i9 14900KS - 320W TDP?

[freakdiablo]

Came across this earlier. I9, 14900KS, tops out at 6.2 Gigstees but at a TDP of 300W+?

It's the P4 days again, where Intel's just throwing power at the problem, isn't it?

 

[freeagent]

If I throw more power at my 5900X I get more performance.. it can do 260w with PBO, so 300w+ is not at all surprising.

Edit:

I have seen over 300w in R23.. that was a big oops.

 

[freakdiablo]

Yeah, but this is out of the box.

 

[freeagent]

Remove the limits on the KS and its 400w+..

 

[spiritedandy]

 

[mackerel]

TDP is not power consumption! The CPU has a TDP of 150W. What this means is if you have a cooler rated at 150W connected to it, you are guaranteed by Intel to get at least the base clocks. The turbo power limits recommended by Intel are 253W and 320W. I don't know if it is still their policy but they didn't consider power limit adjustment by itself to be overclocking.

AMD do a similar thing, if not to such an extreme level. The 7800X3D for example has a 120W TDP but the actual stock power limit is 162W PPT. They do keep the PPT much closer to TDP, typically 1.35x. I don't know AMDs stance on adjusting PPT but on the DDR4 era AMD systems I had, to alter it you'd have to go into the overclocking menu area in bios after getting the overclocking warning.

 

[freeagent]

105w on the box = 143w PPT afaik.. but I don't go by that, I go by what they give me.. because who runs stock

 

[Woomack]

TDP is not power consumption! The CPU has a TDP of 150W. What this means is if you have a cooler rated at 150W connected to it, you are guaranteed by Intel to get at least the base clocks. The turbo power limits recommended by Intel are 253W and 320W. I don't know if it is still their policy but they didn't consider power limit adjustment by itself to be overclocking.

AMD do a similar thing, if not to such an extreme level. The 7800X3D for example has a 120W TDP but the actual stock power limit is 162W PPT. They do keep the PPT much closer to TDP, typically 1.35x. I don't know AMDs stance on adjusting PPT but on the DDR4 era AMD systems I had, to alter it you'd have to go into the overclocking menu area in bios after getting the overclocking warning.

TDP can be translated as a cooler's minimum requirement to run the CPU without thermal throttling. However, for Intel, it's not much above a base clock. For AMD, it's often a declared boost clock (it depends on the CPU). Some years ago, it was closer to the declared specs for Intel. On the latest gen CPUs, you can use 300W+ TDP coolers and see that CPUs with 125W TDP / 253W max turbo, throttle under high load.
It's great that 14900KS is out ... now good luck to anyone finding a cooler that lets it OC without throttling or run higher than a much cheaper 14900K.

X3D chips are different. The 7800X3D has a much higher TDP than real power draw. It seems like AMD designed them to keep the boost clock all the time, and it actually works this way. My 7800X3D runs on even small coolers at about ~88W max under full load, keeping 5.05GHz on all cores. On X3D chips, you can unlock the power limit, and it does nothing. On the other hand, the Ryzen 7600 (non-X) has 65W TDP and ~85W power draw under load.

 

[mackerel]

For AMD, it's often a declared boost clock (it depends on the CPU).

Got a reference to that?

On the latest gen CPUs, you can use 300W+ TDP coolers and see that CPUs with 125W TDP / 253W max turbo, throttle under high load.

It depends on how you interpret throttling, since that is intentional behaviour. If you have a 253W short term power limit for example, it is normal and expected to see a power limit at that value. Likewise a long term limit to 125W. If you don't want to be power limited, you have to remove the power limits. But then you can still be limited due to thermal or current for example. There will be limits somewhere.

X3D chips are different. The 7800X3D has a much higher TDP than real power draw. It seems like AMD designed them to keep the boost clock all the time, and it actually works this way. My 7800X3D runs on even small coolers at about ~88W max under full load, keeping 5.05GHz on all cores. On X3D chips, you can unlock the power limit, and it does nothing. On the other hand, the Ryzen 7600 (non-X) has 65W TDP and ~85W power draw under load.

X3D is a special case since they have to be intentionally limited out of the factory to avoid cooking the cache. Look at the non-3D equivalents to see how much they're limited. 7600 running ~85W sounds like it is on the PPT limiter? If PPT limit is increased does it go higher with sufficient load? (assuming other limiters don't activate)

 

[freeagent]

My 65w 5600X can do 135w, my 58X3D can do 143w, and my 5900X can do 260w, all with PBO and a memory tune.

 

[Kenrou]

"My New Water Block - Time for 14900KS Direct Die"

0:00 Intro
0:27 Determining the base values
2:16 Benchmarks & BIOS for OC
3:51 14900KS delidden
5:44 Removing the indium residue
8:22 Mycro Direct Die water cooler
10:11 Direct Die test values
11:36 OC with Direct Die
12:19 Summary/Conclusion
13:10 Outro

 

[freeagent]

I would still hit it if I had the cash right now. Those big numbers don't scare me because I have not had to deal with them on such a small node

 

[Flamethrower1972]

Came across this earlier. I9, 14900KS, tops out at 6.2 Gigstees but at a TDP of 300W+?

It's the P4 days again, where Intel's just throwing power at the problem, isn't it?

300 Watts is out of this world!!! And if it 100c it will throttle like crazy???

 

[freeagent]

300 Watts is out of this world!!! And if it 100c it will throttle like crazy???

We were doing 300w back on 32nm with an air cooler, 300w on 10nm would suck a little

260w on 7nm sucks

 

[Kenrou]

300 Watts is out of this world!!! And if it 100c it will throttle like crazy???

Peak ~395w on Der8auer's 1st video

 

[Woomack]

Got a reference to that?

I see it in my tests on multiple CPUs and motherboards. Since the declared clock isn't a static value, you have to test it on a specific test platform to find out how it acts. When you run 13700k/14700k, it will show around 3.5-4GHz at an extended high load because it will be limited to 125W. When you run 7800X3D in the same workloads, it will show 5.05GHz all the time and below the power limit (mentioned 88W instead of 120W)... as long as the CPU temp is ~85°C. 7950X or higher X3D will randomly boost 1-2 cores up to their max clock in extended high load (out of the extended power limit window).

It depends on how you interpret throttling, since that is intentional behaviour. If you have a 253W short term power limit for example, it is normal and expected to see a power limit at that value. Likewise a long term limit to 125W. If you don't want to be power limited, you have to remove the power limits. But then you can still be limited due to thermal or current for example. There will be limits somewhere.

Intentional is a balance within a specified frequency window, so between a base clock and turbo/boost clock. Not below the base clock. Throttling is running at a speed below the declared base clock (or other value specified as default). In short, if you have a CPU like 14700K and it drops under 3.4GHz on even a single core, then it's throttling.
If you run diagnostic software like AIDA64 or hwinfo64, it will show you throttling, as it's a CPU-programmed state. I've never seen throttling on new AMD CPUs during regular work. They lower the frequency, but they are still much above the base clock. Intel CPUs show throttling quite often, even in extended mixed load tests.

X3D is a special case since they have to be intentionally limited out of the factory to avoid cooking the cache. Look at the non-3D equivalents to see how much they're limited. 7600 running ~85W sounds like it is on the PPT limiter? If PPT limit is increased does it go higher with sufficient load? (assuming other limiters don't activate)

The 7600 overclocked to 5.3GHz still has 90W, while the 7600 with everything unlocked keeps ~110W after OC. To make it run at a higher wattage, you have to set a higher voltage or lock the voltage at a higher value, but this doesn't help with overclocking, so it's pointless. At stock, it doesn't really matter if you unlock power limiters or not. It will show +/-5W under load.
7950X, as long as you can keep low enough temps, will boost above the specified values at ~170-180W. If you run extended high-load tests, then it will keep 140-150W and ~5.0-5.1GHz. If you do the same on any competitive Intel, they will drop to ~4GHz ~125-150W.

This is why I repeat that Intel added E-cores to compete with AMD in multithreaded tests, as without lower power cores, they can't really beat AMD in any rendering test or anything else that uses all the CPU cores. I guess it's easy to imagine what would be if Intel used 16-24 regular cores from 13/14th gen. Somehow AMD does that.
In games, they stick with 2 cores up to the max boost as numbers sell, and most games are still using only a few cores.

X3D is only frequency-locked, but it overclocks with bclk, and it's not a problem to push it for the next 300MHz+ if you have a motherboard with a separate bclk for the CPU. The only problem can be too high SOC voltage, but everything else can be set above the specs. SOC voltage was locked at the BIOS level a long time ago, and somehow then people noticed it's not really required above 1.25V.

 

[mackerel]

I see it in my tests on multiple CPUs and motherboards.

I guess what I'm after is AMD's definition of what base clock means to them. All I managed to find searching myself is GN ranting about how meaningless TDP is.

 

[EarthDog]

Anand ranted the same. It's the most detailed I've seen on it... and not sure what, if anything has changed.

AMD Zen 3 Ryzen Deep Dive Review: 5950X, 5900X, 5800X and 5600X Tested

www.anandtech.com

Really miss Ian driving that part of the ship...

Reaching for Turbo: Aligning Perception with AMD’s Frequency Metrics

www.anandtech.com