Sandy Bridge vs. NVidia: Video transcoding

[Red Storm]

When I hear about CPU+IGP (they call it CPU+GPU) on the same die, I always get the picture of 2 midgets(AMD &Intel's CPU+IGP) having a slugfest in the ring, while MMA fighters (real GPU's) look on and laugh...

But you have to remember, the midget fights get much, much higher ratings in this bizarre world of yours.

 

[extra]

I have the latest version of Handbrake and Badaboom, where do you see major quality differences? I get a slightly smaller file with Handbrake for portable devices, for high res files it seems to depend on the scene, sometimes Badaboom comes out better, sometimes Handbrake(both mainly revolving around softness issues) with comparable file size.

Interesting, perhaps I will need to try Badaboom again soon (do they have fermi support yet? Last time I checked they didn't). When I last tried it stuff looked quite bad compared to a good handbrake encode. Now, this is for just ripping DVD's and encoding them to have a backup of them. I've not ever tried transcoding them for viewing on a mobile device or something like that. In that case, like say if you wanted to send them to your ipod real fast or something, I could see the GPU encoding speed advantage being an advantage there even with worse quality for sure.

 

[Caribbean Geek]

The GTS450 that launched today runs Wolfenstein at 16x10x4 ~20% faster then 4 32 core servers running it at 12x7x0- considering most people are saying that the GTS450 is priced too high at $130 for its' performance level, quad servers may be *a little* too pricey for their performance

Raytracing isn't the same as rasterizing, talk about spin and BS.

They could choose to render using whatever method they saw fit. Using 4 servers they got smoked by a $130 graphics card, badly.

Making an old game run terrible and look awful? I may be able to find footage of a FX5200 running the game, that may be comparable.

You may want to educate yourself in Raytracing and nVidia's involvement on it because you look disingenuous talking big like that specially on a matter you don't know nothing at all, raytracking is future and look on Youtube the Quake 4 raytracing demo. Even the raytraced car demo ran at less than 4fps on a GTX 480.

http://www.anandtech.com/show/2977/...tx-470-6-months-late-was-it-worth-the-wait-/6

 

[BenSkywalker]

Your statement above is being disingenuous at best. They choose to demonstrate ray tracing. Real time ray tracing in useful applications is a major hurdle and something that many companies are eager to show off for various reasons.

If that were the case, why run it on a dated game and make said game look absolutely terrible- considerably worse then when rasterized? If they wanted to show a decent ray tracing demo they would have used a bit more bounces too, they were very short there as is clearly evidenced by the terrible lighting evident throughout.

Nemesis' point was that he wanted to see what it would require on comparable Nvidia hardware to accomplish the same task (i.e. raytracing that particular game).

Making a ray tracing demo look that bad would take some work, and I guess the easiest thing to say is that it makes Intel look far worse then when they had shown nothing at all. It looks like they are running at best one bounce- I guess someone somewhere may have utilization for such a setup running so poorly, but I'm not familiar with the need.

When I last tried it stuff looked quite bad compared to a good handbrake encode.

However, in reality the picture we've seen here is very different - As well as its fantastic turn of speed, using a fast NVIDIA CUDA-enabled GPU can frequently come close to matching the output video quality seen when transcoding via the CPU, and indeed on occasion it quite easily surpasses it to give us the best image quality of the three videos produced by our testing. Indeed, blow-for-blow we can probably surmise that transcoding through CUDA gave us the most consistent image quality - Not perfect perhaps, but either the best of the bunch or close to it in all of the circumstances in which we tested.

http://www.elitebastards.com/index.php?option=com_content&task=view&id=737&Itemid=29&limitstart=3

When Badaboom first hit it had quite a few issues with quality, not so much now. In that test they ran nV also had nigh idential file size ouput, IME the difference tends to be slightly higher then that to get equal quality(although nothing major, a couple percent or so). I still use Handbrake sometimes for my portable devices as there are times when I'm more worried about file size then anything else, and a few extra percent on a half dozen movies adds up, but overall CUDA transcoding has come a long way. Hell, look at PremiereCS5.

You may want to educate yourself in Raytracing and nVidia's involvement on it because you look disingenuous talking big like that specially on a matter you don't know nothing at all, raytracking is future and look on Youtube the Quake 4 raytracing demo.

I think it was sometime in the late 80s I started using RayTracing. By the mid 90s most people I know had left it behind mainly due to the fact that it sucks, bad. Raytracing isn't the future, it was an ugly past without diffuse that people left behind. Raytracing on certain object flagged for a particular type of reflection can look nice, it is too poor at diffuse lighting to compare even to DX level real time shaders at this point.

 

[Caribbean Geek]

I think it was sometime in the late 80s I started using RayTracing. By the mid 90s most people I know had left it behind mainly due to the fact that it sucks, bad. Raytracing isn't the future, it was an ugly past without diffuse that people left behind. Raytracing on certain object flagged for a particular type of reflection can look nice, it is too poor at diffuse lighting to compare even to DX level real time shaders at this point.

Well, you were the one comparing and stating that the GTS 450 could smoke Sandy Bridge in rendering the game...

BenSkywalker][/B]

They could choose to render using whatever method they saw fit. Using 4 servers they got smoked by a $130 graphics card, badly.

Making an old game run terrible and look awful? I may be able to find footage of a FX5200 running the game, that may be comparable.

Comparing rasterizing with Raytracing wasn't fair, saying that nVidia's GTS 450 in rasterizing could smoke Sandy Bridge which was rendering Raytracing is delusional. And now that I posted a link that shows Sandy Bridge outperforming the fastest nVidia solution in Raytracing then now you downplay its importance, talk about double standards here...

Raytracing alone might not be an ideal solution, but a razterizer/raytracing hybrid rendering should.

 

[taltamir]

Come on, "no market at all"? At the outset, doesn't that seem a bit unlikely? Adobe Premeire CS5 based on the Mercury (cuda) engine is making waves right now with power users of video editing software, which includes encoding.

Because adobe does not sacrifice quality. I wasn't saying that CUDA has no market, I was talking about crude implementation that sacrifice encoding quality for speed, like badaboom encoder.

There is great value and a huge market for hardware that accelerates encode times with 0 loss of quality (commonly known as a faster CPU), there is no market at all for hardware that accelerates encode times at the price of quality.

these are the only major "encoders" that I know of:
1. Companies creating a product
2. Pirates producing a rip.
3. Home users making home movies
4. Video enthusiast home users making home movies

1. not going to compromise the quality unless they can save enough money to justify doing so... and the problem is that CPUs are fast enough that they CAN'T justify it.
2. not going to compromise the quality, nobody on the scene uses badaboom, and nobody is going to.
3. don't care / too cheap / too ignorant to sell badaboom to, the only ones who might be willing to sacrifice quality for somewhat faster encoding speed, but barely understand what is going on. They will use what they get, the key to those is bundling... they will use whatever software comes with their computer / camera / whatever, they will not pay extra.
4. Will not compromise quality.

 

[BenSkywalker]

Well, you were the one comparing and stating that the GTS 450 could smoke Sandy Bridge in rendering the game...

The GTS450 looks much better and runs faster. I consider that smoking SandyBridge in no uncertain terms.

Comparing rasterizing with Raytracing wasn't fair,

Quality and speed, it comes up short. Yes, comparing rasterization is completely fair when it wins both speed *and* quality comparisons by a wide margin. *IF* they show raytracing with noticeably superior quality *then* it becomes unfair to compare them in terms of performance. When they lose both, badly, it is absolutely a fair comparison.

now that I posted a link that shows Sandy Bridge outperforming the fastest nVidia solution in Raytracing

Where was the link of a nV solution rendering a horribly ugly version of an old game? I missed that link. I saw one showing them rendering a model with higher complexity in a frame then Wolfenstein has in an entire level, but that was it.

Raytracing alone might not be an ideal solution

It isn't that raytracing isn't ideal, it's that it's terrible. Awful. Lousy. It is a very poor way to handle rendering real time graphics. The only thing anyone uses it for anymore is to handle reflections in certain situations. It is at this point what should equate to a shader routine used with more robust rendering approaches. If Intel wants to keep being stupid about how they display their technology, I'll keep calling them out on it. Showing you can use quad servers to run a game that looks worse then it did when it came out quite a while ago on dirt cheap consumer hardware just makes you look foolish. If they were going to release a high quality render using ray tracing in real time that would serve some purpose, an outdated game that looked outdated last year that looks inferior to when it game out is *not* a good way to demo it.

 

[taltamir]

Quality and speed, it comes up short. Yes, comparing rasterization is completely fair when it wins both speed *and* quality comparisons by a wide margin. *IF* they show raytracing with noticeably superior quality *then* it becomes unfair to compare them in terms of performance. When they lose both, badly, it is absolutely a fair comparison.

I really have to agree with it... its totally fair to compare if it loses both quality and speed.
The only way it would have been unfair is if you say:
X is better than Y
because X is faster then Y
when X is also lower quality than Y.

 

[Caribbean Geek]

It is a very poor way to handle rendering real time graphics.

Its your personal opinion but I doubt that a giant like Intel will waste money on something awful and useless, also you can't take away the fact that the nVidia's own demo runs much slower on their GTX 480

than the fact that Intel's own demo which ran much faster with a Knights Ferry 128 core card.

Look at this link and see the video, it runs smooth, not blazingly fast though and also looks great.

http://www.pcper.com/comments.php?nid=9247

http://www.youtube.com/watch?v=mtHDSG2wNho

http://bestnewtech.blogspot.com/2010/09/intel-showed-video-game-in-3d-ray.html <<More info of that config is there.

 

[Nemesis 1]

I going to give everyone a time out .

http://www.youtube.com/watch?v=CRMOeMgYESo&search=japan&#37;20library

 

[taltamir]

Its your personal opinion but I doubt that a giant like Intel will waste money on something awful and useless, also you can't take away the fact that the nVidia's own demo runs much slower on their GTX 480

that graph you linked to support your claim? it says "higher is better" (because its measuring framerate), its showing the GTX480 to be 8.78 times faster than the GTX285

 

[Caribbean Geek]

that graph you linked to support your claim? it says "higher is better" (because its measuring framerate), its showing the GTX480 to be 8.78 times faster than the GTX285

Well, if you bothered to read the whole last pages you would know what I was trying to say. If you can call 3.8fps playable, good for you. But 4 Knight Ferry cards (128 cores in total) were able to render a whole Raytracing scene in real time at playable fps, something that the GTX 480 with its 480 cores couldn't do.

 

[Lonbjerg]

Well, if you bothered to read the whole last pages you would know what I was trying to say. If you can call 3.8fps playable, good for you. But 4 Knight Ferry cards (128 cores in total) were able to render a whole Raytracing scene in real time at playable fps, something that the GTX 480 with its 480 cores couldn't do.

You just didn't compare 4(!!!) cards to another single card with different architechture...did you?

If yes...I am amazed at the ignorance.

 

[taltamir]

If you can call 3.8fps playable

I didn't call it playable, you said the GTX480 is slower than the GTX285, it is 8.7 times faster. I do not think it is playable, but it is faster.

 

[Caribbean Geek]

You just didn't compare 4(!!!) cards to another single card with different architechture...did you?

If yes...I am amazed at the ignorance.

Yup, a single card with 480 cores which is slow for Raytracing while 4 cards with 32 cores on each is in order of a magnitude faster, and you look great attacking and trolling here, keep going, if you can't post in a civilized way, don't post at all, the only one who looks ridiculous and uncivilized is you, its just a videocard and technology forum, go outside and enjoy the sunshine.

I didn't call it playable, you said the GTX480 is slower than the GTX285, it is 8.7 times faster. I do not think it is playable, but it is faster.

When did I said that the GTX 480 were slower than the GTX 285? I said that Knights Ferry was faster than the GTX 480 in Raytracing.

 

[taltamir]

When did I said that the GTX 480 were slower than the GTX 285? I said that Knights Ferry was faster than the GTX 480 in Raytracing.

My misunderstand than, sorry.

 

[extra]

When did I said that the GTX 480 were slower than the GTX 285? I said that Knights Ferry was faster than the GTX 480 in Raytracing.

The nvidia ray tracing demo has nothing to do with the intel ray tracing demo. The quality is not remotely comparable, the nvidia design garage demo rendering quality is far, far, *far*, higher. The design garage demo is more comparable to rendering something in a real 3d modeling program.

 

[Caribbean Geek]

The nvidia ray tracing demo has nothing to do with the intel ray tracing demo. The quality is not remotely comparable, the nvidia design garage demo rendering quality is far, far, *far*, higher. The design garage demo is more comparable to rendering something in a real 3d modeling program.

nVidia's car demo is only the car, and some other stuff and its a demo, which is usually very optimized for a specific hardware, looks far better than a game, pushes the limits of the hardware but the scene is very small in size (AMD's Ruby Doublecross is a good example of a techdemo that looks great, runs great but isn't a very big scene) Creating a whole 3D scene of a game with such quality is unfeasible (For example in the X800 old hardware). Intel's demo was a full 3D game (Quality is always lower than a techdemo for the sake of playability and broader audience). But I think we had strayed too far from the thread, but it was interesting nevertheless.

 

[Lonbjerg]

Yup, a single card with 480 cores which is slow for Raytracing while 4 cards with 32 cores ....

I stopped reading any further...you are joking right?