hi,i just built a new pc, after installing windows i tested it by using prime 95,the first time it failed after an hour,the second time after 6 hours,im trying to diagnose it and have come up with 2 culprits
1. using everest home the temp for the chipset is 57 degrees,this is even at idle,i have touched the heatsink on the chipset and yes nearly burnt the tip of my finger! i thought this may cause it to fail
2. the power supply im using is a 500w casecom it only has a 20 pin,whereas the mobo needs a 24 pin, in the manual it says it will run and yes it does but maybe there isnt enough power going to it?
mobo foxconn nf4k8ac
cpu amd 64 3800+
1 gb of known good ram
gigabyte 3d cooler
ok good to know that is what the extra 4 pins are for,no im not overclocking,im considering an aftermarket cooler for the chipset,possibly making one myself from a large heatsink from a mx440 and a small fan
The 24 pin connector is actually just the 20 pin connector with 4 extra pins added on the end. The original 20 pins were unchanged. The older 20 pin ATX power cable only has one 12 volt line. The new 24 pin connector added one line apiece for ground, 3.3, 5, and 12 volts.When you plug a 20 pin cable into a 24 pin connector you're not providing the extra current carrying capacity which may be needed by the motherboard. If your motherboard's current requirements are low enough then it will work properly with only a 20 pin power cabled plugged in. But if the motherboard draws enough current, then you can overheat the 20 pins you're using on the 24 pin connector. Many PCI-Express video cards, even though they have the 6 pin PCI-Express power cable, still draw a substantial portion of their 12 volt load through the PCI-Express slot. The extra 4 pins doubled the current capacity of the 12 volt rail so that one is easy to overload when only using a 20 pin main power cable.
Now, to bring your northbridge into the equation. The northbridge controls memory functions (for Intel chipsets it is where the memory controller is), a level 2 cache communicator and it is the communication center for the CPU and the PCI, PCI Express and AGP lanes. Since you are experiencing greater than normal heat temps for the chipset I wonder if the 20 pin connector/PSU that you are using ( on a 24 pin motherboard) has any influence on this. It is ok to use a 24 pin connector on a 20 pin ATX motherboard but the reverse might not necessarily produce the same results. It is something to check out.
Of course, something else to consider is if the heatsink it seated properly. Since the northbridge does control the graphics functions, having a poorly seated heatsink could adversly affect your PCI Express card.
i removed the chipset heatsink and fan,put the mx440 heatsink on with AS ceramique,i have a 6cm fan on top of it and now everest reports temps at a steady 45-46 degrees,the heatsink no longer feels too hot to touch.
but sadly i ran prime 95 again and it failed after 2 and a half hours. each time it fails it reports it as Fatal error rounding was expected less than 0.5 or something like that
im going to order a power supply anyway just in case,becuase i cant think of anything else that would cuase it to fail. but i will try anyway
1. everest reports the cpu temp changing wildly,it goes from 22 deg c to like 50 then back to 30 or so in the space of a few seconds,i assume its just bad temp detection,looking in the bios on boot up it reports it to be a steady 30 degrees so i think its safe to say that the cpu hsf is on correctly although it is still a doubt
2. i havent ran memtest yet, i know the ram is good becuase it was in my old pc for a year now i feel i should test it to make sure
For your case, whatever PSU you do order, do some reasearch. Especially with the issues you have been concerned with make certain that the PSU has PFC (Power Factor Correction). There are ALOT of PSUs in the 400 - 500 watt range that don't have this and it is often overlooked (much like amps) but, is something that is, actually, pretty important. PFC protects against things like voltage fluctuations and electrical irregularities. There are 3 types of PFC - Active, Passive and none - of which Active is the best.
well i googled my prime95 error, the error i was getting was always the same although happened at random times,sometimes not for 6 hours. looking on other forums including this i see that this error is associated with ram,either being run to fast or timings to tight etc, so i have loosened the timings a bit and am testing again.i may have a psu on the way anyway but im not sure
well my new build has been running prime95 for about 14 hours now,no errors,all i did was change the timings from 3-3-3-8 to 3-4-4-8
now it is stable
wow never knew that timings could affect a pc so much
still new psu on its way soon anyway