News Intel Core i9-12900KS Review: The Fastest Gaming Chip Ever

[gg83]

i9-12900KS + RTX4090 - I feel sorry for the Koalas...

Hahaha! Poor little guys. We're gunna destroy the planet with our current power increase trajectory. 5k watt power supplies will be the norm in three generations.

 

[jacob249358]

Hahaha! Poor little guys. We're gunna destroy the planet with our current power increase trajectory. 5k watt power supplies will be the norm in three generations.

People will need a solar pannel array or a new house to run their pc

 

[Specter0420]

Can we get some specific VR titles tested? There isn't much in the gaming world that can tax single-core performance more than flight simulators like DCS and MSFS in modern VR. Especially when you consider ASW/motion smoothing doesn't work at the distances used in flight sims, it just creates artifacts, so most of us really NEED 90+ FPS minimum.

I see you have a chart labeled "VRMark" but it doesn't have any VR scores included...

 

[NP]

Sorry to say, but this graph is totally useless:

https://imgur.com/x2V8UFZ

It only shows that when the CPU throttles, it throttles. You need a more beefy cooler, not just benchmark the most power hungry processor at static 100C.

 

[TerryLaze]

Sorry to say, but this graph is totally useless:

https://imgur.com/x2V8UFZ

It only shows that when the CPU throttles, it throttles. You need a more beefy cooler, not just benchmark the most power hungry processor at static 100C.

You keep using that word -throttling- I don't think it means what you think it does.
This here is the base performance, it's about 4Ghz all core sustained, what you show is above that, by more than 500Mhz, it's more so it's not throttling, it is showing you the maximum possible with auto turbo default (mobo default, not cpu default) settings and that cooler.

 

[RodroX]

Thats an impresive performance. Then again as the article said, for gaming you need a high level perofroamcne / price GPU to see that small 2.5% gain.

But its always good to see CPU makers fighting each other, thats the only way we consumers get better products at better prices.

Cheers

 

[NP]

You keep using that word -throttling- I don't think it means what you think it does.
This here is the base performance, it's about 4Ghz all core sustained, what you show is above that, by more than 500Mhz, it's more so it's not throttling, it is showing you the maximum possible with auto turbo default (mobo default, not cpu default) settings and that cooler.

Thermal throttling = Cpu not allowing itself to go above specific clocks because it meets a thermal limit. That is what I am talking about. The clearest indicator of that taking place is solid line of clocks at even 100C temp line (or 90C for 5900x).

In order to test performance properly, you need a cooler that can manage that heat so that it is not solid 100C. I have one, Arctic Freezer 420, for which I paid 120euro. So not inhibitively expensive or exotic (like compressors or LN2) by any means.

As it stands, the above graph (NOT the one you linked) illustrates the performance limits of that cooler. Not the performance limits of the cpu.

That is the problem.

 

[TerryLaze]

As it stands, the above graph (NOT the one you linked) illustrates the performance limits of that cooler. Not the performance limits of the cpu.

Yes. That's the point.
They can't show you a super detailed overclocking guide with performance numbers as a normal review.
A normal review is to show what a normal system can do, which means that someone will stick a cooler in it and run it however it will run out of the box and the chart you showed shows exactly that.

 

[NP]

Yes. That's the point.
They can't show you a super detailed overclocking guide with performance numbers as a normal review.
A normal review is to show what a normal system can do, which means that someone will stick a cooler in it and run it however it will run out of the box and the chart you showed shows exactly that.

Nah. A normal review is the kind of review where the test system is built with the requirements of the hardware to be tested in mind. This test bench is not able to benchmark the cpu properly, because it has too weak cooler. A normally priced, but bigger water cooler is needed even for "normal review" of this cpu.

Especially considering the high price premium, this cpu is certainly not intended to be slapped on a mobo with a random cooler and ran at stock settings. If that is the plan, there are literally zero reasons for buying this instead of K or KF (not that there are many reasons to buy this cpu even with a beefy cooler, but that is another thing to discuss).

 

[Phaaze88]

View: https://www.youtube.com/watch?v=oW0aWfxzhyc


So, KS is for someone who wants the best possible performance out of the box without having to even look at OC'ing.
A good K cpu can perform as well or better and yield less power with a little manual tuning, but one will have to hunt for it, or get lucky, as there is a greater swing in silicon quality. KS needs no extra help, aside from some undervolting, and is more consistent for quality.

As someone who's done with OC'ing, KS looks attractive... but I don't need an i9.
Could we get some KS i7s and i5s, Intel?

 

[-Fran-]

Yes. That's the point.
They can't show you a super detailed overclocking guide with performance numbers as a normal review.
A normal review is to show what a normal system can do, which means that someone will stick a cooler in it and run it however it will run out of the box and the chart you showed shows exactly that.

Not to say you're completely wrong, but when you have the CPU pegged at 100°C in a power torture test, you're clearly under-cooling the CPU as it can go higher. The messages have been mixed in your discussion with the other person, so I think you two were arguing different things.

As for my take: the mere fact that, with a proper AIO, this CPU can reach 100°C is bonkers. Never mind the torture test itself, but just getting that high of temperature is just madness. That test tells you something very simple: you can't use 100% of what you're paying for, unless you invest in a very expensive cooling solution. Which, fair enough, but it seems that even with more expensive cooling this CPU still goes up to 100°C rather effortlessly, which is really worrying.

Regards.

 

[TerryLaze]

Not to say you're completely wrong, but when you have the CPU pegged at 100°C in a power torture test, you're clearly under-cooling the CPU as it can go higher. The messages have been mixed in your discussion with the other person, so I think you two were arguing different things.

The thing is that the CPU will always reach 100°C , they touch on that in the article, it's dynamic turbo, you set a power and a temp limit if you want to but if they are not set the CPU will go to 100° unless you way over-cool the cpu.

https://www.intel.com/content/www/us/en/support/articles/000059372/processors.html

Intel® Adaptive Boost Technology (Intel® ABT) is a new feature with 11th Gen Intel® Core™ i9-11900K and i9-11900KF processors that improves gaming performance by opportunistically allowing higher multi-core turbo frequencies. Intel® ABT operates within system power and temperature specifications when current, power, and thermal headroom exist.

"when ... ... thermal headroom exist" = as long as temp not equal to 100° keep pushing higher.

As for my take: the mere fact that, with a proper AIO, this CPU can reach 100°C is bonkers. Never mind the torture test itself, but just getting that high of temperature is just madness. That test tells you something very simple: you can't use 100% of what you're paying for, unless you invest in a very expensive cooling solution. Which, fair enough, but it seems that even with more expensive cooling this CPU still goes up to 100°C rather effortlessly, which is really worrying.

What you are paying for is this, this is what you get within warranty, anything above that is the general risk you take with overclocking on any CPU of any maker, you might get more than others or less.

 

[TJ Hooker]

Sorry to say, but this graph is totally useless:

https://imgur.com/x2V8UFZ

It only shows that when the CPU throttles, it throttles. You need a more beefy cooler, not just benchmark the most power hungry processor at static 100C.

In order to test performance properly, you need a cooler that can manage that heat so that it is not solid 100C. I have one, Arctic Freezer 420, for which I paid 120euro. So not inhibitively expensive or exotic (like compressors or LN2) by any means.

If you click the left/right arrows on the power draw figures, they do have another chart with a 720mm custom loop. The CPU still hits 100C. So no, your AF 420 wouldn't keep it from hitting 100C, and yes, it would require prohibitively expensive/exotic cooling like chiller or LN2. At least when running all core stress tests. I'm sure gaming power consumption, and presumably temps, are much lower.

 

[gruffi]

I'm sure gaming power consumption, and presumably temps, are much lower.

No. It's much more power hungry than 12900K in games and not even close to 5800X3D's gaming efficiency. 12900KS is and remains an absolute joke.

 

[TerryLaze]

No. It's much more power hungry than 12900K in games and not even close to 5800X3D's gaming efficiency. 12900KS is and remains an absolute joke.

Link to the source, the table alone doesn't say much, you can over watt the 12900ks to hell without getting any better performance out of it just to make it look worse than it is.

Also even if this table is legit with good settings, power is still 50-100W lower even in the most extreme cases than what he was talking about so he is still right in what he said, it's still going to be lower and cooler.

 

[KyaraM]

View: https://www.youtube.com/watch?v=oW0aWfxzhyc


So, KS is for someone who wants the best possible performance out of the box without having to even look at OC'ing.
A good K cpu can perform as well or better and yield less power with a little manual tuning, but one will have to hunt for it, or get lucky, as there is a greater swing in silicon quality. KS needs no extra help, aside from some undervolting, and is more consistent for quality.

As someone who's done with OC'ing, KS looks attractive... but I don't need an i9.
Could we get some KS i7s and i5s, Intel?

To be quite honest... a 12700K should be enough for pretty much everything, especially gaming, considering it's only slightly behind the i9 in that regard. And especially especially when gaming in anything above 1080p. At least I'm pretty sure I won't notice a difference if I switched out the CPU only in my 1440p rig

Peoductivity is a different matter, of course. But it's no slouch there, either, so...

 

[gruffi]

Link to the source, the table alone doesn't say much, you can over watt the 12900ks to hell without getting any better performance out of it just to make it look worse than it is.

Source is this. Stop defending this piece of crap. It's ridiculous. 12900KS is a useless space heater. Overpriced, underperforming, inefficient. Period. There is absolutely no reason to get a 12900KS if there is a 12900K. And even the latter one is no good choice for gamers.

 

[KyaraM]

Source is this. Stop defending this piece of crap. It's ridiculous. 12900KS is a useless space heater. Overpriced, underperforming, inefficient. Period. There is absolutely no reason to get a 12900KS if there is a 12900K. And even the latter one is no good choice for gamers.

That's nothing new. Same can be said about AMD, though...

 

[gruffi]

That's nothing new. Same can be said about AMD, though...

No, cannot be said. I don't see any current AMD processor that uses more than 140-150W under full load at stock settings. 12900KS uses up to twice as much. 12900KS is an inefficient piece of crap. A desperate try to counter the 5800X3D gaming beast. Another Intel fail. Gelsinger doesn't seem to be any better than other Intel CEOs. Winning benchmarks on all costs instead of making reasonable products.

 

[jacob249358]

No, cannot be said. I don't see any current AMD processor that uses more than 140-150W under full load at stock settings. 12900KS uses up to twice as much. 12900KS is an inefficient piece of crap. A desperate try to counter the 5800X3D gaming beast. Another Intel fail. Gelsinger doesn't seem to be any better than other Intel CEOs. Winning benchmarks on all costs instead of making reasonable products.

AMD dominates efficiency no doubt about it. Yep, 12900ks was just so intel could hold the marketing scheme of the fastest gaming processor.