Bestand wählen
Merken

Intel BayTrail graphics overview

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Beta
Erkannte Entitäten
Sprachtranskript
it would be the end of the way in which the in the in the in the and in a way that you could do it where thinker once the jump in here many just bonds at working there until open source technology center with but then
just presented and several others who were here today on the and graphics struggle primarily and miscellaneous other GP related stuff inside until but today all be talking about and then tell bishop platform and I'd actually 1 of the cover this last year
but only got approval to talk about anything any detail this year so the push forward on so some this may not be uh totally new to you especially if you've been reviewing what what came out of IDF a couple months ago on nature trail but I'm this is something I'm really excited about because it's so it's so the SOC for us on in a couple ways of scatter and a new CPU the he resigned CPU and and new Judy
relative to a previous SOC products so as the is really interesting and we've also been involved on on the bring up and some of the design work from very early on in the open source anything so it's it's exciting for us from abstracted as well and was just told the driven the when this stuff for all Howard solely with windows that sort of thing so it is and the the and as a new SSC under specifically focus that like the tablet market and convertible market and for not so it's not designed to go into phones
and the the power envelope that you need for that and level of immigration and for that but it does make a freebies entablature it's working arts and fat our our 22 in a process which isn't hard leading edge which is the the bottle stuff that we're we're pushing out now that song 14 and near but our 22 namely processes fricatives spinach or on works well for us the really big I mean it is sperm of really big change both CPU and GPU was us on this seaside only and some analysts of stuff was was a priority from early on and analytics team has has been involved up here are open source team doing Linux drivers is involved in we're happy how around the silicon 1st came back from factories during the initial debugging development of our of our driver stack so then who maybe somewhere around here was that that can grant cues eliciting was there helping and we got the in all 3 games up and running and stuff within a day or 2 of getting stable silicon so that was really exciting the now so you
some of the slides you may think that is really pretty handy camel above well I didn't have these mostly stolen from IDF so hopefully there won't
be too 2 full marketing stuff but the the whole world is the overall architecture of the of the chip so it's is necessary is that what you like the blocks on on but the big changes are the new Adam up 2 quad-core and then the new HD graphics which is then not new from the client side it's basically an Ivy Bridge graphics part but it's been shot down and put onto the SSC so if you're familiar or if you've been following our previous Adam stuff you know that um earlier chips used no the IT block from imagination of the GDR graphic stuff and I'm sure everyone has a lot of love from for their stuff and I will be sad see go but now we've got Genigraphics instead so you have to tolerate them but then the IOS that kind of similar feature set to what you see on a quiet part in others the CPU and a graphics power sharing in the sink sort of ice extensions for security and virtualization that sort of thing is all is all present so it's you know it looks like a river Clancy CPU and actually as similar speed the so I said
these are kind of the key the key changes so the big will exceed you said is going to the on architecture and which of type 1 a 2nd on a graphics going from the US checks a running at a lower frequency to the to the Genigraphics running a higher frequency course is a totally from memory model to solve deferred rendering the fertile between its its immediate and resolutions increased in fact we got on the lecture we go pilot-wave of 16 thanks to the the remote say architecture stuff that day was talking about were able to and change the display cock the costly clock at run time if we need to bump up to an even higher resolution number down with his dramatically increase on some of these cues like on a convertible tablets and notebooks and you get like regular PC-style memory bandwidth which is really great and want to new USB as well so that's that's also important and so you have to wait forever for four-year multi-gigabyte USB
stuff our overall done this is what the kind of basic block diagram looks like and and this is this is a bit different than what you would see on a client platform does not like a QPI bus custom ring fabric that we use only atom side just in within the SOC I'm 1 of the big difference is that we have relative to acquire parts at least us say bridge we've had uh the GPU was able to share in the cash grant way the Lasso cache on the CPU and so they're both the GPU and the CPU or or immigrants in Diamond can share the last will catch on here there the same guy but we do not have that cashiering so each of those dual core chunks of IT has their own and share cash but it's independent of the of the graphics so that makes for an that quite a different the performance characteristics of your tuning for performance on on this part versus say an Ivybridge Bridge laptop to unity under different things because of that lack of cash earnings and so have to be careful how you map and and we that memory and and also how you how you set up the rendering and things like that and it has a so that it has the until she graphics plus the the the quick sink stuff that you see on the client side so I can do the regular video decoding and so encode you see in the quiet space and achieve that with that of the PAQ could engine on these parts of that's available and and used in some cases the this 1 looks really
working like but they cannot take away from this 1 is the the the huge improvements in or performance you get to choose their and the huge difference here is this this is the 1st Adam part this out of order so if you looked at our previous microarchitectures on the CPU side there been in order maybe dual issue now we've get up to 4 cores out of water and the performance is just dramatically improved if you if you use when the old Adams like using an arm platform you know you like I don't wanna do it builds on the set I wanted actually log this machine I wanna just copy stuff onto it and and can pretend it's not there this guy you can log into it and it's like using a Core 2 Duo laptop from a couple years ago and it's like a reasonable speed you don't want stab your eyes out when using this is a huge huge improvement from that from that point you have got the new Architecture extensions also on as a reasonable CPU to deal with in terms of having the feature set used to visual found stuff if you're interested in the the 3 D dates and so this
is some sort of benchmarks we've done comparing to the new Adam CPU side to to some the competition and like I said the the other as the abuser fast onto his but they're still terrible when I mean that we that sorry theory on folks new but I I just I can't I used Adam and like the old generation outermost competitive like it actually be out so alarms and in using a like I don't understand how the armed as it's about like this is a terrible thing to you that I don't wanna use this platform this this new 1 is like it's significantly faster so it's we can see on this respect him so this is like what you know this is recurrent branch basically it's it's way way fast and the power consumption is also lower some these guys you know the the UN this is actually ISO power but you can see that at least the clock rate the the data and yeah saying the take even takes more power on the same benchmark and gets less performance I there's legal disclaimers of the and that I don't know what I'm talking about so we have disagree agree with that
the the the letters of the next big chunk is the is the graphics side so getting rid of imagination and you know adulation is for the false they're tuned to the work in the mall space and antenna fairly power efficient and can achieve high performance and look at the nucleus iPad areas with them pretty amazing things with architectures and it's it's proven
I don't like to work on because it's a it's a closed-source Travis stack has a different architecture than than what we have and by Maurice hours so I'm really happy about this change because we have a fully open source stack and but that's mostly personal preference from from that from Intel perspective I think we have a we have an opportunity here to really prove this architecture and the lower power and will and I think we're doing that in in betrayal select you can see it's only got 4 use like it's a pretty simple chop down from the from the the client-side Ivybridge and so the the performances can scale down on what should do what you'd expect minus the LC shared-memory talked about the and all I can say is In terms of futures just keep your eye on the space it's going to get some pretty interesting this year next year and then I talked about the the power and the video decode stuff it's all available to of
here you can see that the gaming sites this is basically comparing graphics performance and whereas in leadership part maintenance for from a graphics perspective it's it's nearly competitive and to me this is actually pretty impressive given that it is a fairly straightforward chopped down versus like a dedicated we designed for mobile and the fact we did this well and with without a tremendous amount of effort tells me that we have quite a bit of a room to improve and the subsequent parts you she really nice + that an open source that's a pretty exciting yeah the it and if it meets the crap out of all of of
media folks in here that there is there some the acronyms and in different contexts but it does have like the what is it has that the same guy
included decode supported and and transcripts for that you see in the client part so and that's nice for my like teleconferencing or in video calling that sort of thing want you so it's all that stuff is there and that's actually as feature of our of our parts and I probably ought to know more about it but I don't spend much time in the on the display side
that on at 10 by 7 can we see this as a chamber of horrors there are demons hanging people and torturing them that on the display side the the the pipe stuff from the register block on the display of mostly matches what you'd see on a quiet on the client side but when you get to the port side they redesigned everything in the custom a custom implantation of the of the display fire itself that's the Chamber of course it's been a thorn in our side I think the harbor does really take place here and torturing us and it's just that totally program model we have a very low level access to all of the fi characteristics and and all of the filtering and things that could get applied on on the on the DAC the for the phi so it that's been a challenge to get get right across a variety of configurations and but on the plus side we've got maybe here we've got an 4 4 by 4 lanes DisplayPort and we can actually go up past 25 by 16 which was mentioned earlier come I think we're at like 38 by 18 or something like that but so it's it's it's not a it's not a bad display protein and it's relatively low power we have some and power-gating features that we can do like Britain when we should when the pipes go off we can actually particle-hole display engine and like we do on the graphics side but on the Renaissance and when the duties at a we can power that so we have that support on the display so of 2 which is really nice so and that the the low aspect of it is it is a pain that were mostly most of the control so this is what we're
shooting for this is this is a but like a display on idle case you know we're just looking at your screen which are not doing anything from the many when a little bit you can see the graphics is lit up but we're shooting for on Linux is this when the
display goes off we want the whole associated basically often power everything system like tiny microcontrollers and this still weight but this is not a for us through state this is a runtime system states of and that the SSE architecture we have now lets us get to really low power consumption through time great gas are we gonna do a really bother with DL many runtime suspend I Yao S 3 in some cases can still save you a little bit more than a full runtime suspend case so will support both both necessary and sometimes to spend a lot of ways going forward it if we can get parity between 3 tend to spend I think that at some point when may just implement 3 in terms of runtime to spend but and there is a there's a logic and there's an overhead to all of the runtime stuff so as 3 is also simpler from platform implantation of used in this trade offs so upstream status of this we
started pushing these bits way back almost 2 years ago now I landed Initial Graphics stuff and 3 6 and
and the and the fold as 0 experts are a runtime suspend supporters and pushed upstream just yet in the course in the core set of things and graphics that's that's pretty much done we just have to get it to get merged and things really stabilize run 3 times of you start to get these platforms for yourself and don't do is isn't so ordering be in that chamber of horrors of the talk about the display up and down quite have talked about so I don't think any of those super high-res display products are out on the market yet but when they are will be ready for them so you will tend to use those worries R C 6 is is a feature on this to the and it works I know some of you may have had problems with the are 2 6 on betrayal is a different connotation of this on the client side and it's actually simpler and it seemed to be pretty stable the I that some some display powerset and stuff like that talked about some other display features that are actually not betrayal specific the dynamic refresh rate support select if the screen is mostly idle you can actually drop the refresh rate system power on your on the memory bandwidth watchin the Member States of refresh for longer I we have some bugs and again chamber of horrors of the this 1 we just fixed of the day before I cut my flat out here and that that and in German that's for we can have so I'm really so happy that we we fix that so I if you're buying the bidual based platform on Amazon or something now you would run into that light right away that the government controls so that should be fixed in the current kernels and will which that stable don't and then there some other issues like Alan Alan Cox got this to 100 machine with a betrayal T a tablet ask you and and it
had done it's actually maybe base panel on that on his machine but it shows up as a VGA so for some reason and and he was as he was trying to configure it you know it's like I used to work of i fied program energy to do this just so happen that is program GA in such a way that we were driving both VGA and and maybe with the same time instant display
at work so there's still some things we have to work out on the side we have tested it tremendously just upstream monarchs we have on our inventories but so so what is improvments there on the MESA side it's been it's been for a long time since Mr. 9 2 and even before I think there's a lot of performance had room here like the performance 1st for some things as actually looking pretty good and but for other things were like half the speed that we could be so that makes a just can continue improving performance there and Ian is probably not here since we're out last until about 6 30 AM and but you can ask him later when he shows up we on the media
side and it's been supported it in the VA as of the one-to-one release the VP stuff is in progress on the way that the VGA that's a integrated they it it shares some logic in terms of power gating and interrupt support with the other 15 IP block so the there are some my 15 changes required and then there's a a bunch of and driver and retrieve that's that's published an open source but it's an integrated upstream and I don't know if the the people at work and are actually pushing upstream or not might lend staging at some point but the it's it's pretty and we still so I'm working on that there are a few platforms out there
now mostly windows stuff and I I keep refreshing on Amazon until it up because I wanna see that the 110
what's going out but I think those are becoming the next couple of months and or if you go to Shenzhen China You can have that will be 1 right now that's that's been another kind of issue people building word configurations and saying it it's a work and if you so religion the students include them here you can also jump on Wikipedia all there but if you're a new idea something you wanna look there's a a new I knocked the next unit of computing the then tells the cover project for until nice little tiny form factors device and but has this 1 of betrayal there's there's a known of has well new betrayal and so both of those really nice platforms for a for a tiny device like the picture 1 hopefully standards and so that's a fun way to get your atrophy was but with the trail you and then there's this new stuff coming to the sky here keep looking OK we
was going to thing I said you can all this because
and full and the last
so any questions in the and
so you're going to be on the MESA side of the graphics use fully support open GL yes 3 . L and when you bring this over to Android you're obviously going be supporting it there as well will you be bringing desktop open GL to the Android side while being able created within EEG all context will we be rainfall OpenGeo to Android and but that's not something wouldn't stop people from doing but I don't think Google has any plans to to expose the full Geo API through the framework and which is really will be required but if you wanna take of you know build Sarajane with Mesa and and a and enable full geo in a bold and then do any stuff that's going to war with think the uninjured like some additional complexity the gradual wrapper library to to kind of the the feature set but that's something that something different if you to and the antecedent so here but you didn't when ask them to the the other questions people on better to brag about the the about this platform and what's from the thanks thank you so next step is going to be about new uh it's going
Wechselsprung
Intel
Weg <Topologie>
System-on-Chip
Weg <Topologie>
Streuung
Zentraleinheit
Computeranimation
Überlagerung <Mathematik>
Umsetzung <Informatik>
Prozess <Physik>
Weg <Topologie>
Mathematisierung
Analytische Menge
Zentraleinheit
Computeranimation
Übergang
Open Source
Graphikprozessor
Spieltheorie
Bildschirmfenster
Softwareentwickler
Leistung <Physik>
Prozess <Informatik>
Open Source
Einhüllende
Graphiktablett
Biprodukt
Ultraviolett-Photoelektronenspektroskopie
Quick-Sort
Druckertreiber
System-on-Chip
Tablet PC
Faktor <Algebra>
Zentraleinheit
Virtualisierung
Gemeinsamer Speicher
Computersicherheit
Mathematisierung
Ähnlichkeitsgeometrie
p-Block
Zentraleinheit
Ein-Ausgabe
Gerichteter Graph
Quick-Sort
Mehrkernprozessor
Computeranimation
Coprozessor
Intel
Rechenschieber
Reihe
Client
Mereologie
Computerarchitektur
Maßerweiterung
Leistung <Physik>
Umsetzung <Informatik>
Subtraktion
Gemeinsamer Speicher
Weg <Topologie>
Datensichtgerät
Mathematisierung
Zahlenbereich
Bridge <Kommunikationstechnik>
Zentraleinheit
Gerichteter Graph
Mehrkernprozessor
Computeranimation
Videokonferenz
Eins
Intel
Informationsmodellierung
Client
Unterring
Rendering
Regulärer Graph
Blockdiagramm
Notebook-Computer
Datentyp
System-on-Chip
Graphiktablett
Bildauflösung
Stochastische Abhängigkeit
Rechenzeit
Graphiktablett
Frequenz
System-on-Chip
Caching
Festspeicher
Mereologie
Bus <Informatik>
Methode der kleinsten Quadrate
Bandmatrix
Computerarchitektur
Decodierung
Charakteristisches Polynom
Schlüsselverwaltung
Subtraktion
Punkt
Weg <Topologie>
Wasserdampftafel
Mikroarchitektur
Zentraleinheit
Term
Physikalische Theorie
Computeranimation
Intel
Virtuelle Maschine
Notebook-Computer
Maßerweiterung
Benchmark
Leistung <Physik>
Verzweigendes Programm
Bitrate
Quick-Sort
Generator <Informatik>
Menge
Mikroarchitektur
Mereologie
Leistung <Physik>
Speicherabzug
Ordnung <Mathematik>
Simulation
Quilt <Mathematik>
Konvexe Hülle
Weg <Topologie>
Mathematisierung
Varianz
Extrempunkt
Term
Gerade
Raum-Zeit
Kern <Mathematik>
Computeranimation
Videokonferenz
Intel
Flächeninhalt
Einheit <Mathematik>
Perspektive
Decodierung
Computerarchitektur
Leistung <Physik>
Inklusion <Mathematik>
Bit
Web Site
Subtraktion
Open Source
Weg <Topologie>
Graphiktablett
Kontextbezogenes System
Computeranimation
Softwarewartung
Hypermedia
Perspektive
Spieltheorie
Mereologie
Hypermedia
Bildschirmsymbol
Weg <Topologie>
Datensichtgerät
Anwendungsspezifischer Prozessor
Graphiktablett
p-Block
Quick-Sort
Computeranimation
Datensichtgerät
Videokonferenz
Übergang
Intel
Hypermedia
Client
Mereologie
Gamecontroller
Programmierparadigma
Decodierung
Dämon <Informatik>
Konfigurationsraum
Leistung <Physik>
Varietät <Mathematik>
Bit
Gewicht <Mathematik>
Schießverfahren
Punkt
Datumsgrenze
Weg <Topologie>
Datensichtgerät
Laufzeitsystem
Rechenzeit
Mikrocontroller
Physikalisches System
Mathematische Logik
Term
Computeranimation
Gerade Zahl
Computerarchitektur
Overhead <Kommunikationstechnik>
System-on-Chip
Leistung <Physik>
Touchscreen
Aggregatzustand
Expertensystem
Bit
Diskretes System
Datensichtgerät
Potenzmenge
Rechenzeit
Physikalisches System
Biprodukt
Bitrate
Computeranimation
Kernel <Informatik>
Programmfehler
Virtuelle Maschine
Client
Menge
Festspeicher
Trennschärfe <Statistik>
Tablet PC
Vererbungshierarchie
Speicherabzug
Bandmatrix
Leistung <Physik>
Aggregatzustand
Touchscreen
Tropfen
Virtuelle Maschine
Energiedichte
Datensichtgerät
sinc-Funktion
Hypermedia
Optimierung
Computeranimation
Information Retrieval
Punkt
Open Source
Mathematisierung
IRIS-T
p-Block
Ultraviolett-Photoelektronenspektroskopie
Term
Mathematische Logik
Interrupt <Informatik>
Computeranimation
Hypermedia
Verknüpfungsglied
Druckertreiber
Arithmetische Folge
Bildschirmfenster
Leistung <Physik>
t-Test
Computeranimation
Überlagerung <Mathematik>
Weg <Topologie>
Formfaktor
Einheit <Mathematik>
Wechselsprung
DoS-Attacke
Wort <Informatik>
Projektive Ebene
Mini-Disc
Konfigurationsraum
Standardabweichung
Subtraktion
Konvexe Hülle
Gebäude <Mathematik>
Automatische Handlungsplanung
Humanoider Roboter
Kontextbezogenes System
Räumliche Anordnung
Komplex <Algebra>
Framework <Informatik>
Computeranimation
Menge
Einheit <Mathematik>
Offene Menge
Wrapper <Programmierung>
Programmbibliothek

Metadaten

Formale Metadaten

Titel Intel BayTrail graphics overview
Untertitel Discussion of the Intel BayTrail platform and architecture
Serientitel FOSDEM 2014
Autor Barnes, Jesse
Lizenz CC-Namensnennung 2.0 Belgien:
Sie dürfen das Werk bzw. den Inhalt zu jedem legalen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen.
DOI 10.5446/32548
Herausgeber FOSDEM VZW
Erscheinungsjahr 2014
Sprache Englisch

Inhaltliche Metadaten

Fachgebiet Informatik
Abstract Discussion of Intel BayTrail SoC architecture from a graphics perspective, including overview of render engine, display engine, memory architecture characteristics, and current status in Linux. Hopefully the presenter will have some sample platforms for people to play with after the talk. The Intel BayTrail platform incorporates a new, out of order Atom CPU, an Intel HD graphics engine, and various IP blocks to support both tablet and notebook platforms. This talk will give an overview of the new hardware features, and the status of the graphics driver support in Linux.

Ähnliche Filme

Loading...
Feedback