GTX 460 Reviews

What if the hack on ATI matches rendering on XB360?

I'm not saying it does, nor that the hacked IQ is preferable.

Seems NVidia needs to make more of a fuss about this.

 

All of this confusion wouldn't exist, if ATI provided something comparable to Nvidia's "High Quality" setting without having to turn off A.I. altogether. If reviewers maybe started to give ATI some shit about it, something might happen. This has been a problem for at least 3 full GPU generations now, for fucks sake, and it's one of the two reasons why ATI doesn't even enter the picture when I'm making a video card buying decision.

 

your right , ati should hide all optmisations and not allow users to turn them off.

 

This is a common misconception, and the reality is actually worse. Turning off Cat AI does not disable ATI's texture filtering "optimizations." It just kills all the shader ones and for some asinine reason seems to adversely affect video decoding in some situations. Here is how I would rate them (on what the end user actually sees on his screen) currently:

1) NV HQ (since the 8800)
2) NV Quality
3) ATI 5xxx
4) ATI 4xxx
5) Old hardware and "performance" options (they all suck just some more than others)

 

That's kinda what they're doing right now with regard to filtering. What's your point again?

 

nvidia has thier own optimisations that can't be turned off is what my point is. At least ati allows some of them to be turned off.

 

From what I understand, you can turn off more optimizations in the Nvidia drivers than you can do in the ATI drivers, at least without disabling AI altogether, which no one wants to do and which doesn't seem to work anyway. At any rate, the general consensus seems to be that Nvidia's HQ mode results in a better filtering quality than anything ATI has to offer and that's what's important.

 

I got my Sparkle GTX 460 1024mb card today. It's a reference board, blue PCB.

It currently seems to run stable (~ 25 minutes Furmark) at 890/1780 with 1.05V but I'm still testing. Oddly enough, the card doesn't seem to like higher voltages. At 1.087V the card doesn't run at these clock rates, at 1.05 it does - even though GPU temperature is still in the 70°s C with max voltage.

I haven't played around with the memory clocks yet. So far, I'm quite happy with the card. The cooler is amazing compared to the noisy blower of my GTX 260.

Does anyone know how to remove the plastic cover? I tried but nothing worked... I didn't try brute force, though.

Pics:

 

It all flip flops depending on personal bias. When the ati cards had better flitering quality all that mattered was fps . Now that ati has the speed advantage all that matters is filtering quality.

Its the same silly bullshit that keeps going on everytime new cards are released.

 

Yep Yep +1..

 

GPUs have been at a performance level that's more than sufficient for great FPS + excellent image quality for several years now. There is simply no excuse for idiotic filtering shenanigans. The 58xx cards are impressive, their performance is stellar. Why can't we have great filtering, too?

Nvidia gives me the choice. I can have great filtering quality or I can reduce filtering quality and get more FPS. The way this is presented in the drivers is pretty transparent.

It's not about FPS vs IQ, it's about having the choice. ATI doesn't give me a good choice in that matter, they prioritize FPS over filtering quality and force that decision down their customers' throats.

Please don't reduce that to some GeForceFX vs Radeon 9700 debate, where one had performance and the other one had better IQ and fanboys were bashing each others heads in. We can have both now, and it's simply ATI's choice not to provide both. They easily could.

 

Interesting, so the comments offered by Novum and his AF tester is silly bullshit?

The way I see it is there are some posters that are consistent about having quality filtering and allowing flexibility for it.

 

Umm... no. Shimmering is shimmering is shimmering. Personal bias doesn't have anything to do with it. If the output on my screen didn't annoy me, there wouldn't be anything to complain about. At this point it is pretty simple for me, either ATI can fix their @$#!, or my next card will be Nvidia regardless of the price/performance ratio, power, or heat considerations. It has been six years, and I'm tired of waiting.

 

The shroud/HSF is removed as a one piece assembly. It is held on by the four screws of the metal heatsink retention bracket on the back of the card. Once you remove these, the entire shroud/HSF assembly comes off the GPU. You can then access four more screws securing the shroud to the HSF once it's off the card.

Have you tried the 258.96 WHQL drivers? If so, do you observe any issues with stuttering in games, NVCpl working smoothly, panel scaling, wake from sleep?

 

Thanks.

1. Yes
2. I haven't played any games yet, I'm still OC'ing, running benchmarks etc.
3. I've been using nHancer for years, so I don't know how "smooth" the NVCpl ist supposed to be
4. Panel scaling works for me but that really doesn't mean anything. Has anyone actually pinned down the scaling problem? It only seems to affect some people.
5. Wake from Sleep never worked for me, probably some driver issue or something

 

I think the current consensus about ATi's filtering quality is sligtly exaggerated and keeped in this state by some nVidia's fans. I'm not saying, that nVidia's filtering isn't better. It is. But the difference is not as critical as some people admit.

R300/NV30 filtering was different story. R520/G70 were totally different leagues - first one with fast angle independent filtering with quality sampling, the second one slow with heavy angle optimizations and heavy undersampling, which caused moire, shimmering and even distorsion of some fine-detailed textures (missing patterns etc.). Both RV870 and GF100 offers angle independent filtering, none of them produces the horrible moire, which G7x did nor the blurry surfaces of NV30. 10 milions of R8xx GPUs were shipped, but nobody was complaining - till the launch of GF100. GTX470/480 arrived, it was noisy, hot, slower than expected and it offered rather bad value for the money. The only thing, which was slightly better compared to RV870, was texture filtering, so nVidia fanboys decided to overhype it... Now it appears, that the slight difference in TF, which is nowhere close to the difference between R300/NV30 nor R520/G70 is worth the noisy, power hungry and overpriced product...

There's also another important difference. NV30 was capable of better filtering, but its quality was decreased in drivers to keep up with R300 in performance. G70 was capable of better filering, too, but the quality was also decreased in drivers to be on pair with R520's performance. But RV870 was the only product on market. There wasn't any other DX11 GPU or GPU of similar performance. It had no competitor - ATi had no reason to decrease filtering quality to get a few percents of performance. ATi didn't decrease filtering quality even after launch of GF100, so I think it's not a software optimization, but a hardware characteristic. RV870 has 80 texture units - the GPU shouldn't be short of texturing or texture filtering power. Software filtering optimizations also don't help much to decoupled texture units - performance gains are really subtle when compared to the old fixed-pipeline GPUs. I don't believe that the difference in filtering quality could be caused by performance optimisation.

ATi admitted, that they removed about 20% of tranzistors from the original R8xx GPU design to make it manufacturable by TSMC's 40nm process. I think it affected texture filtering, too. It's quite likely, that next generation of GPUs won't be affected by this eleventh-hour decision and texture filtering will be implemented in a bit better way.

 

It's apparent that both ATI and Nvidia use different methods in IQ along with optimizations in which profiles are created. Because of those differences it makes no sense to expect one company to use the exact same techniques of another company because you think they should. Because of this understand IQ subjectiveness should be omitted from performance reviews and use similar control panel settings.

If the reviewer thinks that the IQ of one is better then the other it is their job to prove it in their performance review. However, that slippery slop will be evaluated to see if the reviewer is simply be subjective. Kinda funny IMO, claims are made saying one offers better IQ then another but rarely (if at all) is it shown during the benchmark review. To only find a sparse IQ comparisons using old games like Half Life 2, etc.

So to conclude, control panel settings should be the same regardless of how subjective a person is about them. So that a level playing field is created to show performance results. If by chance IQ differences prevail it is the job of the reviewer to point them out in the review. Not doing so can nullify any claims saying otherwise IMO.

In the end we are right back were we started, looking at reviews wondering if:
-AI was disabled?
-CCC using HQ while nvidia only uses quality?
-Using control panel AF instead of in game AF?
-etc

 

I definitely noticed a lowering of filtering quality, mainly due to shimmering, when I went from my 8800GT to my 5770.