desolator4u

Distinguished
Mar 3, 2011
84
0
18,640
Sup ppl?
I was doing some research, and I found that the GTX 560 Ti has a "Bus width" of 256 Bit and a Bandwidth of 128.27, while the PCI-e 1.0 x16 Slot has a Data Rate of 250 MB/s.
Is the GPU's Bandwith measured in MB/s just like Data Rate of the PCI-e Slot? And if so, does that mean it uses less than the 250 MB/s the PCI-e 1.0 x16 Slot is rated for? Or is the Data Rate of the PCI-e Slot is what is related to the Bus Width of the GPU?

Hope you guys can follow all that...
~Des
 
The GeForce GTX 560 Ti has a Memory Bandwidth of 128 GB/sec. It is the communication bandwidth between the GPU (GF114 Graphics Processing Unit in the GeForce GTX 560 Ti's case) and the graphics memory (1024 MB of GDDR5 in the GeForce GTX 560 Ti's case). This GPU to GDDR communication is done on a separate memory bus on the graphics card not on the PCI Express bus.
 
Even better http://en.wikipedia.org/wiki/PCI_Express

Are you trying to find the bandwidth to find out what is the best card you can get without having a bottleneck? This is not the way to figure that out, the cpu is the main deciding factor on which gpu to get. Yes the pcie bandwidth will hurt performance but not nearly as much if the cpu cant keep up with the gpu. Not to mention the specs of the rest of the pc. The important factor is to have a balanced system.
 

desolator4u

Distinguished
Mar 3, 2011
84
0
18,640


In a sense, yeah. I've already decided on the GTX 560 Ti, but I still wanna know how do you figure out what the speed of the Video Card is, when it is transferring data to and from the MoBo via the PCI Slot? How is it rated? Based on that link and what 4745454b said, it would not be the Memory Bandwidth.
 
I was referring to pcie bandwidth, every part has a bandwidth essentially, I shouldve been more specific. What is the rest of the pc specs? Something just sounds wrong(as in old or low spec) if the mobo only has a pcie 1.0 x16.

Edit: I found a way to measure pcie bandwidth. http://www.evga.com/forums/tm.aspx?m=253215&mpage=1

Edit:Edit: I was just rereading the original post. Just to clarify bus width is like how many lanes a highway has, bandwidth is how fast the cars are moving. Bandwidth=data transfer rate, and you were referring to data rate and bandwidth being different things, they are the same.
 

4745454b

Titan
Moderator
The memory bandwidth of the card needs to be higher then the slot. The CPU will send needed info to the card, but the card has a ton of extra stuff do that the CPU doesn't care about like shading the pixels and apply AA. This requires memory bandwidth.

It sounds like your worried about putting to much GPU into a computer. The (overused) term is "bottlenecking". There isn't really a formula that can be use to figure this out. You basically need to watch a bunch of benchmarks to see where the FPS stop going up with additional GPU power for X resolution.
 

desolator4u

Distinguished
Mar 3, 2011
84
0
18,640
k1114, thanks, but that link involves testing and I don't have the Card as yet. I should ask the same thing there though and see what they say, i've gotten good responses from the EVGA forums before.

Currently I have:
ASUS Striker Extreme 680i MoBo
Q6600 2.4 GHz C2Q. Got the Zalman CNPS 9700 HSF, so gonna OC to at least 3 GHz
2 x 2 GB Corsair 800 MHz Memory
XFX Black Edition 750W PSU

Guess i'll really have to wait til I get the GPU and test it out, but it should work fine with those specs, right? May not be able to utilize the GPU 100% but as long as it's not a waste like 70% of it's full potential or so.

~Des


 


Just to clarify your car analogy: Bandwidth is the space (or lanes) in which the car is traveling (think two-lane country road versus controlled-access super-highway with 10 lanes). Bandwidth=volume.

The data rate (or speed) is different.

Example: AMD HT3 has a single-direction bandwidth of 22 GB/s and at stock a PhenomII processor speeds at 9-10 GB/s (13-14 GB/s when OC'ed).