• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

12700K at 100C after few minutes of Prime95

Joined
May 8, 2016
Messages
1,735 (0.60/day)
System Name BOX
Processor Core i7 6950X @ 4,26GHz (1,28V)
Motherboard X99 SOC Champion (BIOS F23c + bifurcation mod)
Cooling Thermalright Venomous-X + 2x Delta 38mm PWM (Push-Pull)
Memory Patriot Viper Steel 4000MHz CL16 4x8GB (@3240MHz CL12.12.12.24 CR2T @ 1,48V)
Video Card(s) Titan V (~1650MHz @ 0.77V, HBM2 1GHz, Forced P2 state [OFF])
Storage WD SN850X 2TB + Samsung EVO 2TB (SATA) + Seagate Exos X20 20TB (4Kn mode)
Display(s) LG 27GP950-B
Case Fractal Design Meshify 2 XL
Audio Device(s) Motu M4 (audio interface) + ATH-A900Z + Behringer C-1
Power Supply Seasonic X-760 (760W)
Mouse Logitech RX-250
Keyboard HP KB-9970
Software Windows 10 Pro x64
Just offset VCore by -0,05V or -0,1V and check stability.
Decrease all boost frequencies by 100MHz, if it's not stable.
 
Joined
Jan 21, 2021
Messages
55 (0.05/day)
I only meant that running the CPU a high PL2 for periods in the order of a minute or so and then keeping it at a relatively low PL1 for the rest of the computation time makes sense if the intended aim is taking advantage of the "available thermal capacitance". From the datasheet linked earlier:

View attachment 261314

Clearly, they're not referring about core temperature, since it has almost no thermal inertia and reacts immediately to load changes. On the other hand, externally-measured temperatures very close to the CPU itself (e.g. socket temperature as I showed in the image posted the other day) react with speed consistent with the Intel-recommended Tau time.

Of course one might argue that the 240+ watts is an excessively high power draw, even if for short periods. That is a related but different subject.

Say the heatsink is 40 deg C at idle (~15W) and you put 125W to the CPU and the CPU temp went up to 80 deg C.

If the same heatsink (shortly after the CPU was loaded) is 50 deg C at idle (15W) and you put 125W to the CPU at that point, the CPU temp would increase to 90 deg C.

I'm just pointing out they could be talking about this thermal capacitance (I think they are)
 

Rob6502TPU

New Member
Joined
Sep 6, 2022
Messages
10 (0.02/day)
I only meant that running the CPU a high PL2 for periods in the order of a minute or so and then keeping it at a relatively low PL1 for the rest of the computation time makes sense if the intended aim is taking advantage of the "available thermal capacitance". From the datasheet linked earlier:

View attachment 261314

Clearly, they're not referring about core temperature, since it has almost no thermal inertia and reacts immediately to load changes. On the other hand, externally-measured temperatures very close to the CPU itself (e.g. socket temperature as I showed in the image posted the other day) react with speed consistent with the Intel-recommended Tau time.

Of course one might argue that the 240+ watts is an excessively high power draw, even if for short periods. That is a related but different subject.
I see, I assumed you knew that PL2 was allowed for the whole duration of the heavy load, which is why the 100C thermal limit is hit and maintained on well optimised software using the CPU to max.
 
Joined
Jun 22, 2012
Messages
264 (0.06/day)
Processor Intel i7-12700K
Motherboard MSI PRO Z690-A WIFI
Cooling Noctua NH-D15S
Memory Corsair Vengeance 4x16 GB (64GB) DDR4-3600 C18
Video Card(s) Sapphire RX480 Nitro+ 4GB
Storage Samsung 980 Pro 1TB, Crucial MX500 500GB
Case Fractal Define C
Power Supply Seasonic Focus GX-550
Say the heatsink is 40 deg C at idle (~15W) and you put 125W to the CPU and the CPU temp went up to 80 deg C.

If the same heatsink (shortly after the CPU was loaded) is 50 deg C at idle (15W) and you put 125W to the CPU at that point, the CPU temp would increase to 90 deg C.

I'm just pointing out they could be talking about this thermal capacitance (I think they are)

125W would be a load within the processor's stated TDP. If the cooler is only sufficient enough not to cause thermal throttling at the TDP (minimum requirement by Intel), it is virtually guaranteed that the CPU will hit almost immediately the thermal throttling point at a significantly higher load. Even on my Noctua D15S, when a load of say 200W is applied, reported core temperature jumps from ~40 °C to 90°C within 2-3 seconds and reaches the thermal throttling point soon after that.

Cooler temperature however rises much slower than core does. And IHS temperature (Tcase), which unlike core temperature is part of the thermal specifications for lidded CPUs, can be expected to be close that of the cooler near its base, perhaps at a heatpipe. For this, a Tau time of 28~56 seconds as recommended does make sense.

The thermal specifications don't say that the CPU cannot be at the throttling point during the limited time at PL2 (or anyway at a power greater than PL1/TDP), only that core temperatures should be below TJmax when the CPU is at TDP, and that TCase should be below specification values. So the thermal behavior I showed the other day seems acceptable and within specification, in my opinion.

I see, I assumed you knew that PL2 was allowed for the whole duration of the heavy load, which is why the 100C thermal limit is hit and maintained on well optimised software using the CPU to max.

PL2 has always been intended to be limited in duration in the datasheets, not to last for the entirety of the heavy load (a rendering job could last hours or days or more, that seems excessive at PL2).

Alder Lake launch slides did mention that PL1=PL2 brings the best CPU performance, but that doesn't seem to be an official recommendation. So, what to trust? Datasheets or marketing slides?


1663066950947.png 1663067602611.png
 
Joined
Jan 21, 2021
Messages
55 (0.05/day)
125W would be a load within the processor's stated TDP. If the cooler is only sufficient enough not to cause thermal throttling at the TDP (minimum requirement by Intel), it is virtually guaranteed that the CPU will hit almost immediately the thermal throttling point at a significantly higher load. Even on my Noctua D15S, when a load of say 200W is applied, reported core temperature jumps from ~40 °C to 90°C within 2-3 seconds and reaches the thermal throttling point soon after that.

Cooler temperature however rises much slower than core does. And IHS temperature (Tcase), which unlike core temperature is part of the thermal specifications for lidded CPUs, can be expected to be close that of the cooler near its base, perhaps at a heatpipe. For this, a Tau time of 28~56 seconds as recommended does make sense.

The thermal specifications don't say that the CPU cannot be at the throttling point during the limited time at PL2 (or anyway at a power greater than PL1/TDP), only that core temperatures should be below TJmax when the CPU is at TDP, and that TCase should be below specification values. So the thermal behavior I showed the other day seems acceptable and within specification, in my opinion.

I completely agree what you described is in spec.

My point wasn't 125W, my point was the action. Whether that be at 65W or 250W.

I do get why you take issue with me using 125W for the example though - I should've used a value over TDP.

Delta - the difference between two values.

Think of the delta between CPU temperature and heatsink temperature. For example, if 200W creates a 50 deg C temperature delta and the heatsink is cooled to a temperature of 40 deg C at idle, the second you apply 200W, the CPU temperature rises to 90 deg C.

Now think, with 200W, how fast does the heatsink rise in temperature with the fan at 100%? If it's 5 deg C per minute, you have 120 seconds until thermal throttling occurs. As long as so much power isn't added that the delta is 60 degrees (instantly causing the CPU to hit 100 degrees), the CPU is fine thermally - Turbo Boost 3.0 I believe is the one which dictates you can do anything to the CPU up to that point, up to anything that doesn't exceed maximum current obviously

Ideally you have a cooling solution which is able to keep up with any real world load you apply to the CPU so that you never run into thermal throttling.


Side note: For these huge powers that are being allowed lately, I don't think it's a good thing for chip longevity - thermal cycling causes growing and shrinking of the loaded parts of the die, which over time is bad for connections between parts
 
Last edited:
Joined
Jun 22, 2012
Messages
264 (0.06/day)
Processor Intel i7-12700K
Motherboard MSI PRO Z690-A WIFI
Cooling Noctua NH-D15S
Memory Corsair Vengeance 4x16 GB (64GB) DDR4-3600 C18
Video Card(s) Sapphire RX480 Nitro+ 4GB
Storage Samsung 980 Pro 1TB, Crucial MX500 500GB
Case Fractal Define C
Power Supply Seasonic Focus GX-550
What I'm implying above more in general is that when the Intel-recommended PL2 can even be more than 3 times the PL1/TDP value, thermal throttling should be expected when the CPU is boosting above the TDP, in particular if mated with a cooler of its TDP class.

For instance, the 65W (non-k) 12900 has a PL2 of 202W. The previous 65W 11900 had an even higher PL2 of 224W. The boxed coolers they came with couldn't possibly dissipate such power for more than a few instants—much shorter than their Tau time—and would certainly cause the CPU to operate at TJMax until the limited boosting period expires, while CPU power slowly decreases due to the throttling.

However, thermal throttling, when operating under standard settings (dynamic voltages/frequencies), isn't necessarily a critical condition. The CPU will typically very gradually decrease frequencies and voltage to prevent internal temperatures from exceeding the limit, with no perceivable stuttering behavior that perhaps might have used to occur in the past, which made people try to avoid it at all costs. I'm currently running my 12700K with a 90 °C limit (it's configurable within a limited range on Z motherboards), which causes it to thermally throttle earlier and more also with real-world workloads (Blender, etc). There are no practical drawbacks that I can see except obviously that on average frequencies under full load will be lower and decrease quicker.

Of course (again), under standard settings it can be argued that CPU longevity could be negatively affected with frequent intense thermal and power excursions. Constant operation at the thermal limit is definitely not recommended by Intel, but they seem to be ok with these CPUs operating at high power and die temperatures for short periods of time. We'll see in a few years if they will start dying earlier compared to their predecessors.
 

Lei

Joined
Jul 3, 2021
Messages
1,142 (1.14/day)
Location
usually in my shirt
Processor 3900x - Bykski waterblock
Motherboard MSI b450m mortar max BIOS Date 27 Apr 2023
Cooling αcool 560 rad - 2xPhanteks F140XP
Memory Micron 32gb 3200mhz ddr4
Video Card(s) Colorful 3090 ADOC active backplate cooling
Storage WD SN850 2tb ,HP EX950 1tb, WD UltraStar Helioseal 18tb+18tb
Display(s) 24“ HUION pro 4k 10bit
Case aluminium extrusions copper panels, 60 deliveries for every piece down to screws
Audio Device(s) sony stereo mic, logitech c930, Gulikit pro 2 + xbox Series S controller, moded bt headphone 1200mAh
Power Supply Corsair RM1000x
Mouse pen display, no mouse no click
Keyboard Microsoft aio media embedded touchpad (moded lithium battery 1000mAh)
Software Win 11 23h2 build 22631
Benchmark Scores cine23 20000
Just curious for others like myself despite fixing limits that still hit 100c during prime95, what program did you end up using to ultimately determine if the cpu is stable in the long run as a result?
I use Argus monitor. It automatically turns the pc to sleep if cpu reaches a certain temp.
So I can peacefully take shower and leave the PC on.

I have an exe file that sleeps the PC, argus runs it when it hits a high temp (instead of sending me an email lol)
 
Top