Intel Core i9-12900KS review
The intel core i9 12900 ks, didn't take a lot of effort on my part, nice and quick, and it went down very well with the get guru audience. So that's great news minimum work maximum return. By contrast, the full review of this processor has taken a huge amount of effort on my part because it's been a rocky ride, so sit down. What exactly is a core i9-12900KS?
The short answer is it's a slightly faster version of quarter: nine 12 900 k. If you haven't already watched or read our review of the quarry 9 i9 12900k. I suggest you head off and do so because everything in that review, along with intel, z690 chipset, motherboards and ddr5 memory, all of that applies here.
Intel has clearly binned or selected particularly fast processors, they've changed the microcode in the processors and therefore they're identified by the motherboards, and they behave slightly differently for this privilege.
You pay considerably more money and you get hopefully a slightly better product. We'Re talking marginal speed improvements. You can expect to see your ks run at 5.2 gigahertz or cause where your k, you might see, run at five gigahertz, or course we saw ours run at 4.9, gigahertz or cores intel's.
Current architecture has a variety of boosts. You have boosts on top of boosts, so your core i9 12 900 k you'd, probably see one core running at 5.2 or 5.3 gigahertz all cores 49 gigahertz, perhaps 5 gigahertz personally, i've seen quarantine, 1200k run up 4.9 all core and you get flashes of a single core at 5.
But it is momentary, by contrast, the ks. You can expect to see 5.2 gigahertz or cores and 5.5 gigahertz on two cores on paper. It'S a reasonable step forward. It begs the question: why does the ks exist in the first place thing? Is that with the 12900k intel has already got a blazing fast processor, as indeed they claimed at launch, and they were quite correct.
You pay considerably more money and you get hopefully a slightly better product. We'Re talking marginal speed improvements. You can expect to see your ks run at 5.2 gigahertz or cause where your k, you might see, run at five gigahertz, or course we saw ours run at 4.9, gigahertz or cores intel's.
Current architecture has a variety of boosts. You have boosts on top of boosts, so your core i9 12 900 k you'd, probably see one core running at 5.2 or 5.3 gigahertz all cores 49 gigahertz, perhaps 5 gigahertz personally, i've seen quarantine, 1200k run up 4.9 all core and you get flashes of a single core at 5.
But it is momentary, by contrast, the ks. You can expect to see 5.2 gigahertz or cores and 5.5 gigahertz on two cores on paper. It'S a reasonable step forward. It begs the question: why does the ks exist in the first place thing? Is that with the 12900k intel has already got a blazing fast processor, as indeed they claimed at launch, and they were quite correct.
So then we got the 12900 ks which, rather than being the fastest gaming processor, is apparently the fastest desktop processor. Well, that's a funny thing: what exactly is a desktop processor is a thread ripper a desktop processor.
If so, it's certainly got more grunt than this i mean it's got many times more cores, or are we talking about faster clock speeds? Well, if that's the case, the 12 900 k 49 gigahertz all calls it beats amd on clock speeds and those boosts to 5.2 or 5.3 certainly beat amd, and that surely means that intel running at 5.2.Gigahertz all cores on the performance cores and 5.5 gigahertz on two cores is fighting a battle.
If so, it's certainly got more grunt than this i mean it's got many times more cores, or are we talking about faster clock speeds? Well, if that's the case, the 12 900 k 49 gigahertz all calls it beats amd on clock speeds and those boosts to 5.2 or 5.3 certainly beat amd, and that surely means that intel running at 5.2.Gigahertz all cores on the performance cores and 5.5 gigahertz on two cores is fighting a battle.
They'Ve already won what kind of possibly led intel to decide. It was time to bring out a slightly juiced up version of their existing processor, we'll have to just dig into the benchmark results and see what we can find the motherboard i'm going to use for my testing of the ks. Is this gigabyte, z690 aorus master, which i reviewed a few weeks ago and liked a great deal?
The thing is that, since my review of this board, gigabyte has updated the bias as you'd expect a bias that supports the ks processor. We'Ve also had some significant updates to windows 11 and also we've had updates to graphics drivers.
So the first thing i have to do before i can get into the ks is to retest the 12900k at test. Pc gigabyte z690 auras master motherboard ssd is a separate rocket. Nvme 40 processor, core i9, 12900k thermal compound arctic mx4 memory is 32 gigabytes of corsair, ddr5, 5200 graphics card palette, gaming, pro rtx 3080 power supply seasonic, prime tx 850.
The cpu cooler is a corsair h150i elite lcd. The three 120mm fans connected directly to the motherboard and they will rapidly ramp up to 2000 rpm we'll do a quick run of cinebench r23 to remind ourselves what the choir 9 1200k does.
The cpu cooler is a corsair h150i elite lcd. The three 120mm fans connected directly to the motherboard and they will rapidly ramp up to 2000 rpm we'll do a quick run of cinebench r23 to remind ourselves what the choir 9 1200k does.
When it's under full load settles down to 4.9 gigahertz all cores package power Is 211 watts temperature is in the mid 70s and the ambient is currently 20 degrees celsius. Our chart shows the re-tested core. 912 900k has made an improvement from the original review figures.
However, while it looks like a big leap, we're actually talking about three percent so nice to have, but not quite as impressive as it might first appear it's time to plug in the 12900 ks and see how that performs on the gigabyte, z690 oris master core i9-12900 ks correct identified in the bias xmp enabled starting cinebench 23. Ambient is 22 degrees.
However, while it looks like a big leap, we're actually talking about three percent so nice to have, but not quite as impressive as it might first appear it's time to plug in the 12900 ks and see how that performs on the gigabyte, z690 oris master core i9-12900 ks correct identified in the bias xmp enabled starting cinebench 23. Ambient is 22 degrees.
It'S got a little bit warmer in here. Cpu package, power, 260 watts, clock, speed, 5 gigahertz, all cores package power, pulling back to 255 watts cpu package. Temp is 84, you can hear the fans running at full speed, so the clock, speed stabilized at five gigahertz package power just above 250 watts the score from that run of cinebench 23 - is almost identical to my official test result,.
Which is great for repeatability and adds to my disappointment. Although the ks is drawing significantly more power than the k in terms of performance in cinebench 23, there's very little to see for that extra power, i'm going to change from the gigabyte z690 oris master to the msi meg z690 unifi and see whether or not that Helps the situation it's clear.
There's something slightly peculiar going on with the 12900 ks on the gigabyte, so i need to separate motherboard from processor to establish it as the motherboard. That'S doing something odd or the processor. This meg z690 unifies the motherboard.
I used the launch review of the 12900k and it was a good, solid platform. So it's coming out of storage uh having a bias update, i'm slightly disturbed to note the msi bias to support the case is a beta bias rather than a release bias, but hey ho.
So here we go with the meg z690 unifi and beta bias, along with the 12900 ks bias, updated processor correctly recognized xmp enabled we're ready to get testing. Here we go cinebench 23 on the msi meg unifi with the ks. Okay.
That looks like a dismal failure to 40 watt cpu package power temperature straight to 100. Perhaps the problem is neither the motherboard nor what kind of the processor but not the motherboard. Perhaps the problem is the cooling system.
That looks like a dismal failure to 40 watt cpu package power temperature straight to 100. Perhaps the problem is neither the motherboard nor what kind of the processor but not the motherboard. Perhaps the problem is the cooling system.
What i'm doing is replacing the corsair 360 mil aio with a custom loop 360 mil system, cpu block from corsair radiator from ek, fans from fantex pump reservoir, also from uk, and that bracket there is simply to help it stand up to make life easy and the Fittings are from alpha cool.
So there's a good selection of the cooling industry in one happy union here question is: is this gon na tame the ks processor we're getting serious cpu package power 320 and a few watts dropping back to 316 temperatures still, however, 100 celsius, Albeit ambience 23 And i've got a bit of ping pong going on the cpu package going from 100 down to 97.
You can see the power and temperature kind of cycling as it essentially throttles and then regains its poise. We'Ve got the p core cycling between 51 gigahertz and 5.0 gigahertz the e-cores are rock-solid at 4.0 gigahertz cpu package power is just over 300 watts.
Temperatures sitting pretty much at 100, so cooling system clearly is a big part of the equation. The result of the cinebench very close once again to my official test results, which is good for consistency, and now we've got a hint of the narrative.
The new ks clearly requires loads of voltage, loads of power and then runs horribly hot, which means you have to chuck a huge amount of cooling at it. This makes sense, however, we're still not seeing speeds in line with intel's specification. I made a slight error.
The new ks clearly requires loads of voltage, loads of power and then runs horribly hot, which means you have to chuck a huge amount of cooling at it. This makes sense, however, we're still not seeing speeds in line with intel's specification. I made a slight error.
A few moments ago, when i said the cpu demands high voltage and high power, it's probably more accurate to say the motherboard manufacturers seem to be feeding it higher voltage and, as a consequence, an awful lot of power. What happens if we overclock but turn down the voltage core voltage 1.75 and then we want to set load line calibration, let's set it to mode two,
So not quite the highest, but let's just make sure it holds up so overclocking, both the p cores and the E chords are the the e causer just a sideshow and we are holding the voltage down, and here we go again with cinebench r23 cpu package power around 245 watts. All the p calls humming away at gigahertz, faster clocks and less power.
Hurrah package temperature is creeping up, however. Power and clock speeds are holding steady, overclocking, the ks or undervolting. The ks, at any rate, looks like a good move that little piece of overclocking has worked well, as you can see in our charts we're making progress with the ks.
The cinebench score is looking good. Power is under control and the cpu temperature is slightly less terrifying than it was on auto. With that success under our belts, let us bump up the p core ratio, one more time to 5.3 gigahertz, and here we go. This may well be the final run of cinebench r23.
The cinebench score is looking good. Power is under control and the cpu temperature is slightly less terrifying than it was on auto. With that success under our belts, let us bump up the p core ratio, one more time to 5.3 gigahertz, and here we go. This may well be the final run of cinebench r23.
Thank goodness package power to 50 watts, clock speeds, 5,3 gigahertz on the p cores 4.1 on the e cores. That'S all looking promising and we have an ambient 23 degrees 23.3 as it happens, and the cpu temperature is just about to hit 90 And there we have our results: cinebench r23, 5.3 gigahertz on all the p cores 41 gigahertz on the e-cores, an impressive score and all done on 250 watts of cpu package power.
I'M sure you realize i had no intention of spending nine minutes discussing the behavior of this processor running cinebench r23. My original plan, as i'm sure you will have realized, was to show you the 12900k in action switch over to the ks.
I'M sure you realize i had no intention of spending nine minutes discussing the behavior of this processor running cinebench r23. My original plan, as i'm sure you will have realized, was to show you the 12900k in action switch over to the ks.
Look at the extra performance, the higher clock speeds, and then we could get into a good long discussion about gaming and power draw and temperatures and such like, instead of which, as you saw, it went totally sideways. So here we are 10 minutes into my video and i'm not even touched on the 5.5 gigahertz behavior.
So it seems like the kind thing is to show you a little bit of b-roll of this processor running with two cores at 5.5 gigahertz, which certainly looks impressive. It doesn't make a huge difference to the actual performance, but 5.5 gigahertz is a very large number. Let'S get into the benchmarks: cinebench 15 single core.
So it seems like the kind thing is to show you a little bit of b-roll of this processor running with two cores at 5.5 gigahertz, which certainly looks impressive. It doesn't make a huge difference to the actual performance, but 5.5 gigahertz is a very large number. Let'S get into the benchmarks: cinebench 15 single core.
You can see that at the top we have the ks running on auto because it can use that 5.5 gigahertz clock speed, but it is a tiny step up from the overclock ks, which is locked at 5.3, gigahertz or indeed 5.2 gigahertz, very few points in it, but ks on auto wins: cinebench r15 multi-core, as you would expect the overclocked ks at 5.3 gigahertz beats out the ks at 5.gigahertz. Interestingly, the ryzen 9 5950x is very slightly behind.
In this showing intel has beaten amd in cinebench r15. Realistically we can call it a dead heat, but that's good going and then, as we look down the charts, we can see the ks on auto first on the msi meg unifi and then on the gigabyte. Auras master trailing behind the core: i9 12 900 k ryzen 9 5900x, the ryzen 7, 5800x.
Moving on to cinebench 23 single core. It'S a clean sweep for the ks at the top of the chart and again it's the auto behavior at 5.5 gigahertz that wins the day: cinebench r23 multi-core the overclocked ks at 5.3 gigahertz at the top of the charts, with a score that is frustratingly close to 30 000. We'Re not quite there.
Moving on to cinebench 23 single core. It'S a clean sweep for the ks at the top of the chart and again it's the auto behavior at 5.5 gigahertz that wins the day: cinebench r23 multi-core the overclocked ks at 5.3 gigahertz at the top of the charts, with a score that is frustratingly close to 30 000. We'Re not quite there.
However, and as you go down the chart, there are very few surprises in store: babco crossmark, the intel friendly test, so the ks on auto wins by a very small head over the overclocked ks blender 3. classroom. Sadly, we don't have recent figures for the ryzen 9 5950x in this particular test, because blender is updated.
Last time we ran that processor was on 2.9 and that means that amd's big boy is not represented in these charts. At the top, we have the overclock ks at 5. gigahertz, followed by the ks at 5.2. Behind that we have the k overclocked at five gigahertz beating out the ks on auto, and if you look at the top of the chart.
You'll see that really there's very little to separate those processors. Handbrake h.264 conversion the ks tops the chart and when it's overclocked, you simply cannot beat it for performance. It'S the same story with the h.265 conversion in handbrake ks at the top of the charts clock.
Speed is what it's all about 3d mark times by. This is just the cpu element of the test. Once again, it's all about clock, speed. The ks beats out the k, but you will see there's not a lot at the very top of the chart to separate one processor from the other gaming, far cry 6 at 1080p. As we know, 1080p gaming is about processor rather than graphics, and the ks on the meg unify on auto does very nicely.
If you go down to third place, you'll see an anomaly which is the ryzen. 7. 5800X 3d has a very high average speed, while the minimum is not that impressive. So i'm telling you that ks did particularly well and the 3d is a fly in the ointment. Far cry 6 at 1440p.
It'S the chaos is at the top of the charts, regardless of whether it's on auto or overclocked. Again, if you look down the page, you'll see, the 5800x 3d sticks out very slightly high average low minimum. Realistically, all of these processors are doing a very fine job.
At 1440p marvel's guardians of the galaxy at 1080p, we don't have a huge amount of data to work on here. You can see that the ks beats out the k, just as you'd expect, and that overclocked is better than auto.
At 1440p marvel's guardians of the galaxy at 1080p, we don't have a huge amount of data to work on here. You can see that the ks beats out the k, just as you'd expect, and that overclocked is better than auto.
There is a healthy amount of separation between the different settings, and this is telling us that raw power makes a lot of difference, but all of these processes are running over 200 fps, which is clearly huge by contrast, guardians of the galaxy at 1440p there's essentially nothing To choose between them watch dogs, legion at 1080p.
We expect 1080p gaming to be all about the processor and look what's at the top of the chart that ryzen 7 3d again, not a lot to choose between them. But the win goes to amd, followed by the overclocked ks. Surprisingly, it's the 5.2 gigahertz overclock just beating out the 5.3, which suggests that 5.2 is completely stable. 53 might be a fraction wobbly watch dog's legion at 1440p.
Well, it's all over the place. Here. We'Ve got the ryzen 7 at the top of the chart, the regular ryzen 7. That is, then: we have the ryzen 9, followed by some core 9 ks figures followed by the ryzen 7 3d, not a lot to choose between any of these processors, but the order here looks almost random and then we come on to the technical tests, so cpu Temperature,
This is delta temperature over ambient, which was around about 20 degrees, give or take it varied from 19 to 21. During the testing you can see, the coolest processor was the regular quarter, nine 12 900 k and then we get into the ks on aio cooling, followed by the ks overclocked at 5.gigahertz on custom loop.
Then we have the ks overclocked at 5.2 gigahertz on custom loop, because those two processor settings around exactly the same temperature. Then we have the ks on the msi meg unifi, with custom loop, cooling running on auto, but with manual vehicle 245 watts, as you can see. In the note there next up.
Second, from the hottest on the chart, we have the ks on the meg unifi on auto 305 watts. That'S 60 watts difference from resetting the v core and you can see that it makes 8 celsius difference to the cpu package temperature. If you're buying a ks.
Second, from the hottest on the chart, we have the ks on the meg unifi on auto 305 watts. That'S 60 watts difference from resetting the v core and you can see that it makes 8 celsius difference to the cpu package temperature. If you're buying a ks.
You will seriously want to consider setting the v core manually and then the hottest in the chart is the core: i9 1200k overclocked with aio cooling, 81 degrees. There add on 19 for ambient we're talking 100 celsius for the package, and then we have power consumption. This is cpu package power, not system power draw.
You want to add on 250 to 300 watts for the graphics and the system about 550 watts at the wall. Socket in many of the cases down at the bottom of the charts we have amd, which is, we know much more efficient than intel, then we have the auto core. I9 12 900 k pulling 213 watts for the package. 12 900 ks overclocked at 5.2. Gigahertz takes 241 watts, setting the v core manually on the ks and leaving it on auto 245 watts, ks overclocked at 5.3 gigahertz,
You want to add on 250 to 300 watts for the graphics and the system about 550 watts at the wall. Socket in many of the cases down at the bottom of the charts we have amd, which is, we know much more efficient than intel, then we have the auto core. I9 12 900 k pulling 213 watts for the package. 12 900 ks overclocked at 5.2. Gigahertz takes 241 watts, setting the v core manually on the ks and leaving it on auto 245 watts, ks overclocked at 5.3 gigahertz,
It's in the same ballpark 248 watts the ks on the gigabyte on auto 250 watts and then the ks on auto on the meg Unifi 305 watts that difference between 250 watts and 305 watts on auto, that's all down to the different buyer settings. What do i think of the intel core? I9? 12900 ks pros the good points stunning performance does really well in our test charts.
The other good point, the z690 platform with ddr5 it's a good platform decent. I o high performance, seems solid and stable cons, the bad points, so much power and so much heat. This processor tops the charts in terms of performance and in terms of the bad stuff bias. Support is patchy. I'Ve got a degree of sympathy here for, in this instance, gigabyte and msi.
I can't speak at this moment for as rock and for a zeus and the other manufacturers. This is a single processor which is obviously based very very very closely on the 1200k, which was already a hot and juicy processor. I'Ve got no idea how much development time the motherboard manufacturers have to do the bias to support this processor.
The other good point, the z690 platform with ddr5 it's a good platform decent. I o high performance, seems solid and stable cons, the bad points, so much power and so much heat. This processor tops the charts in terms of performance and in terms of the bad stuff bias. Support is patchy. I'Ve got a degree of sympathy here for, in this instance, gigabyte and msi.
I can't speak at this moment for as rock and for a zeus and the other manufacturers. This is a single processor which is obviously based very very very closely on the 1200k, which was already a hot and juicy processor. I'Ve got no idea how much development time the motherboard manufacturers have to do the bias to support this processor.
They must surely be thinking to themselves. The potential market for this processor is absolutely tiny. After all, the core i9, you wouldn't expect to sell in huge numbers in the first place, and this is a niche niche niche product and they're, probably thinking to themselves realistically anyone's spending.
The money for this processor is going to be using proper cooling and they're going to tweak the settings manually, they're not going to run it on auto, and it seems to me the auto settings are just something to get the motherboard working with this processor and then It'S over to you.
I wish they'd actually said as much uh and hadn't sort of listed it just uh without a warning note in a way msi kind of covered themselves by calling it a beta bias because it needs help, so buyer support is patchy.
That'S another way of saying, for goodness sake, get in there and sort out the settings, your processor and finally, the value for money argument is awful. Well, yes, i mean you're paying a huge premium for this processor.
I think we know why it costs so much, but hey ho. If you're a true enthusiast, the extra money, it's a mere detail. You probably spent a fortune on your pc graphics cards, cost the absolute earth high-end motherboards are expensive.
I think we know why it costs so much, but hey ho. If you're a true enthusiast, the extra money, it's a mere detail. You probably spent a fortune on your pc graphics cards, cost the absolute earth high-end motherboards are expensive.
The cooling system you're going to require to cool this processor, that's loads of money. What is a few hundred pounds between friends.