Intel Plans back
Board: Board index ‹ Ray tracing ‹ Links & papers
(L) [2007/06/02] [Phantom] [Intel Plans] Wayback!A dutch technology website has an interesting article on Intel's future plans:
[LINK http://tweakers.net/nieuws/47792/Intel-onthult-details-over-Larrabee-processor-gpu.html]
They point to a presentation by an Intel guy for the Michigan State University, for which the slides are online:
[LINK http://bt.pa.msu.edu/TM/BocaRaton2006/talks/davis.pdf]
Ray tracing is explicitly mentioned. One Teraflops of 64bit computing power, available in highly parallel x86 cores... I think I'll freeze my plans to convert to GPU ray tracing. [SMILEY :)]
(L) [2007/06/02] [Phantom] [Intel Plans] Wayback!A dutch technology website has an interesting article on Intel's future plans:
[LINK http://tweakers.net/nieuws/47792/Intel-onthult-details-over-Larrabee-processor-gpu.html]
They point to a presentation by an Intel guy for the Michigan State University, for which the slides are online:
[LINK http://bt.pa.msu.edu/TM/BocaRaton2006/talks/davis.pdf]
Ray tracing is explicitly mentioned. One Teraflops of 64bit computing power, available in highly parallel x86 cores... I think I'll freeze my plans to convert to GPU ray tracing. [SMILEY Smile]
_________________
--------------------------------------------------------------
Whatever
(L) [2007/06/02] [greenhybrid] [Intel Plans] Wayback!Leider spreche ich kein niederländisch, auch wenn ich nur 10km entfernt lebe [SMILEY ;)]
(for those not speaking the dark language [sidenote: I've read once the japans use game-titles like "Zweihänder" or similar because of the dark sound of the german language]: "A pity me idiot speaks no dutch, esp. regarding I'm living near the border")
Hmm, where exactly do they mention Ray Tracing? Quick Search thru website and a fast glimpse over the paper (with all of their exclamation marks, how frenetic they use "!", phew) gave me no results?
(L) [2007/06/02] [greenhybrid] [Intel Plans] Wayback!Leider spreche ich kein niederländisch, auch wenn ich nur 10km entfernt lebe [SMILEY Wink]
(for those not speaking the dark language [sidenote: I've read once the japans use game-titles like "Zweihänder" or similar because of the dark sound of the german language]: "A pity me idiot speaks no dutch, esp. regarding I'm living near the border")
Hmm, where exactly do they mention Ray Tracing? Quick Search thru website and a fast glimpse over the paper (with all of their exclamation marks, how frenetic they use "!", phew) gave me no results?
_________________
[LINK http://greenhybrid.net/ greenhybrid.net]
Real Men code Software Graphics.
(L) [2007/06/02] [tbp] [Intel Plans] Wayback!Even if i can't read any of those weird saxon languages (in this particular instance, a kind of vowel explosion) i can still recognize when someone from the marketing department is having a seizure. This time there's no pastels, so we known it's not from Apple.
80 core here, zeta flop there...
_________________
May you live in interesting times.
[LINK https://gna.org/projects/radius/ radius] | [LINK http://ompf.org/ ompf] | [LINK http://ompf.org/wiki/ WompfKi]
(L) [2007/06/02] [ingenious] [Intel Plans] Wayback!>> greenhybrid wrote:Leider spreche ich kein niederländisch, auch wenn ich nur 10km entfernt lebe
Aaah, ich bin sogar kein Deutscher, aber wenn man English und Deutsch kennt, ist es schon nicht so schwer auch Niederländisch zu versthehen . Zum Beispiel:
[IMG #1 Image]
Wenn du das nicht verstanden hast:
[IMG #2 Image]
I love this language [SMILEY :)]
It's so wonderful to listen it on the street. Do you know how Vincent Van Gogh sounds in Dutch? Weee [SMILEY :lol:] [SMILEY 8)]
[IMG #1]:
![[IMG:#0]](images/c1e346cab67d2595eb91af4c63d318795525472c2f53cf0cb314dcc04f4c8244.jpg)
[IMG #2]:
![[IMG:#1]](images/bd00104dbe7ac6b4bc79fbe0d0aa3db96c92687bcedc91329dfdc62186a5e625.jpg)
(L) [2007/06/02] [greenhybrid] [Intel Plans] Wayback!>> tbp wrote:[lazy_quote]Saxon[/lazy_quote]
ahh, the Saxon:
[img width=320 height=250]http://upload.wikimedia.org/wikipedia/en/6/69/Old_Saxony1_in_western_empire_843.jpg[/img]
>> ingenious wrote:greenhybrid wrote:Leider spreche ich kein niederländisch, auch wenn ich nur 10km entfernt lebe
Aaah, ich bin sogar kein Deutscher, aber wenn man English und Deutsch kennt, ist es schon nicht so schwer auch Niederländisch zu versthehen . Zum Beispiel:
...
Wenn du das nicht verstanden hast:
...
I love this language
It's so wonderful to listen it on the street. Do you know how Vincent Van Gogh sounds in Dutch? Weee
Jupp, diese Sätze verstehe ich auch noch (Rauchen ist so eine schlechte Marotte von mir [SMILEY :|])
...
back to that commercial language:
I assume you come from BeNeLux, too?
(L) [2007/06/02] [ingenious] [Intel Plans] Wayback!>> greenhybrid wrote:Jupp, diese Sätze verstehe ich auch noch (Rauchen ist so eine schlechte Marotte von mir )
Tell me about it [SMILEY :)]
>> greenhybrid wrote:back to that commercial language:
I assume you come from BeNeLux, too?
Actually from Bulgaria. [SMILEY :)]
But I study in Saarbruecken.
(L) [2007/06/02] [Phantom] [Intel Plans] Wayback!>> greenhybrid wrote:Hmm, where exactly do they mention Ray Tracing? Quick Search thru website and a fast glimpse over the paper (with all of their exclamation marks, how frenetic they use "!", phew) gave me no results?
It's on the very last slide, ray tracing is mentioned as a killer app.
Sorry about the Dutch, I just wanted to state where I got a link to those slides.
(L) [2007/06/02] [ingenious] [Intel Plans] Wayback!>> Phantom wrote:greenhybrid wrote:Hmm, where exactly do they mention Ray Tracing? Quick Search thru website and a fast glimpse over the paper (with all of their exclamation marks, how frenetic they use "!", phew) gave me no results?
It's on the very last slide, ray tracing is mentioned as a killer app.
And on slide 16 [SMILEY :wink:]
(L) [2007/06/03] [Ho Ho] [Intel Plans] Wayback!I'd say ray tracing is definitely on Intel roadplan. Why else does it do so much research on it?
Also, as ray tracing takes less memory bandwidth than rasterizing and delivering enough bandwidth to CPU socket is much more difficult than to deliver it to GPU soldered on a PCB together with RAM, ray tracing seems to be the best option to use.
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.
Jan L.A. van de Snepscheut
(L) [2007/06/03] [ingenious] [Intel Plans] Wayback!Ho Ho,
I just love your signature!
(L) [2007/06/03] [greenhybrid] [Intel Plans] Wayback!>> ingenious wrote:greenhybrid wrote:Jupp, diese Sätze verstehe ich auch noch (Rauchen ist so eine schlechte Marotte von mir )
Tell me about it
Hmm, not much to tell: I smoke self-rolled cigaretes [SMILEY :D] ...., [SMILEY :rolleyes:]
>> ingenious wrote:Ho Ho,
I just love your signature!
Love is such a wonderfull abstraction of physics [SMILEY :D]
(L) [2007/06/04] [Phantom] [Intel Plans] Wayback!Yay, a smokers forum.
(ME TOO ME TOO)
By the way, this gotta be the most undisciplined thread I have ever seen on this forum. No respect for the topic at all.
_________________
--------------------------------------------------------------
Whatever
(L) [2007/06/06] [tbp] [Intel Plans] Wayback!Lingua franca follow-up [LINK http://www.tgdaily.com/content/view/32282/137/]
_________________
May you live in interesting times.
[LINK https://gna.org/projects/radius/ radius] | [LINK http://ompf.org/ ompf] | [LINK http://ompf.org/wiki/ WompfKi]
(L) [2007/06/06] [Ho Ho] [Intel Plans] Wayback!Am I the only one who noticed that since its publication a few pages have been removed from the original PDF? Most notably pages 17 and 31.
Luckily I had the original one in my browser cache and I've uploaded it [LINK http://hoho.bafsoft.net/data/konquerort0Lnyb.pdf here]
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.
Jan L.A. van de Snepscheut
(L) [2007/06/07] [Ho Ho] [Intel Plans] Wayback![LINK http://www.xbitlabs.com/news/video/display/20070607063459.html Intel May Enter Graphics Chip Market in Several Quarters.]
Intel Rumoured to Release First Graphics Chip in Years in Early 2008
I wonder if and how much will those things have in common with Larrabee [SMILEY :)]
(L) [2007/06/07] [Ho Ho] [Intel Plans] Wayback![LINK http://www.xbitlabs.com/news/video/display/20070607063459.html Intel May Enter Graphics Chip Market in Several Quarters.]
Intel Rumoured to Release First Graphics Chip in Years in Early 2008
I wonder if and how much will those things have in common with Larrabee [SMILEY Smile]
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.
Jan L.A. van de Snepscheut
(L) [2007/06/12] [Ho Ho] [Intel Plans] Wayback!An interesting find, I think. Though a bit short text [SMILEY :)]
[url=http://209.85.135.104/search?q=cache:26wgc1l8aj4J:www.computerpoweruser.com/editorial/article.asp%3Farticle%3Darticles/archive/c0609/67c09/67c09.asp%26guid%3D+Q%26A+With+Jerry+Bautista&hl=et&ct=clnk&cd=1&gl=ee&client=firefox-a]Q&A With Jerry Bautista
The Future Of CPUs & GPUs [/url]
>> CPU: How soon will ray tracing on a CPU overtake graphics on a GPU?
JB: Our horizon is three to five years. We are at about a three-year point in terms of showing that these things can be done.I sure hope Larrabee has something interesting for us [SMILEY :)]
PS!
If the original link doesn't work for some reason try [LINK http://www.google.ee/search?q=Q%26A+With+Jerry+Bautista&ie=utf-8&oe=utf-8&aq=t&rls=org.mozilla:en-US:official&client=firefox-a this] and read the cached version. Direct link requires you to buy a subscription.
(L) [2007/06/12] [Ho Ho] [Intel Plans] Wayback!An interesting find, I think. Though a bit short text [SMILEY Smile]
[LINK http://209.85.135.104/search?q=cache:26wgc1l8aj4J:www.computerpoweruser.com/editorial/article.asp%3Farticle%3Darticles/archive/c0609/67c09/67c09.asp%26guid%3D+Q%26A+With+Jerry+Bautista&hl=et&ct=clnk&cd=1&gl=ee&client=firefox-a Q&A With Jerry Bautista
The Future Of CPUs & GPUs ]
(L) [2007/09/18] [Ho Ho] [Intel Plans] Wayback!Bunch of information about Intel future plans: [LINK http://www.anandtech.com/tradeshows/showdoc.aspx?i=3101&p=1]
Unfortunately very little about Larrabee. Though they do talk about 8-core Nehalem coming in H2 2008 that has SMT support [SMILEY :)]
(L) [2007/09/18] [Ho Ho] [Intel Plans] Wayback!Bunch of information about Intel future plans: [LINK http://www.anandtech.com/tradeshows/showdoc.aspx?i=3101&p=1]
Unfortunately very little about Larrabee. Though they do talk about 8-core Nehalem coming in H2 2008 that has SMT support [SMILEY Smile]
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.
Jan L.A. van de Snepscheut
(L) [2007/09/18] [lycium] [Intel Plans] Wayback!yup, i was also just reading about that on techreport: [LINK http://techreport.com/discussions.x/13232]
btw, amd seems pretty doomed: [LINK http://techreport.com/articles.x/13224]
(L) [2007/09/21] [Ho Ho] [Intel Plans] Wayback![LINK http://www.pcper.com/article.php?aid=455]
Some talk about Intel and ray tracing (I hope image linking works)
[IMG #1 Image]
In about two years there should be first versions of Larrabee and updates to Nehalem. I wonder which of those are they going to use for RT.
[IMG #1]:
![[IMG:#0]](images/14a919e548b968a94e0cf5b2ce79976ba0712ce034e95af08a263eb480eb464a.jpg)
(L) [2007/09/21] [Ho Ho] [Intel Plans] Wayback![LINK http://www.pcper.com/article.php?aid=455]
Some talk about Intel and ray tracing (I hope image linking works)
[IMG #1 ]
In about two years there should be first versions of Larrabee and updates to Nehalem. I wonder which of those are they going to use for RT.
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.
Jan L.A. van de Snepscheut
[IMG #1]:
(L) [2007/09/22] [jogshy] [Intel Plans] Wayback!I think a prototype could appear in late 2008. We could get the 16 core version by 2009. Then the 32 and finally the 48 cores by the start of 2010.
Probably will use 16-way SIMD registers ( instead of the current 4 )... but perhaps a scalar achitecture like the G80 would be better...
Intel is working with Daniel Pohl to create a raytracing API. They also bought Havok so physics will be available too. We can think about Larrabee as the combination of three elements: A Clearspeed coprocessor(gpgpu) card + a graphics card(Radeon,GeForce) + a physics card(Ageia PhysX). For gpgpu we could use OpenMP or Ct ( [LINK http://download.intel.com/pressroom/kits/research/Flexible_Parallel_Programming_Ct.pdf] ).
The raytracing/GI API will include functions to trace rays(hit/closestHit) and to gather photon map samples(GatherPhotons). It could work with dynamic scenes using "local space" B-KD trees.
I hope will be DX10 compatible(rasterization, zbuffer, blah blah). Perhaps Microsoft after seeing the raytracing/GI and physics API will need to modify DirectX to include gpgpu, physics and raytracing... what will yield us to DX11.
Well... if can make realtime raytracing for an HD monitor and also to perform global illumination then will be very welcome! ( but I doubt the processing power could be there! ). I'm very skeptic about Larrabee... and wonder why NVIDIA and ATI did not started to think about HW-accelerated raytracing yet...
_________________
I know I ask too much... but i'm on panic mode and there is no panic button to press [SMILEY stick out tongue]
(L) [2007/09/22] [Phantom] [Intel Plans] Wayback!Jogshy, where did you get this info? Especially the info about numbers of cores I haven't seen anywhere yet?
_________________
--------------------------------------------------------------
Whatever
(L) [2007/09/22] [lycium] [Intel Plans] Wayback!i doubt he can name any sources or confirm any of this stuff; nevertheless, it is interesting to read :)
(L) [2007/09/22] [jogshy] [Intel Plans] Wayback!>> lycium wrote:i doubt he can name any sources or confirm any of this stuff; nevertheless, it is interesting to read
Some links with extra info:
The start of all ( Larrabee Development Group )
[LINK http://www.xbitlabs.com/news/video/display/20060816030933.html]
[IMG #1 Image]
About core roadmap:
[LINK http://community.zdnet.co.uk/blog/0,1000000567,10005497o-2000331777b,00.htm]
[LINK http://translate.google.com/translate?u=http%3A%2F%2Fpc.watch.impress.co.jp%2Fdocs%2F2007%2F0611%2Fkaigai364.htm&langpair=ja%7Cen&hl=en&safe=off&ie=UTF-8&oe=UTF-8&prev=%2Flanguage_tools]
[LINK http://forum.sakhtafzar.com/showthread.php?t=984]
[LINK http://vga.zol.com.cn/58/583048.html]
[LINK http://www.overclockers.ru/hardnews/25985.shtml]
Larrabee silicon die(unconfirmed):
[LINK http://www.puissance-pc.net/cartes-graphiques/les-cartes-graphiques-intel-pour-2008.html]
16 way SIMD(512 bits registers):
[LINK http://techreport.com/discussions.x/12237]
About Ct:
[LINK http://download.intel.com/pressroom/kits/research/Flexible_Parallel_Programming_Ct.pdf]
[LINK http://download.intel.com/pressroom/kits/research/poster_Ct_Data_Parallel_Programming.pdf]
[LINK http://blogs.intel.com/research/2007/09/the_many_flavors_of_data_paral.html]
About Havok adquisition:
[LINK http://arstechnica.com/news.ars/post/20070917-intel-picks-up-gaming-physics-engine-for-forthcoming-gpu-product.html]
About Daniel's raytracing:
[LINK http://www.pcper.com/article.php?aid=455]
[LINK http://www.theinquirer.net/?article=41858]
[LINK http://www.intel.com/cd/software/products/asmo-na/eng/302910.htm] ( non-free?, notice the "and new functions for ray-tracing and rendering." )
--- Now, a bit off-topic but related ---
Intel Polaris ( 80 core cpu, TeraScale, Larrabee's parent ):
[LINK http://www.theinquirer.net/default.aspx?article=40469]
[LINK http://www.youtube.com/watch?v=aiARDNp66Cw]
[LINK http://www.youtube.com/watch?v=TAKG0UvtzpE]
[LINK http://www.intel.com/research/platform/terascale/TeraScale_whitepaper.pdf]
IBM Cell raytracing iRT engine & SaarCOR FPGA & ARTVPS & Avalon ( Intel's hardware competency )
[LINK http://gametomorrow.com/blog/index.php/category/cell/]
[LINK http://www.alphaworks.ibm.com/tech/irt]
[LINK http://graphics.cs.uni-sb.de/SaarCOR/DynRT/DynRT.html]
[LINK http://www.artvps.com/page/109/raybox.htm] ( I asked them for a SDK... they don't have it.. only supports Renderman, Maya and 3dsmax using propietary drivers & software )
[LINK http://www.pixelution.co.uk/Products/ARTVPS_raybox.html]
[LINK http://www.schwarzers.de/project_features.php.htm]
Future Microsoft DirectRT (raytracing) and Direct Physics projects ( unconfirmed ):
[LINK http://research.microsoft.com/users/brianme/ray.htm]
[LINK http://www.xbitlabs.com/news/multimedia/display/20060620235215.html]
[IMG #1]:
![[IMG:#0]](images/6164e27fdaabd43cebc30e9720cedf0589230c9723fd208692c04241fde0cefa.jpg)
(L) [2007/09/22] [Phantom] [Intel Plans] Wayback!Wow. [SMILEY :)] So all that info is actually simply available. Problem is, several of the things you mention have been whispered to me at several occasions, and it happened before that some of that 'secret info' was on the street the next day...
Anyway, thanks very much for this huge list! I'll need to feed that French Larrabee info to babelfish, but that die foto is simply awesome.
(L) [2007/09/22] [Phantom] [Intel Plans] Wayback!Wow. [SMILEY Smile] So all that info is actually simply available. Problem is, several of the things you mention have been whispered to me at several occasions, and it happened before that some of that 'secret info' was on the street the next day...
Anyway, thanks very much for this huge list! I'll need to feed that French Larrabee info to babelfish, but that die foto is simply awesome.
_________________
--------------------------------------------------------------
Whatever
(L) [2007/09/22] [jogshy] [Intel Plans] Wayback!>> Phantom wrote:but that die foto is simply awesome.
Bah!
[LINK http://www.tomshardware.fr/images/3d/20020425/die.jpg]
It's from an Intel 740 [SMILEY :cry:]
On the other hand was logical.. I doubt Intel has tapped out any Larrabee sample yet
(L) [2007/09/22] [lycium] [Intel Plans] Wayback!phoar! *tries to remove firmly lodged foot from mouth*
thanks for those great links :D
(L) [2007/09/23] [jogshy] [Intel Plans] Wayback!Looooool!
[IMG #1 Image]
...a fake perhaps?
Well, whatever Intel does with the Larrabee's raytracing API finally, I hope:
1) The API should be free and publically available... I see they added raytracing to their Integrated Performance Primitives 5.2 currently... but that package is not free...( costs 200$ and I'm not sure if that includes Daniel Pohl's raytracing work ).
2) The API should include these functions:
- Trace. Test if a ray intersects any geometry in the scene. It should accept a C "callback" function so the user can abort ( get first hit boolean for shadows ), get the closest hit(primary rays, reflection) or get all the hits(for transparency).
- GatherPhotons. Get the nearest photon map samples in disk using a maximum radius distance.
- GatherPoints. Like the Renderman one... fires rays in an hemisphere distribution ( good for ambient occlusion ) delimited by an angle.
3) Should support both dynamic and static scenes. I think will be possible to perform "local space" kd-trees for non-deformable objects and a B-KDTree or BIH for deformable meshes?
4) Should support custom shaders and lighting models.
5) Should support triangles, quads, subdivision surfaces and nurbs.
6) Should contain built-in functions for vectors, matrices, quaternions, BLAS, FFT, ect
What do you think/expect?
[IMG #1]:
![[IMG:#0]](images/b9ac292dd06a23a980c135a54007ea5feb14aa3023bac7d6f318284b7e65839c.jpg)
(L) [2007/09/23] [jogshy] [Intel Plans] Wayback!Looooool!
[IMG #1 ]
...a fake perhaps?
Well, whatever Intel does with the Larrabee's raytracing API finally, I hope:
1) The API should be free and publically available... I see they added raytracing to their Integrated Performance Primitives 5.2 currently... but that package is not free...( costs 200$ and I'm not sure if that includes Daniel Pohl's raytracing work ).
2) The API should include these functions:
- Trace. Test if a ray intersects any geometry in the scene. It should accept a C "callback" function so the user can abort ( get first hit boolean for shadows ), get the closest hit(primary rays, reflection) or get all the hits(for transparency).
- GatherPhotons. Get the nearest photon map samples in disk using a maximum radius distance.
- GatherPoints. Like the Renderman one... fires rays in an hemisphere distribution ( good for ambient occlusion ) delimited by an angle.
3) Should support both dynamic and static scenes. I think will be possible to perform "local space" kd-trees for non-deformable objects and a B-KDTree or BIH for deformable meshes?
4) Should support custom shaders and lighting models.
5) Should support triangles, quads, subdivision surfaces and nurbs.
6) Should contain built-in functions for vectors, matrices, quaternions, BLAS, FFT, ect
What do you think/expect?
_________________
I know I ask too much... but i'm on panic mode and there is no panic button to press [SMILEY stick out tongue]
[IMG #1]:
(L) [2007/09/23] [lycium] [Intel Plans] Wayback!three more distributions: phong, cosine-weighted (diffuse) and specular (conductor/dielectric).
screw realtime, let's go after pixar's renderman ;)
(L) [2007/09/23] [playmesumch00ns] [Intel Plans] Wayback!How about just a user-supplied distribution? prman supports that after all [SMILEY ;)]
(L) [2007/09/23] [playmesumch00ns] [Intel Plans] Wayback!How about just a user-supplied distribution? prman supports that after all [SMILEY Wink]
(L) [2007/09/23] [Ho Ho] [Intel Plans] Wayback!If I understood correctly then in [LINK http://softwarecommunity.intel.com/isn/Community/en-US/forums/thread/30240197.aspx this] thread one Intel worker sais that Larrabee will be availiable next year.
Also I checked out the non-commercial version of Intel integrated performance primitives. From the quick look I couldn't find anything about ray tracing from the headers and other files. I couldn't read the .pdf files as my box is actually pretty much dead and I can only use plain X without most of the programs, if I find anything from the PDF's I'll report later, in few hours probably
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.
Jan L.A. van de Snepscheut
(L) [2007/09/23] [jogshy] [Intel Plans] Wayback!>> playmesumch00ns wrote:How about just a user-supplied distribution? prman supports that after all
Yep, for the Gather functions would be very desireable to allow the user to make custom distributions. To have some standard ones like uniform, cosine weighted, etc will be good too
>> Ho Ho wrote:Also I checked out the non-commercial version of Intel integrated performance primitives. From the quick look I couldn't find anything about ray tracing from the headers and other files
See the ippmqr.pdf and ippr.h files... functions ipprIntersectMO_32f, ipprIntersectEyeSO_32f, ipprIntersectAnySO_32f, ipprCastEye_32f, ipprCastReflectionRay_32f, ipprCastShadowSO_32f, ipprTriangleAccelInit, ipprKDTreeBuildAlloc, etc...
Anybody know how fast are those functions? Can manage million of polygons meshes?
Btw, I hope intel makes this free for everybody once Larrabee is released... because 200$ of the IPP + the CGPU cost seems to be a lot...
(L) [2007/09/28] [moris] [Intel Plans] Wayback!Hello,
I just tried RT with IPP 5.2 Trial but I can only build the kdTree (ipprKDTreeBuildAlloc). It takes about 6s on my Core 2 Duo 2Ghz to build Conference. After, a memory failure is raised
when I want to launch primary rays (ipprIntersectEyeSO_32f).
I am very curious about these performances. Somebody knows more than me ?
_________________
GC
If you don't understand me, perhaps there is nothing to understand
(L) [2007/09/28] [Xela] [Intel Plans] Wayback!>> moris wrote:Hello,
I just tried RT with IPP 5.2 Trial but I can only build the kdTree (ipprKDTreeBuildAlloc). It takes about 6s on my Core 2 Duo 2Ghz to build Conference. After, a memory failure is raised
when I want to launch primary rays (ipprIntersectEyeSO_32f).
I am very curious about these performances. Somebody knows more than me ?
RT IPP started few years ago when we transfered to IPP group our old codebase (pre-MLRTA). At this point, I advised them not to do this, they ignored it.
RT (unlike codecs, for example) is a rapidly developed area mostly done in Academia, and with all new advances it doesn't make sense to release binaries.
Source code would do better, but IPP management didn't like this idea either.
(L) [2007/09/28] [moris] [Intel Plans] Wayback!Thank you Xela for this precision. Now, I have to undelete my RT code... [SMILEY :)]
(L) [2007/09/28] [moris] [Intel Plans] Wayback!Thank you Xela for this precision. Now, I have to undelete my RT code... [SMILEY Smile]
_________________
GC
If you don't understand me, perhaps there is nothing to understand
(L) [2007/10/02] [pers] [Intel Plans] Wayback!Hi, Moris.
I’m one of IPP developers and I’ll be glad to help you with our product if you still wish to evaluate it. Please visit the official support forum at [LINK http://softwarecommunity.intel.com/isn/Community/en-US/forums/1274/ShowForum.aspx] and describe you problem with intersectors in more detail. At first glance you didn’t initialize all IpprIntersectContext members properly...
I do agree with Xela on topic of rapid RT development. In fact I believe this rapidness makes it a disruptive technology thus limiting its acceptance in commercial products. Still it is added to our product in hope that not everyone need to reinvent a bicycle and as it’s in a state of early 2006 it’s still very good at performance (better then the Xela’s prototype). Please give it a try and maybe you’ll like it. I just can suppose you didn’t initialize all members of IpprIntersectContext structure: IppBox3D_32f and IpprTriangleAccel.
(L) [2007/10/02] [pers] [Intel Plans] Wayback!Hi, Moris.
Im one of IPP developers and Ill be glad to help you with our product if you still wish to evaluate it. Please visit the official support forum at [LINK http://softwarecommunity.intel.com/isn/Community/en-US/forums/1274/ShowForum.aspx] and describe you problem with intersectors in more detail. At first glance you didnt initialize all IpprIntersectContext members properly...
I do agree with Xela on topic of rapid RT development. In fact I believe this rapidness makes it a disruptive technology thus limiting its acceptance in commercial products. Still it is added to our product in hope that not everyone need to reinvent a bicycle and as its in a state of early 2006 its still very good at performance (better then the Xelas prototype). Please give it a try and maybe youll like it. I just can suppose you didnt initialize all members of IpprIntersectContext structure: IppBox3D_32f and IpprTriangleAccel.
(L) [2007/11/02] [jogshy] [Intel Plans] Wayback!Today Qimonda released their first GDDR5 memory samples. TheINQ says Intel perhaps have received some for Larrabee.
Says the chips can do 3x the performance of a DDR3 due to the new capabilities like write data to multiple chips at the same time.
[LINK http://www.theinquirer.net/gb/inquirer/news/2007/11/02/gddr5-samples-ship]
Also heard they are experimenting with XDR 2:
[LINK http://www.channelregister.co.uk/2005/07/07/rambus_xdr_2/]
and 1Terabyte/second Rambus RAM
[LINK http://www.tgdaily.com/content/view/34964/113/]
[LINK http://www.dailytech.com/Rambus+Aims+for+TerabytePerSecond+Memory+Bandwidth/article9771.htm]
(L) [2007/11/27] [jogshy] [Intel Plans] Wayback!The other day Intel CEO confirmed LArrabee to get more than 10 cores:
[LINK http://www.computerworlduk.com/technology/hardware/processors/in-depth/index.cfm?articleid=957]
>> Q:How many cores will Larrabee have?
A: I can't comment on details about the first product. It's sufficient to say "more than 10", which is what we define as the boundary between multicore and many-core. It's better to think of it as a scalable architecture family, with varying numbers of cores depending on the application.
Btw, today Intel bought Neoptica:
[LINK http://www.xbitlabs.com/news/video/display/20071126140151_Intel_Quietly_Acquires_Graphics_Software_Firm.html]
[LINK http://xtreview.com/addcomment-id-3817-view-Intel-purchased-Neoptica-company.html]
[LINK http://www.neoptica.com/]
see [LINK http://www.neoptica.com/NeopticaWhitepaper.pdf] for detail about this company technology.
This company also got an excellent k-buffer transparency paper at
[LINK http://www.sci.utah.edu/~bavoil/research/kbuffer/]
the company was basically created by 3d offline render developers + some NVIDIA ex-employees.
(L) [2007/11/27] [jogshy] [Intel Plans] Wayback!The other day Intel CEO confirmed LArrabee to get more than 10 cores:
[LINK http://www.computerworlduk.com/technology/hardware/processors/in-depth/index.cfm?articleid=957]
(L) [2007/11/28] [lycium] [Intel Plans] Wayback!i think matt pharr is well on the road to world domination ;) he cannot keep a company of his own for more than a few months without some major player buying it up, and then there's the pbrt money... good going!
that k-buffer paper looks really interesting (it was posted here before i think?), but i suppose it's already patented for commercial use in 3d accelerators... also, i'm not sure how surfaces parallel to the camera's view direction would be handled (e.g. one can think of situations in which you see the edge of an object, and what's behind it needs to be handled properly), but i should check the paper in more detail.
back