3570K and 75C gaming temps

[Don Karnage]

According to the screenshot they supply. Its running at 4.5Ghz at 101C. Meaning no throttle at 101C.

Plus I have tested with my own 3570K. So I know it doesnt throttle until 105C.

Same here. My 3570K throttled at 105C

 

[cmdrdredd]

My 3570k @ 4.7Ghz runs upwards of 90-91c with max IBT and it holds on to 4.7Ghz like nothing is happening.

I leave it on 4.6Ghz though

 

[Idontcare]

http://www.tomshardware.com/reviews/ivy-bridge-overclocking-core-i7-3770k,3198-2.html



http://hothardware.com/Reviews/Intel-Core-i73770K-Ivy-Bridge-Processor-Review/?page=3

I doubt tom's was seeing thermal throttling. They probably saw over-current protection doing its thing because they were being silly with the voltages on their OC and mistook the throttling as being thermal throttling.

Thermal throttling occurs at TJmax, not lower.

Over-current throttling can occur at any temperature, even at LHe temperatures, provided the over-current levels are left to their default values in the BIOS and the "over-current protection" setting is left enabled in the BIOS.

 

[cmdrdredd]

I doubt tom's was seeing thermal throttling. They probably saw over-current protection doing its thing because they were being silly with the voltages on their OC and mistook the throttling as being thermal throttling.

Thermal throttling occurs at TJmax, not lower.

Over-current throttling can occur at any temperature, even at LHe temperatures, provided the over-current levels are left to their default values in the BIOS and the "over-current protection" setting is left enabled in the BIOS.

Yes this is true, my motherboard has a setting where I can adjust the value for "CPU Current Capability". If I set it to a number, say 100% and it goes over that then it will throttle. Default for my motherboard is 140% to allow overclock headroom.

 

[bononos]

There are very exacting socket and surrounding component specs for a reason.

Do you happen to know where it is exactly in the datasheets? I don't remember seeing anything like that.

........
If Intel has failed to properly characterize the capability of its products to endure thermal cycling then to be sure the upper temp spec will only exacerbate the problem, but lowering the upper temp spec will not negate such a problem either.

The mobo makers are required to design their products such that they can properly operate in whatever conditions the other components are operating.

If a mobo maker creates a mobo that supports IB but the mobo fails if the IB CPU itself gets to 105C then that is a problem created by, and owned by, the mobo maker.

Nothing prevents the mobo maker from making shoddy mobo's, the onus is on the consumer to avoid known shoddy mobo makers if you intend to push the components to the limit. And this is true of all electronics, right? We don't buy a 750W rated PSU for $15 and actually expect it to function if tasked with supporting a 700W load. We know we bought a shoddy PSU for $15 so we keep its load under 300W.
.........

Ok. Only relying on the 'avoid shoddy mb brands' isn't exactly comforting in a very concentrated market for mb and cpu manufacturers. I have a hard time believing Intel cpus can survive normal lifespans running just below the redline (like the poster above me said).

 

[Ferzerp]

Do you happen to know where it is exactly in the datasheets? I don't remember seeing anything like that.

e.g. http://www.intel.com/content/www/us/en/processors/core/3rd-gen-core-lga1155-socket-guide.html

 

[bononos]

e.g. http://www.intel.com/content/www/us/en/processors/core/3rd-gen-core-lga1155-socket-guide.html

Yeah I've already seen that doc before but I didn't find Intel thermal guidelines for surrounding components (around the cpu) as per your earlier statement:

There are very exacting socket and surrounding component specs for a reason.

 

[poohbear]

what effect does the higher temperature have on the thermal paste? Most thermal paste out there was designed for cpus running in the 45-65c range. Does the thermal paste degrade quicker and hence need more frequent replacement if i'm running 85c+ everyday?

 

[Ferzerp]

Yeah I've already seen that doc before but I didn't find Intel thermal guidelines for surrounding components (around the cpu) as per your earlier statement:

It tells exactly what heat the proc may be giving off. The motherboard has the be able to handle that. There are no surprises for the mobo makers. If they don't build to handle that spec, it's all on them.

 

[bononos]

It tells exactly what heat the proc may be giving off. The motherboard has the be able to handle that. There are no surprises for the mobo makers. If they don't build to handle that spec, it's all on them.

I don't dispute that. What I was interested in was your statement about 'exacting standards for surrounding components' so I was expecting some very precise guidelines. Something like max operating temps within boundries of concentric circles from the cpu socket or approved components like caps/vrms or their specs which the document does NOT have.

 

[Idontcare]

I don't dispute that. What I was interested in was your statement about 'exacting standards for surrounding components' so I was expecting some very precise guidelines. Something like max operating temps within boundries of concentric circles from the cpu socket or approved components like caps/vrms or their specs which the document does NOT have.

Yeah, they don't exist. Intel just expects the mobo makers to wing it.

Kinda like their TJmax spec, they know their chips will die if people actually let them get that hot, and they could have avoided that by simply putting TJmax even lower such that whatever temp it was set at would mean the chip was OK operating at those temps, but instead of doing that they just threw a dart and picked a TJmax value at random.

Its all true, you can't find an internet document that proves me wrong, so it must be true.

 

[tweakboy]

That is insane 100c load 34c idle,,,,,, What HSF you using. Did you apply thermal did you install it properly. Your @ 1.27v and your getting 100c full load. Something doesn't sound right. gl

 

[Ferzerp]

Yeah, they don't exist. Intel just expects the mobo makers to wing it.

Kinda like their TJmax spec, they know their chips will die if people actually let them get that hot, and they could have avoided that by simply putting TJmax even lower such that whatever temp it was set at would mean the chip was OK operating at those temps, but instead of doing that they just threw a dart and picked a TJmax value at random.

Its all true, you can't find an internet document that proves me wrong, so it must be true.

I missed your sarcasm tag until the second sentence

 

[IntelUser2000]

Thermal throttling occurs at TJmax, not lower.

Yep, I stand corrected. It may be just wrong measurements then.

 

[ShintaiDK]

That is insane 100c load 34c idle,,,,,, What HSF you using. Did you apply thermal did you install it properly. Your @ 1.27v and your getting 100c full load. Something doesn't sound right. gl

Did you read the thread at all? Its from Tomshardware

 

[Magic Carpet]

Kinda like their TJmax spec, they know their chips will die if people actually let them get that hot, and they could have avoided that by simply putting TJmax even lower such that whatever temp it was set at would mean the chip was OK operating at those temps, but instead of doing that they just threw a dart and picked a TJmax value at random.

What about the military grade processors, are they the same, you reckon?

 

[darkfalz]

I think some people are a little nuts. What does 4.4~4.5 get you that ~4.0 doesn't get you? 10% extra speed when most games are already GPU limited? I undervolt and overclock my 3570k to 4.0 Ghz. Temps barely get above 50 during gaming and this is with the stock cooler. To get 4.4 I needed a pretty steep increase in voltage / LLC and I think my temps were in the 90s. I don't care if the tjMax is 105, I don't want it in the 90s. I already got a 10% boost for free (undervolted in fact) do I really need another 10% so I can use 40% more power and run 40% hotter?

 

[tweakboy]

Guys, while the Tjunction might be rated 105C, Intel recommends that the heat sink solution be designed so it stays well under that. In the datasheet they give an example of 10 degrees lower, which is 95C.

In previous Intel processors, throttling of the Base clock happened at 80-85C.

I can vouch for that.

 

[cmdrdredd]

I think some people are a little nuts. What does 4.4~4.5 get you that ~4.0 doesn't get you? 10% extra speed when most games are already GPU limited? I undervolt and overclock my 3570k to 4.0 Ghz. Temps barely get above 50 during gaming and this is with the stock cooler. To get 4.4 I needed a pretty steep increase in voltage / LLC and I think my temps were in the 90s. I don't care if the tjMax is 105, I don't want it in the 90s. I already got a 10% boost for free (undervolted in fact) do I really need another 10% so I can use 40% more power and run 40% hotter?

I don't know what you're running but if you see 90c in games even Battlefield 3, you have a problem. I don't even hit 90c during prime.

My Q9550 hit 85c during prime and never throttled either...

 

[TheInternal]

I'm currently hitting a peak of 74C on one of my cores (Core#3) while running LinX at 4.5GHz with the motherboard (ASUS P8-Z77-V Pro) on my desk, using an aftermarket air cooler. CPU Core voltage seems to be peaking at 1.256V. I have the "CPU Voltage" set to 1.235 in the ASUS TurboV EVO application that came with my P8z77-V Pro. However, while running LinX, my desktop does go black sometimes... not entirely sure why... seems to always happen if I have the ASUS app open. The start menu / quicklaunch stays around though. it's odd. Not sure if it's a sign of instability or just something the program does.

I guess I'm still not entirely clear on what "throttling" is... is it when the voltage jumps from 1.0xx volts idle to 1.264 volts at 4.5GHz overclocked?

 

[RaistlinZ]

^^ I think it refers to the CPU speed automatically dropping when it reaches a certain temperature, typically over 100C or more.

 

[darkfalz]

I don't know what you're running but if you see 90c in games even Battlefield 3, you have a problem. I don't even hit 90c during prime.

Nah, those were IBT Very High temps with the stock cooler @ 4.4. I doubt gaming would be as high - my point is that I'd rather run 30 degrees cooler and lose a measley 10% speed that cook my chip just for e-penis bragging rights or have to spend another $30-50 on an aftermarket cooler.

 

[excaliburxvii]

Nah, those were IBT Very High temps with the stock cooler @ 4.4. I doubt gaming would be as high - my point is that I'd rather run 30 degrees cooler and lose a measley 10% speed that cook my chip just for e-penis bragging rights or have to spend another $30-50 on an aftermarket cooler.

I know I'm resurrecting an old thread, but I'm pretty much stuck. I agree with you, and so far I was able to get my chip up to 4.2 GHz on stock cooling, with all stock settings except for the turboboost clock. Can I ask what your settings are? 4.0 is where I'd like to settle but I lowered it to 3.8 (Even with 4 cores active.) when it got to almost 90C (4.2 and 4.0. 4.0 seemed to heat up faster than 4.2 for some reason). I haven't overclocked in a while and 90C still seems a bit high to me. I'd be happy under 85. (This is with Prime95.) Just trying to get a little extra performance for my GTX 780.

 

[BonzaiDuck]

I know I'm resurrecting an old thread, but I'm pretty much stuck. I agree with you, and so far I was able to get my chip up to 4.2 GHz on stock cooling, with all stock settings except for the turboboost clock. Can I ask what your settings are? 4.0 is where I'd like to settle but I lowered it to 3.8 (Even with 4 cores active.) when it got to almost 90C (4.2 and 4.0. 4.0 seemed to heat up faster than 4.2 for some reason). I haven't overclocked in a while and 90C still seems a bit high to me. I'd be happy under 85. (This is with Prime95.) Just trying to get a little extra performance for my GTX 780.

Scouring the web for answers to questions and it brought you here, eh? Welcome to Anandtech forums.

Seems like topics overlap between forums, but this cooling issue is CPU-specific.

High temperatures will slowly age your CPU, but less critical for sudden damage than high voltage.

IB shouldn't be that much different than my SB-K chip. You can attempt to lower high-load temperatures in increments. Diamond TIM paste or liquid metal "ultra pro" may win you more than 5C degrees improvement over Arctic Silver, depending on the choice and the thermal wattage. Lapping the processor IHS to make it flat and get rid of the nickel-plating is worth a few C degrees, as is eliminating the nickel plate on the heatsink or waterblock's copper base. Ducted fan strategies for high-end air-ccooling heatpipe coolers is worth a few more C degrees.

But for the IB cores, the tedium and risk of de-lidding, together with a liquid-metal TIM seem to be the formula for temperature improvements as great as 20C.

Finally, I may have seen the same observation on an IB overclocking guide, but at least for my SB-K processor (which would fit in your motherboard), at least one stock voltage setting is excessive: PLL voltage. This an be lowered by maybe 0.15V for both better stability and lower temperature, even if the difference in temperature is somewhere between 1C and 2C.

You might also take pains to disable the Intel HD iGPU if you aren't using it. Try to avoid enabling "PLL Overvoltage," [not the same as the voltage setting mentioned above.] That will also increase temperatures. There are other settings for voltage regulation and phase control, in which the more "extreme" settings will add more temperature.

 

[excaliburxvii]

Thanks for the reply and the welcome, I've been checking out AnandTech for reviews and benchmarks for a while and I was kind of surprised when I found out I hadn't registered yet. I'm not looking for any crazy overclocks, just a stable bump up. I was always pretty sketchy about lapping, I know it works I just don't know if I trust myself with it, lol. I could solve the temp problems just by grabbing an aftermarket HSF, but it's not that big of a deal to me anymore. The fact that there are thermal pastes 5C cooler than AS5 just proves that I've been out of the game for a while, lol. But other than lapping/upgrading my thermal paste, I should step the voltage down 0.15v, disable the integrated graphics, and make sure Internal (CPU) PLL Overvoltage is disabled?

Edit: Also, during some searching I found someone claim that disabling Internal PLL Overvoltage would break hibernation, do you know if that's true?

Edit #2: Disabled integrated graphics and Internal PLL Overvoltage, didn't touch any voltages, and now I'm running Turboboost at 4.0GHz, topping out at 91C/90C/87C/87C. That seems decent to me - I don't need this chip to last a decade. Is there anything glaringly wrong here? I can provide more info if you need it.

Specs:
GIGABYTE GA-Z77X-UD4H
Core i5-3570K
CORSAIR XMS 8GB (x2, for 4GB x4) 1600, 11-11-11-30



Edit: Also, just curious, could lowering the voltage affect the performance of the CPU? I might just be imagining it but it seemed like the 4 processes in Prime95 ran more in-step with stock voltage. No blue screens or anything.