Intel Core i5-11400 Runs Without a Cooler Under Thermal Camera

[AleksandarK]

Famous chip photographer Fritzchens Fritz has always surprised us with some awesome die shots of the latest processors. Today, he has prepared another interesting surprise for all technology enthusiasts. Mr. Fritz has managed to run Intel's Core i5-11400 "Rocket Lake" processor without any type of colling solution, and use a thermal camera to capture what is happening inside the silicon. As the Rocket Lake design is impossible to run at any low-power setting, the author has made some changes to get a sustained run from the CPU. For starters, he set the operating clock speed to the constant 800 MHz, with iGPU, AVX, and HyperThreading disabled. The VCCSA was offset by -0.200 mV and the memory speed was lowered to DDR4-1333 speed.

The results? Well, the CPU has managed to run some tests without a cooler, and the thermal camera shows us just how the CPU works. As a CPU core gets in use, a thermal camera picks it up and we can see a core sort of spiking. Its temperature increases and it becomes distinctive from the rest of the die. After some time, the CPU became unusable, which is to be expected given that Rocket Lake's power-hungry design managed to survive quite a long time without any sufficient cooling.

You can check out the YouTube video below and see the magic happen.

View at TechPowerUp Main Site

 

[Fouquin]

Going by the CB R15 number alone; 800MHz RKL with every possible performance option dragged to the floor is as fast as a 3.2GHz Athlon 64 from 2007. When he turned AVX+HT back on that went up to being on par with 3GHz Sandy-Bridge; or 4.2GHz Conroe (65nm Core 2) and Phenom II. We've made a lot of progress even if it doesn't feel like it because of the incremental improvements.

 

[Haile Selassie]

Of course it runs. Every modern CPU with integrated heat spreader can run without a heat sink fairly easily.

IHS prevents chip from cracking due to thermal dilatation prematurely but other than that there is no mechanical constraint.

 

[Gmr_Chick]

I watched that video and thought it was really neat to see the instructions bouncing around from core to core.

 

[Lianna]

Comparing CB R15 with
https://www.techpowerup.com/268501/amd-renoir-processor-run-without-any-cooling

i5-11400: ST 106, MT 116 // destroyed
R3 4300U: ST 124, MT 327 // stable

Heatspreader area is probably not that different, but maybe in a laptop it's more likely to throttle safely.
Would be great to compare with Tiger Lake U or Ice Lake U...

 

[Vayra86]

I ran a Haswell Pentium without cooler. 80C+ but did just fine in my HTPC until I discovered the fan cable wasn't connected and push pins were halfway down.

Heck I ran Borderlands on it. It throttled a bit and then I found out

 

[AusWolf]

It's interesting that HWinfo reported 100 C temperatures across the cores, while the thermal camera barely showed 70-75 at the IHS. Thermal conductivity between the die and IHS must be pretty low.

 

[Wirko]

I'm surprised to see how bad the IHS is at spreading heat - it's easy to not just recognize individual cores but also see each one's shape quite accurately.

Also, why does the area surrounding the die seem hotter than the die itself? The effect is the same on chips that have no IHS at all, for example the Ryzen 4300U.

 

[lexluthermiester]

We've made a lot of progress even if it doesn't feel like it because of the incremental improvements.

I'll agree with this. Modern CPU's are incredibly advanced even when compared to just 5 years ago.

 

[InVasMani]

To be fair it had a heatspreader so it just had a very terrible and low profile heatsink. I wonder what would happen if the test was ran agains after a delid with liquid metal applied!? Just how much would the thermal image change. It seems rather pointless, but with the thermal image it does give a good idea of where the most heat concentration on the chip is.

 

[AusWolf]

I'll agree with this. Modern CPU's are incredibly advanced even when compared to just 5 years ago.

Absolutely. It was never apparent with the 5-10% performance uplifts between generations, but when I ran Cinebench on the 3rd gen Core i5 in my HTPC just to compare it to my Ryzen 3 3100... oh my God!

People always complain about the current gen not bringing enough improvement compared to the previous one. They just forget about the fact that you aren't meant to upgrade with every generation nowadays.

 

[TheinsanegamerN]

I'm surprised to see how bad the IHS is at spreading heat - it's easy to not just recognize individual cores but also see each one's shape quite accurately.

Also, why does the area surrounding the die seem hotter than the die itself? The effect is the same on chips that have no IHS at all, for example the Ryzen 4300U.

Well intel paste is not known for its efficiency after all.

To be fair it had a heatspreader so it just had a very terrible and low profile heatsink. I wonder what would happen if the test was ran agains after a delid with liquid metal applied!? Just how much would the thermal image change. It seems rather pointless, but with the thermal image it does give a good idea of where the most heat concentration on the chip is.

See I'm curious what this chip could do with one of those chipset heatsinks from frozen CPU on it. Give it some (barely) actual fins and see how much higher it can crank. For science.

 

[The red spirit]

Going by the CB R15 number alone; 800MHz RKL with every possible performance option dragged to the floor is as fast as a 3.2GHz Athlon 64 from 2007. When he turned AVX+HT back on that went up to being on par with 3GHz Sandy-Bridge; or 4.2GHz Conroe (65nm Core 2) and Phenom II. We've made a lot of progress even if it doesn't feel like it because of the incremental improvements.

You are wrong. It wasn't at 800Mhz, it was left to run with single core up to 2.6GHz base clock speed and then reached 67 points. With 800MHz it would be much slower than it was and likely beaten by Athlon 64.

 

[TheDeeGee]

It's interesting that HWinfo reported 100 C temperatures across the cores, while the thermal camera barely showed 70-75 at the IHS. Thermal conductivity between the die and IHS must be pretty low.

We need delided CPUs again, with a raised area around the Die to prevent coldplate wobble.

Just like what you see with Nvidia GPUs these days, why wouldn't that be possible for a CPU?

 

[AusWolf]

We need delided CPUs again, with a raised area around the Die to prevent coldplate wobble.

Just like what you see with Nvidia GPUs these days, why wouldn't that be possible for a CPU?

You mean like this?

NVIDIA GeForce RTX 3090 Specs

NVIDIA GA102, 1695 MHz, 10496 Cores, 328 TMUs, 112 ROPs, 24576 MB GDDR6X, 1219 MHz, 384 bit

www.techpowerup.com

Graphics cards come with coolers attached, which makes the warranty process a lot simpler in most cases. With CPUs, there's a million options for cooling, and you basically need an aftermarket cooler in most cases anyway, so the chance of cornering a die would be much higher with no IHS on it.

Or something else, I'm just guessing. It would sure be nice to see, though.

 

[Darmok N Jalad]

I'm surprised to see how bad the IHS is at spreading heat - it's easy to not just recognize individual cores but also see each one's shape quite accurately.

Also, why does the area surrounding the die seem hotter than the die itself? The effect is the same on chips that have no IHS at all, for example the Ryzen 4300U.

Maybe because the cores can power gate, but the area surrounding the cores cannot? They are likely what is feeding the voltage to the cores. Just a wild guess.

 

[iO]

No IHS, the CPU was obviously delidded.

 

[InVasMani]

No IHS, the CPU was obviously delidded.

It's much easier to see in the video than the screenshot with the infrared image. Seems they were spraying something non conductive on it in video periodically too as well.

 

[AusWolf]

Another thing that's interesting to note is the presence of disabled cores. Intel has previously used smaller dies with fewer cores for lower-end parts, but with this generation, every SKU seems to get the full die only with parts disabled. Yields must be lower than they used to be.

 

[Wirko]

No IHS, the CPU was obviously delidded.

In this case, I shall correct my previous post - the video is very unsharp. Maybe it's because of the Sony camera. Fritz's video of Ryzen 4300U was taken with a dedicated infrared camera and is much sharper.

Seems they were spraying something non conductive on it in video periodically too as well.

Canned air?

Maybe because the cores can power gate, but the area surrounding the cores cannot? They are likely what is feeding the voltage to the cores. Just a wild guess.

Hmm, no. The die is not that big, so the outer area that is at 60 - 75°C most of the time must be the substrate.

Here's my wild guess: it's the copper wires and power planes on the various layers of the substrate that are feeding large currents to the entire chip. They don't dissipate very much power and that power isn't concentrated in hot spots, but these conductors aren't cooled either, so they can reach quite high temperatures.

 

[InVasMani]

Does canned air show up on infrared!? I mean that's fascinating if it does, but it looked like a liquid. That's the last thing that would've of crossed my mind and why would they spray canned air?

 

[Caring1]

Does canned air show up on infrared!? I mean that's fascinating if it does, but it looked like a liquid. That's the last thing that would've of crossed my mind and why would they spray canned air?

Odd question as it wasn't an infra red camera, it was a thermal camera that detects heat.
Canned air would show up in that case as it's cooler and would lower the temperature.

 

[Wirko]

Another thing that's interesting to note is the presence of disabled cores. Intel has previously used smaller dies with fewer cores for lower-end parts, but with this generation, every SKU seems to get the full die only with parts disabled. Yields must be lower than they used to be.

That's not new or unexpected, Intel always had a small number of silicon die variants in each generation but these were used in CPUs with 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, you name it ... cores.

Edit: Even the i5-10600K/KF is built on a 10-core die.

 

[Paganstomp]

for a minute there I thought they were going to be talking about Famous Amos chocolate chip cookies.

 

[purecain]

For some time I thought that all of the cores in a multi threaded CPU were accessible at the same time.

After being a PC Pro for 15yrs it was studying data science when I really gained a good understanding of how a CPU works and how the multiple cores are accessed.

This is what gives GPU's the advantage with certain tasks as the same instructions can be run in parallel (as we all know), where as a CPU only accesses one core at a time... all be it very quickly.

It still has to use the cores individually. It was fascinating seeing each of the cores being used and the speed each core was accessed after the one before. Brilliant video!!!