Intel Prepares Pre-Binned Core i9-12900KS Processors Clocked at 5.2 GHz

[AleksandarK]

According to the latest round of rumors coming from tech media VideoCardz, Intel could be preparing an answer to AMD's 3D V-cache in the form of... pre-binned Core i9-12900KS? As per the report, Intel could be making a pre-binned, pre-overclocked Core i9-12900KS processor with an all-core turbo boost frequency of 5.2 GHz. This alleged clock speed will push the processor to some fantastic heights and increase the overall performance of the regular Core i9-12900K processor. With AMD's Ryzen processors with 3D V-cache incoming, Intel has prepared this solution to keep up with the increasing pressure from AMD.

So far, we don't know the specific requirements for Core i9-12900KS to reach 5.2 GHz. However, we assume that Voltage needs a big boost, making cooling and power supply requirements increase. This special edition Alder Lake design should launch around the same time frame that AMD reveals its 3D V-cache enabled Ryzen processors, so Intel doesn't let AMD steal the performance crown.


View at TechPowerUp Main Site

 

[T1beriu]

Speculation?! It's a rumor! Big difference.

 

[AleksandarK]

Speculation?! It's a rumor! Big difference.

Indeed! Fixed.

 

[Ja.KooLit]

recommended PSU would be minimum 1kw?

 

[TheinsanegamerN]

I'd rather see them bring back thermal velocity boost for a single core to hit 5.3-5.4 GHz. In multicore the standard 12700k and 12900k are incredible already, but to date the best out of box experience for single threaded games is the 10900k.

 

[Chrispy_]

This is entirely pointless for the overwhelming majority of even high-end enthusiasts with money to burn, right?

Even if Intel bin one that's stable at 5.2GHz it's going to need 400W from the socket minimum. Most reviews and forum posts seem to suggest that 1.45V is kind of mandatory for anything over 5GHz all-core, and @W1zzard hit 404W (full system) with his sample at just 5.0GHz and 1.4V. Binning can only get you so far and 5.2 all-core could very well require 500W+ from the socket, limiting it to those absolutely retarded $1000 motherboards given to all the streamers and vloggers in the desperate hope that they can provide enough exposure to snag some sales to the 0.01% for their 9000% profit, ultra-flagship motherboards.

Maybe I'm becoming cynical in my old age but when a 5950X Zen3 is boosting at 142W and the stock 12700K pulling ~200W from the socket is basically matching a heavily-overclocked 12900K in performance for the majority of workloads, you realise that these chips are just vanity objects to massage the fragile ego of the show-offs. The 12900KS' time in the spotlight before something better comes along is likely to be too short to justify the purchase on anything other than "I MUST HAVE THIS SHINY NEW BAUBLE".

If someone gave me a 12900KS for free, I'd sell it and buy a 12700K so that I could get 95% of the same result in a $200 motherboard with some DDR4. It only has any value whilst it's the king of the hill and that might even be under threat right now with Vermeer S (5950X with 3D vCache) already in mass production and expected within the next couple of months. Intel may have the clockspeed advantage and competitive IPC but they only go to 24 threads in the i9 and only 8 of the cores are P-cores. The older Zen3 tech will likely brute force the win with 16 "P-cores" and 32 threads, so long as the IPC gains from the vCache are even half of what was promised.

 

[Deleted member 185088]

This is entirely pointless for the overwhelming majority of even high-end enthusiasts with money to burn, right?

Even if Intel bin one that's stable at 5.2GHz it's going to need 400W from the socket minimum. Most reviews and forum posts seem to suggest that 1.45V is kind of mandatory for anything over 5GHz all-core, and @W1zzard hit 404W (full system) with his sample at just 5.0GHz and 1.4V. Binning can only get you so far and 5.2 all-core could very well require 500W+ from the socket, limiting it to those absolutely retarded $1000 motherboards given to all the streamers and vloggers in the desperate hope that they can provide enough exposure to snag some sales to the 0.01% for their 9000% profit, ultra-flagship motherboards.

Maybe I'm becoming cynical in my old age but when a 5950X Zen3 is boosting at 142W and the stock 12700K pulling ~200W from the socket is basically matching a heavily-overclocked 12900K in performance for the majority of workloads, you realise that these chips are just vanity objects to massage the fragile ego of the show-offs. The 12900KS' time in the spotlight before something better comes along is likely to be too short to justify the purchase on anything other than "I MUST HAVE THIS SHINY NEW BAUBLE".

If someone gave me a 12900KS for free, I'd sell it and buy a 12700K so that I could get 95% of the same result in a $200 motherboard with some DDR4. It only has any value whilst it's the king of the hill and that might even be under threat right now with Vermeer S (5950X with 3D vCache) already in mass production and expected within the next couple of months. Intel may have the clockspeed advantage and competitive IPC but they only go to 24 threads in the i9 and only 8 of the cores are P-cores. The older Zen3 tech will likely brute force the win with 16 "P-cores" and 32 threads, so long as the IPC gains from the vCache are even half of what was promised.

It is pointless because (for gaming at least) the GPU is by far the most important component.

 

[W1zzard]

thermal velocity boost

Yeah I wonder why they removed it in the first place, I tried to get an answer out of Intel but that part of the question was ignored

 

[Dr. Dro]

This is entirely pointless for the overwhelming majority of even high-end enthusiasts with money to burn, right?

Even if Intel bin one that's stable at 5.2GHz it's going to need 400W from the socket minimum. Most reviews and forum posts seem to suggest that 1.45V is kind of mandatory for anything over 5GHz all-core, and @W1zzard hit 404W (full system) with his sample at just 5.0GHz and 1.4V. Binning can only get you so far and 5.2 all-core could very well require 500W+ from the socket, limiting it to those absolutely retarded $1000 motherboards given to all the streamers and vloggers in the desperate hope that they can provide enough exposure to snag some sales to the 0.01% for their 9000% profit, ultra-flagship motherboards.

Maybe I'm becoming cynical in my old age but when a 5950X Zen3 is boosting at 142W and the stock 12700K pulling ~200W from the socket is basically matching a heavily-overclocked 12900K in performance for the majority of workloads, you realise that these chips are just vanity objects to massage the fragile ego of the show-offs. The 12900KS' time in the spotlight before something better comes along is likely to be too short to justify the purchase on anything other than "I MUST HAVE THIS SHINY NEW BAUBLE".

If someone gave me a 12900KS for free, I'd sell it and buy a 12700K so that I could get 95% of the same result in a $200 motherboard with some DDR4. It only has any value whilst it's the king of the hill and that might even be under threat right now with Vermeer S (5950X with 3D vCache) already in mass production and expected within the next couple of months. Intel may have the clockspeed advantage and competitive IPC but they only go to 24 threads in the i9 and only 8 of the cores are P-cores. The older Zen3 tech will likely brute force the win with 16 "P-cores" and 32 threads, so long as the IPC gains from the vCache are even half of what was promised.

Yeah, that is how I feel as well. Maybe we have gotten old. I will be quite content to flip my 5950X for its S version, but only because it is so hassle-free. There is equally little reason to and not to, truth be told if I wasn't about to give this processor to my brother for an upgrade, I probably would not bother.

Alder Lake's performance is solid, but the extreme power consumption and heat dissipation requirements quickly make it unviable for the climate I live in.

 

[ir_cow]

More or less even if you get a pre-binned 5.2 All-Core, it is going to take at least 1.35v and that is 350+ watts under Prime95. Good luck cooling it. My limit is 300~ with a waterblock. of course I could delid and go LM, but still can't absorb much more heat.

 

[seth1911]

5 GHz looks like the limit why i should buy a xyz Z Board for a K CPU to oc it for about 2, 3, 400 MHz.

Yeah kinda of trolling i buy a non K CPU on a B Board set the Turboclock to 5.2 GHz and thats it.



Or i go in the past may 10 years ago and u could OC a CPU via FSB/BCLK to its limit without K Cpu or Z chiptset,
and the oc was imrepssive like 45% higher performace via OC.

If there no news about new designs about semiconductor, x86 will be dead in the future.


Let me say one Thing:
2011 a dualcore with HT did go up to 5.4 GHz via Air Cooling
2021 a xyz Core hit it marks via Watercooling with 5.6 GHz

We are at the End

 

[Vayra86]

It is pointless because (for gaming at least) the GPU is by far the most important component.

It is pointless for MSDT.

ADL is a nice PoC for Intel, but realistically they still have overburdened Core at its heart, still trying to keep it from being EOL, which it has been since Skylake.

For mobile, sure. Maximize perf within limited power budget, and high versatility are key. But for MSDT its entirely pointless. Its marketing trying to kill Ryzen, which it really cant.

5 GHz looks like the limit why i should buy a xyz Z Board for a K CPU to oc it for about 2, 3, 400 MHz.

Yeah kinda of trolling i buy a non K CPU on a B Board set the Turboclock to 5.2 GHz and thats it.



Or i go in the past may 10 years ago and u could OC a CPU via FSB/BCLK to its limit without K Cpu or Z chiptset,
and the oc was imrepssive like 45% higher performace via OC.

If there no news about new designs about semiconductor, x86 will be dead in the future.


Let me say one Thing:
2011 a dualcore with HT did go up to 5.4 GHz via Air Cooling
2021 a xyz Core hit it marks via Watercooling with 5.6 GHz

We are at the End

ARM is going to hit the same walls. The node and the metal are the same things, and if ARM is going to have featureset parity, it will be the same thing. Its like inventing a wheel, good luck making it not a circle.

 

[InVasMani]

Do these come pre-binned with liquid nitrogen?

 

[phanbuey]

More or less even if you get a pre-binned 5.2 All-Core, it is going to take at least 1.35v and that is 350+ watts under Prime95. Good luck cooling it. My limit is 300~ with a waterblock. of course I could delid and go LM, but still can't absorb much more heat.

There are 12900K chips that hit 5.3 with ease under 1.3v - my 12600k does that as well. A binned chip won't need that much voltage to hit that clock.

I really wish they would move away from single core boost though. Almost all games now days use all your cores, so it would be nice if they boosted based on watts, not on core load.

 

[WorringlyIndifferent]

I really wish they would move away from single core boost though. Almost all games now days use all your cores, so it would be nice if they boosted based on watts, not on core load.

This is absolutely not true. Games are still overwhelmingly dominated by single core performance, and will be for the foreseeable future. Programming for multithreading is already hard enough, and on top of that, the gains from going multithreaded are almost never worth the massively increased expense and time spent programming it.

Multithreading *exists*, yes, and it is put to good use by putting lower priority operations on other cores to free up compute time on the primary core. But you are essentially never limited by the performance of the secondary cores - the core where the main game engine is running, is the limiting factor.

 

[phanbuey]

This is absolutely not true. Games are still overwhelmingly dominated by single core performance, and will be for the foreseeable future. Programming for multithreading is already hard enough, and on top of that, the gains from going multithreaded are almost never worth the massively increased expense and time spent programming it.

Multithreading *exists*, yes, and it is put to good use by putting lower priority operations on other cores to free up compute time on the primary core. But you are essentially never limited by the performance of the secondary cores - the core where the main game engine is running, is the limiting factor.

Correct... but what I am saying is your BOOST is dictated by how MANY cores are loaded. If the secondary cores are loaded with less important crap, then your boost isn't going to 5.1 or 5.2.

If you look at the clock graphs for a 12900k when running modern games, it sits at all core boost, even though those secondary threads aren't intensive loads (at all), and the chip could realistically be running at 5.3 - 5.4 without breaking thermal or wattage limits due to the low load of those ancillary threads.

So your main thread of the game is running at 100% and could definitely use more clock, but because the game spawns 4-16 threads your chip won't boost past however many simultaneous threads are in flight.

 

[WorringlyIndifferent]

Correct... but what I am saying is your BOOST is dictated by how MANY cores are loaded. If the secondary cores are loaded with less important crap, then your boost isn't going to 5.1 or 5.2.

If you look at the clock graphs for a 12900k when running modern games, it sits at all core boost, even though those secondary threads aren't intensive loads (at all), and the chip could realistically be running at 5.3 - 5.4 without breaking thermal or wattage limits due to the low load of those ancillary threads.

So your main thread of the game is running at 100% and could definitely use more clock, but because the game spawns 4-16 threads your chip won't boost past however many simultaneous threads are in flight.

Oh gotcha gotcha, I see what you're saying. True for Ryzen CPUs too, and really annoying. Current boost algorithms make the quoted single core max boost basically never actually happen in real life.

 

[Dr. Dro]

This is absolutely not true. Games are still overwhelmingly dominated by single core performance, and will be for the foreseeable future. Programming for multithreading is already hard enough, and on top of that, the gains from going multithreaded are almost never worth the massively increased expense and time spent programming it.

Multithreading *exists*, yes, and it is put to good use by putting lower priority operations on other cores to free up compute time on the primary core. But you are essentially never limited by the performance of the secondary cores - the core where the main game engine is running, is the limiting factor.

ADL-S may dominate raw IPC, but thus far the practical outcome of it vs. Zen 3 has been negligible at best with RTX 3090 hardware. I would place a safe wager on Vermeer-S closing this gap again and with relative finesse so to speak (aka won't be guzzling 400 watts of power to do so). Besides, if you want to have the strongest single-core performance, you don't buy a processor like the 5950X or the 12900K to begin with.

The high core count comes at a cost, the power budget per-core diminishes with every core added, even though binning keeps that from some extent, you won't be getting even the best 5950Xs doing 4.8 all-core like a 5600X will on most motherboards unless you're running an Aorus Xtreme or a C8DH/C8E with heavy-duty cooling anyway... my personal sample does ~4450MHz 32T on my B550-E Strix with a 360mm AIO (160A TDC/190A EDC cap) and it's a chip that refuses to go below -2 all-core CO, from all I can tell, it's a god tier bin chip.

 

[InVasMani]

I'd think the 5950x /129000K could get better single core performance, but you'd need to tweak it more to do so.

 

[looniam]

wonder what the price premium will be.

i suspect intel thinks there's money left on the table since silicon lottery closed up shop.

 

[RandallFlagg]

Oh gotcha gotcha, I see what you're saying. True for Ryzen CPUs too, and really annoying. Current boost algorithms make the quoted single core max boost basically never actually happen in real life.

So have you guys ever tried setting an all core turbo speed high, and power limits low? Like a 10900K at 5.4Ghz all core with a 180W max turbo power limit.

I've been thinking about this problem of only having one or two cores loaded, while preventing sudden spike overheating in the occasional all core load, is why I ask. I see what you're describing all the time not just in games but daily use, 12 vCores at 10% and 4 at 80-100%.

 

[Ruru]

And what's the real world performance difference? With clocks that high, a few hundred MHz difference is something you'll see only with synthetic benchmarks.

 

[dicktracy]

Weak! The competition does single core 5Ghz for a few seconds using 1.5v and superior TSMC 7nm process!

 

[AnarchoPrimitiv]

And what's the real world performance difference? With clocks that high, a few hundred MHz difference is something you'll see only with synthetic benchmarks.

Agreed, I've been demanding it for years, but SOMEBODY has to do some research into how large a framerate difference has to be before a typical human being can actually distinguish the difference, e.g. At 142 FPS, a 5% increase would mean 149 FPS, and I highly doubt very many human beings can distinguish that difference, if at all....so it'd be nice to have an actual experiment prove it, so that when a new CPU comes out, we can say with confidence that unless it outperforms another CPU in framerates by at least 15%, it's pointless. It'd be nice to know that people bragging over a 6% average difference in framerates are just wasting their time....conclusively.

 

[Ruru]

Agreed, I've been demanding it for years, but SOMEBODY has to do some research into how large a framerate difference has to be before a typical human being can actually distinguish the difference, e.g. At 142 FPS, a 5% increase would mean 149 FPS, and I highly doubt very many human beings can distinguish that difference, if at all....so it'd be nice to have an actual experiment prove it, so that when a new CPU comes out, we can say with confidence that unless it outperforms another CPU in framerates by at least 15%, it's pointless. It'd be nice to know that people bragging over a 6% average difference in framerates are just wasting their time....conclusively.

My opinion is basically that if you put a high end Intel and AMD systems next to each other with a same GPU, nobody would know the differences without an OSD/FPS counter.