Originally Posted by Drumphil
er, last time I checked the nforce was the ONLY chipset that had a functional AGP/PCI lock, so for most people the FSB speed does effect the RAM speed.. (and agp and pci). You can decide wether or not to lock the AGP/PCI on the nforce boards I think..
And please stop using double negatives.. Im still not exactly sure what you are trying to say.
anyways, your test will test how FSB effects benchmark performance, not GPU performance. And increasing FSB will increase the IPC for the CPU, but the IPC rate for the GPU will remain the same, because the GPU has its own ram.
Ummm since we're talking about strictly GPU related tests, Would that not effect Memory bandwith? Since FSb itself has no affect on GPU cept MEmory Bandwith, You Could theoretically change memory Speeds to Run async to prove or disprove what you are saying.
Since you are saying that FSB has nothing to do with Improved GPU performance, But only memory bandwith, That could be easily tested, With just altering memory Speeds, And Running them Aynsc from the FSB, If its helping, Youd be able to see it, Verses a FSB alteration, Specially in 2003 with Game Test 2, 3 and 4, which are very much shader limited.
Your Point is that Memory Speed is affecting GPU performance more than FSB, This is easy to test, You can run your Memory Out of sync From FSB at higher or Lower frequency than the FSB itself, since They are "GPU" Limited, As you say, The CPU Latency wont make a difference and the GPU will benefit from the increased memory bandwith on GPU limited situations,
And 3dmark2003 Game test 2 3 and 4 are GPU limited, Its a pretty simple test. because FSB always improved even my shader Limited Games.
P.S. Most Nforce 2 boards auto Lock PCI, Very few allow you to unlock PCI bus, And AGP is pretty much controlled independently.