Question Raptor Lake - Official Thread

[Hulk]

Since we already have the first Raptor Lake leak I'm thinking it should have it's own thread.
What do we know so far?
From Anandtech's Intel Process Roadmap articles from July:

Built on Intel 7 with upgraded FinFET
10-15% PPW (performance-per-watt)
Last non-tiled consumer CPU as Meteor Lake will be tiled

I'm guessing this will be a minor update to ADL with just a few microarchitecture changes to the cores. The larger change will be the new process refinement allowing 8+16 at the top of the stack.

Will it work with current z690 motherboards? If yes then that could be a major selling point for people to move to ADL rather than wait.

 

[Just Benching]

The temperature is an internal property of the CPU, the hottest of the many sensors the CPU has inside. It says little about the overall CPU stress or total heat production.

Wattage is something that the CPU cooler must deal with. In the situation you are limited by a cooler, you must set the power draw limit to what the cooler can handle, and not the temperature limit.

Setting the temperature limit is useful if you for example want to prolong the life of the CPU, when you do not believe that 100°C limit by Intel is reasonable.

From my experience most coolers can easily deal with the 13900k, it's really easy to dissipate heat from that chip. My u12a peaked at 260-270w with the 12900k, with the 13900k it can do 330+ watts.

The whole overheating shaenanigans started from hwunboxed. That guy managed to throttle the 13900k at 240watts with a 360 AIO (!). He even managed to throttle his 13600k with a 360 aio....He was even hitting 87C at 144 watts!! I don't want to blame him for bias, but he is definitely doing something wrong

 

[igor_kavinski]

?
Setting a lower temperature limit will make you throttle sooner.

Dynamic frequency scaling shouldn't be as drastic as thermal throttling. I think thermal throttling is the one that will ruin game experience.

 

[igor_kavinski]

From my experience most coolers can easily deal with the 13900k, it's really easy to dissipate heat from that chip. My u12a peaked at 260-270w with the 12900k, with the 13900k it can do 330+ watts.

That means the 13900K has a better heatspreader with possibly better TIM.

 

[Just Benching]

That means the 13900K has a better heatspreader with possibly better TIM.

I think its mostly the bigger die - more cores. But both were pretty easy to cool, I've no idea why the internet went crazy about it.

 

[igor_kavinski]

I think its mostly the bigger die - more cores.

Yeah bigger die would give more surface area for heat to escape through but the extra cores would also add additional heat. Something about the heat transfer between the die and IHS has improved considerably for the newer i9 to dissipate extra 60W with the same cooler.

 

[Hulk]

Of course, that's what im saying, different workloads will end up with different CPU temperatures even at similar wattages.

This is absolutely true and I can provide another example. I can run CB R23 at 230W and temps stay below 86C. 200W in Vegas Pro during a Render+ will smoke the CPU to 100C in 10 seconds. Not every workload stresses the core(s) the same way. A certain workload could only stress 4 or 5 cores, but really activate all structures in them and cause really high temps with perhaps 125W (think AVX512).

Rather than have a PL2 with a tau it would be much smarter to have TL1 and TL2 with a tau for TL2.
By "TL" I mean temperature limit.

So for example you could set TL1 to 75C for long term loads but TL2 to 90C for 10 seconds for bursty workloads.

As process size gets smaller, while the amount of heat energy to be removed may decrease a bit, the area by which this heat is removed seems to be decreasing at a greater rate, at least for Intel. The only way to keep up is to increase the thermal transfer coefficient. We're pretty much at the limit in that area, outside of delidding, so going sub ambient or perhaps a really good custom loop would be the only way to cool such a CPU at all power levels on all applications.

Anyway, the point is that hotspots are an issue, which means the best way to control core temps is to match your cooling solution by temperature and not power since as Herald astutely posted "different workloads will end up with different CPU temperatures even at similar wattages."

I have to admit it's a very smart observation and I wish I'd thought it through myself.

 

[Hitman928]

From my experience most coolers can easily deal with the 13900k, it's really easy to dissipate heat from that chip. My u12a peaked at 260-270w with the 12900k, with the 13900k it can do 330+ watts.

The whole overheating shaenanigans started from hwunboxed. That guy managed to throttle the 13900k at 240watts with a 360 AIO (!). He even managed to throttle his 13600k with a 360 aio....He was even hitting 87C at 144 watts!! I don't want to blame him for bias, but he is definitely doing something wrong

HWUB’s 13900k throttled at over 300W with a 360 aio because their motherboard allowed it to consume as much out of the box as do several others. They discuss it in the video and even a 360mm aio is not enough to handle such power dissipation.

 

[Just Benching]

HWUB’s 13900k throttled at over 300W with a 360 aio because their motherboard allowed it to consume as much out of the box as do several others. They discuss it in the video and even a 360mm aio is not enough to handle such power dissipation.

If you watch his KS review, he has a side by side with the 13900k, and it shows the 13900k hitting 100c at 257 watts which is just not possible.

Here is the link, go to 4:28

You can also check his 13600k which also throttled, hitting 100c at 180watts with a 360AIO as well.

He is obviously doing something wrong.

 

[Kocicak]

HWUB’s 13900k throttled at over 300W with a 360 aio because their motherboard allowed it to consume as much out of the box as do several others. They discuss it in the video and even a 360mm aio is not enough to handle such power dissipation.

Again, I will post a video of my 13900K consuming 300W after 10 minutes of Cinebench being mostly under 90°C with A SMALL 240 AIO.

These CPUs are generally very easy to cool. However I believe that there might be some samples out there with a very thick layer of TIM. I also believe that some reviews were affected by early bioses which pushed too high voltage in the CPUs.

 

[Hitman928]

If you watch his KS review, he has a side by side with the 13900k, and it shows the 13900k hitting 100c at 257 watts which is just not possible.

Here is the link, go to 4:28

I think you are seeing a slight delta between the reported temp an power, that's all. In that clip, he is showing the 13900k sustaining 280W and the temp is bouncing between 98C and 100C. There's a very brief break in the work load and the temp and power dip significantly and then start to rise again. There's a very brief shot of the temp spiking with the power at 257W but most likely the power sensor is just polling slower than the temp sensor so the power is actually higher, it just hasn't updated yet in the program. If it played for 1 second longer, you'd see the power number update too.

In HWUB's favor, their setup seems to be able to sustain higher clocks at the same power/temperature as other testers, so their results aren't bad at all.

You can also check his 13600k which also throttled, hitting 100c at 180watts with a 360AIO as well.

He is obviously doing something wrong.

I agree on the 13600k though, there seems to be something wrong there. It does seem like their 13600k is pushing higher frequencies so most likely their board is just pushing crazy voltages out of the box to maximize frequency with little effect.

 

[Hitman928]

Again, I will post a video of my 13900K consuming 300W after 10 minutes of Cinebench being mostly under 90°C with A SMALL 240 AIO.

These CPUs are generally very easy to cool. However I believe that there might be some samples out there with a very thick layer of TIM. I also believe that some reviews were affected by early bioses which pushed too high voltage in the CPUs.

That's very impressive, are you delidding or anything? I haven't seen anyone be able to sustain that amount of power on a AIO before. Which model do you have?

 

[IEC]

Extreme OCers who go through multiple chips for binning purposes to find their golden sample are obviously going to have better thermals because it will take less voltage to sustain the same clocks versus a poorly binned specimen.

And if a motherboard's stock setting is to go full speed ahead and exceed 300W... that should be called out as well. Because most users won't alter default motherboard behavior and that's going to push the chip into the hard to cool territory with minimal gains in performance.

 

[Harry_Wild]

Other then the 13900 65W CPU, everything else is either marked up or out of stock! Still waiting for the 13600 to be release and will pay whatever to buy it!

 

[Kocicak]

That's very impressive, are you delidding or anything? I haven't seen anyone be able to sustain that amount of power on a AIO before. Which model do you have?

That was stock CPU with ordinary thermal paste. It may have by chance thin TIM layer and also it did not have any warping issues.

Just yesterday I bought 13600K, which BTW runs comfortably at 5500 MHz, but this particular CPU in a new MB mount HAS THE WARPING ISSUE.

So the temperature is affected by:

• Power draw.
• Heat removing capacity of the cooler.
• Quality of the silicone.
• Correct voltage supplied by MB.
• Thickness of the TIM layer.
• Flatness of the IHS by itself.
• Flatness of the cooler contact plate.
• Combination of IHS and CPU mount, which may result in the warping problem and cause too thick layer of thermal paste in some areas.
• Improper cooler mounting.

It is a lot of stuff that can go wrong.

BTW the CPU was power limited to draw "only" 300W. The situation may have been different if I let it loose and draw what is could (340W).

 

[Hitman928]

That was stock CPU with ordinary thermal paste. It may have by chance thin TIM layer and also it did not have any warping issues.

Just yesterday I bought 13600K, which BTW runs comfortably at 5500 MHz, but this particular CPU in a new MB mount HAS THE WARPING ISSUE.

So the temperature is affected by:

• Power draw.
• Heat removing capacity of the cooler.
• Quality of the silicone.
• Correct voltage supplied by MB.
• Thickness of the TIM layer.
• Flatness of the IHS by itself.
• Flatness of the cooler contact plate.
• Combination of IHS and CPU mount, which may result in the warping problem and cause too thick layer of thermal paste in some areas.
• Improper cooler mounting.

It is a lot of stuff that can go wrong.

BTW the CPU was power limited to draw "only" 300W. The situation may have been different if I let it loose and draw what is could (340W).

Which model AIO do you have?

 

[IEC]

Other then the 13900 65W CPU, everything else is either marked up or out of stock! Still waiting for the 13600 to be release and will pay whatever to buy it!

You realize the 13600 non-K is based on Alder Lake silicon and not Raptor Lake silicon, right?

i5 13600 non-K ARK
i5 13600K ARK

The fact that no other posters have pointed this out is also a travesty.

Notice the total L2 cache of 11.5MB vs 20MB and the official DDR5-4800 support vs DDR5-5600... clearly indicating it is based on older Alder Lake silicon.

 

[Kocicak]

Which model AIO do you have?

Just an ordinary 240 Arctic liquid freezer. BTW the room was cca 16-18°C. So in 21°C room you could see higher temperature.

 

[Just Benching]

That's very impressive, are you delidding or anything? I haven't seen anyone be able to sustain that amount of power on a AIO before. Which model do you have?

My u12a can do 330-350 watts

 

[Hitman928]

My u12a can do 330-350 watts

For 30 seconds?

 

[Just Benching]

For 30 seconds?

For hours. The 13900k is pretty easy to cool.

Also, 30 seconds or 500 minutes doesn't make a difference when it comes to air cooling. They get saturated in a matter of seconds, long tests are relevant for watercooling cause it has way more thermal mass. An aircooler will hit it's max temp pretty fast

 

[Hitman928]

Just an ordinary 240 Arctic liquid freezer. BTW the room was cca 16-18°C. So in 21°C room you could see higher temperature.

That makes more sense, that's a pretty cold ambient. Most room temps are going to be significantly warmer.

 

[Hitman928]

For hours. The 13900k is pretty easy to cool.

Also, 30 seconds or 500 minutes doesn't make a difference when it comes to air cooling. They get saturated in a matter of seconds, long tests are relevant for watercooling cause it has way more thermal mass. An aircooler will hit it's max temp pretty fast

I have strong doubts you can accomplish well over 300W cooling with that cooler.

Tower heatsinks with heatpipes saturate a lot faster than water, but it's not instantaneous. If you are maxing out CPU temp within a couple of seconds with your cooler, then there's no way you're not having significant throttling.

 

[Just Benching]

I have strong doubts you can accomplish well over 300W cooling with that cooler.

Tower heatsinks with heatpipes saturate a lot faster than water, but it's not instantaneous. If you are maxing out CPU temp within a couple of seconds with your cooler, then there's no way you're not having significant throttling.

View attachment 75503

I don't know, haven't tried running hour long 350 watt workloads - cause obviously it's pointless. If you wanna some specific workload at over 300watts I can run it and post proof.

 

[Hitman928]

I don't know, haven't tried running hour long 350 watt workloads - cause obviously it's pointless. If you wanna some specific workload at over 300watts I can run it and post proof.

Blender?

Edit: And no pumping in winter air to get temps down

 

[Just Benching]

Blender?

Edit: And no pumping in winter air to get temps down

Nope, my ambient is 21-22