General questions about GPU's

[Megazord552]

Hello! I'm new to the whole computer building scene, but I've been reading a lot and watching LinusTechTips in order to get an idea of how to look at things.

I have a some questions about graphics cards:

1) Just to clear my terminology, are GPU's, graphics cards, video cards all the same thing?

2) What are factors that affect performance of a GPU (would be helpful if each one is explained in brief)? (stream processors, RAM, CUDA cores etc) Do higher numbers always mean better?

3) How does VRAM differ from the RAM on my motherboard?

4) Does VRAM add linearly over SLI/Crossfire? (2 GTX 980 Ti's in SLI = 2*6 = 12GB?) and would the performance be comparable to a card with a similar VRAM? (eg. 2 980 Ti's in SLi ~ 1 Titan X?

I'm a total n00b at this. Still learning the ropes here, hopefully I can get a rig going in the near future,

Thanks in advance for answers!

[AlwaysFSX]

1. Yup

2. I skip, too much for my fried brain to go over

3. 

4. No, but it might with DirectX 12.

[C0LL0SS0S]

Well technically, the GPU is the chip that sits in graphic and video cards, but its used the same mostly,

You are better of with just looking at benchmarks, but normally more is better, but not always says much about performance

Vram is ALOT faster than RAM

Vram does not add up in SLI, rumours say it will be like this in DX12 but for now nope,

[NeatSquidYT]

1) No, but yes. A GPU is a graphics processing unit. This could be integrated, on board or in video card form. A graphics card, video card and VGA are all the same things, as long as VGA stands for video graphics adapter, not the video output.

2) Stream processors and CUDA cores are accelerators for workstation uses iirc, and VRAM is used to cache textures etc., but apart from that, I don't really know either.

4) No it doesn't.

[dadasmithywinkle]

1) Yep, all the same. Just different names.

2) Kinda like comapring an AMD CPU to an Intel CPU, they all are quantifiable measurements but not always a sign of performance

3) VRAM transfers memory at far fast speeds. The amount of data sent to the screen determines how much VRAM you need. The amount of data used by processes determines how much RAM you need.

4) As of now it doesnt not. If you have two 980ti's, you only get a slight boost i performance but no extra video memory.

 

No worries and welcome to the forums!

[minibois]

-snip-

1) Yes. VGA is sometimes also a term used for it, but those ones you mentioned are the most common.

2) Real life benchmarks show what GPU is better or worse. The specs do not directly increase or decrease fps.

VRAM is sometimes important, but still not always. Sometimes they give low end cards a high amount of vram which they wont be able to use  (because they are too weak)

3) GDDR5 vs. DDR3. LTT did a video on this on their 'TechQuickie' YouTube channel.

GDDR5 is basically just faster and for video cards.

4) No. Not at the moment.

2 4GB GPU's in SLI means 4GB usable VRAM. 

And in games, the 980 Ti seems to perform about equal to the Titan X

[Bananasplit_00]

1 GTX980TI is basicly the preformance of a titan X anyways

[NeatSquidYT]

1. Yup

2. I skip, too much for my fried brain to go over

3. 

4. No, but it might with DirectX 12.

1. Not really

2. VRAM doesn't have to be GDDR.

[Comessy]

1) Just to clear my terminology, are GPU's, graphics cards, video cards all the same thing?

Yes

2) What are factors that affect performance of a GPU (would be helpful if each one is explained in brief)? (stream processors, RAM, CUDA cores etc) Do higher numbers always mean better?

Arcitecture is the main factor. Higher numbers within one arcitecture means more performance

3) How does VRAM differ from the RAM on my motherboard?

Vram is used by GPU, ram used by CPU. Not interchangable. Vram is also MUCH faster

4) Does VRAM add linearly over SLI/Crossfire? (2 GTX 980 Ti's in SLI = 2*6 = 12GB?) and would the performance be comparable to a card with a similar VRAM? (eg. 2 980 Ti's in SLi ~ 1 Titan X?

No, but might in the future

[AlwaysFSX]

1. Not really

2. VRAM doesn't have to be GDDR.

Same terminology, and it's the only thing that's different vram than on a motherboard.

[NeatSquidYT]

Same terminology, and it's the only thing that's different vram than on a motherboard.

GPU isn't just in card form, and that doesn't make any sense

[AlwaysFSX]

GPU isn't just in card form, and that doesn't make any sense

It works..

[Megazord552]

Thanks guys! I'm going to do a bit more digging before I finalise anything. Really appreciate all the help! Will get back if I need to know more/share progress

[Circs]

Hello! I'm new to the whole computer building scene, but I've been reading a lot and watching LinusTechTips in order to get an idea of how to look at things.

I have a some questions about graphics cards:

1) Just to clear my terminology, are GPU's, graphics cards, video cards all the same thing?

2) What are factors that affect performance of a GPU (would be helpful if each one is explained in brief)? (stream processors, RAM, CUDA cores etc) Do higher numbers always mean better?

3) How does VRAM differ from the RAM on my motherboard?

4) Does VRAM add linearly over SLI/Crossfire? (2 GTX 980 Ti's in SLI = 2*6 = 12GB?) and would the performance be comparable to a card with a similar VRAM? (eg. 2 980 Ti's in SLi ~ 1 Titan X?

I'm a total n00b at this. Still learning the ropes here, hopefully I can get a rig going in the near future,

Thanks in advance for answers!

 

1. A graphics and video cards are the same thing. The GPU is technically the processor sitting in the middle of the card. These chips are not sold loose and the terms are sometimes used interchangeably.

 

2. Broadly speaking you have: Unified Shaders, Texture Mapping Units, and Render Output Units. These are comparable only across a single GPU family. (Don't use it to compare a GTX 600 series to a GTX 900 series or to compare AMD and Nvidia products.) CUDA is just an Nvidia trademark referring to groupings of these components I mentioned and Stream Processors are the AMD equivalent. Generally speaking more is better.

 

VRAM is the RAM for your graphics card. Think of it as the desk your GPU does its work at. Again as a rule more is better, but just like a ridiculously huge desk once you've got enough space more doesn't really matter.

 

3. VRAM generally uses a different memory geometry that makes it faster for the specific type of work it does in a graphics card. Regular system memory is used on some low end cards, but its performance is lower as it is more optimized for general computing. Generally more is better, but past a certain point you hit diminishing returns.

 

4. At this point the answer is no. DirectX 12 and its competitors plan to bring support for this feature in the future, but don't build now with the hope they'll backport it to your card. Even if they do we haven't seen how it'll perform and if any compatibility issues may arise.