Please explain gpu

Okay guys
Please explain what the bus speed does, memory, shador clock, engine clock, core clock does?
Does bus speed affect memory?

 

So complicated...i'll come back later and post it all if no one beats me to it. I gotta get some stuff done first.

 

Bus Speed - Consider this your video card's processor. The faster it is, the faster it can crunch numbers to do calculations for rendering. A higher bus speed also means increased heat output just like on a cpu.

Memory - Your gpu's RAM for storing data and images and whatnot that it needs to do its job. The amount needed depends on application type and the performance of the gpu.

A higher amount of memory does NOT mean a faster video card. Think of it this way - You have two hands. I give you a suitcase to store stuff in. You can hold this with one hand. I give you another one. You now have 2 suit cases, one for each hand. I now give you an additional 2, bringing you up to 4 suitcases. But you can only use 2 at a time, making the other 2 rather useless unless you put down the current two to pick up the other two, which takes additional time and energy.

For games, 512mb is pretty much the gold standard for a decent card. 1GB is overkill for all but the highest performers. Most midrange cards like a 9600 or 3650 have 512 or 256MB.

Memory Type - A little clarification on the two main types of gpu ram (hereby called vram). There is ddr2 vram and ddr3 vram. Much like normal ram, ddr3 vram is faster than ddr2 vram. However unlike in desktops the performance differences between ddr2 and ddr3 ram is colossal. Due to the sheer number of calculations and processes that a gpu runs, the ability to very quickly store and then retrieve this information from memory very rapidly is crucial to all graphics applications.

DDR3 is much faster than DDR2. I believe all 256-bit cards use DDR3 vram now. Many midrange 128-bit cards are using ddr3 vram too. A ddr3 version of a card will run circles around the ddr2 equivalent. The 8600gt with ddr3 was noticeably faster than the ddr2 version. And the 8700gt which was a 512mb ddr3 equipped 8600gt with a higher bus speed was quite the performer for a 128-bit card.

Shader/Engine/Core clock - ...Okay I have to look this stuff up but as usual, higher rates = better. This information is generally unimportant to know until you want to overclock your video card. The other specs are more...visible. Basically when looking at the specs sheet for a laptop the video card's memory size, memory type, and name are given. The card's name denotes it's bus speed and memory

Oh and here's some extra info

The GPU "bit" number - Okay So I used the phrases 256 bit and 128 bit without explaining them properly first so I'll do an explanation here. It's a little complicated but you should be able to understand it. Information in computers can be broken down and broken down until it is nothing more than 1's and 0's - binary. If you know binary, then you know how to write the numbers in binary since it's all based on powers of 2. The power denotes how many memory locations each bit has. 2^7 = 128 so 7 memory locations = 128 bit. and 2^8 = 256 so 8 memory locations generates a 256 bit. . As you can see, the 256 bit card has an extra memory location over a 128 bit card which alows for greater amounts of data to be transferred per bit.

A 256-bit card has the ability to send and receive larger amounts of data than a 128-bit card since there are extra memory locations. All the clock speed, memory speed, and memory storage in the world can't overcome a limit in bit-width. This is why when given the choice between a lower clocked 256-bit card and a higher clocked 128-bit card, you go with the 256 every time!

 

Thanks i gave you a rep.

I cant find my shader clock for my mobility 3450. Its not listed in gpu z or notebookcheck.

 

Memory Bus: the interface between the GPU and video memory, the larger or "wider" the better. Usually come in these sizes: 64 bit, 128 bit, 256 bit. It's a reflection of how many bits of info are moving per clock cycle. This x the memory clock speed + the type of memory being used give you the memory transfer rate. Of course, the higher the better.

Memory Speed: the number of cycles per second that RAM transfers data. The faster the better of course. True memory speed is usually half the amount said, as DDR means "double data rate" where data transfers twice per clock cycle (I believe that's the right explanation).

Video Memory (VRAM): This is memory (or RAM) used for video and display purposes. There are many solutions out there. Usually are of the DDR2, GDDR2 or GDDR3 type.

Types of video memory solutions:
-Shared/Integrated: all of the video memory is actually system RAM memory (SRAM) that has been allocated to be used for video purposes. Because this is system RAM used for not just video, it's slower than dedicated video RAM since more than just video information is being transfered at once and has to share the bandwidth with the computer's other memory needs. Easily the lowest end type of video memory solution. The faster the system RAM, the better.

-Mixed/Hybrid: This type has dedicated VRAM for only video purposes as well as shared SRAM in order to speed up things a good bit. Not as prevelant as it used to be, it was a cheap way to have some dedicated VRAM without being too expensive but still having a fairly large total video memory pool. Most hybrid set ups have 64 MB of dedicated VRAM plus a much larger amount of SRAM allocated to video usage (192 MB or larger) to create the total video memory pool.

-Dedicated: This is the best solution for any graphics set up as all the video memory is dedicated to graphics use only and is actual RAM only for video. There is no need to regulate a shared video memory pool either hence the fastest type of memory system compared to the other 3.

Core clock/engine clock: the number of cycles per second in the "ingoing and outgoing parts" of the GPU. In newer GPUs from Nvidia (8xxx series and onward), the core clock and shader clocks are independent. All pre-8 series GPUs from Nvidia and all GPUs from ATi have non-independent shader and gpu clocks. The higher the clock speed means more cycles per second hence faster of course.

Shader clock: clock cycle speed of the shader units in 8 series and newer Nvidia GPUs. ATi GPUs since the 2xxx series have the same speed in the shader and main core GPU areas, and they are negligable because of that. Nvidia's shader units are usually clock much higher than the main GPU components. The faster, the better of course.



The graphics solution is dependent on all these components.

Typical graphics solutions you might see:

Very low end/integrated:
Any Intel graphics solution qualifies as such. Have no dedicated video memory. Decent for photo and video editing. Not good for gaming, but much older games can scale with them easily. The best integrated solutions out their is ATi/AMD's Radeon HD 3200 and Nvidia's GeForce 9400M. They are actually quite capable but held back by the lack of dedicated VRAM, however some laptops with the Radeon HD 3200 actually have a small amount of dedicated VRAM with shared memory to create hybrid graphics memory systems.

Low end dedicated/hybrid: you won't see to many hybrid systems but there is one particular one out there but the Radeon HD 3200 is sometimes in this category, especially in HPs since they use both dedicated and shared memory for a hybrid solution. Typical low end dedicated graphics have 128 or 256 MB of dedicated video memory most usually with 64-bit memory buses. Typical GPUs in this category are the Radeon HD 3450/3470 and Nvidia GeForce 9200/9300. You can lightly game on these with newer games as they won't have the ability to run newer games at full graphics, but at lower settings they work fine. If all you want is to video and photo edit, these are very good solutions for that kind of work. However medium end dedicated graphics are usually best as they have much larger dedicated VRAM pools for that work.

Medium end dedicated: Video graphics solutions such as the Radeon HD 3650/Nvidia GeForce 8600/9500/9600 and usually have 128-bit memory buses with 256 or 512 MB of dedicated video memory. 256 MB of memory is enough for these solutions when it comes to gaming, the 512 MB of memory is good if you want to video edit. As far as gaming goes, these are most certainly capable of it, typically running well with a mix of medium and high settings in games. If you're wondering, medium end GPUs such as the ATi Radeon HD 3650 and Nvidia GeForce 8600 GT are quite on par with the 360 graphically. If you're certainly worried about decent gaming on a laptop, look into a medium end graphics equipped system. You can find very well priced machines with medium end graphics, hence they are pretty popular with gamers who want to spend not too much money.

High end dedicated: Video graphics solutions such as ATi Radeon HD 3850/Nvidia GeForce 8800/9800. Will have 512 MB of video memory or more. Usually have the graphics power to run any game at high/max settings. If you're a hardcore gamer get a system with this kind of graphics or better if you highly value having the best graphics experience better. GeForce 9800M GS system is actually twice as powerful as the 360s graphics system. Almost always have 256-bit memory buses and GDDR3 video memory. Some have GDDR5.

 

I'd figure that high-end is anything with a 256-bit memory bus. The difference between the other categories is a little more grey, imho.

 

The shader clock is the same as your GPU clock. In ATi GPUs, they don't have seperate clock domains like Nvidia chips.

And I edited that last bit (lol bit) for ya Delta

 

This is a good read: http://www.notebookreview.com/default.asp?newsID=4563

 

Guessing thats a typo?

 

LOL
I was about to point out the same thing

But apart from that, great thread
All the posters that have contributed, great job..........

I've learned quite a bit more about my card

 

fixed my typo

Would like to point out the difference between a few video cards based on some 3dmark06 scores I ran using mine and some friends' computers. In parenthesis is the processor

Asus C90 w/ 512mb ddr2 8600gt - ~3900(desktop E6420 with OC to processor)
Dell 1520 w/ 256mb DDR2 8600gt - ~3400 (T7500)
Dell 1505 w/ x1400 64MB DDR - uuuhhhh...I don't want to embarass this computer too much so lets just go with very low (T7200)
Gateway p7811-FX w/ 512mb 9800GTS - ~9000 (p8400)
Dell XPS 1530 w/ 256mb DDR3 - ~4800 (p8300, he might've done some OCing of the card, he uses rivatuner and loads a profile at start but this should be stock settings.)

Now notice a few things here. My 8600gt has twice the memory of the dell 1520 but I only got a higher score because I had a much faster processor than him and thus boosted my score a little over a mobile processor. The xps 1530 on the other hand with ddr3 memory and a mobile processor easily trumped both ddr2 8600gt's and his cpu has a much lower clock than mine did.

Now look at the gateway compared to the xps 1530. The p8300 and p8400 are very similar in speeds so the main differences will all come down to the gpu. The 9800gts has 512mb of ddr3 and is a 256-bit card. The 8600gt has 256mb of ddr3 and is a 128-bit card. The result? The 9800 is about twice as good. That's a pretty good example of all of this coming together, or at least as close as I can make it.

Oh I should also mention before anyone points this out that the gateway and xps are at 4GB of ram the rest were at 2 GB so that was a small factor in this but 3dmark scores are predominately decided by cpu and gpu.