Difference between vRAM and shared regular RAM?

Functionally, the RAM on a graphics card is typically the same as the RAM used by your OS. However, the fact that the graphics card has DEDICATED RAM means that it will use that RAM first before looking to the system's RAM for help.

Ideally, you use your system RAM for your OS and applications, and let the GPU use it's own RAM. Having the separation of responsibility is just better for all involved.

Now, if you want to talk about the differences between CPUs and GPUs, then you're getting into a far more complicated conversation

I'm not sure, but isn't there also a speed issue? If the GPU uses VRAM, it doesn't have to go out on the main memory bus and transfer information back and forth to main memory. I think it should be faster in talking to the dedicated memory to which it is hard-connected.

mkrishnan said:

I'm not sure, but isn't there also a speed issue? If the GPU uses VRAM, it doesn't have to go out on the main memory bus and transfer information back and forth to main memory. I think it should be faster in talking to the dedicated memory to which it is hard-connected.

Click to expand...

I don't know for absolute certain, but I would imagine that's at least somewhat true. I think you're talking about a difference that's measured in microseconds from one to the other, though....(again, just a guess)

http://en.wikipedia.org/wiki/Ddr3
In short the system memory is DDR2 whereas the vram is DDR3.
DDR3 is faster than DDR2.

I didn't see anything on apple's site that suggests the video card can use system ram.

skubish said:

I didn't see anything on apple's site that suggests the video card can use system ram.

Click to expand...

Actually, I was going to ask about this also and forgot. There are many Centrino-based notebooks out there that have hybrid setups where there is a "dedicated" GPU and an "integrated" one, and it does allow for some kind of hybrid memory allocation system where there is VRAM and there is also the ability to use regular memory for the GPU. However, I don't think I've seen other documentation on the Macbook Pro having these features?

Video card RAM is sometimes GDDR or DDR3 so it could be different type of RAM from the machine - also it runs on a dedicated memory bus from the GPU to the VRAM, so it is not subject to the machines memory controller and contention from all the other RAM accesses of the computer. And depending on the video card, the speed and width of memory access can be higher than the computers main memory could achieve.

OTOH, a high end video card with 256 Mb or 512 Mb of dedicated high speed VRAM costs more than a Mac Mini...

mkrishnan said:

I'm not sure, but isn't there also a speed issue? If the GPU uses VRAM, it doesn't have to go out on the main memory bus and transfer information back and forth to main memory. I think it should be faster in talking to the dedicated memory to which it is hard-connected.

Click to expand...

That's the gist of it. I'm a computer engineer and I recently took a Computer Architecture class in which we developed our own processors. There's good example from that class that can help out a lot for this discussion. The example uses CPU's, but it's really the same principle.

Memory is what slows down our processors. If we could instantly get data to our processors whenever they fetched it (went to get it to use), we would have a lot more computing power. Basicly, a computer has to fetch the information it wants. To get the information, it goes through several levels of memory:

Hard Drive
Ram
Level 3 cache
Level 2 cache
Level 1 cache
register

In most chips, the register can be puled from without delay, but the other levels take time. Here's an example chart. This isn't exact, I just made up the figures. The numbers are the clock cycles it takes to fetch piece of information (called a word in computer terms)

registers - 0
L1 cache - 2
L2 cache - 4
L3 cache - 7
Ram - 20
HDD - 43

As you can tell, when you have higher amounts of "closer" memory it really helps the system. When a computer finds it needs something, it will have less to travel to get it.

With GPU memory, it works the same way. I don't know how much of a difference it is, but it will take longer to go to the Computer Ram than to use it's own on board Ram.