Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
780i
Pros:
·SLI (if you care)
·Works with Penryn out of the box
Cons:
·Retarded Bridge chip design for check-mark PCIe 2.0 Support.
- PCIe 2.0 Bridge chip connected to the PCIe 1.0 bus.... see a problem here?
·Southbridge contains PCIe 16x 1.0 slot... Note that the SB link isn't really designed for that + all other relevant IO.
·Really is just a 680i with a higher price tag and a whole bunch of things thrown onto it. See the SNL iRaq skit.
·Poor Latency due to bridge chips and shoehorned SB PCIe lines FTW
It really makes me wonder why they didn't just put an HT link on the NB (a la Apple/IBM's G5 northbridge) and connect it to a "southbridge" that incorporated the PCIe 2.0 lanes and IO hardware. They would have had to engineer two chips instead of 4 (counting the 790) and one chip would have just been the 680i + HT link circuitry and possibly minus it's own PCIe circuitry. The "southbridge" could have ended up being their next AMD chipset as well.
Is that true? I did not know that. New (bad) info for me. Time for another evaluation of choices2) Intel chipsets don't underclock memory. So if I'm running a 500 FSB then slowest I can run the memory is 1:1 and that is 1000. I gota have pretty good memory to overclock too the high FSB levels. But a Nvidia chipset unlinks so that I can "affectively" run memory and CPU asyn (and there is really minimal performance hit for doing so).
i don t have any problem to run my system over 1:1 memory devider with my ix38 quad gt ... take a look !!!
....
500 fsb at 1:1 i think the cpu will be the hard thing to push to 500 fsb not the memory !!!
Understand Evilsizer but how do you tell if it is CPU or motherboard hitting the wall short of playing musical CPUs? And he appears to be running a dual, not quad, but assume it might happen there too?
I guess, I will be choosing the x38. The 780i is the upgraded version of 680i, just a few extra things huh?
And why is the CPU hard to push above 500? Just drop the multiplier. I would think it is primarily a matter of the motherboard being able to do 500. Please explain, I'm always willing to learn.
By the way, 3916 very nice.
OK, I think most of the common points have been made and I agree with them.
BUT a few technical things I don't understand about the march to X38 around here (and note that I am running a X38 because the 680i doesn't run the Q6600 well).
1) if I have a CPU with a lower multiplier then I understand I need a higher FSB for a reasonable overclock but with something like a Q6600 and its 9 cpu multiplier, a 400 FSB does fine. I don't need a 500 FSB overclock and in fact ain't sure what I'd do with it if I could do it. If I run two tests with CPU at same speed but a different FSBs the performance doesn't improve for the higher FSB (the CPU speed is the controlling factor).
2) Intel chipsets don't underclock memory. So if I'm running a 500 FSB then slowest I can run the memory is 1:1 and that is 1000. I gota have pretty good memory to overclock too the high FSB levels. But a Nvidia chipset unlinks so that I can "affectively" run memory and CPU asyn (and there is really minimal performance hit for doing so).
The thing I like about the Nvidia chipsets has been the ability to unlink CPU and memory performance and tweak both to the max. I've tested this capability heavily and don't see any performance penality in taking that route. Personally I don't care about SLI and most likely never run SLI.
If I've missed something please help me understand. I am painfully aware that I don't know everything. BUT please remember I'm not interested in what your uncle's cousin's brother's best friend says they can do. Show me the real test results.
Merry Christmas to ALL!!!
2) Intel chipsets don't underclock memory. So if I'm running a 500 FSB then slowest I can run the memory is 1:1 and that is 1000. I gota have pretty good memory to overclock too the high FSB levels. But a Nvidia chipset unlinks so that I can "affectively" run memory and CPU asyn (and there is really minimal performance hit for doing so).
I gotta call shenanigans here. I have a G33 based board, and it offers a 1.6 memory multiplier. I can run the FSB at 500, and the memory at 1.6x=800mhz.