningless until we have a review of both parts.

And right on Rand!
Being a Fanboy gives us something to talk about!
If we had nothing to talk about, where would this forum be?

 

[AA0]

actually, the GF4 no longer supports a lot of options the GF3 does, including npatches at all. The performance hit for it is massive in the drivers. There is a list somewhere of the changes.

Thanks for someone else mentioning the nvidia driver situation. I have always preferred ATI drivers because when they fail, its not so bad. They don't kill my whole system, where as when nvidia drivers fail, they go down big, locking up the whole system isn't uncommon. Much like the reason I use Opera over IE for my browser.

 

[Rand]

<< actually, the GF4 no longer supports a lot of options the GF3 does, including npatches at all. The performance hit for it is massive in the drivers. There is a list somewhere of the changes.

Thanks for someone else mentioning the nvidia driver situation. I have always preferred ATI drivers because when they fail, its not so bad. They don't kill my whole system, where as when nvidia drivers fail, they go down big, locking up the whole system isn't uncommon. Much like the reason I use Opera over IE for my browser. >>



GF4 did away with support for n-patches?
Very interesting indeed, I don't suppose you recall where you read about this?

 

[Kerouactivist]

I have owned both geforce and radeon based products.

I do know that the ATI radeon card I have now definitely has some issues with windows xp and counterstrike even with the latest drivers and fixes. (I am very disappointed by this). Next time I upgrade I think I will go with an nvidia based card unless I stop hearing or at the very least myself stop having so many problems with their drivers. IMHO

 

[LuDaCriS66]

<< I have owned both geforce and radeon based products.

I do know that the ATI radeon card I have now definitely has some issues with windows xp and counterstrike even with the latest drivers and fixes. (I am very disappointed by this). Next time I upgrade I think I will go with an nvidia based card unless I stop hearing or at the very least myself stop having so many problems with their drivers. IMHO >>



I have visual problems with my Gainward GF3 Ti200 in CS too.. personally, I think CS looks way better on my old Radeon board than my GF3.

 

[FlamingPanther]

Link to the best review i've read concerning the 8500's Visual Quality
http://www.tech-report.com/reviews/2001q4/vidcards/index.x?pg=1

 

[AA0]

took a while, found it.
digit-life has it.

"The drivers of the NV20 and NV25 do not support hardware tessellation of smooth surfaces (HOS based on RT-Patches). When a card doesn't support N-Patches on a hardware level the API tries to emulate them using RT-Patches. It makes operation of N-Patches very slow. NVIDIA thus had to disable the RT-Patches so that games supporting N-Patches won't be too slow."

some other things I found odd...
NV25 doesn't support indexed matrix blending like the NV20 as the shaders can help organize flexibly any schemes of matrix blending.

NV25 doesn't allow compressing 3D textures. It is a bad drawback of the drivers or the chip. At the same time the OpenGL drivers from NVIDIA have their own 3D texture compression format.

There are a lot of other things that just seem like they didn't do aswell as they could, maybe there was a hardware limit or something, doesn't make sense that they didn't bother to upgrade the features to DX 8.1

 

[Rand]

<< took a while, found it.
digit-life has it.

>>



Thanks a lot, I appreciate your taking the time to find it for me.

 

[tazdevl]

<< phooie, and I just bought an 8500, when the GF4 comes out and the R300 on the way.) >>



R300 hasn't even gone to tape out yet. Wouldn't exactly say it's on the way.

bthorny, the only people that are having issues with CS are those that don't know what they're doing or haven't bothered to search for answers. There are 2 patches for CS you can download from sierra.com. A half life patch and CS patch. Once I installed the second my overflow issues stopped.

The other problem is refresh rate related. ATI drivers lock the refresh rate @ 60hz. I found a great program called RefreshLock. Small memory footprint and doesn't mess up your registry like refreshfix. Be sure to download your monitors .inf file and install it via the device manager. Once you do that, you'll be able to play CS @ 120 hz instead of 60.

Let's hear about all these problems. I'm sure that there are simple solutions out there. Rage3D is a, ATI enthusiast site, lot ofl folks are very knowledgeable about ATI products. Post your issues there if you aren't getting the answers you need here. FYI the CS team did a crappy job with Truform... type ati_npatch 0 in the console to turn it off.

 

[BFG10K]

Those drivers were beta drivers from ATI. They fixed bugs, but it killed the speed, the next official release is supposed to fix the speed problems with the betas.

I hope that is the case and I wait with anticipation to see if it happens.

 

[FlamingPanther]

R300?
whoops, sorry,
I meant to say R250

(but they are making an R300, right?)

 

[moocat]

Can someone clarify something for me concerning the 'lightspeed' memory controller?

Which layer of software is manipulating the memory controller? Is it at the drivers level or in the game code? My impression is that unless software is optimized to take advantage of the impressive flexibility of the lightspeed controller you aren't going to see any performance gains.

I could be completely off on my understanding of how the controller works so please enlighten me if I am.

 

[BFG10K]

It happens at the driver level just like Z-occlusion culling; it's totally transparent to the games.

The drivers optimise requests that require memory bandwidth by spreading the load across the controllers in a more even fashion.