Hardware Analysis Community Forums http://www.hardwareanalysis.com/content/topic/36358/ 2009-01-07T23:29:19-05:00 http://www.hardwareanalysis.com/content/topic/36358/ http://media.hardwareanalysis.com/halogo.gif text/html 2004-12-28T22:51:31-05:00 Sean B http://www.hardwareanalysis.com/content/topic/36358/#219535 It would rape bandwidth, motherboard realestate, and latency..... so it's out of the question.<br /> <br /> I did come up with a concept I'm going to mail to Intel that would put them on top in the gaming hardware industry. Unfortunately, they don't accept proposals via e-mail for legality.... so it might take a while. text/html 2004-12-28T21:44:15-05:00 James Marlin http://www.hardwareanalysis.com/content/topic/36358/#219507 Are you suggesting that video cards start using a system similar to how L2 cache' used to be on the main board and not the CPU? With the very early ones being slot, instead of BG?<br /> Basically that's how it would work, because the new concept your referring to (I think), is where manufactures are going to reduce the on board video card memory to a cache' and have the card use all the extra PCIe bandwidth to fetch information from system memory. This will reduce video card costs quite a lot.<br /> <br /> Ok just off the top of my head<br /> Will this sort of concept reduce the on board video bandwidth any? I would assume it would, since at the very least there would be over head using a slot based system as apposed to BG.<br /> It's been proved with the move to on die L2 that it's the speed and bandwidth as much as the amount of memory where cache' performance is concerned, so it might not help any, and could hurt.<br /> <br /> Plus what about weight and cooling?<br /> <br /> Still it's an interesting idea, over all. text/html 2004-12-28T21:13:18-05:00 Sean B http://www.hardwareanalysis.com/content/topic/36358/#219500 Slightly lower latency, but then you have to weigh in the realestate on the card. text/html 2004-12-28T21:07:30-05:00 JOhn Bobbitt http://www.hardwareanalysis.com/content/topic/36358/#219495 I agree with you, but my initial post was in regards to the new Radeons using the PCI-X bus to leverage system RAM as video memory. In that case, it would still be faster to have DIMMs local to the card than on the system board, going over the bus, no? text/html 2004-12-24T00:26:24-05:00 Sean B http://www.hardwareanalysis.com/content/topic/36358/#217963 The embedded memory is directly pipelined ball grid array ram.<br /> <br /> If you were to throw it on a DIMM, it would defeat the purpose.<br /> <br /> <br /> Besides, software doesn't even take advantage of the memory graphics cards have now, so why do you think 1GB would be any better? text/html 2004-12-23T05:25:09-05:00 JOhn Bobbitt http://www.hardwareanalysis.com/content/topic/36358/#0 Here's my question - why dont we have the option anymore to put our own memory on video cards nowadays. Old video cards usedta let us do that, and now they are using system memory for video cards across the PCIX bus, so why not give us slots on our video cards where we can drop our own 256, 512, or 1GB DIMMS?