I am hoping someone out here can help me on this. I just built a new system, specs are: EVGA X58 SLI LE, 6 gig corsair Dominator RAM(1.65v 1600), EVGA 295 GTX, i7 920 (cooled with Corsair H50)& Ultra X4 1200W PSU.
After putting the system togther I ran it for 2 days without issue. However, immediately after I tried to OC the CPU all heck broke loose. After hitting save and boot in the CMOS the system shut down (like it should). But, when it tried to reboot the system started for about a second then shut down again. It went into this endless loop of comming on and going off. They only way to stop it was to shut off the toggle switch on the PSU. Interesting enough it I let the machine rest for about 30 min after which system would boot normally. (no BIOS changes ever took effect). So I started to think it had to be one of 4 problems. 1.Bad PSU 2.BAD mobo 3. BAD memory 4. BAD CPU. When I did get the system running I checked the BIOS under PC HEALTH the 12 rail was running at 12.55 v. On the upper side of the (+/-) 5% of allowed. At any rate I sent it back and got a new one that is now running at 12.3 v. Still alittle on the high side but should not cause this kind of issue. I called EVGA (which by the way are awesome to deal with) after much testing including using a barebones config and testing 1 stick of ram at a time determined that it may be a BAD BIOS chip. So they set me up with a RMA and sent me a new mobo. After getting the new board the problem persisted. I called EVGA agian and they said the mobo I returned checked out perfect. So at this point I ruled out the PSU and MOBO, next I tried to swapped out the Video card (i figured this was a long shot) with a NVIDIA 9800 GT, still no change in the system. Called EVGA again and concentrated on the memory and CPU as the prim suspects. In the BIOS the tech had me check the RAM voltage, which was fine at 1.66v. We tested each stick of RAM. Only one stick allowed the system to post, but only once. An important note to make here is they system will run fine when it is stone cold. Only after the system runs for 10min and you attempt a restart will the cycling take place. So somehow someway it has to be a thermal problem, right? The CPU at max load (stock speed 2.66) only got up to 27C.(gotta love that H50) So I'm left thinking at this point it has to be the memory controller on the CPU. I have a new CPU and Memory coming in the next few days. Does anybody else have any clue on what is going on here? I have been building systems for 20 years and never seen anything this before.
This seems really weird. It does look a lot like it could be a thermal issue, but then again maybe not because the first time this happened you said you let it sit for 30 minutes and it still wasn't booting. What would happen if you tried to run it with the side panel off and a house fan pointing towards the case? If it still doesn't work after that it might be safe to throw the heat issue out the window.
When you tried to OC did you over volt it at? Maybe you fried something in the chip. Have you tried to RMA that yet?
What I did not mention is that I also ran the system outside of the case on box it came in to be sure there was not a grounding issue with the case. In and outside of the case the problem persisted. I did not over volt the CPU. I was starting with some simple and easy OC'ing first. This is what I did:
Bios settings For i7 920 2.66 stock to 3.83 overclock
1. CPU Feature
a. Set QPI Frequency Selection Set to: 800
i. Intel speed step
ii. Turbo Mode Function
2. Frequency/Voltage Control
a. CPU Uncore Frequency Set to: 3205Mhz (16x)
3. CPU Clock Ratio
a. Set to: 19x
4. CPU Host Frequency
a. Set to: 200Mhz
5. Memory Feature
a. Memory Control setting : Enable
b. Memory frequency: 1067 Mhz/2:8 can bump it up to 1333 to 1600
6. Voltage control
a. QPI PLL Core: 1.300v
Hmmm.....this is a bit of a doozie . When it boots from the single stick of RAM do you use the same one each time, or cycle them? I'm thinking that possibly you have one stick that works and the other two do not. Other than that I can't really think of anything else to test, and I'm assuming you've already done that memory thing....Sorry to just be wasting your time
Yeah, I've done cycled thru all three sticks. Interestingly enough when the system was running I ran Memtest86+ on them. They checked out perfect. So that is why I'm leaning to a bad memory controller on the CPU. I'm telling ya I've been working on this thing for a month. I'm about ready to punt and start allover with a new config all together.
The system is up and running. After all that. I started over and begain with the PSU. The second PSU was not compatabile with the system. I was using an Ultra X4 1200. I swapped it out with the ANTEC 1200 and now the system is stable and running great. Next I'm going to OC and see where that leads me.