Got my i5-760 running at 4Ghz(20 X 200)@1.25vcore and using my Xigmatek 1283 HSF I'm getting idle temps of 39 but Prime takes the load to over 77 and my system freezes in a couple of minutes. Don't want to mess with changing coolers so I figured I'd back down on the OC a bit but wondered what is causing the heat mostly.
What raises CPU heat most when OCing?
- Thread starter perdomot
- Start date
Spikesoldier
Diamond Member
- Oct 15, 2001
- 6,766
- 0
- 0
upping frequency or vcore is going to increase your heat. usually increase in frequency correlates with an increase in vcore.
on the quad core i5's, a good thermal sweet spot seems to be around 3.7.
on the quad core i5's, a good thermal sweet spot seems to be around 3.7.
CPU heat is proportional to CPU temperature is proportional to CPU power-consumption.
Power-consumption is proportional to the cube of the operating voltage and linearly with the clockspeed.
http://forums.anandtech.com/showpost.php?p=27459192&postcount=20
If you want to decrease power/heat/temps then do what you can to reduce Vcc first and foremost, that may mean lowering clockspeeds so a lower Vcc is doable or improving temps (better cooler) so that a lower Vcc is doable.
tweakboy
Diamond Member
that's high temp for only 1.25 vcore and no hyperthreading. Wonder if you have bad airflow.
Anyways, I bet if you had better airflow (if that's a good cooler) or got a better cooler (if you have good airflow) then you'd be able to maintain your 4ghz overclock.
When you're on the edge of stability, high thermals can make you unstable. Electrons get jittery and excited and start doing things they're not supposed to.
My case is currently open with the HSF pointing towards the rear exhaust fan. Room temp is 40F. A review on Youtube had my cpu at 4Ghz priming at 61C with a vcore of 1.25. Had to raise mine to 1.275 to get stability and its idling at 38-39C. When I used my video encoder temps got up to 71 when I reduced the speed to 3.6Ghz. My HSF is a good one so I'm not sure what to do.
While not necessarily a factor for the OP. Is there documentation on power consumption vs operating temperature on various semiconductors? Specifically 65nm and down. Obviously silicon doped with different materials will behave differently, and possibly other factors may affect this such as overly leaky transistors, but I'm wondering if there is any info on the internet of this.
Your idle temps even sound too high for 40F ambient. Since you have the case open, direct a fan at it and get some more of that cold air across it and see how much difference it makes. Too much thermal compound maybe?
My system has frozen a couple of times so far with the Event Viewer saying it was a Event 41 Kernel Power. Anyone know what might be causing this? I was at idle when the last one hit.
I decided to reset everything in the bios and while I was thinking of what to do next, I noticed that cpu-z was showing the cpu going at 3300 at default voltage. Decided to just change the bclk to 160 which is a little more and nothing else and rebooted. Mobo raised the vcore to 1.3 and idle temps are about 36-37 and when I encoded some video temps never got above 67C which is good so far. Gonna try the 150 bclk and see if vcore drops.
- Sep 28, 2005
- 21,086
- 3,592
- 126
What QPI/VTT voltage are you using right now for 200 Bclk? The secret to stable OCs and low temps with these chips is to use sufficient QPI/VTT so you can use LESS CPU vcore.
Also, your last post about your system raising the vcore to 1.30...how does that do that on its own? The only way that happens is if you have the XMP memory profile selected, which will in some cases (like my Gigabyte) cause your mobo to choose preset settings. Make sure you are in full manual mode on CPU vcore and not on Auto. Ditto for DRAM voltage and settings -- input the main 4 timings manually.
- Sep 13, 2008
- 8,253
- 3,142
- 146
also, keep in mind the memory controller is also on the chip on the newer CPUs, so to up frequency, you up blck, which means needing to up uncore voltage, unless you have an unlocked. Upping uncore voltage, either for higher bclk, faster uncore, or faster memory, will also heat up the CPU, in addition to vcore.
RaistlinZ
Diamond Member
- Oct 15, 2001
- 7,470
- 9
- 91
You might need a bit more juice to get 4.0 stable. But your temps don't seem to be causing it to freeze. It could just be that your chip doesn't do 4.0 stable, that's all.
My CPU can do 4.00 but will freeze up at 4.05 during Prime, even if I crank up the voltage to 1.35v. Nothing I can do about it.
My CPU can do 4.00 but will freeze up at 4.05 during Prime, even if I crank up the voltage to 1.35v. Nothing I can do about it.
I used the following voltages for the 4Ghz OC: QPI=1.35, PCH=1.14, CPU PLL=1.9. I did set the CPU and RAM maually. I did some searching on Google and these numbers seem to be the most commonly used. What I don't get is the heat issue. According to this review, their Noctua cooler kept temps around 60C but my Xigmatek 1283 which is supposed to perform about 2 degrees warmer can't keep it below 80 while running Prime: http://www.overclock3d.net/reviews/cpu_mainboard/intel_i5_760_review/2
Maybe I need a better fan on there?
It sounds like you're at the limit of your chip. Raising QPI/VTT beyond 1.35v is not advised, and if you are getting 77C temps at 1.25v CPU vcore, I'd say that's probably it for you.
Do you have Load line calibration enabled ?
Yeah I got it enabled. The new Gigabyte H55 mobo I switched too seems to be having problems too. I tried going to 4Ghz without running any benches to stress it and it froze and now anything above blck 150 freezes the system when I do stuff like Youtube. Might have damaged something. I need a good OC mobo recommendation if you guys have any.
I'm not sure where you'd find it on the net, but this is undergraduate level stuff for EEs. There are many textbooks that talk about this.
Off the top of my head, you'd have some changes in electron mobility and charge concentration. In CMOS, the power hit comes in the switching and those two factors will affect the switching time.
Hacp
Lifer
- Jun 8, 2005
- 13,923
- 2
- 81
IIRC, frequency vs power is linear while voltage vs power is quadratic. So upping the voltage will have a larger effect than upping the frequency. I am not an EE student I just somehow remember seeing that relation on some website.
Modelworks
Lifer
- Feb 22, 2007
- 16,240
- 7
- 76
Right now power consumption is the big topic of most of the major players. By major players I don't mean intel and amd but companies like Renesas, Marvell, Realtek, TI, that make embedded processors. There is a sort of competition going on where they are trying for the very lowest possible power and heat production to process 1 instruction. Along with that there is shrinking of the die but a good part of the power reduction comes from being able to do more with less transistors by changing the architecture of the core.
The information on how that is done is widely available because each of these companies likes to show off to the others what they came up with. All you need to do is download datasheets. Some of the interesting ones.
http://am.renesas.com/products/mpumcu/rx/Documentation.jsp#
The most common approach to lowering heat is to put the processor to sleep when it doesn't have to be doing something. I know people will think of the HLT instruction or windows throttling down a processors speed but that doesn't come close to what these upcoming processors are doing.
They essentially turn completely off if no instruction needs to be processed that nanosecond. Going from full clock speed to no clock speed . It isn't a case of processor only needs to run at 50% to do the current task , so lets run it at a reduced clock like in windows. Instead it is run the instruction at full clock and then do nothing at all, not even a single clock until it is needed again, then go straight back to full clock speed until again it is not needed and drops to zero MHZ. With the way pc are designed that isn't possible because they are evolved from the older designs. If they did use the technology your pc cpu would consume no electricity while you looked at your screen reading something. It wouldn't start to consume again until you typed a key or moved the mouse or something had to be done.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 23K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter