• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Imagination update from CES - Ray tracing slated for Q4

Ah so now it will run but only slowly, rather couldn't run at all. nice to see you being consistent. :rolleyes:
I am talking generically about various scenes. Some scenes cannot be rendered, some are rendering super slow. Sorry if I wasn't clear.



lol the irony overload from your opening paragraph. :)

So games that are available in identical forms on other platforms and a tech demo like this one: https://www.youtube.com/watch?v=jRr-G95GdaM

Keep up the delusion old chap.
The only funny thing is you think there is irony when there isn’t any. How are those games remotely identical on other platforms? I even listed all the graphics that got removed. How is removing tons of high end graphics on K1 versions classed as identical? How is the K1 version running at vastly reduced graphics identical? Can you show me the K1 running any of those large list of games at full graphics? The answer to that is no as it cannot.

As for that tech demo from K1 its rubbish, looks about 2 generations out of date. Graphically that tech demo is simple and easy to render. It’s a tiny map, simple graphics, no AI, limited objects, only 2 moving objects. It’s basically a single corridor map with 2 moving objects and some fancy lights. Remember how I said the K1 is limited by how many moving objects it can deal with? Can you show me a highly detailed K1 game or app with lots of objects on a decent size map?

Compare the two tech demos. We have K1, on a tiny map, no AI, no player interaction, heavily scripted and 2 moving objects in a simple mostly empty corridor, low polygon count, low map detail no real interaction.

Against PowerVR with an entire map, 100’s of AI that interact with the player, 3000+ moving objects that interact, 1000’s more still objects and all the same graphic light effects all on a much larger and highly detailed map. Yet you want to call them the same? There is light years difference between the two.

Keep calling me delusional and living in fairy tale thread as much as you like.
Just remember me when you start seeing hardware ray tracing in devices and think about how wrong you was and how much money you could have made. Its also funny how you call me delusional let you ignored all the facts and called reduced graphics identical.
 
...aaand here we reach the stage of a Pottsey thread where he plucks facts out of thin air to suit his argument and rubbishes the competition.

https://www.youtube.com/watch?v=zfyTQxU2oE4

I'm out.
You mean we reach the part of the thread when you realize you are wrong so you do a runner like every other time. As soon as any type of proof comes along you do the same thing every single time. The problem with your lies is everyone can see they are lies.

I cannot work out if you genuinely don’t understand what’s being posted or are just trying to argue even though you know you are wrong. Linking to what looks like 2 year old graphics tech demos doesn’t prove anything but it suggest to me you genuinely don’t understand so are resorting to name calling. We had games engines like that link you posted 2 years ago https://www.youtube.com/watch?v=Ru-Uw4jwxQw

Did you even watch that video? He said 30,000 polygons in that scene and only 5 or 6 AI's compared to 100's in high end games. The top end high graphic mobile games use 1.3million polygons and are far more highly detailed. 30,000 polygons puts us in the land of games from 2 years ago. Say I am making stuff up as much as you like as it doesn't change the facts that its true and anyone who understands the subject can check.

Anyway I seem to realize you are wrong and are leaving the thread, perhaps we can get back to talking about ray tracing and how it might change the market.
 
Last edited:
I, for one, actually hopes Pottsey is correct as it will provide some disruption to the duopoly we currently have in graphics.

The one thing I'm not sure about it how raytracing suddenly makes games look good. You might be able to suddenly get more accurate lighting, but there still all the other issues to contend with (AI, textures, LoD, game logic etc).
 
You mean we reach the part of the thread when you realize you are wrong so you do a runner like every other time. As soon as any type of proof comes along you do the same thing every single time. The problem with your lies is everyone can see they are lies.

No, it's the point where you ignore reality and fact and any evidence that does not fit the story you have built in your head. Attempting to engage with you beyond this point is utterly pointless as no matter what anybody now says, if it does not fit your narrative it is either dismissed as a lie, you make up some random facts with no basis in reality or you just trash it and make out that everything was done by PVR years ago.

We did this in the Tegra thread as well, I presented fact and benchmark numbers, you put your fingers and your ears and basically went 'la-la-la PVR is bestest because I say so'. Where can actual debate go from there?

If you wish to take this as winning, be my guest, everyone else can watch the videos and make their own mind up.

As for RT, have you not been saying PVR will conquer the market with their magic technology for years now? It always seems to just be 12 months away. Also, pro tip; NV hold a lot of IP in this area. https://developer.nvidia.com/optix
 
Last edited:
I have been highly sceptical in the past but it is beginning to look quite promising. The Unity integration in particular is a positive sign.

What they are really talking about is hybrid ray-tracing with deferred renderers, at least initially. This article explains what they are doing pretty well: http://gamasutra.com/blogs/Alexandr...tical_techniques_for_ray_tracing_in_games.php

I'm not an expert in raytracing but this is how the numbers could break down:
720p * 8 rays per pixel = 921,600 * 8 = 7,372,800 rays
7,372,800 * ( 4 lights * 4 rays per light ) = 117,964,800 rays per frame
Given they said they could do 300M ray per second in 2014 (http://techreport.com/news/26178/powervr-wizard-brings-ray-tracing-to-real-time-graphics) they must be compromising somewhere to get the 30fps that they claim they can achieve. The obvious place to compromise is the number of light sources per pixel and number of ray per light, but if you compromise there then you lose most of the advantages of the technique over rasterization. 4 lights sources is already very conservative, as is using only 8 rays per pixel.

Or perhaps there is some hidden optimisation I'm missing, perhaps using inter-frame coherency to reduce the number of pixels that need recasting.

There is also the issue of resolution. Every increase in pixel count requires at least an 8 fold increase in the number of rays required, realistically more like 16-32 for good results, so even just 1080p seems a way off.
 
Last edited:
Or perhaps there is some hidden optimisation I'm missing, perhaps using inter-frame coherency to reduce the number of pixels that need recasting.
There was a really good video that explained the hidden optimisation you missed. I will try and dig it out. Not got time right now on a lunch break.



No, it's the point where you ignore reality and fact and any evidence that does not fit the story you have built in your head. Attempting to engage with you beyond this point is utterly pointless as no matter what anybody now says, if it does not fit your narrative it is either dismissed as a lie, you make up some random facts with no basis in reality or you just trash it and make out that everything was done by PVR years ago.
Did it ever stop to think that perhaps PVR did do it years ago and that’s why I am not finding it impressive now years later?

The thing is I am not ignoring reality and facts. You are misunderstanding what is written then blaming me due to your own lack of knowledge in the area. My so called random facts got backed up with links from devs, white papers, specs and other sources. Just because you don’t understand something it doesn’t means it’s made up or not true. I kind off feel sorry for you as you don’t seem to realise all the things you just accused me of doing you are doing yourself. Just how are the links and evidence I posted made up by me? How can you not realise you are doing the fantasy thing you say I am doing?

“We did this in the Tegra thread as well, I presented fact and benchmark numbers, you put your fingers and your ears and basically went 'la-la-la PVR is bestest because I say so'. Where can actual debate go from there?”!
That’s not what happed at all. I pointed out a flaw in your argument but you didn’t seem to understand the technology talk and then you went off into that fantasy mode you seem to do. I am trying to have a real debate but it’s hard when every time anything get little technical and you don’t understand it you seem to fly of the wall. The 2 videos you posted are 2 of many examples I could give. I put up a valid counter points with technical facts and technical evidence. You ignored the evidence and went back into this fantasy story. Why are you doing that?



“As for RT, "have you not been saying PVR will conquer the market with their magic technology for years now? It always seems to just be 12 months away. Also, pro tip; NV hold a lot of IP in this area. https://developer.nvidia.com/optix
When did I ever say 12months away? Haven't I been saying since 2010 something around 5 to 7 years away, admittedly I thought it would be closer to 5 then 7 but I gave a range.

Optix is so far away from real time hardware ray tracing it’s not even worth looking at. It’s well over a factor of 100x slower, x100 more expensive. Optix is what 1 frame every 30seconds? Has massive bandwidth needs, massive memory needs. Its years behind the stuff PowerVR are showing. It’s like comparing software rendering to hardware rendering. I have not seen any NV patents that solve the real time hardware ray tracing problems. I have never seen NV pull off real time ray tracing. What makes you think NV can do that when there is no evidence? At least no evidence I have seen.
 
Last edited:
I have been highly sceptical in the past but it is beginning to look quite promising. The Unity integration in particular is a positive sign.

What they are really talking about is hybrid ray-tracing with deferred renderers, at least initially. This article explains what they are doing pretty well: http://gamasutra.com/blogs/Alexandr...tical_techniques_for_ray_tracing_in_games.php

I'm not an expert in raytracing but this is how the numbers could break down:
720p * 8 rays per pixel = 921,600 * 8 = 7,372,800 rays
7,372,800 * ( 4 lights * 4 rays per light ) = 117,964,800 rays per frame
Given they said they could do 300M ray per second in 2014 (http://techreport.com/news/26178/powervr-wizard-brings-ray-tracing-to-real-time-graphics) they must be compromising somewhere to get the 30fps that they claim they can achieve. The obvious place to compromise is the number of light sources per pixel and number of ray per light, but if you compromise there then you lose most of the advantages of the technique over rasterization. 4 lights sources is already very conservative, as is using only 8 rays per pixel.

Or perhaps there is some hidden optimisation I'm missing, perhaps using inter-frame coherency to reduce the number of pixels that need recasting.

There is also the issue of resolution. Every increase in pixel count requires at least an 8 fold increase in the number of rays required, realistically more like 16-32 for good results, so even just 1080p seems a way off.


Not got time to rematch all these but if they are the videos I remember they might explain a lot for you. One of them talked about how they cut down the amount of rays needed.

Practical Techniques for Ray Tracing in Games
http://www.gdcvault.com/play/1020688/Practical-Techniques-for-Ray-Tracing

New Techniques Made Possible by PowerVR Ray Tracing Hardware
http://www.gdcvault.com/play/1020741/New-Techniques-Made-Possible-by

The below link has the SDK for download and one of the ray tracing demos.
http://www.imgtec.com/powervr/powervr_openrl_sdk.asp

Once you have the SDK or just the run time installed. Download http://cdn.imgtec.com/OpenRLSDK/Open...id-Example.zip for a demo based around shadows.

EDIT: https://www.youtube.com/embed/LyH4yBm6Z9g
 
Last edited:
I must be watching the wrong Youtube channels because I don't recall anyone spending time at the Imagination booth. Journalists and reporters are either anti British or didn't think there was anything worth reporting on.
 
There was a really good video that explained the hidden optimisation you missed. I will try and dig it out. Not got time right now on a lunch break.




Did it ever stop to think that perhaps PVR did do it years ago and that’s why I am not finding it impressive now years later?

The thing is I am not ignoring reality and facts. You are misunderstanding what is written then blaming me due to your own lack of knowledge in the area. My so called random facts got backed up with links from devs, white papers, specs and other sources. Just because you don’t understand something it doesn’t means it’s made up or not true. I kind off feel sorry for you as you don’t seem to realise all the things you just accused me of doing you are doing yourself. Just how are the links and evidence I posted made up by me? How can you not realise you are doing the fantasy thing you say I am doing?

“We did this in the Tegra thread as well, I presented fact and benchmark numbers, you put your fingers and your ears and basically went 'la-la-la PVR is bestest because I say so'. Where can actual debate go from there?”!
That’s not what happed at all. I pointed out a flaw in your argument but you didn’t seem to understand the technology talk and then you went off into that fantasy mode you seem to do. I am trying to have a real debate but it’s hard when every time anything get little technical and you don’t understand it you seem to fly of the wall. The 2 videos you posted are 2 of many examples I could give. I put up a valid counter points with technical facts and technical evidence. You ignored the evidence and went back into this fantasy story. Why are you doing that?




When did I ever say 12months away? Haven't I been saying since 2010 something around 5 to 7 years away, admittedly I thought it would be closer to 5 then 7 but I gave a range.

Optix is so far away from real time hardware ray tracing it’s not even worth looking at. It’s well over a factor of 100x slower, x100 more expensive. Optix is what 1 frame every 30seconds? Has massive bandwidth needs, massive memory needs. Its years behind the stuff PowerVR are showing. It’s like comparing software rendering to hardware rendering. I have not seen any NV patents that solve the real time hardware ray tracing problems. I have never seen NV pull off real time ray tracing. What makes you think NV can do that when there is no evidence? At least no evidence I have seen.

No Pottsey, no you really don't. You make wild claims that you have real trouble backing up. The last time you started talking about how amazing Apple's new API was you could only present Apple's very own press as evidence (hardly a reliable source). When I came back with benchmark and power use numbers from a reliable source (Anandtech), you either ignored them, or attempted to move the goalposts in order to make them seem invalid. Attempting to argue any point regarding this with you is a complete waste of time as you refuse to acknowledge any facts that do not fit the hilariously pro PVR narrative you have built.

Hell, you are doing the same thing in the post I have quoted, making up accusations and mistruths as to why any technology competing with PVR is inferior.

You debating style is to make wild claims about PVR, ignore any counter point/make up more 'facts', then attack the person making them when you cannot find a credible source to back up your claims.

Oh and..
https://www.youtube.com/watch?v=h5mRRElXy-w
A real time NV Ray Tracing demo from 3 years ago. I wonder how much more advanced it is now. :)
 
The real problem with ray tracing is that given the finite computing power available in any particular transistor/power budget, you will always be able to get superior performance from rasterization than with ray tracing. So it will always be a trade off between performance and accuracy.
 
The second one in particular is pretty informative. The first one covers includes some info on the optimisations they use for decreasing the per-light ray count if you get get through the guys nervous delivery.
 
“No Pottsey, no you really don't. You make wild claims that you have real trouble backing up.”
Can you point out one place in this thread where I did that? I can point outs lots of places you have made wild claims with no backup. Like when you insist that games like Modern Combat 5 have the same graphics on Tegra as PowerVR. Even though I listed all the graphics removed on Tegra version backed up with a link.



“The last time you started talking about how amazing Apple's new API was you could only present Apple's very own press as evidence (hardly a reliable source).”
So all the big names like Unity, Frostbite and Epic and all the others are not reliable sources in your mind! What is a reliable source then?



“Hell, you are doing the same thing in the post I have quoted, making up accusations and mistruths as to why any technology competing with PVR is inferior.” “
You need to take a step back and look at your own posts. Can you point out any made up accusations or mistruths I made? I bet not. I directly pointed out accusations and mistruths you made and you ignored them. What does that say about you? Almost every single time I posted a fact backed up with evidence you resorted to name calling, accusations and mistruths. Look at the videos you posted I made perfectly reasonable comments. The polygon counts are even mentioned in the video. Look at what you did in response. Just because I post something technical you don’t understand it doesn’t mean I am making it up.

There is a reason why I think PVR technology is superior. Do you even know what a tile based deferred rendering architecture is? Let me link a really simple screenshot to make it clear for you
http://images.anandtech.com/reviews/video/Kyro2/figure3.gif
Conventional 3D is what Tegra/NVidia render. The right is what PowerVR render. If you don’t think a tile based deferred rendering architecture is superior for mobile can you please explain why? Do you even understand everything this means?

Since you like anandtech this might help http://www.anandtech.com/show/4686/samsung-galaxy-s-2-international-review-the-best-redefined/15

http://images.anandtech.com/doci/4008/Z-Cull.png
IMR also known as AMD and NVidia render the full car and the tree, even though part of the car is occluded. PowerVR only render what we see. That means PowerVR can render the same imagine for less power which means more battery life and less memory and less bandwidth needed all of which are in limited supply in mobile. So yes I think PowerVR is the superior method and more efficient and I think that for good reason. Do you understand why I think that? Do you disagree and if so why?



“You debating style is to make wild claims about PVR, ignore any counter point/make up more 'facts', then attack the person making them when you cannot find a credible source to back up your claims.”
You do realise don’t you that the only person to do the above is you. That statement fits you perfectly not me. Case in point I listed the graphics effects removed from Tegra games, backed myself up with links, dev posts. You then told me I am making up facts, have no credible source to backup claims, and starting attacking me and going into that fantasy of yours. The only person to not use facts or credible source was you. You started the personal attacks as well. So step back and take a look at your posts.



“The real problem with ray tracing is that given the finite computing power available in any particular transistor/power budget, you will always be able to get superior performance from rasterization than with ray tracing. So it will always be a trade off between performance and accuracy. “
Are you aware the big name at the GPU conferences have GPU engineers stand up and say the very opposite of what you just said. What you said is wrong as many effects require less computing power to do via ray tracing and look better via ray tracing then rasterization. Take shadows as one example. Ray tracing has superior performance for less computing power to render shadows.

That is why PowerVR are doing a hybrid card. The effects which are done better on rasterization are done via that GPU. The effects that are better via ray tracing are done by that unit.



“I must be watching the wrong Youtube channels because I don't recall anyone spending time at the Imagination booth. Journalists and reporters are either anti British or didn't think there was anything worth reporting on. “
You must be watching the wrong channels as there was a lot of videos. Want me to link some? http://hexus.net/tv/show/2015/01/Imagination_Technologies_explains_merits_of_mobile_PowerVR_GPUs is just one of many exmples I could give.
 
Last edited:
The second one in particular is pretty informative. The first one covers includes some info on the optimisations they use for decreasing the per-light ray count if you get get through the guys nervous delivery.
I take it I got the right video then about how they pulled off the reduced per-light ray count . Didn't have time to re-watch it was worried I got the wrong link.


Oh and..
https://www.youtube.com/watch?v=h5mRRElXy-w
A real time NV Ray Tracing demo from 3 years ago. I wonder how much more advanced it is now. :)
Well going by GPU shows it looks like they are not much past the basics level in that video.
 
Thanks for link Pottsey I take a look when I get home.

BTW we don't allow hotlinking of image files, you need to get those images re-hosted on photoshack or imgur otherwise the mods will wield the hammer of doom!
 
Sigh.. Pottsy...

You read something added 1+1 and came up with 3. You quote vendor PR puff pieces as reliable source information and use wildly out of date articles to back up your claims. I could have linked to a Nvidia PR puff piece where Tim Sweeny claims the K1 is at a level he did not believe mobile would be at for another 3-4 years, but unless it's from an unbiased source it cannot be trusted. tldr; If it's from a vendor PR video it's a suspect source until proven by third parties.

As for your crazy idea that PVR have some magic technology for not rendering objects that are not visible on screen leagues ahead of anybody else... Did you even bother to read and digest the 4 year old Anandtech article? If you did you will have noted that both NV and AMD have been using technology to drastically reduce the overdraw for years as well. Not as efficient as the PVR solution, but as we have seen from available benchmarks putting K1 ahead both solutions have their advantages.

Also please link me up to where people are saying that RT is less computationally and power intensive than traditional rasterization. At the same visual fidelity RT will always be more computationally intensive. Lets see what some bloke who works for Facebook has to say on the matter: http://arstechnica.com/gadgets/2013...rd/?comments=1&post=23723213#comment-23723213 some interesting comments about overdraw reduction in traditional IMR systems as well. :)

For your inevitable PVR cheerleading response please refer back to my previous posts in this thread for my reply.

peace
 
Last edited:
Back
Top Bottom