• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Graphic card bandwidth sufficient?

bharatwaja

Senior member
I dont understand one thing...

PCI express 2.0 is listed to have a bandwidth of 16GB/s (correct me if I'm wrong)...

Where as a 8800GT has a bandwidth throughput of 57.2 GB/s

how is it even possible?
I mean, there should be some serious slowing down/lag right??

What is the concept of all this....
 
The bandwidth of pci express 2.0 is irrelevant to a card's performance, because it's only involved in CPU->GPU communication. The bandwidth that really matters is the card's video memory bandwidth, which is completely separate from pci-e, and that's where you get 57.2 GB/s.
 
thank you... I was extremely confused...

But still, if this is the case, how much of the PCI e 2.0 bandwidth will a 8800GT card use?

and... could you explain a lil more on video memory bandwidth? is it internal to the card without any relation to the PCIe 2.0 Bandwidth?
 
Ok....PCI express is the means of transport for all the information to the card (textures, geometry, shaders....ect ect) which then sits in the video memory. This memory is also home to the frame buffer and is where all the work is actually done. The bandwidth of the video memory needs to very high so that it can keep the GPU fed with information. So, very simply, pci express gets the information and the bandwidth listed for the card is how fast the card uses the information once it has it
 
Correct, video memory has it's own bus connecting it to the gpu, and most of the workload utilizes that bus. In general, video games try to use the local video memorey as much as possible, and avoid hitting the system memory. That's why all the components of a scene, like textures, geometry, and shaders get stored in the video memory. The pci-e bus is only used in uploading that data to the video card the first time the data gets requested, and in sending driver commands to the gpu. The bandwidth offered by pci-e 1.0 is plenty for that purpose.
 
very informative... thank you....

Still one question remains... jus for the sake of information.... How much bandwidth of PCIe 2.0 does a 8800GT use?? is there a place to find out this for the various cards available on the market???

But from what you have told now, that questions seems like totally dependant on the game/program that uses the card right? i/e/ it could vary for different games, cause they differ in graphic requirement and usage... right?
 
It doesn't matter how much it uses - not even an 8800 Ultra or a 2900 XT could max out a PCI-e [1.1] x16 link... In fact, I'm not even sure if they could max out PCI-e x8 - the 7900 cards couldn't even max out an AGP x8 link... and since the 3850/70 are rumored to be coming out in AGP, it's not likely that they use much more bandwidth than an AGP x8 link. PCI-e 2.0 is just a gimmick to get you to buy new motherboards/cards. By the time it's even close to being maxed, we'll be on some new slot with a zillion times the bandwidth.
 
Back
Top