The concept is the same of a cluster, but it is in a single processing unit. A cluster requires several processing units, this is the main difference (or else we cound consider an I7 a cluster, but anyway).
This is all i know how to explain, but basically the GPU is specialized in several areas where a normal CPU isn`t and would suffer to emulate. Also the calculation of vector and matrix operations (which are the most common in Graphics) are really faster because of, again, the speciallized architecture for this. I don`t have a link to any white paper that i can remember right now, but please have a look at this link of wikipedia, that may help you understand a little more about GPUs and then you`ll clearly see the reason why it`s almost impossible to emulate through software at a decent speed. http://en.wikipedia.org/wiki/Graphics_processing_unit 2010/6/18 Katrina Payne <fullmetalhar...@nimhlabs.com> > Thanks ZuM > > I apologise, I may have been misinterpreting the TLA, of GPU to mean > "Graphics > Processing Unit"... I guess I must be interpreting it wrong, as what you > described was a cluster, not a unit. > > Though, this is not the first time a hardware or software term was fairly > misleading. > > Thank you, as now I understand it is not a unit, but a cluster of units. > > Sorry for the issues, I may have caused in my misinterpretation of what the > TLA indicated. > > Back to your regularly scheduled mailing list. > > ~Katrina > > On Friday, June 18, 2010 06:25:03 am ZuM wrote: > > Hi, > > > > The major differences are that while a CPU has one processing core (or > > today, two processing core), any older GPU has quite a lot of processing > > units that run at a very low speed but are specialized in certain areas, > > this way the load is really divided. Also just a comment, several labs > are > > starting to use GPU to do MAJOR calculations specially because of their > high > > paralel processing oportunity. > > > > And for instance Shader calculation is a very expensive method, but for > > instance most of the more modern cards have several shader processing > units > > in which their architecture and also their basic functionality are > optimized > > to do this shader calculation that would make a CPU cry. > > > > The major differences Katrina between the two is that a CPU is a generic > > processing unit, which contain a lot of commands and functionalities, > while > > a GPU contains quite a lot of VERY specialized processing cores. This > > differences is what makes it be really faster than using Software > rendering, > > specially since some with Software rendering you would need to emulate > some > > features that are basic nowadays with the GPUs, and do so much > calculations > > in a serial way, while trying to maintain a decent 3d output. So again i > > say, the most important feature of the GPUs is their paralel programming. > > > > 2010/6/18 Katrina Payne <fullmetalhar...@nimhlabs.com> > > > > > That is good, I never asked for a lecture anywhere. > > > > > > If anything, I asked for some terms I could use in Google. A mantra or > some > > > such concept I guess. > > > > > > I also asked for text books and white papers. > > > > > > No where did I ask for a lecture. > > > > > > I even flat out stated many times that what you were to explain would > be > > > too > > > long for an email to fit, so the idea that I would be asking for a > lecture > > > is > > > ludicrous. > > > > > > Now then, rather than saying this obvious statement, how about you just > > > tell > > > me some google search mantra to use (as really, my mind is blank for > any > > > keywords I could use that would get me anything meaningful), a set of > text > > > books, or maybe a white paper? > > > > > > I do not appreciate feeling like I am being talked down to. > > > > > > It is also statements like this below, that make me think you are > > > disregarding > > > my suggestions, as, if you were not disregarding them, you would have > just > > > gave those text book titles, or some keywords to search with, or at the > > > very > > > least a basic white paper. > > > > > > Had it been some standard of programming, I could attack ISO, RFC, > Working > > > Draft, Recommendation, Best Practices, Primer or Howto to something > like > > > "GPU". > > > > > > In this case, I do not expect any of those to work. > > > > > > Now then, please, stop talking to me like a fool, and blantantly > > > disregarding > > > stuff I have asked for, and brushing me off as somebody expecting a > lecture > > > over > > > a mailing list--or some such retarded notion you have gotten into what > I > am > > > asking for here. > > > > > > Thank you > > > > > > ~Katrina > > > > > > On Friday, June 18, 2010 05:58:37 am Bob Somers wrote: > > > > Katrina, I'm not giving lectures on computer graphics here. Google > has > > > > all the information you asked for. If you'd like, I can also > recommend > > > > some graphics textbooks which would clear things up > > _______________________________________________ > To unsubscribe, edit your list preferences, or view the list archives, > please visit: > http://list.valvesoftware.com/mailman/listinfo/hlcoders > > _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlcoders
