Graphic Cards help me on my seminar on GPU

sagar.sdm

sagar.sdm

Disciple
Namasthe to all dudes..
(pardon me if this is the wrong place to post this thread)

I'm pursuing final sem of engineering and I have chosen Graphics Processing Units (GPU) as my seminar topic, which is a part of the curriculum.

So, If you have any links that can be good source for me to refer... or if there's any new technology advancement in the same domain, which is good to include in my seminar.. or rather any suggestions/help are all welcome..

Hoping to get enlightened :! from you TE geeks :cool2:

Thanks in advance.
 
get ATI details from muzux...and nvidia details from sumulation3..ask them to debate the issues abt these cards...note down the points..u r set for 50/50...
 
thanks all for the quick reply..

@ cricfan.. muzux and sumulation3.. TE members right ? will they help me out with this.. ?

please help me with the links if possible..

Thanks again !
 
^^ One WAS a TE member... never mind, what do you want to know about GPU? Also, please tell me from where you are doing your engineering.

I have no engineering background but as you prepare for seminar, I will too learn a few things in the process.
 
cricfan said:
get ATI details from muzux...and nvidia details from sumulation3..ask them to debate the issues abt these cards...note down the points..u r set for 50/50...
Click to expand...

LOL.
And convince them to prepare the content for your seminar as well...
 
@ morgoth.. well bro I'm quite comfortable with their installation and to get them working wonders in gaming... but now that I need to give a seminar on the same... I need to know its basic working principle.. till the latest advancement in the same domain.. and need to get to know those things which will add weight to my seminar.. hope you get what I mean..

I have some basic idea.. but not so much so to put it on paper or rather to speak out the same to my lecturers.... for which I need some help ..
 
sagar.sdm said:
@ morgoth.. well bro I'm quite comfortable with their installation and to get them working wonders in gaming... but now that I need to give a seminar on the same... I need to know its basic working principle.. till the latest advancement in the same domain.. and need to get to know those things which will add weight to my seminar.. hope you get what I mean..

I have some basic idea.. but not so much so to put it on paper or rather to speak out the same to my lecturers.... for which I need some help ..
Click to expand...

If you just know how to install GFX cards and that you can play games on them then there will be a lot of ground we will need to cover. As I told you earlier, I have no engineering background and I am sure that I am not the best person to provide you with inputs regarding GPU techology. But given that you seem to have very little info about GPUs, at least I can give it a try. Start with: http://www.techenclave.com/guides-and-tutorials/a-basic-guide-to-graphics-cards-130281.html

This is a a very basic guide to GPUs that I wrote to help a newbie decide about his first GFX card and it has no info about what really goes into a Graphics Processing Unit like RV770. It's more of a buying guide. But let us begin with it and then we will follow it up with basic architecture of GPUs.
 
morgoth said:
If you just know how to install GFX cards and that you can play games on them then there will be a lot of ground we will need to cover. As I told you earlier, I have no engineering background and I am sure that I am not the best person to provide you with inputs regarding GPU techology. But given that you seem to have very little info about GPUs, at least I can give it a try. Start with: http://www.techenclave.com/guides-and-tutorials/a-basic-guide-to-graphics-cards-130281.html

This is a a very basic guide to GPUs that I wrote to help a newbie decide about his first GFX card and it has no info about what really goes into a Graphics Processing Unit like RV770. It's more of a buying guide. But let us begin with it and then we will follow it up with basic architecture of GPUs.
Click to expand...

Done with your basic-guide. Shall we move to the basic architecture of GPUs ?

Thanks for the your patience and concern dude..
 
yup.. better u keep the seminar specific to the technology rather than stressing on individual achievements from manufacturers..

start wid how a gpu...as a whole...helps the system wid graphics processing..

then go into details on what technologies have been involved to make that process faster n better.

put a little insight on the future of gpus n technologies xpected...

do include little notes about dx11 n cinema 2.0 ..include pics n vids for the same n tat will act as extra spice.

although avoid stressing too much on it...

also include more info on gpus used professionally...

finally u can finish wid discussion on gpgpu n or sumthin other exciting upcoming technologies..
 
sagar.sdm said:
Done with your basic-guide. Shall we move to the basic architecture of GPUs ?
Click to expand...

yes we can...

A GPU has following important components:

1- PCB with all the circuitry that connects the different components

2- Output connectors (VGA, DVI, HDMI, S-Video, Display port etc). To know more about each, fire up Wikipedia.

3- GPU. As TomsHardware once wrote it is the heart of a graphics card. The calculations take place here.

4- Memory (VRAM). It stores the calculated frames/data

5- Power regulation. The area with all the capacitors etc. Responsible for supplying clean power to GPU and memory

6- Cooling which consists of active components like fans and passive components like heatsinks, heatpipes.

It is the GPU which is of primary concern to you... and me. Now, before we move ahead here is a basic glossary of what goes inside a GPU:

Vertex processors: vertex processors are components on the graphics processor designed to process shaders that affect only vertices

Pixel Processors (a.k.a. Pixel Shader Units)

A pixel processor is a component on the graphics chip devoted exclusively to pixel shader programs. These processing units only do calculations regarding pixels.

Unified Shaders

ertex geometry and pixel shader code structures will be functionally similar but have dedicated rolls. DirectX 10 required and it's now it is a big thing with GPGPU

Texture Mapping Units (TMUs)

Textures need to be addressed and filtered. This job is done by TMUs that work in conjunction with pixel and vertex shader units.

Raster Operator Units (a.k.a. ROPs)

The raster operation processors are responsible for writing pixel data to memory. The speed at which this is done is known as the fill rate.

Memory bus: Memory Bus

The memory bus is one of the most important aspects of memory performance. A modern graphics card's memory

bus can range from 64 bits to 256 bits, and in some cases, it can be 512-bits wide. As the bus width increases, so does the amount of data that it can carry per cycle, and that is very important for performance.
Click to expand...

Reference:

Graphics Beginners' Guide, Part 1: Graphics Cards : Graphics Card Fundamentals - Review Tom's Hardware

Graphics Beginners' Guide, Part 2: Graphics Technology : Graphics Card Technology - Review Tom's Hardware

Graphics Beginners' Guide, Part 3: Graphics Performance : Graphics Performance Dissected - Review Tom's Hardware

Once you are through with them, we will delve more into GPU architecture and how they work :)
 
madnav said:
yup.. better u keep the seminar specific to the technology rather than stressing on individual achievements from manufacturers..

start wid how a gpu...as a whole...helps the system wid graphics processing..
then go into details on what technologies have been involved to make that process faster n better.
put a little insight on the future of gpus n technologies xpected...
do include little notes about dx11 n cinema 2.0 ..include pics n vids for the same n tat will act as extra spice.
although avoid stressing too much on it...

also include more info on gpus used professionally...
finally u can finish wid discussion on gpgpu n or sumthin other exciting upcoming technologies..
Click to expand...

Thanks Madnav.. If you provide me with links which explains in brief more of the underlined quotes it would be more helpful. I have the links for the rest you have mentioned.
 
sagar: I'd have hunted down a few links and helped you, but the fact is that I'll be offline for about two or three weeks starting tomorrow. Very quickly though, you won't get much out of enthusiast / review sites. At least, not enough for a paper in engineering.

Your best bet is to start asking at dev forums. Intel is actually a good start. Graphics cards are about both hardware and software, so if you don't understand one part, you won't understand the other.

At a basic hardware level, it's a heavily paralleled CPU architecture with a very fast local cache, so it's not that much different, from, say, a RISC computing cluster. The outputs are simply DACs (or encoders, for DVI and DP/HDMI outputs). Everything else is power and connections. That's pretty much it, and development over the years has focused on more parallelism, faster caches and wider data paths. So it's not the hot topic it looks like - the beauty is in the drivers, really. They translate commands into action, and understanding that will also be critical.

Have fun, I'll check back in a few weeks' time to see how your paper went :)

All the best.
 
cranky said:
sagar: I'd have hunted down a few links and helped you, but the fact is that I'll be offline for about two or three weeks starting tomorrow. Very quickly though, you won't get much out of enthusiast / review sites. At least, not enough for a paper in engineering.

Your best bet is to start asking at dev forums. Intel is actually a good start. Graphics cards are about both hardware and software, so if you don't understand one part, you won't understand the other.

At a basic hardware level, it's a heavily paralleled CPU architecture with a very fast local cache, so it's not that much different, from, say, a RISC computing cluster. The outputs are simply DACs (or encoders, for DVI and DP/HDMI outputs). Everything else is power and connections. That's pretty much it, and development over the years has focused on more parallelism, faster caches and wider data paths. So it's not the hot topic it looks like - the beauty is in the drivers, really. They translate commands into action, and understanding that will also be critical.

Have fun, I'll check back in a few weeks' time to see how your paper went :)

All the best.
Click to expand...

Thanks for your input dude. But now that you have told me that software plays equally important role as the hardware.. I'm a bit worried.. I'm comfortable with the hardware part with almost zero knowledge about the software part.

I will have to explain something about the software part in the seminar. So, if possible please provide me a link which gives a basic overview of the same..

Thanks again !
 
morgoth said:
yes we can...

A GPU has following important components:

1- PCB with all the circuitry that connects the different components

2- Output connectors (VGA, DVI, HDMI, S-Video, Display port etc). To know more about each, fire up Wikipedia.

3- GPU. As TomsHardware once wrote it is the heart of a graphics card. The calculations take place here.

4- Memory (VRAM). It stores the calculated frames/data

5- Power regulation. The area with all the capacitors etc. Responsible for supplying clean power to GPU and memory

6- Cooling which consists of active components like fans and passive components like heatsinks, heatpipes.
It is the GPU which is of primary concern to you... and me. Now, before we move ahead here is a basic glossary of what goes inside a GPU:
Reference:

Graphics Beginners' Guide, Part 1: Graphics Cards : Graphics Card Fundamentals - Review Tom's Hardware

Graphics Beginners' Guide, Part 2: Graphics Technology : Graphics Card Technology - Review Tom's Hardware

Graphics Beginners' Guide, Part 3: Graphics Performance : Graphics Performance Dissected - Review Tom's Hardware
Once you are through with them, we will delve more into GPU architecture and how they work :)
Click to expand...

Done with the tom's hardware links you had given bro... next step ?
 
^^ Ah, yes!

More we will delve into it. But you know, soon I am going to hit the dead-end. I told you that I am not an engineer and you will have to source either books or find someone who is into the field to know more about magic of 3D rendering because whatever I read or know is sufficient for an amateur hobbyist like me but professionals need to know a lot more.

I can only tell you that a GPU has this number of transistors and how they are stacked. I can tell you a little bit of theory about GPU hardware and what are the steps involved in creating 3D images. But how transistors do their magic once they are fed power can only be learned from someone who has a solid background in this field :)

But that is later, right?

Today, we are going to see how exactly a GPU looks like. So far in gaming, Rasterization is the preferred method of 3D rendering and GPUs are designed to be raster monsters. CPUs are good at Ray Tracing (though the task in itself is so computing heavy that * feasible* real time ray tracing is not possible for now) and are "still" used by Hollywood production houses to create animated films/scenes.

Now that you have gone through Tom's articles, you know what ROPs or TMUs do and what is the use of memory or memory bus. So it's time to see how they are really stacked inside a die and how they process instructions:

For an example we will take RV770, the architecture from AMD/ATI on which HD 4850 and 4870 are based.

here is the die shot of RV770. It's the silicon fabricated on 55nm process at TSMC:





Of course, unless you are a GPU engineer this doesn't give you much idea about the chip. But this diagram will give a little more clear picture:



This shows the various engines/components that are responsible for various functions inside the GPU. 16 ROPs/RBEs are in the four engines at the bottom. Each of them has 4 ROPs.

another pic:



Each SIMD core (there are ten cores) has 80 (arrayed in 4+1 config) Unified shader processors and each of them is connected to one TMU engine that has 4 TMUs each.

Today's reading: ATI Radeon HD 4800 Graphics Architecture: Long Anticipated Revenge? - X-bit labs

Terascale Graphics – ATI's new RV770 GPU - 3D Graphics, Audio & HDTV by ExtremeTech

These two articles explains in (somewhat) details the RV770 architecture and will help you understand the basic GPU architecture.

This will give you the basic idea about the GPU architecture and hardware. Next we will see how it is used to create complex 3D scenes. Note the use of "we" because soon you and me are going reach one platform and them you will to find someone else who can assist you further or walk alone ;)
 
well, i had gfx design as a course in one semester.. Probably they might ask you why you need a GPU? then you may have to explain how a GPU helps in rendering.. what is Dithering, Anti-aliasing(different alogorithms), Anisotropic Filtering etc. etc....well it all depends how are your profs there :p .. if they like to screw people then better you get your basics right..
 
Back
Top