I have a 13900K and while this seems interesting on the surface I wonder if these parts differ from the normal 13900K in that they simply have two really great cores that can reach 6GHz or all of the cores are better than the best 13900K? If it's the former than it's pretty much useless because it is so rare that you have an application that only uses 1 or 2 cores and allows the ramp up to 5.8GHz (or 6 in this case). But that latter could mean that at lower frequencies this is a cooler, more efficient part vs. the 13900K. The extra $120 to get a really great binned part could be worth it in that case.
Unfortunately we have no information at this point in time. The increase in TDP makes me think this is just a more "leaky" part that can hit higher clocks at the expense of efficiency. That would be a shame because this would be completely a PR scheme.
There is some serious silicon lottery with Raptor lake. I built a few alder lake systems and their voltages were pretty similar to each other. I've built 4 13600K's so far and one ran at around 1.2V max, another at 1.3, and a junk chip that needed 1.4V at stock settings. Difference of up to 10C on the same NHD15 cooler on the best to the worst chip.
Pneumothorax, I've noticed this too! In fact I believe this is why many boards overvolt the K chips by default. For example, the MSI Z690-A and Asus Tuf Z690 both run non-K chips at stock relaxed voltage while K chips are +0.100 offset by default, without the offset even being configured. This obviously causes the chips to run hotter even at idle, and there have been no negative effects when I set the offset manually to - and voltage offset to 0.100
I wonder if board makers are aware of the silicon lottery and are padding the stability of their platforms by providing more voltage. This kind of sucks for consumers who don't realize this because you will run higher temps and shorted boost\turbo without oversized cooling.
While I realize the TDP is different between non-K and K chips, the voltages are not, so there is no other logical reason. Peak power draw should be determined by amp's being pulled, not voltage. There is no reason for any of these chips to be running at 1.4v-1.45v stock when the non-K (and even previous 12th gen K) run at 1.2v stock.
I thought it was motherboard/bios issue, but they were all eventually tested on the same motherboard and CMOS was reset between each CPU. They definitely have very different core VIDs due to silicon lottery.
CLCs are a joke. They don't have long-term thermal capacity because the radiator IS the reservoir. Any cheap $40 Thermalright or BeQuiet twin tower cooler will outperform even a 360mm radiator CLC without a reservoir because eventually the fans wont keep up with coolant thermal runaway (unless you have enormous airflow.)
Heatpipes have all the benefits and none of the drawbacks of a CLC. The primary benefit that your air cooler will last decades while your CLC will last, if you are lucky, a few years before the cheap-ass non-replaceable pump abruptly fails.
I don’t get your aversion to CLCs. All cooling systems are for sustained loads eventually limited by its ability to dissipate heat into the environment, reservoir or no reservoir. And if your plan is to have a reasonably sized chassis, a CLC is arguably the most efficient way to move heat. Any air-based solution is going to require much better airflow through the chassis and likely a larger, clumsier chassis as a consequence.
As for stability, I have a Corsair H100 thats been going pretty much 24/7 for the last ten years on an old 5960x. Also I have several NZXT rigs that have been running flawlessly for the last four years.
Judging by your comment you’re obviously an experienced builder, but I just want to share that your experience with CLCs is not representative of everyone.
H100 lasted 10 years? Yeah right. The best fan in the market wouldn't last that long, let alone the motor and liquid in a CLC. CLC is overrated and trouble prone. If any CPU needs more than a NH-D15 to cool properly, that CPU is inefficient. In other word - garbage.
>The best fan in the market wouldn't [10 years] I'm not sure what you're using, but I've got *plenty* of fans that I've been running 24/7 for over 10 years for sure, including all of the fans in the i5 2400 computer I'm typing this on. Sure, I've got fans that have died as well, but in my experience that's very much the exception rather than the norm.
I have an IFX-14 with the og thermalright YT140 from 2009 that has been running 24/7 most of its life and has no indications of failing soon. I have nice filters for my lianli v2000 case and minor dust. I have it mounted vertically now with the fans pull/pushing up because the mounting kits only allows that config on AM4 after using it with a intel 920, intel 3700k and now ryzen 3800.
I have a H105 AIO which I bought in 2014. It's still going and has been my main cooler since. I now use it on a 3900X and have zero thermal or performance issues. It is closely monitored. There's an air filter that is cleaned when it gets manky, but no other maintenance. I've been doing this a long time and I actually can't remember when I last had a fan failure personally. I've seen it happen and I've replaced them on others but never personally. I've run a PSU with a full time fan in for something like 9 years before exchanging it. The guy I gave it to is still running it.
When this AIO begins to show signs of failure (I've got alarms set up now, as it's making me twitchy), I'm undecided whether I'll go for another AIO or a large air cooler.
AIOs definitely have their place. As do air coolers. Calm down.
CLCs are valid options. There are just build quality variations with them. The good quality ones, mounted properly, will last more than 2yrs. More like 7yrs with minimal maintenance. With the heat output of CPUs these days, CLCs and custom water cooling will be the only option with top end CPUs.
A 360mm or 420mm radiator has more heat capacity than a dual tower 140mm air cooler, so IDK what your complaint there is.
Do CLCs require more care with how it's oriented in the case when compared to air coolers? yes. Are there vast build quality discrepancies between different brands and models? yes.
Are CLCs better at cooling than air coolers? Actually, yes. Just be sure to get a 280mm, 360mm, 420mm, or 480mm sized one. There is not a single air cooler that can keep up with the sheer volume and surface area of larger CLCs.
Unfortunately DH15 cannot keep up with the heat density of the Intel 11th gen + parts essentially anything over 300W is harder and impossible to keep it tamed. Only 360mil+ CLC can do the job. Note heat density, the high output of heat coming from Intel 11th gen and up is very very high and it will saturate even a CLC pretty fast. i9 12th gen and i7, i9 for 13th gen are impossible to cool on an Air cooler.
Noctua should release a successor. The problem with CLC is microfins, over the time it will become clogged and water will become messy, although it depends on how much usage you are doing to see how fast it can go bad. So that is definitely a thing. If anyone plans to keep a machine for decades, then Water cooling is no go. No matter open or close the microfins will be clogged. Open loop you can clean but it needs a lot of maintenance. CLC no such option, any leak or such = whole system will be down.
I want zero risk esp when I'm dealing with WD Gold drives in my machine so I went with a DH15 Chromax for my i9 10900K no fuss. The annoyance of mounting and GPU cannot be removed easily nor Cooler is a mega downside also unfortunately I use ASUS's DIMM.2 else no issues with DRAM clearance for 2nd fan, so I had to use NFA12x25mm.
So all this complexity many do not want and go with CLC only.
CLC is complex. A tower cooler with a fan is as simple as it gets. And cheaper tower coolers (or even $100 Noctua's that are similarly priced to a low-end CLC) will beat them 10 out of 10 times in almost every way, while lasting "forever."
I've never seen a CLC last 7 years. The pumps are generally rated at 25,000 hours. That's 2.5 years and anything beyond that you are on borrowed time. A general google search will reveal the average lifespan is 1-3 years.
9 years is anecdotal experience. Thousands of reviews between Newegg, Amazon, and just typing into google "how long to CLC's last" will give you an average window of 3 years.
If you actually got 9 years of continuous use out of a CLC, it must have been one of the early Asetek models with a full ceramic pump. Those were rated for 100,000 hours. The only CLC I know of ever made with such a rating. You'd know because the pump\waterblock is cylindrical and quite tall.
You can tame a Alder/Raptor lake for a NHD15, but you do leave some performance on the table. I usually have to power limit them to <200 watts to have some noise control. There is still some throttling however.
Don't confuse opinions with facts. CLC's have some downsides, as do normal coolers. But what you are talking about, The amount of heat you can get rid of continuously, CLC are superior. They would be better if they where thicker and made fully from copper, even better would be if the CPU side base would be silver (Oh I mis the start of this century). CLC's are mostly hampered by pattents and the dimentions of cases.
I suspect Intel is keeping tight lipped about the "lottery" because things are special than they really are. It's possible Intel adopted AMD's "best core" strategy and simply binned the chips to find two cores that could hit 6GHz reliably (with overhead, etc) but we won't know that for a fact until silicon is reviewed and it's apparent the same two cores are chosen to hit 6GHz opposed to 'random' cores.
There is inherently nothing different between all these chips, they are all made the same. Just some will clock higher and they save those to sell for more. The 200MHz and $120 for an "S" chip is 100% just ego.
Why do you think that? 13900K is already overclocked to crazy high frequencies. 13900KS should go even higher since these are binned chips, meaning 13900KS can reach the same speed of 13900K at a much lower power usage.
Agreed, either overclocking the 6GHZ chip will make it either unstable, too hot, and then over-heat causing speed down throttling. This assumes you don't run up against the power maximum. If there is room for over-clocking it will be small amount like 5%, before you hit a power maximum wall.
Intel relies on Turbo Boost Max 3.0 (i.e. favored cores) to get their peak clockspeeds. So that means only a couple of cores are validated to run at the highest clockspeeds.
7950X3D is -700MHz. How is that going to demolish this?
These are high core count chip suitable for multi-threaded workloads. In every multi-threaded benchmark, the KS chip should demolish the X3D, instead. Even in most low thread counts applications including many AAA games this chip should beat X3D easily due to higher single threaded performance. It is only the simulation-heavy games, where having a larger L3 cache will be somewhat more important than clock speed.
Because clockspeed isn't everything. The 13900KS has some serious thermal issues. 13900KS is barely faster than the 13900K in a variety of benchmarks, because a 420mm CLC can't keep it cooled. HW Unboxed has a review out already, and it really isn't looking too good for the 13900KS.
The 3D V-cache isn't going to give the 7950X any lead in productivity benchmark as we already saw with the 5800X3D, which performed worse than the 5800X due to the lower clock speeds. In gaming, the 13900KS isn't using all of its power so it's going to be nowhere near its thermal limit. The 7950X3D can still outperform it but not because of any thermal limits on the 13900KS but purely because of the added cache.
Pretty sure AMD's announcement for the 7950X3D said it was 5.70GHz; same as current 7950X CPU. They didn't have to sacrifice any clockspeed for the 3D V-Cache this time around.
-700MHz does not equal less performance. Clock frequency has never translated to performance. AMD has a superior architecture at the moment and the Zen CCD's are on a superior TSMC 5nm process to Intel 7 (which is 10nm but in reality similar to TSMC 7nm)
Intel always bins their i9 processors. Always the case since 9th, ASUS started putting prediction ratings on the top mobos still 10th gen stats for bins were all over the place and many were just junk. However the i9 10850K was even a low bin which could not hold 5GHz+ on all cores unlike the top SKU and then with 11th gen the whole backport killed it as even top bin CPUs were having worst IMC and poor scaling.
With 12th gen P and E (nonsense) to the mix the Silicon binning saw some upgrades as Intel put different bins on E and P and effective Silicon rating too ultimately on 12900K and KS the aggressive bins and top end up as KS limited run. Also to mention, ASUS VID output is solid feature but their Silicon Prediction still does not stick accurately as probably ASUS samples and then changes their algorithm as some of these processors change their prediction overtime with BIOS updates. ASUS killed the scam artists who overcharge for Intel binned processors like that SL and Derbaur. MSI also started using Force rating for the Silicon binning by the way. Although ASUS started providing IMC binning too lol. This is DDR5 folks which is just snake oil at best tbf, still something for the folks to burn their cash on I guess.
Anyways from 13th silicon binning took more upgrades, as with Raptor Lake most of the 13900K end up top rated bins as checked by Igor's lab samples. I've also seen on forums that avg i9 13900K ends up on a top spot. This KS is aggressive bin just like 12th gen, basically only the top of the top. But $700 ? Rip off esp when LGA1700 is getting one last final refresh as Raptor Lake Refresh which probably is like Haswell Devils canyon bearing Base and Boost clock boost. So blowing that much cash now and end up getting EOLed in mere months is not good. Anyone with LGA1700 should wait for RPL-SR and then upgrade, also buy that ILM socket holder because Intel destroyed their engineering expertise with such hamfisted socket bending nonsense.
If not for E cores rubbish and ILM failure design, I would have got myself a Z790 Apex and run Windows 7 on it with RTX3090Ti but had to settle for a 10900K on Z590 Apex.
I don't think anyone is surprised at the existence of the KS model, just eager to laugh or balk at hot, expensive Intel parts with minimal gains.
AMD's X3D is where the action is in 2023. That could be flawed as well if the asymmetric approach isn't handled properly. However, AMD has to rip off that band-aid eventually, since they are likely to pursue hybrid CPUs combining Zen 5 and Zen 4C cores, for example (which would be another discrepancy of cache and clocks, but also IPC).
Not that Hybrid again for AMD. I have mentioned countless times. AMD official slides and Financial report slides, both never mentioned that Zen 4C is for consumers. Bergamo is not a consumer product.
AM5 does NOT have Hybrid garbage. It does not exist. AMD's future of x86 big little does not make sense. Intel 13900K is the proof that Intel needs the little cores to accelerate MT performance because look at the power and the heat density with just 8P cores. Intel is at limit, that is the reason why they had to go hybrid design.
PLUS Intel does not have MCM expertise, for that reason their SPR XEON 4th gen was delayed by 1.5 years. AMD on the other hand gained that from 2017. Which is why you see MI300 the pinnacle of MCM design.
Again AMD's processor design of Zen uArch is newer. Intel Core is a derivative of Sandy Bridge design of RING bus and they are just using the same fundamental which is why it is called Intel CORE. Intel hired Jim Keller but due to some BS drama inside with high execs and Jim Keller they ruined it and also Intel hired most of the old staff who designed the foundation of Intel Core, Glenn Hinton on a new uArch redesign. Until that happens Intel CORE will not go away and Intel will face Big little BS always also they have that BGA advantage on Big Little junk so they won't stop making them.
Finally AMD hamfisted the X3D processors. It's worthless. I expected X3D to be a refresh not a stupid "GAMING" PR sticker slapped over it. The TDP is reduced, Base clocks neutered (AMD always reigned supreme in this as Intel base is pathetic 3-3.5GHz while AMD Zen 4 is ultra fast 4.7GHz), Boost clocks neutered. Hamfisted the 12C and 16C parts by making it have single CCD V-Cache and depend on Windows scheduling nonsense. Pure garbage vs Intel's Thread Director HW level layer. Finally the worthless 7950X3D won't imprve anything on MT performance as it got regressed. There's no action here.
Which is why Intel also went with RPL-S Refresh because it's pointless to fight Zen 4X3D with 14th gen as it does not have ANYTHING except gaming garbage. And that gives Intel 14th gen / LGA1851 to have 15th gen vs Zen 5 only as AMD does not plan Zen 6 on that AM5 socket at best 5X3D again a stunted CPU refresh for those GAMERS.
so is Intel the way to go for the next few years or so vs AMD AM5 path? i got a 12900K+z690A motherboard at the moment and trying to decide which split in the road to take.
My biggest concern with the 79XXx3d chips is windows 11 scheduler. It’s still having problems even with the 13900K with the 13600K scoring better at times due to less cores. It’s going to be a bigger issue making sure the app goes to the correct CCD
Please, please hire an editor, Anandtech. Thankfully the spelling errors aren't common but just general grammar/wording makes many articles unnecessarily hard to read on the first pass.
Regardless, it isn't too much effort for the author to re-read short posts like these for errors or grammar before hitting publish. You don't necessarily need a copy editor for news posts; most large blogs out there don't have one and do just fine.
The owner is Future plc (https://en.wikipedia.org/wiki/Future_plc) which bought Anandtech's prior owner Purch. Future also owns Tom's Hardware and other various thingies, but I wouldn't call this particular article clickbait.
Typically I do some level of copy-edit pass on all front-page articles. However due to the timing of the Intel announcement (we only received a short heads-up), I wasn't available to do it ahead of publication for this one.
I've since gone back and done a full copy-edit pass.
People buy new Nvidia GPUs that are super expensive, they also buy these because it has bigger numbers... There always are enough moro... I mean people who don´t have money problems.
This post states a contradiction: - "...and is pre-binned through a unique selection process to ensure the Core i9-13900KS's special edition status for their highest level of frequency of 6 GHz in a desktop chip out of the box..." - "Intel hasn't specified if the Core i9-13900KS is a special binned part"
Which is it? More importantly, why is speculation passed off as fact?
Not to be pedantic, but 6.0 GHz is not guaranteed, isn't it? How can they call it first of anything? Base clock is 3.2 GHz which they actually guarantee.
They could guarantee it with enough asterisks. Like if you run the thing without a cooler, it's clearly not going to boost to 6 GHz. If you have a properly installed cooler rated for the correct TDP, then it can boost 200 MHz higher than the average similarly situated 13900K because of better binning.
How long will any of its cores turbo at 6 GHz? I think that was one of the big issues with AMD's 9590 scam product. 5 GHz for how long... milliseconds?
Who is the intended buyer? If you need crazy cpu power, wouldn't a Threadripper destroy any i9 in that? And in gaming.... It's hard to imagine someone spending double to get 1% more, since at 4k your GPU will quickly be the bottleneck. But even so, Then wait for that 3D on-board memory Zen4 chip? If I was intel, I would focus on competing against those low power Zen4's. They are still to expensive I think.
"Judging from end-user reports of overclocking the normal Core i9-13900K, we're left with the impression that very few of those chips could hit 6 GHz with ambient air cooling. So the i9-13900KS appears to be a very thin bin"
or intel is snatching them up for..... binning.......
My main system CPU generates 150W max. Its Noctua NH-U12A cooler can barely keep this from thermal throttling at 91 degC, in a very high-airflow (7-fan) case at 72 degC ambient.
This is one of the best air coolers available; and I won't even consider liquid cooling.
So I won't be running any CPU above 150W. That's why I'm most interested in how they compare when capped at around this point.
It get’s a bit noisy though when encoding video. Perfectly fine for gaming however. I’ve built multiple Alder/Raptor systems with the NH15D and the i5’s run great, but the i7/i9 can definitely overpower the NH15D at stock motherboard (unlimited PL) settings. I usually cut down the PL1/PL2 to around 200watts depending how much noise my customers are willing to put up with.
Overall it is a disappointing demonstration for a top of the crop sample right in the Intel lab (under full control of its engineers). It hits 1.523v VID for running 1 thread of 7-Zip benchmark and makes no good use of its massive power headroom.
My own 13900K sample is "just good", runs at 20% below stock power and still manages to only use 1.3v for the same load and then beat both the video sample and published Geekbench scores. That is because of the massive 60/60/56...56 drop used at stock the moment any background third thread/core becomes active during the benchmark (which often happens).
Instead of polishing their crown jewel, they just put in in a cardboard box and add an "S" using black marker.
O how the mighty have fallen. +200Mhz on 5.8Ghz ... IF there is thermal headroom, and ONLY on favoured cores ... using 320W ... for $110 extra? Is this a subtle homage to The Onion?
I struggle to see any scenario whatsoever where a two-core +200Mhz boost is going to make any perceptible difference whatsoever. That includes niche things like scientific modeling, extreme gaming, and real-time trading. So who, exactly, is the market for this?
Unfortunately, there will always be people with more money than sense, so I expect that some Greater Fools will be along shortly to buy this nonsensical product, thus vindicating Intel's current strategy of "MOAR power! MOAR!" and damn the voltage-frequency curve. 320W for the dubious pleasure of having TWO cores with +200Mhz. Holy Netburst, Batman, here we go again.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
100 Comments
Back to Article
Hulk - Thursday, January 12, 2023 - link
I have a 13900K and while this seems interesting on the surface I wonder if these parts differ from the normal 13900K in that they simply have two really great cores that can reach 6GHz or all of the cores are better than the best 13900K? If it's the former than it's pretty much useless because it is so rare that you have an application that only uses 1 or 2 cores and allows the ramp up to 5.8GHz (or 6 in this case). But that latter could mean that at lower frequencies this is a cooler, more efficient part vs. the 13900K. The extra $120 to get a really great binned part could be worth it in that case.Unfortunately we have no information at this point in time. The increase in TDP makes me think this is just a more "leaky" part that can hit higher clocks at the expense of efficiency. That would be a shame because this would be completely a PR scheme.
Pneumothorax - Thursday, January 12, 2023 - link
There is some serious silicon lottery with Raptor lake. I built a few alder lake systems and their voltages were pretty similar to each other. I've built 4 13600K's so far and one ran at around 1.2V max, another at 1.3, and a junk chip that needed 1.4V at stock settings. Difference of up to 10C on the same NHD15 cooler on the best to the worst chip.Samus - Thursday, January 12, 2023 - link
Pneumothorax, I've noticed this too! In fact I believe this is why many boards overvolt the K chips by default. For example, the MSI Z690-A and Asus Tuf Z690 both run non-K chips at stock relaxed voltage while K chips are +0.100 offset by default, without the offset even being configured. This obviously causes the chips to run hotter even at idle, and there have been no negative effects when I set the offset manually to - and voltage offset to 0.100I wonder if board makers are aware of the silicon lottery and are padding the stability of their platforms by providing more voltage. This kind of sucks for consumers who don't realize this because you will run higher temps and shorted boost\turbo without oversized cooling.
While I realize the TDP is different between non-K and K chips, the voltages are not, so there is no other logical reason. Peak power draw should be determined by amp's being pulled, not voltage. There is no reason for any of these chips to be running at 1.4v-1.45v stock when the non-K (and even previous 12th gen K) run at 1.2v stock.
Pneumothorax - Friday, January 13, 2023 - link
I thought it was motherboard/bios issue, but they were all eventually tested on the same motherboard and CMOS was reset between each CPU. They definitely have very different core VIDs due to silicon lottery.meacupla - Thursday, January 12, 2023 - link
Reading and watching reviews, the 13900KS runs into thermal issues while on a 240mm CLC.The results are that it is barely any better than the 13900K.
Samus - Friday, January 13, 2023 - link
CLCs are a joke. They don't have long-term thermal capacity because the radiator IS the reservoir. Any cheap $40 Thermalright or BeQuiet twin tower cooler will outperform even a 360mm radiator CLC without a reservoir because eventually the fans wont keep up with coolant thermal runaway (unless you have enormous airflow.)Heatpipes have all the benefits and none of the drawbacks of a CLC. The primary benefit that your air cooler will last decades while your CLC will last, if you are lucky, a few years before the cheap-ass non-replaceable pump abruptly fails.
MarcusMo - Friday, January 13, 2023 - link
I don’t get your aversion to CLCs. All cooling systems are for sustained loads eventually limited by its ability to dissipate heat into the environment, reservoir or no reservoir. And if your plan is to have a reasonably sized chassis, a CLC is arguably the most efficient way to move heat. Any air-based solution is going to require much better airflow through the chassis and likely a larger, clumsier chassis as a consequence.As for stability, I have a Corsair H100 thats been going pretty much 24/7 for the last ten years on an old 5960x. Also I have several NZXT rigs that have been running flawlessly for the last four years.
Judging by your comment you’re obviously an experienced builder, but I just want to share that your experience with CLCs is not representative of everyone.
sonny73n - Saturday, January 14, 2023 - link
H100 lasted 10 years? Yeah right. The best fan in the market wouldn't last that long, let alone the motor and liquid in a CLC.CLC is overrated and trouble prone. If any CPU needs more than a NH-D15 to cool properly, that CPU is inefficient. In other word - garbage.
Dolda2000 - Saturday, January 14, 2023 - link
>The best fan in the market wouldn't [10 years]I'm not sure what you're using, but I've got *plenty* of fans that I've been running 24/7 for over 10 years for sure, including all of the fans in the i5 2400 computer I'm typing this on. Sure, I've got fans that have died as well, but in my experience that's very much the exception rather than the norm.
Byte - Monday, January 16, 2023 - link
I have an IFX-14 with the og thermalright YT140 from 2009 that has been running 24/7 most of its life and has no indications of failing soon. I have nice filters for my lianli v2000 case and minor dust. I have it mounted vertically now with the fans pull/pushing up because the mounting kits only allows that config on AM4 after using it with a intel 920, intel 3700k and now ryzen 3800.philehidiot - Monday, January 16, 2023 - link
I have a H105 AIO which I bought in 2014. It's still going and has been my main cooler since. I now use it on a 3900X and have zero thermal or performance issues. It is closely monitored. There's an air filter that is cleaned when it gets manky, but no other maintenance. I've been doing this a long time and I actually can't remember when I last had a fan failure personally. I've seen it happen and I've replaced them on others but never personally. I've run a PSU with a full time fan in for something like 9 years before exchanging it. The guy I gave it to is still running it.When this AIO begins to show signs of failure (I've got alarms set up now, as it's making me twitchy), I'm undecided whether I'll go for another AIO or a large air cooler.
AIOs definitely have their place. As do air coolers. Calm down.
jakky567 - Tuesday, January 17, 2023 - link
Yep! I refuse to go to water cooling.meacupla - Friday, January 13, 2023 - link
CLCs are valid options. There are just build quality variations with them. The good quality ones, mounted properly, will last more than 2yrs. More like 7yrs with minimal maintenance.With the heat output of CPUs these days, CLCs and custom water cooling will be the only option with top end CPUs.
A 360mm or 420mm radiator has more heat capacity than a dual tower 140mm air cooler, so IDK what your complaint there is.
Do CLCs require more care with how it's oriented in the case when compared to air coolers? yes.
Are there vast build quality discrepancies between different brands and models? yes.
Are CLCs better at cooling than air coolers? Actually, yes. Just be sure to get a 280mm, 360mm, 420mm, or 480mm sized one. There is not a single air cooler that can keep up with the sheer volume and surface area of larger CLCs.
Silver5urfer - Friday, January 13, 2023 - link
Unfortunately DH15 cannot keep up with the heat density of the Intel 11th gen + parts essentially anything over 300W is harder and impossible to keep it tamed. Only 360mil+ CLC can do the job. Note heat density, the high output of heat coming from Intel 11th gen and up is very very high and it will saturate even a CLC pretty fast. i9 12th gen and i7, i9 for 13th gen are impossible to cool on an Air cooler.Noctua should release a successor. The problem with CLC is microfins, over the time it will become clogged and water will become messy, although it depends on how much usage you are doing to see how fast it can go bad. So that is definitely a thing. If anyone plans to keep a machine for decades, then Water cooling is no go. No matter open or close the microfins will be clogged. Open loop you can clean but it needs a lot of maintenance. CLC no such option, any leak or such = whole system will be down.
I want zero risk esp when I'm dealing with WD Gold drives in my machine so I went with a DH15 Chromax for my i9 10900K no fuss. The annoyance of mounting and GPU cannot be removed easily nor Cooler is a mega downside also unfortunately I use ASUS's DIMM.2 else no issues with DRAM clearance for 2nd fan, so I had to use NFA12x25mm.
So all this complexity many do not want and go with CLC only.
Samus - Friday, January 13, 2023 - link
CLC is complex. A tower cooler with a fan is as simple as it gets. And cheaper tower coolers (or even $100 Noctua's that are similarly priced to a low-end CLC) will beat them 10 out of 10 times in almost every way, while lasting "forever."I've never seen a CLC last 7 years. The pumps are generally rated at 25,000 hours. That's 2.5 years and anything beyond that you are on borrowed time. A general google search will reveal the average lifespan is 1-3 years.
Oxford Guy - Friday, January 13, 2023 - link
My H50 (I think it was called), exploded — destroying a motherboard and video card and covering the inside of a case with whitish splatter.philehidiot - Monday, January 16, 2023 - link
I'm definitely on borrowed time then... 9 years...Samus - Monday, January 16, 2023 - link
9 years is anecdotal experience. Thousands of reviews between Newegg, Amazon, and just typing into google "how long to CLC's last" will give you an average window of 3 years.If you actually got 9 years of continuous use out of a CLC, it must have been one of the early Asetek models with a full ceramic pump. Those were rated for 100,000 hours. The only CLC I know of ever made with such a rating. You'd know because the pump\waterblock is cylindrical and quite tall.
Pneumothorax - Saturday, January 14, 2023 - link
You can tame a Alder/Raptor lake for a NHD15, but you do leave some performance on the table. I usually have to power limit them to <200 watts to have some noise control. There is still some throttling however.corinthos - Sunday, January 22, 2023 - link
Would you leave less/no performance on the table with 7950x and NHD15 instead of 13900K?Foeketijn - Saturday, January 14, 2023 - link
Don't confuse opinions with facts. CLC's have some downsides, as do normal coolers. But what you are talking about, The amount of heat you can get rid of continuously, CLC are superior. They would be better if they where thicker and made fully from copper, even better would be if the CPU side base would be silver (Oh I mis the start of this century). CLC's are mostly hampered by pattents and the dimentions of cases.Samus - Thursday, January 12, 2023 - link
I suspect Intel is keeping tight lipped about the "lottery" because things are special than they really are. It's possible Intel adopted AMD's "best core" strategy and simply binned the chips to find two cores that could hit 6GHz reliably (with overhead, etc) but we won't know that for a fact until silicon is reviewed and it's apparent the same two cores are chosen to hit 6GHz opposed to 'random' cores.Obviously speculation.
Byte - Friday, January 13, 2023 - link
There is inherently nothing different between all these chips, they are all made the same. Just some will clock higher and they save those to sell for more. The 200MHz and $120 for an "S" chip is 100% just ego.escksu - Thursday, January 12, 2023 - link
This has to be the first 6ghz in retail. Having said that I don't think there is any headroom for oc.m53 - Thursday, January 12, 2023 - link
Why do you think that? 13900K is already overclocked to crazy high frequencies. 13900KS should go even higher since these are binned chips, meaning 13900KS can reach the same speed of 13900K at a much lower power usage.Smartoverclockjohn - Friday, January 13, 2023 - link
Agreed, either overclocking the 6GHZ chip will make it either unstable, too hot, and then over-heat causing speed down throttling. This assumes you don't run up against the power maximum. If there is room for over-clocking it will be small amount like 5%, before you hit a power maximum wall.TIGOS - Thursday, January 12, 2023 - link
150 watts... just thatSamus - Thursday, January 12, 2023 - link
Preshott is back baby!shabby - Thursday, January 12, 2023 - link
Why does that promotional video only show 2 p-cores running? Shouldn't it show all p-cores with only two hitting 6ghz?krazyfrog - Thursday, January 12, 2023 - link
As the guy says in the video, he chose 7-Zip specifically because it lets him choose the number of threads and he's limited them to two.Timur Born - Sunday, January 15, 2023 - link
1Ryan Smith - Friday, January 13, 2023 - link
Intel relies on Turbo Boost Max 3.0 (i.e. favored cores) to get their peak clockspeeds. So that means only a couple of cores are validated to run at the highest clockspeeds.nandnandnand - Thursday, January 12, 2023 - link
Nice +200 MHz. Get ready for 7950X3D to demolish it.m53 - Thursday, January 12, 2023 - link
7950X3D is -700MHz. How is that going to demolish this?These are high core count chip suitable for multi-threaded workloads. In every multi-threaded benchmark, the KS chip should demolish the X3D, instead. Even in most low thread counts applications including many AAA games this chip should beat X3D easily due to higher single threaded performance. It is only the simulation-heavy games, where having a larger L3 cache will be somewhat more important than clock speed.
meacupla - Thursday, January 12, 2023 - link
Because clockspeed isn't everything.The 13900KS has some serious thermal issues.
13900KS is barely faster than the 13900K in a variety of benchmarks, because a 420mm CLC can't keep it cooled.
HW Unboxed has a review out already, and it really isn't looking too good for the 13900KS.
krazyfrog - Thursday, January 12, 2023 - link
The 3D V-cache isn't going to give the 7950X any lead in productivity benchmark as we already saw with the 5800X3D, which performed worse than the 5800X due to the lower clock speeds. In gaming, the 13900KS isn't using all of its power so it's going to be nowhere near its thermal limit. The 7950X3D can still outperform it but not because of any thermal limits on the 13900KS but purely because of the added cache.meacupla - Thursday, January 12, 2023 - link
IDK, maybe you should go look at the clock speeds for 7xxxX3D again?https://www.anandtech.com/show/18709/amd-unveils-r...
NextGen_Gamer - Thursday, January 12, 2023 - link
Pretty sure AMD's announcement for the 7950X3D said it was 5.70GHz; same as current 7950X CPU. They didn't have to sacrifice any clockspeed for the 3D V-Cache this time around.cchi - Thursday, January 12, 2023 - link
True only tot the die without 3DVSamus - Thursday, January 12, 2023 - link
-700MHz does not equal less performance. Clock frequency has never translated to performance. AMD has a superior architecture at the moment and the Zen CCD's are on a superior TSMC 5nm process to Intel 7 (which is 10nm but in reality similar to TSMC 7nm)JKflipflop98 - Friday, January 13, 2023 - link
Why, AMD is so superior that's it's slower in all per-core benchmarks across the board! GO AMD!lopri - Monday, January 16, 2023 - link
Clock frequency has never translated to performance, are we living in the same universe?supdawgwtfd - Monday, January 16, 2023 - link
He is saying that you can'took at the clock speed of 2 different architectures and assume the one with higher clock speed is better.History shows this to be incorrect.
Byte - Monday, January 16, 2023 - link
5800x3D can still beat it in certain games. A 7800x3D is 500MHz faster than that.nirolf - Thursday, January 12, 2023 - link
Since when 20% more TDP is considered “slightly higher”?This looks more like a Prescott era effort trying to stay ahead of the X3Ds.
bernstein - Thursday, January 12, 2023 - link
Yeah, so much this!Silver5urfer - Thursday, January 12, 2023 - link
Intel always bins their i9 processors. Always the case since 9th, ASUS started putting prediction ratings on the top mobos still 10th gen stats for bins were all over the place and many were just junk. However the i9 10850K was even a low bin which could not hold 5GHz+ on all cores unlike the top SKU and then with 11th gen the whole backport killed it as even top bin CPUs were having worst IMC and poor scaling.With 12th gen P and E (nonsense) to the mix the Silicon binning saw some upgrades as Intel put different bins on E and P and effective Silicon rating too ultimately on 12900K and KS the aggressive bins and top end up as KS limited run. Also to mention, ASUS VID output is solid feature but their Silicon Prediction still does not stick accurately as probably ASUS samples and then changes their algorithm as some of these processors change their prediction overtime with BIOS updates. ASUS killed the scam artists who overcharge for Intel binned processors like that SL and Derbaur. MSI also started using Force rating for the Silicon binning by the way. Although ASUS started providing IMC binning too lol. This is DDR5 folks which is just snake oil at best tbf, still something for the folks to burn their cash on I guess.
Anyways from 13th silicon binning took more upgrades, as with Raptor Lake most of the 13900K end up top rated bins as checked by Igor's lab samples. I've also seen on forums that avg i9 13900K ends up on a top spot. This KS is aggressive bin just like 12th gen, basically only the top of the top. But $700 ? Rip off esp when LGA1700 is getting one last final refresh as Raptor Lake Refresh which probably is like Haswell Devils canyon bearing Base and Boost clock boost. So blowing that much cash now and end up getting EOLed in mere months is not good. Anyone with LGA1700 should wait for RPL-SR and then upgrade, also buy that ILM socket holder because Intel destroyed their engineering expertise with such hamfisted socket bending nonsense.
If not for E cores rubbish and ILM failure design, I would have got myself a Z790 Apex and run Windows 7 on it with RTX3090Ti but had to settle for a 10900K on Z590 Apex.
Wereweeb - Thursday, January 12, 2023 - link
I was about to ask "Wow how come you can write so much nonsense with that much confidence" and then I saw the username. Checks out.Silver5urfer - Thursday, January 12, 2023 - link
If you do not have a single point to mention what's wrong why even bother ?JKflipflop98 - Friday, January 13, 2023 - link
Everything you said is wrong. Every. Word.Silver5urfer - Friday, January 13, 2023 - link
Everything is 100% correct on dot. You fail to produce what is wrong clown.nandnandnand - Friday, January 13, 2023 - link
I don't think anyone is surprised at the existence of the KS model, just eager to laugh or balk at hot, expensive Intel parts with minimal gains.AMD's X3D is where the action is in 2023. That could be flawed as well if the asymmetric approach isn't handled properly. However, AMD has to rip off that band-aid eventually, since they are likely to pursue hybrid CPUs combining Zen 5 and Zen 4C cores, for example (which would be another discrepancy of cache and clocks, but also IPC).
Silver5urfer - Friday, January 13, 2023 - link
Not that Hybrid again for AMD. I have mentioned countless times. AMD official slides and Financial report slides, both never mentioned that Zen 4C is for consumers. Bergamo is not a consumer product.AM5 does NOT have Hybrid garbage. It does not exist. AMD's future of x86 big little does not make sense. Intel 13900K is the proof that Intel needs the little cores to accelerate MT performance because look at the power and the heat density with just 8P cores. Intel is at limit, that is the reason why they had to go hybrid design.
PLUS Intel does not have MCM expertise, for that reason their SPR XEON 4th gen was delayed by 1.5 years. AMD on the other hand gained that from 2017. Which is why you see MI300 the pinnacle of MCM design.
Again AMD's processor design of Zen uArch is newer. Intel Core is a derivative of Sandy Bridge design of RING bus and they are just using the same fundamental which is why it is called Intel CORE. Intel hired Jim Keller but due to some BS drama inside with high execs and Jim Keller they ruined it and also Intel hired most of the old staff who designed the foundation of Intel Core, Glenn Hinton on a new uArch redesign. Until that happens Intel CORE will not go away and Intel will face Big little BS always also they have that BGA advantage on Big Little junk so they won't stop making them.
Finally AMD hamfisted the X3D processors. It's worthless. I expected X3D to be a refresh not a stupid "GAMING" PR sticker slapped over it. The TDP is reduced, Base clocks neutered (AMD always reigned supreme in this as Intel base is pathetic 3-3.5GHz while AMD Zen 4 is ultra fast 4.7GHz), Boost clocks neutered. Hamfisted the 12C and 16C parts by making it have single CCD V-Cache and depend on Windows scheduling nonsense. Pure garbage vs Intel's Thread Director HW level layer. Finally the worthless 7950X3D won't imprve anything on MT performance as it got regressed. There's no action here.
Which is why Intel also went with RPL-S Refresh because it's pointless to fight Zen 4X3D with 14th gen as it does not have ANYTHING except gaming garbage. And that gives Intel 14th gen / LGA1851 to have 15th gen vs Zen 5 only as AMD does not plan Zen 6 on that AM5 socket at best 5X3D again a stunted CPU refresh for those GAMERS.
corinthos - Sunday, January 22, 2023 - link
so is Intel the way to go for the next few years or so vs AMD AM5 path? i got a 12900K+z690A motherboard at the moment and trying to decide which split in the road to take.Pneumothorax - Tuesday, January 17, 2023 - link
My biggest concern with the 79XXx3d chips is windows 11 scheduler. It’s still having problems even with the 13900K with the 13600K scoring better at times due to less cores. It’s going to be a bigger issue making sure the app goes to the correct CCDachfoo - Thursday, January 12, 2023 - link
Please, please hire an editor, Anandtech. Thankfully the spelling errors aren't common but just general grammar/wording makes many articles unnecessarily hard to read on the first pass.Wereweeb - Thursday, January 12, 2023 - link
Whoever owns Anandtech is letting it die, honest and intelligent articles doesn't make money like clickbait does.krazyfrog - Thursday, January 12, 2023 - link
Regardless, it isn't too much effort for the author to re-read short posts like these for errors or grammar before hitting publish. You don't necessarily need a copy editor for news posts; most large blogs out there don't have one and do just fine.PeachNCream - Thursday, January 12, 2023 - link
The owner is Future plc (https://en.wikipedia.org/wiki/Future_plc) which bought Anandtech's prior owner Purch. Future also owns Tom's Hardware and other various thingies, but I wouldn't call this particular article clickbait.Ryan Smith - Friday, January 13, 2023 - link
Typically I do some level of copy-edit pass on all front-page articles. However due to the timing of the Intel announcement (we only received a short heads-up), I wasn't available to do it ahead of publication for this one.I've since gone back and done a full copy-edit pass.
bernstein - Thursday, January 12, 2023 - link
Why the fuck is a ~3% overclocked cpu “ highly anticipated”???!!! And who would even think of paying 20% more for it?haukionkannel - Friday, January 13, 2023 - link
People buy new Nvidia GPUs that are super expensive, they also buy these because it has bigger numbers... There always are enough moro... I mean people who don´t have money problems.dillonnotz24 - Thursday, January 12, 2023 - link
This post states a contradiction:- "...and is pre-binned through a unique selection process to ensure the Core i9-13900KS's special edition status for their highest level of frequency of 6 GHz in a desktop chip out of the box..."
- "Intel hasn't specified if the Core i9-13900KS is a special binned part"
Which is it? More importantly, why is speculation passed off as fact?
Ryan Smith - Friday, January 13, 2023 - link
It's a binned part. Intel just isn't saying anything in particular about the binning metrics this time around.I've changed up the text to clarify that point.
ballsystemlord - Thursday, January 12, 2023 - link
Meanwhile, people are crying about AMD having a 230W PPT on their highest end parts.lopri - Thursday, January 12, 2023 - link
Not to be pedantic, but 6.0 GHz is not guaranteed, isn't it? How can they call it first of anything? Base clock is 3.2 GHz which they actually guarantee.haukionkannel - Friday, January 13, 2023 - link
the fastest core can go up to 6 MHz.. if you have really powerful cooler.Kvaern1 - Monday, January 16, 2023 - link
You don't need much cooling when you're only working a few cores.nandnandnand - Friday, January 13, 2023 - link
They could guarantee it with enough asterisks. Like if you run the thing without a cooler, it's clearly not going to boost to 6 GHz. If you have a properly installed cooler rated for the correct TDP, then it can boost 200 MHz higher than the average similarly situated 13900K because of better binning.don0301 - Friday, January 13, 2023 - link
Can it run Crysis?WestPole - Thursday, January 19, 2023 - link
That comment never fails to slip in every now and then… 😁Gothmoth - Friday, January 13, 2023 - link
nonsense product....Oxford Guy - Friday, January 13, 2023 - link
How long will any of its cores turbo at 6 GHz? I think that was one of the big issues with AMD's 9590 scam product. 5 GHz for how long... milliseconds?Foeketijn - Saturday, January 14, 2023 - link
Who is the intended buyer? If you need crazy cpu power, wouldn't a Threadripper destroy any i9 in that?And in gaming.... It's hard to imagine someone spending double to get 1% more, since at 4k your GPU will quickly be the bottleneck. But even so, Then wait for that 3D on-board memory Zen4 chip?
If I was intel, I would focus on competing against those low power Zen4's. They are still to expensive I think.
Iketh - Saturday, January 14, 2023 - link
"Judging from end-user reports of overclocking the normal Core i9-13900K, we're left with the impression that very few of those chips could hit 6 GHz with ambient air cooling. So the i9-13900KS appears to be a very thin bin"or intel is snatching them up for..... binning.......
nandnandnand - Saturday, January 14, 2023 - link
Nice catch. It's almost as if everyone knew there would be a 13900KS all along, lol.Arbie - Saturday, January 14, 2023 - link
My main system CPU generates 150W max. Its Noctua NH-U12A cooler can barely keep this from thermal throttling at 91 degC, in a very high-airflow (7-fan) case at 72 degC ambient.This is one of the best air coolers available; and I won't even consider liquid cooling.
So I won't be running any CPU above 150W. That's why I'm most interested in how they compare when capped at around this point.
Arbie - Sunday, January 15, 2023 - link
* 72 degF ambient, not 72 degC. Someday - possibly before the Sun explodes - we may be able to edit comments here.Kvaern1 - Monday, January 16, 2023 - link
12700K + NH15D here. Run within it's offical specs the CPU is no match for the Noctua.Pneumothorax - Tuesday, January 17, 2023 - link
It get’s a bit noisy though when encoding video. Perfectly fine for gaming however. I’ve built multiple Alder/Raptor systems with the NH15D and the i5’s run great, but the i7/i9 can definitely overpower the NH15D at stock motherboard (unlimited PL) settings. I usually cut down the PL1/PL2 to around 200watts depending how much noise my customers are willing to put up with.Kvaern1 - Tuesday, January 17, 2023 - link
Agreed. My BIOS auto settings also allows it to draw a stupid amount much power for no good reason but that can be remedied as you write.corinthos - Sunday, January 22, 2023 - link
would it be better to go 7950x or 13900k when capped at 150w?Timur Born - Sunday, January 15, 2023 - link
Overall it is a disappointing demonstration for a top of the crop sample right in the Intel lab (under full control of its engineers). It hits 1.523v VID for running 1 thread of 7-Zip benchmark and makes no good use of its massive power headroom.My own 13900K sample is "just good", runs at 20% below stock power and still manages to only use 1.3v for the same load and then beat both the video sample and published Geekbench scores. That is because of the massive 60/60/56...56 drop used at stock the moment any background third thread/core becomes active during the benchmark (which often happens).
Instead of polishing their crown jewel, they just put in in a cardboard box and add an "S" using black marker.
Carmen00 - Monday, January 16, 2023 - link
O how the mighty have fallen. +200Mhz on 5.8Ghz ... IF there is thermal headroom, and ONLY on favoured cores ... using 320W ... for $110 extra? Is this a subtle homage to The Onion?I struggle to see any scenario whatsoever where a two-core +200Mhz boost is going to make any perceptible difference whatsoever. That includes niche things like scientific modeling, extreme gaming, and real-time trading. So who, exactly, is the market for this?
Unfortunately, there will always be people with more money than sense, so I expect that some Greater Fools will be along shortly to buy this nonsensical product, thus vindicating Intel's current strategy of "MOAR power! MOAR!" and damn the voltage-frequency curve. 320W for the dubious pleasure of having TWO cores with +200Mhz. Holy Netburst, Batman, here we go again.
Foeketijn - Tuesday, January 17, 2023 - link
And even the same guy on the steering wheel.Can't believe, they do it again.
dizzynosed - Tuesday, January 17, 2023 - link
Can I use this chip in Asus TUF B760-M DDR4 mobo ? Will it have the same speed as in Z790 mobo?