Intel Burn Test fails

Silver5656

Limp Gawd
Joined
Feb 10, 2004
Messages
238
so my stock i7 3770k fails the intel burn test sometimes. I am stock with a super mega cooler and dual ap-15's. Temps never get higher than 63 on the highest temp core and 58 on the lowest temp core. Any ideas?
 
Most likely RAM. IBT will detect memory errors as well as CPU instability, so it can be at times hard to differentiate the two if you don't know beforehand your memory is 100% perfect.
 
I ran memtest on each stick of ram individually for 3 hours each, and then all 4 (16gb) last night for 18 hours and no errors
 
although i noticed on the ram that i set it on the bios to 1.35v (9/9/9/24/1 @ 1866) but it shows at 1.5v on cpu-z and easytune
 
i just ran linx and the Gflops is 58ish.....seems low? and yes i have sp1 installed
 
Try setting the ram to 1.5v 1600 9-9-9-28 and see if it works. 1866 IS technically overclocking the memory controller.
 
so I should set my ram to stock? Is that just for the burn test? (ie if it passes then i can change the ram again?) I ask because stock the ram is 11/11/11/28 1600 @ 1.35v (samsung 30nm)
 
I ran memtest on each stick of ram individually for 3 hours each, and then all 4 (16gb) last night for 18 hours and no errors

memtest won't stress the ram like ibt ... memtest will only catch dud ram pretty much, but not slightly flaky memory timings.

Try 1333 9-9-9 2T @ 1.55 as a sanity test or something.

Lol no you can't test the ram at slow speed, and then crank it back up and assume that all is well. It should pass all the time at your desired settings.
 
for some reason I cant get the ram to register at 1.35v (which is its stock voltage at 1600 11/11/11/28) through the bios
 
i mean i can set the dram voltage to 1.35 but it will read on the bios and on cpuz and easytune as 1.5
 
i will have to check on the mobo (gigabyte z77 ud5h). I ran linx for 20 runs and got 89 GFlops using 11/11/11/28 1600 @ 1.5 and it was fine. I guess its time to fine tune
 
how many runs of linx and IBT are considered "stable"? Also, linx tests mostly cpu while IBT tests cpu/ram?
 
how many runs of linx and IBT are considered "stable"? Also, linx tests mostly cpu while IBT tests cpu/ram?

Entirely up to you, but I would say 5 is good for some quick testing, then follow it up with 20 or so. And LinX and IBT are just two front-ends for the same Linpack testing.
 
IBT uses LinX: essentially, IBT is LinX in a nicer user interface.

They do test the RAM. The amount of RAM they test is dependent on the size of the problem you set.

I use HCI Memtest to test my memory, it seems to catch errors quicker than Memtest86+.

The amount of runs I would do is dependent on the size of the runs. If you're running 16gb problem sizes, then 1-2 runs to test for initial stability, then 10-20 runs for final stability would be enough. If you're doing 1-2gb problem sizes, then at least 40. Basically, for quick tests, you want to run it for at least 5-10 minutes. For final stability, about 2-3 hours.
 
I ran memtest on each stick of ram individually for 3 hours each, and then all 4 (16gb) last night for 18 hours and no errors

Memtest is completely useless. A stick would have to ave very serious errors for Memtest to throw an error
 
Back
Top