Intel 10-Core Comet Lake-S CPU Could Suck Up To 300W

[TrigrH]

1 hour ago, porina said:

For context, my 8086k (6 core) running Prime95 6x128k FFT with uncapped power took about 120W. Scaling that to 10 cores (assuming same clocks) would give 200W. The all core clock is 4.3 GHz. If Intel have increased the all core boost even at 10 cores, that could result in the higher claimed consumption.

My 8086k easily does 200w+ Maximuxs X Apex running full phase mode, max current, 1.4v, 5.2ghz full avx. UPS system load at idle: 230w  

[porina]

1 hour ago, thorhammerz said:

Curious as to how much power your particular sample uses on a 4.9 GHz all-core OC? That would paint a better apples-to-apples comparison of per-core-scaling power consumption.

Yes and no. Overclockers will generally aim to find the lowest voltage for a given clock that appears stable. This is very different from Intel setting a higher voltage so it'll be stable under a wider set of conditions.

 

I missed it earlier, but OP shows the new CPU to run 4.7 GHz all core, which is higher than the 4.3 GHz all core of the 8086k. That will also be another good chunk of power used.

5 minutes ago, TrigrH said:

My 8086k easily does 200w+ Maximuxs X Apex running full phase mode, max current, 1.4v, 5.2ghz full avx. UPS system load at idle: 230w  

Overclocked anything goes... my sample will also do 5.2 non-AVX but I never Prime95 (AVX) tested it.

[Bombastinator]

3 hours ago, Harry P. Ness said:

https://www.tomshardware.com/news/intels-new-desktop-processor-draws-too-much-power

AMD cpu suck up to 300watts

 

Intel cpu suck up 300watts

 

300watts power consumption on 14nm comes with a feature to cook your egg like GTX 480 did in the past. But wait, our cpu can cook faster than GTX 480 up to 25%! In addition, our 10 core cpu can be a portable heater too. 

300w is outside the air cooling performance envelope.  It’s going to need water for absolute top performance.  Big 6 pipe air coolers can hit 200w no big deal and some can do 250.  300w is out of range though.  Maybe put them on larger silicon even though they don’t need it just to get better heat transfer?  Something needs to be done.

[The Sloth]

Time for a Canadian winter build, GTX 480 paired with 10-Core Comet Lake-S CPU. 

[thorhammerz]

54 minutes ago, porina said:

I missed it earlier, but OP shows the new CPU to run 4.7 GHz all core, which is higher than the 4.3 GHz all core of the 8086k. That will also be another good chunk of power used.

I misread the article on my first skim as well, but 4.7 GHz should be the 9900K, whereas the new 10900K "supposedly" runs at 4.9 GHz all-core.

[mr moose]

2 hours ago, Tech_Dreamer said:

 

 

Thought i'd never say this about Intel my whole life, But stop selling 2nd grade cut down product when your competition is close to 2 steps ahead in performance.

That's what we've been saying about AMD for so long now.  It actually feels nice to see performance swing start to turn. 

 

With that said, people still need to understand that these things aren't exactly directly comparable,  Intel can boost all cores to a higher speed but still fall down in some benchmarks while  AMD can only boost X number of cores and still consume the same power.  Etc etc etc.  Unless we are careful how we compare,  in many ways we are only fooling ourselves.

[JustWantTech]

AMD 7nm => 300 watts

Intel 14nm => 300 watts

[schwellmo92]

5 hours ago, porina said:

AMD behaviour is different, in that by default the PPT limit is honoured by the mobo by default. While there isn't a 10 core CPU to compare against, it's more than the 8 a single CCD supports so let's assume for now the nearest to compare is a 3900X. Although it is 105W TDP, the actual PPT limit is 142W. By default, it wont take more than 142W. If you enable PBO, that removes the power limit and is more comparable in operating state to Intel's. I have no idea what power you can draw under that condition. My attempts at testing using the 6 core 3600 were thermally limited since the thermal density is higher than on Intel, so I couldn't push the power limiter before the thermal limiter kicked in. There is also another difference to note here, in that I believe PBO is considered by AMD to be an overclock. Please correct if necessary. As another data point, when I was overclocking a 1700 (8 cores) it was drawing 180W at 4.0 GHz. Higher powers are not surprising if you push the CPUs enough.

Even with PBO enabled my 3600 doesn’t go past the stock PPT limit lol (88w).

[schwellmo92]

5 minutes ago, huilun02 said:

How many cores would a 300W AMD part have from the latest Zen iteration?

A Ryzen 9 3950X (16c/32t) at 1.4v under full load draws 270w~ for comparison. That is with it pushed to the maximum, which I guess is what Intel is doing. If you look at the Threadripper 3990X (64c/128t) at stock configuration that only draws 300w~.

[DildorTheDecent]

My 5820K at 4.9GHz would pull 400W during Cinebench and FS Physics.

 

10 cores at those clocks pulling 300W (worst case scenario I think) isn't too bad at all. Motherboards that's don't suck could do it at a slightly lower power draw too.

 

[sazrocks]

6 hours ago, porina said:

.AMD behaviour is different, in that by default the PPT limit is honoured by the mobo by default. While there isn't a 10 core CPU to compare against, it's more than the 8 a single CCD supports so let's assume for now the nearest to compare is a 3900X. Although it is 105W TDP, the actual PPT limit is 142W. By default, it wont take more than 142W. If you enable PBO, that removes the power limit and is more comparable in operating state to Intel's. I have no idea what power you can draw under that condition. My attempts at testing using the 6 core 3600 were thermally limited since the thermal density is higher than on Intel, so I couldn't push the power limiter before the thermal limiter kicked in. There is also another difference to note here, in that I believe PBO is considered by AMD to be an overclock. Please correct if necessary. As another data point, when I was overclocking a 1700 (8 cores) it was drawing 180W at 4.0 GHz. Higher powers are not surprising if you push the CPUs enough.

 

I have to wonder, if a solution for this would be for mobo manufacturers to enforce a PL2 value suitable for their mobo's capacity to deliver power. PL2 is the short term boost power limit level that enthusiast boards traditionally ignore. Overclockers who want to go all the way can then pick those supporting the higher PL2 values.

My 3900x with PBO on pulls about 185w under p95 small FFT. Hits ~88C with my NH-D15 (Strix X470-F, CLU between IHS and the cooler).

6 hours ago, spartaman64 said:

idk how i feel about having to buy a noctua nh-dh15 cooler for a consumer cpu

Not even the NH-D15 would be able to handle this beast, its top rating is 250w. 300w is minimum 360 rad territory.

[Bombastinator]

What I want to know is how much power does it take when it’s run at, say 180w, which I’m randomly picking as a convenient number that I feel I can likely keep temps below 80c at with big air? (For all I know the number could be as high as 220w) and how fast is it single core compared to a similarly priced AMD offering boosted to that wattage?  I might not even go for the 10 core. 8 core multi is likely enough I think.

[SeraphicWings]

Who cares if it needs a nuclear reactor to power when it can reach aBoVe 5 JigAhErTz f0r gAmiNg

[Letgomyleghoe]

6 hours ago, Reytime said:

The last AMD cpu I used was the Athlon II X4 635 Black Edition, maybe this year I should try AMD again 

Devin?

[Letgomyleghoe]

I get 300w on a cpu

But not a 10 core CPU...

Maybe try 12+ next time intel

[Letgomyleghoe]

1 hour ago, DildorTheDecent said:

My 5820K at 4.9GHz would pull 400W during Cinebench and FS Physics.

 

10 cores at those clocks pulling 300W (worst case scenario I think) isn't too bad at all. Motherboards that's don't suck could do it at a slightly lower power draw too.

 

Holly shit, thats more than half of my whole systems power draw

[williamcll]

So what air coolers that fits the LGA 1200 can handle 300 Watts?

[porina]

8 hours ago, thorhammerz said:

I misread the article on my first skim as well, but 4.7 GHz should be the 9900K, whereas the new 10900K "supposedly" runs at 4.9 GHz all-core.

Doh! You're right, but doesn't change what I wrote earlier much...

 

6 hours ago, schwellmo92 said:

Even with PBO enabled my 3600 doesn’t go past the stock PPT limit lol (88w).

On mine Cinebench R15 was enough to take it to PPT limit of 88W. When I unlimited it, it didn't change, presumably as I was on the limit of my cooling at that point and it was also hitting the thermal limiter.

 

5 hours ago, huilun02 said:

How many cores would a 300W AMD part have from the latest Zen iteration?

Because AMD actually enforce a power limiter at stock, you can pick as many cores as you like. As a short answer, I'd go with "around 32" given the 3950X is 142W PPT so you can have at least two of those, or the 3970X is also pretty close to that ball park.

 

If Intel similarly enforced a power limit, they can still claim a high single core boost speed like AMD and have more cores running at lower power. The article claimed a PL2 value (short term boost) of 250W for example. This is more likely enforced on OEM systems, but mobos used in enthusiast self builds tend to set PL2 unlimited.

 

4 hours ago, sazrocks said:

My 3900x with PBO on pulls about 185w under p95 small FFT. Hits ~88C with my NH-D15 (Strix X470-F, CLU between IHS and the cooler).

Not even the NH-D15 would be able to handle this beast, its top rating is 250w. 300w is minimum 360 rad territory.

I can't remember where the thermal limit is on Zen 2 but you're probably limited by that as much as the power. Zen 2 runs rather hot even at its lower power.

 

4 hours ago, Bombastinator said:

What I want to know is how much power does it take when it’s run at, say 180w

It takes 180W when it is taking 180W. I guess the intent of the question is more how much clock, and therefore performance, do you lose when running at lower power. You can pick you performance vs efficiency curve depending on where you run.

 

2 hours ago, williamcll said:

So what air coolers that fits the LGA 1200 can handle 300 Watts?

The 300W value is the "can take up to" value, not usually the 24/7 operating state. If system manufacturers enforce the PL1 and PL2 values it doesn't need a 300W cooler, just one rated for the TDP. Enthusiast mobos generally default to unlimited PL2, so in the worst case, a load can take more than rated PL2. If you're building a high end system, you can pair it with high end cooling if you want the most out of it.

 

Worst case workload is probably Prime95 small FFT like workloads. This is actually an interest of mine, as I use other software which shares the same math library as Prime95. I wouldn't pick a consumer Intel system today if I was buying another one to run this workload. Zen 2 is a bit more efficient if we limit to consumer CPUs, and Cascade Lake-X will destroy Zen 2 in this type of application thanks to AVX-512 and is now much more price competitive than Skylake-X.

[justpoet]

OK, so…context.  300w sucks for this level of "consumer" part and is downright embarrassing.  The worst part though is board partners complaining about it and having problems making it stable across the motherboard lineup since they were surprised by the expected draw (not to be confused with draw when people do serious overclocking).

 

But, when you look at the HEDT market, 300w worth of CPU is really not insane.  Also, when you look at graphics cards, even something as commonly discussed as the 1080TI could peak around 300W and a 2080TI can peak around 350w.  So, really, they're just blurring the lines more between "desktop" and "HEDT", but going with more clock speed vs more cores due to die size at their process node.

 

Will it be easy to cool?  Nope.  Will it be possible though, even on air?  Well…those graphics cards exist as air cooled cards, so…yeah.

 

Note: Graphics card power draw taken from tomshardware reviews, since they were the first to pop up in a quick search for graphics card power draw.

[5x5]

I am reminded of that 2002 video of someone cooking salami on a pentium 4. Can we do that but with pasta and a pan this time?

[porina]

1 hour ago, justpoet said:

Will it be easy to cool?  Nope.  Will it be possible though, even on air?  Well…those graphics cards exist as air cooled cards, so…yeah.

See my previous reply, it's not like everyone will be running Prime95 on it 24/7. 300W isn't likely to be reached in most workloads. System builders will be running it on power limit. This is more a theoretical problem for enthusiasts.

 

1 hour ago, justpoet said:

Note: Graphics card power draw taken from tomshardware reviews, since they were the first to pop up in a quick search for graphics card power draw.

On nvidia cards you can use monitoring software to see the card power. Stock 980Ti for example can hit 250W under a suitable load, and factory OC likely will exceed that. AMD cards only report GPU power so can't be as easily observed in that way.

 

7 minutes ago, 5x5 said:

I am reminded of that 2002 video of someone cooking salami on a pentium 4. Can we do that but with pasta and a pan this time?

I'm not sure these CPUs get hot enough to cook meats, other than warming it up. It does get hot enough to make egg white "cook" though, so if you like a runny fried egg...

[0head]

Source: Guru3D

 

Intel at this point hasnt announced their new 10th gen main stream CPUs. having a 300W TDP next to AMDs offerings seems like a darn good reason why that might be.

 

Quote

While Intel did not give an official explanation, Germany based and well respected Computerbase notes that industry sources revealed Intel does not want to launch the new procs just yet because they seeing power consumption exceeding 300W, and are trying to optimize their design before launching them to the market.

 

This apparently is just a rumor based on chatter at the moment but it does seem pretty believable. The article also mentions that the 300W TDP is higher than what manufacturers have been told to aim for and could create some issues for motherboards/performance.

[Curious Pineapple]

2 minutes ago, gabrielcarvfer said:

A few more years of 14nm++++++++++++++++++++++++ before 10nm+ or 5nm.

Don't you mean 14nm######################, you need to stack the +'s together so more fit on the box

[spartaman64]

It was rumored a while back also but maybe it's more concrete now? If it's true then they not only delayed it but lied about it also ... I wonder if that violates any sec regulations

[.Apex.]

2 minutes ago, 0head said:

This apparently is just a rumor based on chatter at the moment but it does seem pretty believable. The article also mentions that the 300W TDP is higher than what manufacturers have been told to aim for and could create some issues for motherboards/performance.

It's believable since they're trying to push the CPU's to their max potential because of how far AMD is ahead, which then it could exceed 300W, the 9900K alone can pull 250W at 5Ghz, so increasing the core count by 25% is gonna get you +25% power consumption.