Kitz Forum
Computers & Hardware => PC Hardware => Topic started by: Chrysalis on April 06, 2014, 04:13:35 PM
-
So as below.
been getting weird results.
Tests I did are.
all 4 dimms memory consistency errors, removed all overclocking and set dimms back to low speed, still errors.
2 dimms, pass
other 2 dimms in same 2 dimm slots as first 2 dimms, pass
repeat test in other 2 dimm slots, fail
at this point thought slam dunk board issue.
one dimm in first of bad 2 dimm slots, pass
same dimm in 2nd supposedbly bad dimm slot, pass
ok whats going on
2 dimms in supposedbly bad dimm slots (diff 2 dimms to the pair used with failed test), pass
inserted other 2 dimms in what thought good dimm slots so now 4 total again, fail
removed one dimm so 3 dimms in place, pass.
reinstated cpu overclock and put ram back to XMP speed, pass
so I have now out of the case what may be a bad dimm.
about to do a new test now swapping out one of the 3 dimms with what I think is the bad one. if that fails I will then put the good one back in and bad out but test using the unused dimm slot, if that passes I will assume bad dimm. - both these tests failed so this is a mess :(
So basically no consistent results which give clear cause of problem.
-
Would just like to suggest the usage of Memtest86+ (http://www.memtest.org/). :)
-
I'll second that. There's nothing to beat it for thoroughness.
-
I changed the ram timings to lower values, silly thing to do but still, i can only use 1 of the 2 ram slots now as when i try to use both the computer wont boot, with or without overclock, both ram sticks are still perfect, but can only use 1 of the sticks. ??? Must have damaged the mb
-
burakkucat said: Would just like to suggest the usage of Memtest86+
I second that extremely good at weeding out randomly failing memory.
Loading said: Must have damaged the mb
You normally have to go some to do that.
I assume you had the AC lead out of the PSU whilst removing and adding the memory? At this point it may also be worth doing a factory default on the bios.
Whilst not common these days are the contacts in the memory slot the same color as the pads on the memory? IF they are'nt electrolysis can occur causing corrosion of the contact resulting in a bad joint.
Have you checked the Mobo for blown electrolytic caps?
-
I assume you had the AC lead out of the PSU whilst removing and adding the memory?
I always have the power cable plugged in, but the socket and the switch on the power supply turned off, that way you can discharge static safely by touching the metal of the case.
-
I assume you had the AC lead out of the PSU whilst removing and adding the memory?
I always have the power cable plugged in, but the socket and the switch on the power supply turned off, that way you can discharge static safely by touching the metal of the case.
Thats the best way Ronski.
Run memtest86+ on each dimm you would need to let it to run for 5 - 6 passes on each dimm to catch the faulty one & yes I know it takes and age for each pass, once all dimms have been checked and no errors then it points the finger at the motherboard and don't forget the Ram modules also needs to be close to the same clock frequency for Dual Channel to work.
-
It may also be worthwhile using the Windows Memory test (http://www.sevenforums.com/tutorials/715-memory-diagnostics-tool.html) in addition to Memtest 86+, I have heard of rams passing memtest, but failing the WIndows test. There are also various types of test within Memtest, some of which seem to catch certain types of error more than the standard test.
-
Unfortenatly I cannot get consistent results to point to one component.
eg. when testing one dimm at a time they all fine.
When testing one dimm slot at a time they all fine.
So its using a certian combination, loading the slots in certian ways is triggering the errors.
After some research it has become apparent haswell is very finnicky with ram to the point that corsair have made specific dimms for haswell stating "optimised/designed for haswell".
I ended up with buying 2 8 gig dimms and so far on my testing they ok. Will likely sell the 4 4gig dimms to offset the cost.
Also memtest86+ didnt pick up errors on many of the combinations, hci memtest is much better at finding errors from my experience.
-
Faulty Ram modules will sometimes pass the test in memtest86+ on the first run thats why 5 - 6 passes is required, this has happened to me many of times with memtest86+ it depends on the severness of the fault on the dimm, the faults seem to show up better when the dimm is stressed via higher temperatures.
-
the dual dimm setup I am using now seems fine.
the 4 dimms are fine in the old rig as well as when tested seperatly, they are not faulty. I even spent 2-3 hours getting the old rig in an operational state to specifically test them.
personally I dont rate memtest86+ at all, last night it passed with 17 passes when left for many hours. hci memtest then failed within 10 minutes :)
hci is superior as its testing with the OS up and running, its a more realistic mode which also generates more stress.
-
Using all four slots does put more stress on the motherboard, I remember when I was over clocking my current rig people said it was easier to get better results with only two sticks of memory.
Out of curiosity what's the specs of your new rig?
-
asus z87-plus 1504 bios
haswell i5 4670k currently I set turbo to 4.3ghz all 4 cores using adaptive voltage offset of 0.05v - cooled by arctic freezer i30 cooler
16gig ram via 2 8 gig vengeance dimms 1600CL9
evga 760GTX SC ACX
fractal R4 case.
onboard intel nic
carried over following from old rig.
coolermaster silent pro M 600W psu
asus xonar D2X soundcard
samsung 830 128 gig ssd
2x WD red 3TB hdd
WD green 2tb hdd - disabled auto head parking to preserve longevity
WD 640gig black hdd
samsung 750gig spinpoint f1 hdd
pioneer dvr-216d dvd burner
I agree on the overclocking and 4 dimms but sadly this wouldnt even work right at stock clocks/voltages. What stopped me returning the board is the plain hassle involved in removing all components and rebuilding again, this cooler is a job and a half to fit.
-
Nice setup, should certainty be quick, hope you've got your ram problem sorted now.
-
yeah thats fine now, sorting small isues on audio card.
-
Nice setup, should certainty be quick, hope you've got your ram problem sorted now.
It seems to be TOO quick for the HG12 modem.
It now looks like I'll have to slow down my programs just to keep Chrys's fantastic setup fed with modem stats data :lol: :lol:
-
I been told by the overclocking community I have a wicked cpu, I have barely touched the voltage and have 4.3 clocks. So might be even faster soon if I feel like pushing it more :)
most people are going 1.2v or higher to get 4.3-4.4 I am on 1.13v.
-
Go on you know you want to ;D
I built a 6 core Intel system last year for a freind, I had that running at 4.5Ghz, but the voltage was bumped up quite a bit.
-
Its all overkill Chry unless your doing some 3D ray tracing for a Movie :) I get by on a Dual Core system with 4GB DDR3 dual channel :o it's all down to my optimization.
-
You be surprised, firefox frequently saturates my cpu (it can only utilise one core).
Plus I do use various apps, heavy load etc. that I need the grunt for.
-
most people are going 1.2v or higher
I haven't really tried over clocking as in the past when doing so was just starting out whilst you could get the frequency to report it was higher usage never backed the figures up!
Still increasing the voltage doesn't make sense in my mind. The whole point of using lower voltage was to reduce the time it takes to go from 0 to 1?
-
I will never go to 1.2v, I am not after a extreme overclock, I want sane temperatures. IF I only managed to get 4.0 with this overclock then its probably where I stay. I just got lucky, I may try 4.4 (but only at the same voltage).
Also adaptive voltage keeps the same voltage when the cpu is not in turbo mode, so eg. if the pc is idle for 10 hours, then those 10 hours its the same as stock clocks/voltages, it will idle at 800mhz and use about 0.1v. Adaptive voltage is a new thing with haswell.
-
Adaptive voltage is a new thing with haswell.
No it isn't, my Ivy Bridge has adaptive voltage based on core speed (actually based on multi). At 1.6ghz it runs at 0.82v, at 4.4ghz it runs at 1.25v
-
Adaptive voltage is a new thing with haswell.
No it isn't, my Ivy Bridge has adaptive voltage based on core speed (actually based on multi). At 1.6ghz it runs at 0.82v, at 4.4ghz it runs at 1.25v
adaptive voltage is not voltage been reduced at idle
its voltage been changed from full stock to overclocked speeds.
is 1.6ghz your stock or idle speed?
-
1.6ghz is idle, 4.4ghz is full (overclocked) speed, stock speed is 3.5ghz. I have speedstep enabled.
-
ok so compare your voltage from 3.5 to 4.4, if that changes then its adaptive.
-
not seen it at exactly 3.5 yet, but at 3.0 it was 1.00V
-
anything below 3.5 is fluctuations due to speedstep and c1e.
-
After refurbishing the old pc to test win10/8.1 I discovered that pc uses the ram at a lower speed. The ram is supposed to be able to run at 1600 but it was only running at 1333 in that pc, of course it still errored out at 1333 in this pc, but I can confirm it tests ok and no instability in the other pc, so it will remain somewhat a mystery and I am putting it down to a compatibility issue for now.
-
I had something like that when I added the nominal full amount of RAM to a motherboard .
At the frequencies concerned the clock will be an overtone crystal i.e if it is a 9th O/T 100 MHz crystal say, it is also resonant at 7/9 of 100 MHz, 5th at 5/9, etc etc and with insufficient load isolation can jump to a lower O/T.
It could be poor design of either motherboard or RAM. I ended up settling for 3/4 of the maximum to get nominal speed.