Printable View
there are other fets that look roasted if not toast as well on the edge of the last pic...Quote:
Originally Posted by INFRNL
seems that they got really hot... as in 125C :S
if you would have been there while it happened, holding your ear over the board, you might have heard a quiet "hot hot hot HOT HOT HOTTTT!!!" before it blew up ^^
have you guys seen this?
http://www.tomshardware.com/reviews/...d,2436-15.html
asrock msi and ecs all have problems with their pwms burning on 1156 boards, and at measily vcore of 1.3-1.4v...
those are value boards, sure... but i havent heard of this happening since amds 140W issues...
and yes, of course those boards dont have beefy pwms... but they should shut down and not burn...
it looks like thre is definately something wrong with the 1156 power design guide...
i highly doubt all these mb makers are breaking the spec or guidelines from intel...
Hmm, and what will happen when they implement over current protection in bios as they say? The board will just shut off when I use 1.36+V and that's it? Basically I will save the board from burning but I will be heavily limited
That problem is down to no heatsinks on the FET's for one thing. Some of those PWM circuits are limited to around 120-140amps tops. When you then use LLC and ramp the processor frequency with VID it's not good at all (without a heatsink and active cooling of those FET's near their peak current handling its a disaster waiting to happen). Those boards are best used for sub 3.6Ghz overclocks in the long run. The vendors don't add too much to boards in this price bracket to encourage product segregation and to justify the existence of higher end boards.
That's not quite truth...Asrock P55-PRO has heatsink on mofets and Msi has to even if is small...Quote:
That problem is down to no heatsinks on the FET's for one thing.
http://media.bestofmicro.com/P55-Mai...-225678-13.jpg
And even so for example Asrock VRM is in configuration 8+2 phases and 2 mofets per each phase.....Why sould't be these enough? Or 1.3v> the power consumption of 750/860 increase exponentialy? ....
And IIRC, LLC is not part of the intel spec which is why I never use it.Quote:
Originally Posted by raju
Depends on how well the FET's are contacting the heatsink, and if the sink is actively being cooled (thermal runaway issues can happen if contact is poor and also if LLC is causing oscillation). Also need to factor the current handling per FET. Upto 4GHz most CPU's will not pull more than 140w to VCC (some may pull a bit higher but not all).Quote:
Originally Posted by xdan
Depends on how that LLC is designed. Some VRM solutions can support a tighter droop specification than 0.05v. So long as peak overshoot does not breach the applied VID for longer than Intel spec. On the cheapo boards LLC implementations are generally outside spec at higher current draw, but people call the board crap if the vendor does not include some kind of droop control.Quote:
Originally Posted by chew*
Err should I be concerned about this? Just reading this thread is making me pretty nervous - I have a UD3R by the way. I've been rather busy, so I'm currently at 3.4Ghz (stock vcore) on my i5. What's the deal - should i be restricting vcore/VTT to prevent socket burn?
did you meassure this?Quote:
Upto 4GHz most CPU's will not pull more than 140w to VCC (some may pull a bit higher but not all).
i highly doubt that... my 920 D0 pulls more than 130W at stock speed with all 8 threads loaded... im 100% sure but unfortunately dont have a watt meter...
A fully loaded Xeon W3540 running Linx at 4.45GHz, EPS draw is around 21.5 amps. Around 250W of total draw on EPS 12V - 15~20% switching loss, 40w to VTT, around 160w to VCC. Do you think a 870 @ 4GHz pulls more to VCC than a Xeon W3540?Quote:
Originally Posted by saaya
Pretty much what I come up with also. It's quite common to take a reading and forget about the inherent losses. I see the same thing with load testing screenshots. Using the value a Kill-A-Watt shows and not taking into consideration the loss across the variac.Quote:
Originally Posted by raju
i dont know about an 870... :shrug:Quote:
Originally Posted by raju
where do you get the 15-20% switching loss though?
the pwm engineers at foxconn told me the load efficiency of the p6t, gigabyte x58 extreme and BR are all around 90%
im sure the "24phase" thing from gigabyte has a worse efficiency though...
but this is a 1366 cpu, so i guess you meassured it on the evga 1366 digital pwm? if the efficiency is around 90% you get 25W more for vcc, around 185W...
even if the efficiency is really only 80-85%, i wouldnt be surprised if the low temperature is making quite a diference here, and current draw at higher temps is notably diferent.
Quote:
Originally Posted by saaya
Most PWM circuits have an efificieny of 85% at the very best. The 4.45Ghz measurement was not on a cascade but water cooled at ambient temps. THe 870's CPU's do not pull more than 1366 CPU's to VCC - I have checked. What have you measured? Nothing I suspect :rolleyes:
At high loads, a good motherboard supply will peak at or barely below 90% efficiency.
Measured at ambient temps using an Intel stock cooler on an 870 running Linx - 15.7 amps peak (186w total current draw) running Linx 8 threads 64 bit with 4 GB memory at 1800MHz. Take 30w-40w for VTT considering Intel's 25 amp or so stock spec (not considering any other sub rail like PCH etc) , then even considering a 90% efficiency you get 140W to VCC. :) Peak PWM efficiency is usually deliverd at around half load handling of the circuit btw, which is why I generally cite around 85% given modulation of loads etc..
Cinebench full thread test only draws 12.9 amps from the EPS line.
my new P55-GD80 have a LOTES socket... but i have a problem. When I have Vcore over 1,49V board has shut down and phase led display show 0. I think this is over current protect? this is no good ... :shakes:
//btw i7 860 @ 4200MHz 1,44V full load 8 threads with linx no problem for socket
Ok so looking at the last page I see someone saying that one of his boards had LOTES, the other had a Foxconn. Is there a possibility that all mobo manuf. will release updated versions with better sockets in the future? Then there's discussion of LLC being a factor as well. Am I getting this right? I haven't enabled LLC and am leaving Intel-spec vdroop in place for the moment...
I usually pop boards faster than Kirstie Alley can down a gallon of ice cream but I have over 800 hours on each of those boards without a problem. 400 hours of that is full load overclocks in a case with only the power supply providing airflow exhaust.Quote:
Originally Posted by saaya
So, either he had some really bad luck or something else is going on. Granted I have kept my voltages under 1.4 real with air cooling, but not a single hiccup yet. That said they are all back in cases right now for another round of testing.
To top it off, my burnt pad problem (it happens on air too) has only occurred on the top end boards so far. :confused:
IF he uses a "not that good PSU" to feed 12V on the VCore Mos-fets, this could happen...."Spikes" will travel through the PSU to the Mos-fets and cause them failure..... ;)
I have ONLY TWO PHASE circuit here and I can draw EASY ~225Watts.... ;)
Also I want to inform you something about efficiency on them.....The hotter the go, the LESS efficiency you get out of them.... ;)
You'll get LESS efficiency is the Mos-Fet works without a heatsink at - ex. 95*C and you get more if it works at ex. 50*C.... :)
Quote:
Originally Posted by hipro5
Yes that is right, as the FET temps increase RDS-on will change and this will affect the efficiency.
Hmmm........I like you M8..... :up: :DQuote:
Originally Posted by raju
I hope you don't mind, but hipro are you retired? :yawn: I didn't saw any new benches from you and I really like your stuff
Overclocking is bad, mkay?
Carry on finking :)Quote:
Originally Posted by Russian