Worst and best graphics tech ever?

[SirPauly]

FXAA is welcomed based on it does help and in the context of no other enhancements available. It also is welcomed for higher resolutions where a gamer may choose the lower hit of FXAA. Compared to traditional enhancements like multi-sampling or super-sampling, FXAA, may not compare with a moving environment but don't tend to look at FXAA as a replacement but more-so an enhancement for flexibility.

Stereo3d is welcomed because playing a 3d accelerated game on a 2 plane of existence, well, it's about time gaming moves beyond the restricted 2d plane, one may imagine. It's not about forcing 3d stereo but offering a choice for 3d stereo. So grateful that nVidia and even AMD are trying to create awareness and trying to improve upon this experience.

Improved Physics is welcomed based on three areas that may be improved upon -- fidelity, realism and improved game-play.

 

[SirPauly]

The most surprising and best quality was with the Voodoo 5 and their FSAA quality for its time. AA was in its infancy in 2000, but the AA quality from 3dfx was incredible with-in a moving screen.

With anything new, there were detractors and gamers clamoring AA was a gimmick and only high resolution is needed, when AA was simply a choice back then. Never understood gamers clamoring and complaining about a choice one has when they don't have to use the feature.

AA has matured with immense flexibility from transparency, super-sampled, multi-sampling, various AA enhancements and shader based techniques and still innovation left to evolve with more choice and flexibility.

 

[Anarchist420]

My GTX680 came out after all my games were released. Following your paradigm, I guess I can’t play any of my 129 installed games unless they get patched to contain code for the GTX680?

A patch wouldn't necessarily be necessary. The drivers could be updated, or shaders (or a combination of shaders and hardware) could be used to emulate the old hardware.

The codes are actively added to shipping drivers by the driver programmers, and they function in shipping games. That sounds like support to me.

I don't know why the person said that. Perhaps I misunderstood.

You mean like consoles? Oh wait, you wanted them to die: http://forums.anandtech.com/showthre...0392&highlight=

I know.

You ever written any code? Hardware abstraction and APIs provide numerous advantages over programming hardware directly, not just for games.

No, I'm too dumb to do so. However, the API has gotten so big because of IP and that has basically been a barrier to entry for those who wish to become GPU creators.

You’ve repeatedly started topics about items in the OP, e.g. railing against texture compression: http://forums.anandtech.com/showthre...5026&highlight=

No, what I meant was I've never started a thread asking what others thought was the best and worst graphics tech before. I just decided to offer what I thought was the best and worst graphics tech.

I'm sorry if I made you so angry as it wasn't my intention.

 

[PrincessFrosty]

I'm not a huge fan of FXAA either, or any post processing Anti-Aliasing in fact.

The problem I have with them is that the actual technical cause of Aliasing, when you break it down to it's simple form, is you're presenting a low number of samples of the scene to the viewer.

More traditional Anti-Aliasing like SSAA and MSAA solve this problem by taking a greater number of samples of the scene and using that to average the colour, you'll never get the same quality out of FXAA because it's not taking more samples, at least not in conjunction with another form of AA.

We fixed the transparency issue with Adaptive and Transparency variants of MSAA which work well although somewhat more expensive in terms of GPU power, a good trade off IMHO.

Thing is we've seen almost no serious improvement in graphics in the last 5 years, we're stuck with developers still designing games for the console generation and we have an absolute ass load of GPU power to spare, so further killing quality by using lite versions of AA seems like a waste of time to me.

We're stuck in a generation of games that don't challenge our PC hardware even slightly, it's a generation where we can hook up 3x 1920x1080p monitors in eyefinity to a single card and play the latest games with a good frame rate and actually keep a straight face when we do so. The last thing we need is features that lower quality and improve FPS.

 

[exar333]

Best: adaptive MSAA, DX9
Worst: SLI/ CF, DX11 tessellation

Adaptive MSAA because it provides near the quality of supersampling at very low performance penalty over regular MSAA.

DX9 because its the most widely used API offers great performance and allows easy development on both console and PC. It can still look good as witcher 2 will prove.

SLI/ CF because of MS and poor driver support.

DX11 tessellation because it requires very powerful hardware and the returns are not very obvious vs simply increasing geometry complexity.

MS doesn't even write CF/SLI profiles.

 

[Homeles]

MS doesn't even write CF/SLI profiles.

Micro-stutter.

 

[Kippa]

Best : 3dfx VoodooII, it was my very first gaming gfx card and it made Quake 1 look and run excellent. I wish I had kept that card after I upgraded it.

Worst : Geforce 4 MX.

 

[exar333]

Micro-stutter.

Duh.

Agree 100%.

 

[BenSkywalker]

I wasn't aware that SGi made consumer grade graphics processors other than what they did for the N64.

I guess my browsers have bugs, no matter how I load up this page the OP asks about graphics technology. Didn't see him asking about fanboi shout outs for products they ripped off from other people and then put into the market resulting in catastrophic failure(and there really isn't anything else you could call the VSA 100). BTW- The SGi team that did the graphics chip for the N64(ArtX) also made the Radeon 9700 Pro. You may have heard of that one

 

[SirPauly]

VSA-100 was indeed a failure but it certainly offered a higher quality AA. 3dfx misread the market and paid a heavy price.

 

[BFG10K]

A patch wouldn't necessarily be necessary. The drivers could be updated, or shaders (or a combination of shaders and hardware) could be used to emulate the old hardware.

Okay, so everyone directly programs for the GTX680 (for example), and future hardware emulates it via drivers.

What will that achieve? How do we take advantage of new features and hardware if developers are constantly targeting an older part by directly programming it?

If I constantly write SSE assembler, how will my code ever benefit from SSE2 or SSE3?

A far smarter option is to write high level code (e.g. C++) and let the compiler/driver target the hardware. That relieves the developer of the burden and allows code to automatically take advantage of the features of different platforms.

That’s what an API and abstraction does, and that’s why it’s good.

I know.

There’s a device that already does what you want. It uses an OpenGL derivative and is directly programmed at the hardware level. Furthermore it has no involvement with Microsoft or DirectX, and it also emulates older hardware to some degree.

The PS3 console is the perfect device for you, yet you want consoles to die. Your arguments are inconsistent and illogical.

However, the API has gotten so big because of IP and that has basically been a barrier to entry for those who wish to become GPU creators.

So your solution is to punish developers by forcing them to directly program hardware?

Also if I’m targeting a GTX680 directly, how will a start-up benefit from this? Does that start-up implement emulation for a GTX680 in their drivers? If so, how will they take advantage of their own native features?

Ironically something like DirectX will help a start-up because all they need is DirectX drivers and games will automatically run on their new hardware. But you’re railing against APIs and abstraction.

I'm sorry if I made you so angry as it wasn't my intention.

I’m not angry, are you?

 

[Anarchist420]

So your solution is to punish developers by forcing them to directly program hardware?

No, I don't see how they'd be forced to do anything due to not having patents. The IHVs could get together to make their own API or they could still use DX.

What will that achieve? How do we take advantage of new features and hardware if developers are constantly targeting an older part by directly programming it? If I constantly write SSE assembler, how will my code ever benefit from SSE2 or SSE3? A far smarter option is to write high level code (e.g. C++) and let the compiler/driver target the hardware. That relieves the developer of the burden and allows code to automatically take advantage of the features of different platforms. That’s what an API and abstraction does, and that’s why it’s good.

More double precision shader power will help with older things. I know it's more transistors and it may be slower, but I think it's time for blending and depth hardware to die... just my humble opinion though

If I'm not mistaken, Kepler and the 7970 only do 1/2 fill rate RGBA16FP blending (although in fairness, the 7970 has 64 ROPs). Blending hardware may be more efficient than blending through double precision shaders, but then again it may not be... it depends on the application.

I’m not angry, are you?

No

 

[Homeles]

(although in fairness, the 7970 has 64 ROPs).

It's actually got 32.

 

[Anarchist420]

It's actually got 32.

Does the 7990 have 64 then?

 

[Homeles]

Yeah.

 

[BFG10K]

No, I don't see how they'd be forced to do anything due to not having patents. The IHVs could get together to make their own API or they could still use DX.

This makes absolutely no sense whatsoever.

More double precision shader power will help with older things. I know it's more transistors and it may be slower, but I think it's time for blending and depth hardware to die... just my humble opinion though

I think there’s some confusion here - games don’t use DP. At all.

The 57xx line (for example) doesn’t even support DP.

 

[Anarchist420]

The 57xx line (for example) doesn’t even support DP.

I knew that.

I think there’s some confusion here - games don’t use DP. At all.

I know that current games don't, but double precision would certainly be helpful for more accurate emulation.

 

[f1sherman]

The problem I have with FXAA is that it often doesn't do a good job with edges and it's being promoted in favor of rotated grid sampling

100% agree on FXAA.

It's a ludicrous technology. It gets pushed as moving forward with IQ, when it is a step backwards. haha.

Another thing is FXAA somehow got misunderstood by a few as a high-end solution for AA instead of what it really is, the fastest possible approximate solution with no-to-bad quality.

--Timothy Lottes--