

QX8650 vs Q6600 & XFx nForce 680i SLI LT
#1
Posted 10 July 2012 - 07:49 PM
I obtained a QX8650 (Intel Quad Extreme 3.0 GHz) used, popped it in, and found through testing that my system is very unstable under load if the 3rd and 4th cores are enabled. It doesn't appear to be a heat issue -- the first 2 cores are stable while running up to 70C (well under the max temp most OCers discuss for this processor: 75C), while the 3rd and 4th are having problems at ~60C.
By "very unstable" I mean the system reboots within a few seconds (gaming) or minutes (calculating pi to 4M digits) of putting a heavy load on either the 3rd or 4th core (playing a game, calculating pi). The system doesn't give me any problems with just 1 core, or having the first 2 cores enabled. I can run for hours, calculating pi or playing a game.
I upgraded the BIOS, but they only went up to P08 for this board (despite the non-LT boards having BIOS revisions much more recent, and up to P33). Didn't help.
My 800W PSU should be putting out plenty, esp since I'm only running one video card (previously was running 2 x 9600; now running 1 x MDT GTX 570).
After narrowing down the problem to the 3rd and 4th cores, I spent hours online researching. I saw *one* post regarding this exact symptom, where the guy hadn't narrowed down the issue already like I had, and there was one reply that said, seeming out of the blue and with no further explanation: "you may have to disable the 3rd and 4th cores to stabilize the system."
I saw a couple more posts about this processor and the EVGA version of this board, that suggested maybe this particular processor may be hitting its "FSB Hole" right at the spec settings on the 680i chipset.
Does anyone know of a known incompatibility, or have any specific experience, information or recommendations for this hardware?
I'm going to try underclocking it first (1333 -> 1200), to see if it stabilizes, per one recommendation I found online. If it does, that might be a good indication I'm hitting the FSB Hole, and that OC'ing will work. I may not currently be able to cool it enough to OC, as this is a very hot processor.
If I can't get a stable system, I'm thinking of going back to the Q6600, and OC'ing it. I hear good things about OC'ing it; most people can achieve 3.0 GHz with inexpensive air cooling, and 3.4 (or more) with expensive air cooling or liquid cooling. Many OCers seem to think the Q6600 is better than the QX8650, in fact.
#2
Posted 10 July 2012 - 08:38 PM
but do you get interrupted clock cycle blue-screen when it reboots or what?
stock instability is the sign of a dieing CPU, you can always try to put a tiny bit more voltage in to the CPU see if that helps.
i have never come across a FSB hole but i have heard of people who have, down is not the way to go under-clocking would solve Manny problems so you wont know if that's the case with you cpu so you should go up, just remember to set your ram dividers.
but im just guessing without your system info
almost forgot an FSB hole if you have one is a motherboard issue not a cpu issue so even if you got another CPU it would have the same issue because its the Bord that doesn't like the FSB settings
Edited by Scilya, 10 July 2012 - 08:41 PM.
#3
Posted 11 July 2012 - 03:49 AM
If you can't get the Q6850 stable, go back to the Q6600. OC'ed it as high as you can while keeping voltage below 1.4 and temps below 70'c under full load. For everyday use, keep your temps under 60'c.
#4
Posted 11 July 2012 - 09:55 AM
The problem is the FSB holes are VERY random per board and CPU combo, and often the same board and CPU will work just fine above the hole or below it, just not at it. No amount of voltage tweaking is going to fix this either.
Finally the 680 boards had problems with newer socket 775 cpus. They were pushed out the door to work with the original core2's and were picky as hell even then with a core2 x6800 and the qx6700 q6600, many of them didn't play nice with the 6850 and quad versions of that line, and outside of the striker extreme didn't really work at all with the 45nm quads at all (problems you're reporting were typical) and even then couldn't be OC'd. The design was actually a board flaw which is why nvidia pushed out the 780 boards.
Roll back your FSB from 1333 to 1066 and then raise your clock rate via the unlocked multiplier, just be aware that the boards aren't designed to properly power the 1333 FSB procs and aren't going to work with the 45nm procs. For that you'll need a 7 series board.
#5
Posted 11 July 2012 - 11:21 AM
For now I'm back to the Q6600. If I get bored some day, I may see if there are software programs that will allow me to manually adjust the BIOS settings. However, I think I'll just be building a new system some day.
And now I know: stay the heck away from XFX -- their support site is horrible!
#6
Posted 11 July 2012 - 11:29 AM
Haydon Jurai, on 11 July 2012 - 11:21 AM, said:
For now I'm back to the Q6600. If I get bored some day, I may see if there are software programs that will allow me to manually adjust the BIOS settings. However, I think I'll just be building a new system some day.
And now I know: stay the heck away from XFX -- their support site is horrible!
Keep in mind that many 680 boards could not OC newer core2/quad variants, it's just not possible.
If the q6600 is in what options are you missing?
#7
Posted 11 July 2012 - 11:39 AM
silentD11, on 11 July 2012 - 11:29 AM, said:
If the q6600 is in what options are you missing?
All it gives me is the ability to change the CPU multiplier, some of the voltages, and the RAM settings.
I do believe the LT ("light") board was marketed without OC settings, so if I had them, it was probably that I'd somehow obtained a reference BIOS previously instead of an XFX BIOS.
(And Windows wouldn't boot with altered voltages on the QX6850, btw; I was hoping it might have just needed a bit more power after seeing some suggestions along those lines in various forums for the processor and chipset.)
#8
Posted 11 July 2012 - 11:53 AM
Haydon Jurai, on 11 July 2012 - 11:39 AM, said:
All it gives me is the ability to change the CPU multiplier, some of the voltages, and the RAM settings.
I do believe the LT ("light") board was marketed without OC settings, so if I had them, it was probably that I'd somehow obtained a reference BIOS previously instead of an XFX BIOS.
(And Windows wouldn't boot with altered voltages on the QX6850, btw; I was hoping it might have just needed a bit more power after seeing some suggestions along those lines in various forums for the processor and chipset.)
No, the LT boards can OC same as the full 680i boards. Granted they didn't have as many options, especially when it came to memory, but you can OC just fine. I take it you've never OC'd something before?
The CPU multiplier is how you'd OC the QX, but for the Q6600 you'd OC it with the RAM settings. If you're using DDR2 800 than set the speed, timings, and voltage of the memory manually under the RAM timings and then make sure the CPU's FSB is locked to memory and not de-linked or asyemeterically timed, it should OC perfectly fine right there and DDR2 800 speeds are fairly safe for FSB holes.
What's type of memory do you have while you're at it?
#9
Posted 11 July 2012 - 11:59 AM
#10
Posted 11 July 2012 - 12:04 PM
Clay Pigeon, on 11 July 2012 - 11:59 AM, said:
Quasi correct, nforce for intel was ALWAYS crap, including the 500 series. For AMD they work just fine. The problems were not just the bumps, that's another issue. The SATA controller was borked, the power design only worked properly for generation 1 core2 chips, and the memory controller was full of holes.
Nforce 680a, aka AMD, works just fine. One of the reasons I moved back over to AMD is after going from core2 x6800 striker extreme, to massive problems with Quads on the Striker Extreme, to striker II forumla, to striker II extreme, they never fixed it. Crosshair II forumla with 940 BE, rock solid from the start.
#11
Posted 11 July 2012 - 12:36 PM
Barbaric Soul, on 11 July 2012 - 03:49 AM, said:
If you can't get the Q6850 stable, go back to the Q6600. OC'ed it as high as you can while keeping voltage below 1.4 and temps below 70'c under full load. For everyday use, keep your temps under 60'c.
I went from the x6700 to a QX6800 i paid all of $300 for my QX6800
What Revision is the CPU? there are 3 revs of that cip that i know of... one Rev sucks at overclocking or must be kept cool or it crashes
#12
Posted 11 July 2012 - 02:34 PM
#13
Posted 12 July 2012 - 07:07 PM

I have never OC'd, but I did a fair bit of research on novice OC'ing over the course of several days. (I won't be toying with any advanced timings adjustments or pencil-mark hacks any time soon.)
Anyway, in the BIOS I cannot unlink FSB settings, nor the northbridge to the southbridge, and most other options are missing that I see described in the forums for the 680 chipsets. (Like I said, I thought they were there, but they certainly aren't now.)
I did find I can OC from a software app, but only via the dynamic BIOS settings. In other words, the settings are not retained after a reboot. Currently using nVidia's Performance tab in their Control Panel.
So far I tested the Q6600 and 2.5 and 2.6 with success -- cool and stable. I'd initially tried to jump straight up to 3 GHz (lots of documentation on the recommended settings), but it locked up immediately, so I may have hit my FSB hole.
On the positive side of the OC settings not being saved in the BIOS is that I can test, and if my system locks up or is overheating, all I have to do is reboot it; no need to move the CMOS jumper to clear the settings.
Once I've dialed in my idea settings, I'm sure I can locate a software OC utility that'll let me load the settings with a switch, or from the command line, so I can load them on startup, or on demand via batch file. If I have to pay a few bucks for it, it might be worth it, once I know it will be useful.
#14
Posted 13 July 2012 - 06:48 AM
Don't write it off as an FSB hole just yet, 3.0ghz with a 6600, provided it's a g0 is a conservative OC on nvidia boards and fairly safe. You might need to raise the voltage though. Go up till it's not stable and then raise the voltage a bit, see if that locks it in.
#15
Posted 18 July 2012 - 11:31 PM
1 user(s) are reading this topic
0 members, 1 guests, 0 anonymous users