News Leaked Intel Core i9-14900KS spec sheet all but confirms 6.2GHz boost clock

[Admin]

A leaked spec sheet has revealed that Intel is preparing to release a new flagship for its Raptor Lake Refresh CPU lineup, dubbed the i9-14900KS, sporting a clock speed of 6.2GHz.

Leaked Intel Core i9-14900KS spec sheet all but confirms 6.2GHz boost clock : Read more

 

[rluker5]

It gets tiring to remind:
1. Number of cores changed +
2. Amount of cache per core changed +
3. How the cores communicate with each other changed +
4. Ring/LLC speed no longer limited by enabled e-core speed
5. An across the board +500mhz clockspeed increase at the same node
=/= a refresh if Zen is not on refresh #4

 

[ingtar33]

It gets tiring to remind:
1. Number of cores changed +
2. Amount of cache per core changed +
3. How the cores communicate with each other changed +
4. Ring/LLC speed no longer limited by enabled e-core speed
5. An across the board +500mhz clockspeed increase at the same node
=/= a refresh if Zen is not on refresh #4

IT GETS EVEN MORE TIRING TO REMIND
same IPC as the 12000 series. set them all to 4ghz, and you get the same core performance within 1%
the 12000 series to 13000 series to 14000 series are all the same chip, just refined process, and more cores.
claiming this is anything but, is disingenuous to say the least. the difference in performance between these refreshes came from increased core count and increased clocks, not from improved design)

-----

Finally zen 1 -> zen 2 == 15% ipc increase
zen 2 -> zen 3 == 30% ipc increase
zen 3 -> zen 4 == 30% ipc increase
zen 4 -> zen 5 == ??? leaks are claiming 30%-50% ipc increase

thats not a refresh. the lone refresh in there was zen 2.5, when AMD released a small refresh between 2 and 3 (confusingly numbered the 3000 series, zen 3 was 5000)

finally this new 14900ks is just piledriver 9000 series lunacy. they're pushing 400W through that thing to get those 6ghz clock speeds. it's just a factory overclocked 14900k

 

[rluker5]

IT GETS EVEN MORE TIRING TO REMIND
same IPC as the 12000 series. set them all to 4ghz, and you get the same core performance within 1%
the 12000 series to 13000 series to 14000 series are all the same chip, just refined process, and more cores.
claiming this is anything but, is disingenuous to say the least. the difference in performance between these refreshes came from increased core count and increased clocks, not from improved design)

-----

Finally zen 1 -> zen 2 == 15% ipc increase
zen 2 -> zen 3 == 30% ipc increase
zen 3 -> zen 4 == 30% ipc increase
zen 4 -> zen 5 == ??? leaks are claiming 30%-50% ipc increase

thats not a refresh. the lone refresh in there was zen 2.5, when AMD released a small refresh between 2 and 3 (confusingly numbered the 3000 series, zen 3 was 5000)

finally this new 14900ks is just piledriver 9000 series lunacy. they're pushing 400W through that thing to get those 6ghz clock speeds. it's just a factory overclocked 14900k

I know you've heard things from AMD sales pitches, but here are the first search results I've found:

from: https://www.techspot.com/review/2552-intel-core-i9-13900k/ and that is ignoring the architectural changes that made e-cores useable in games, and also ignored the addition of 8 more e-cores. If e-cores were enabled as default and as these CPUs are tested for reviews the difference would be a multiple of what you see here. These differences had to be ignored to isolate the architectural changes to a portion of the chip which ignored the larger architectural changes that happened elsewhere. I know, not that impressive, but you can say the same about Ryzen from: https://www.techspot.com/article/2143-ryzen-5000-ipc-performance/ And this included all architectural improvements, not just a part.

They both had architectural changes so neither are refreshes. The 14th gen clearly is a refresh. I'm saying the 13th isn't. You can even look at die shots. Why does the 12th, 13th look different if they are the same? It isn't that hard to understand.

 

[ingtar33]

i really am just sighing right now. read your own sources. the problem is you're being bamboozled by graphs designed to exaggerate differences visually...
here i'll help you with the math.
259/245 == 5%
213/208 == 2%
215/208 == 3%
190/187 == 1.5%
165/159 == 3%
121/115 == 5%
average IPC uplift (6 games) from 12000 to 13000? 3%

you cannot seriously consider a 3% uplift in IPC to be anything more than a refresh.

------

now then, though this is an 8 game average, here is the ryzen examples you gave.
from Zen 2.5 to Zen 3 == 22%
from zen 2 to zen 2.5 == 14% (and this was the refresh mind you, i'm pretty sure the difference wasn't this big)
from zen 1 to zen 2 == 2% (and i know for a fact this ipc gain was significantly larger, so i don't know what games they were testing to get just a 2% uplift, must have been a title or two with almost no difference or a significant negative for some reason to get this result)

so what point are you trying to make? you proved the nearly 0% uplift in ipc from 12000 to 13000, and proved the significantly larger jumps between zen architectures (which you were trying to claim were refreshes), there is something weird about the games they used for the zen comparison though, as those numbers don't look right. for example, I'm pretty sure there wasn't a 15% improvement from zen 2 to 2.5... just as i'm pretty sure the ipc uplift from 1 to 2 was much larger then 2%, like significantly larger. But those are just small details, and in the end they don't matter for the overarching argument.

I am willing to bet if you hunt down 14000 locked to 5ghz for those same titles you'd see basically the same results as the 13000, with a result within the margin of error. I know this because the result of 3% ipc gain from 12000 to 13000 is basically the same number i saw for 12000 to 14000. it was under 5% for sure (I can't recall the exact number but it was definitely less then 5%)

 

[TerryLaze]

finally this new 14900ks is just piledriver 9000 series lunacy. they're pushing 400W through that thing to get those 6ghz clock speeds. it's just a factory overclocked 14900k

IT GETS EVEN MORE TIRING TO REMIND
that 6.2Ghz is for one single core and they don't need 400W to get one single core to 6.2Ghz.
The end user needs 400W, if not more, to get all 8 p-cores to 6Ghz+ but that's because it is physically possible to do, unlike ZEN where you just can't do that no matter how much power you put through them,mainly because they catch on fire waaaaay waaaay before reaching 400W.

 

[NinoPino]

IT GETS EVEN MORE TIRING TO REMIND
that 6.2Ghz is for one single core and they don't need 400W to get one single core to 6.2Ghz.
The end user needs 400W, if not more, to get all 8 p-cores to 6Ghz+ but that's because it is physically possible to do, unlike ZEN where you just can't do that no matter how much power you put through them,mainly because they catch on fire waaaaay waaaay before reaching 400W.

I'm unable to remind nothing to nobody but... 6.2GHz is the main reason of the 400W, single core or not. The fact that Zen cannot reach such frequencies are a design choice as it is a design choice of Intel to go to 6Ghz.
I cannot understand your point, are you asserting that Intel have a better CPU of AMD ? Or that 14900 is not so power hungry ? Or that Intel process node is better than TSMC ? (of course I disagree on all the points).
I often wonder who are the typical buyers of 1x900KS versions, professionals ? gamers ? overclockers ? arctic people ?
For every category there is a better choice. 😀

 

[hotaru251]

if intel's history is to go by...you'll rarely ever see these for mroe than 5seconds & only if you delid your cpu.

 

[rluker5]

i really am just sighing right now. read your own sources. the problem is you're being bamboozled by graphs designed to exaggerate differences visually...
here i'll help you with the math.
259/245 == 5%
213/208 == 2%
215/208 == 3%
190/187 == 1.5%
165/159 == 3%
121/115 == 5%
average IPC uplift (6 games) from 12000 to 13000? 3%

you cannot seriously consider a 3% uplift in IPC to be anything more than a refresh.

------

now then, though this is an 8 game average, here is the ryzen examples you gave.
from Zen 2.5 to Zen 3 == 22%
from zen 2 to zen 2.5 == 14% (and this was the refresh mind you, i'm pretty sure the difference wasn't this big)
from zen 1 to zen 2 == 2% (and i know for a fact this ipc gain was significantly larger, so i don't know what games they were testing to get just a 2% uplift, must have been a title or two with almost no difference or a significant negative for some reason to get this result)

so what point are you trying to make? you proved the nearly 0% uplift in ipc from 12000 to 13000, and proved the significantly larger jumps between zen architectures (which you were trying to claim were refreshes), there is something weird about the games they used for the zen comparison though, as those numbers don't look right. for example, I'm pretty sure there wasn't a 15% improvement from zen 2 to 2.5... just as i'm pretty sure the ipc uplift from 1 to 2 was much larger then 2%, like significantly larger. But those are just small details, and in the end they don't matter for the overarching argument.

I am willing to bet if you hunt down 14000 locked to 5ghz for those same titles you'd see basically the same results as the 13000, with a result within the margin of error. I know this because the result of 3% ipc gain from 12000 to 13000 is basically the same number i saw for 12000 to 14000. it was under 5% for sure (I can't recall the exact number but it was definitely less then 5%)

First of all not all of the architectural changes were accounted for in the comparison of 12th to 13th gen, the p-cores were isolated. If the e-cores were run the ring would have been 1ghz slower and ram latency about 5ns higher at the same clocks for Alder with e-cores enabled, and the time to communicate to fewer e-cores was much higher in Alder. And you don't get a universal 10% clockspeed increase with just a refresh. You get chips overlapping, like my ordinary 13900kf being able to run stock 14900k speeds if I OC it. There are no Alder chips that can approach their Raptor counterparts without LN2.

And if physical changes to the number of cores, cache and communication lines don't count? If only IPC counts then Meteor Lake is also a refresh of Raptor Lake, Zen 4 is also a refresh of Alder Lake and Zen 3 is a refresh of a refresh of Skylake.
Why aren't you claiming that AMD ripped off the Skylake arch with Zen 3? Probably because there are differences other than IPC.

 

[TerryLaze]

I'm unable to remind nothing to nobody but... 6.2GHz is the main reason of the 400W, single core or not. The fact that Zen cannot reach such frequencies are a design choice as it is a design choice of Intel to go to 6Ghz.
I cannot understand your point, are you asserting that Intel have a better CPU of AMD ? Or that 14900 is not so power hungry ? Or that Intel process node is better than TSMC ? (of course I disagree on all the points).
I often wonder who are the typical buyers of 1x900KS versions, professionals ? gamers ? overclockers ? arctic people ?
For every category there is a better choice. 😀

13900ks uses 253W just like the 13900k, just like the 14900k, and the 14900ks will also be 253W.
Single core 6.2Ghz will be much much much lower than 253W
The highest you can make the 14900k go in single core is 54W (and it also shows that the 14900k already uses 400W+ if you let it) and even if the KS is going to be higher it will only be by a little bit, it's not going to go from 54W to 400W ........

 

[CmdrShepard]

Wake me up when Intel changes focus to power efficiency because this is getting ridiculous.

I suspect I'll need cryo sleep until then.

 

[Amdlova]

My IPC today is Balanced power on windows, 35W CPU locked and Four Mighty cores.
And yes running all games and programs without any drawbacks.

 

[NinoPino]

13900ks uses 253W just like the 13900k, just like the 14900k, and the 14900ks will also be 253W.
Single core 6.2Ghz will be much much much lower than 253W
The highest you can make the 14900k go in single core is 54W (and it also shows that the 14900k already uses 400W+ if you let it) and even if the KS is going to be higher it will only be by a little bit, it's not going to go from 54W to 400W ........

For 13900KS you say 253 W but...
Tom's Hardware wrote 320 W : "Intel Core i9-13900KS Review: The World's First 6 GHz 320W CPU"
HwUpgrade review : 80 W in single core 350 W at stock and 400 W with power limit removed.

For 14900KS, the test unit used 409 W at stock speeds (it seems the turbo have a maximum limit of 56 seconds).

 

[TerryLaze]

For 13900KS you say 253 W but...
Tom's Hardware wrote 320 W : "Intel Core i9-13900KS Review: The World's First 6 GHz 320W CPU"
HwUpgrade review : 80 W in single core 350 W at stock and 400 W with power limit removed.

For 14900KS, the test unit used 409 W at stock speeds (it seems the turbo have a maximum limit of 56 seconds).

Yes, they are unlocked CPUs and can go above 253W , nothing nowhere nobody said that the 14900ks was running at stock, mobo stock maybe but that's just auto overclocking.
Actually the article itself says that.

It's worth mentioning that these chips only pull these crazy high-power numbers with the default profiles in motherboard firmware. If you enforce Intel's specified power limits manually in BIOS, these chips will consume far less power (at the cost of multi-core CPU performance). Enforcing default power limits can not only save power and make the chips run more efficiently, but it can also improve stability for chips that have lost the silicon lottery.

 

[CmdrShepard]

My IPC today is Balanced power on windows, 35W CPU locked and Four Mighty cores.
And yes running all games and programs without any drawbacks.

You don't need four cores to run Solitaire or Candy Crush.

 

[Amdlova]

You don't need four cores to run Solitaire or Candy Crush.

2560x1080 max 60 fps locked helldivers 2

 

[ingtar33]

and if physical changes to the number of cores, cache and communication lines don't count? If only IPC counts then Meteor Lake is also a refresh of Raptor Lake, Zen 4 is also a refresh of Alder Lake and Zen 3 is a refresh of a refresh of Skylake.

no, my point was the design changes from 12th to 13th to 14th gen were so minor they accomplished zero IPC uplift (or close enough) from 12th to 14th gen. it's all just one big refresh.

obviously design changes can result in a new chip, but when the underlaying architecture is the same, and the ipc doesn't change it looks like a refresh to pretty much everyone. from what i recall intel described 13th gen as a refinement of 12th, even intel wouldn't claim it was a new chip, just a refinement, a minor one at that. what performance gain there was for 13th gen was mostly from clock speed improvements. same for 14th gen.

Why aren't you claiming that AMD ripped off the Skylake arch with Zen 3? Probably because there are differences other than IPC.

well skylake is a radically different chip design; claiming it's a refresh when it's not made by the same company and doesn't even have the same design or chip performance envelope is sorta silly.


Finally one more point. those arguing intel is "ahead" because of their 6ghz chips you do know how silly that sounds in the wake of piledriver right? I mean the 9000 series for piledrive was basically marketed the same way "look we have 5ghz!, intel doesnt! we're faster!" yet everyone knew this was deceptive to say the least. For one the intel chips running on 65-80W and 4.5ghz out performed the AMD Piledriver at 5.5ghz and 300W. for two, no one liked heating their appt with their cpu.

so pointing to 6.2ghz single core isn't really impressing that many people.

 

[TerryLaze]

no, my point was the design changes from 12th to 13th to 14th gen were so minor they accomplished zero IPC uplift (or close enough) from 12th to 14th gen. it's all just one big refresh.

obviously design changes can result in a new chip, but when the underlaying architecture is the same, and the ipc doesn't change it looks like a refresh to pretty much everyone. from what i recall intel described 13th gen as a refinement of 12th, even intel wouldn't claim it was a new chip, just a refinement, a minor one at that. what performance gain there was for 13th gen was mostly from clock speed improvements. same for 14th gen.

Is that a bad thing for you?!
Intel makes advancements in the nodes and instead of raising prices like TSMC does (they did it like 3-4 times in the last few years) they just release a new gen at the exact same prices, which also makes the older gens cheaper in retail even though they are basically the same thing.
So the customer wins in at least two ways.

Finally one more point. those arguing intel is "ahead" because of their 6ghz chips you do know how silly that sounds in the wake of piledriver right? I mean the 9000 series for piledrive was basically marketed the same way "look we have 5ghz!, intel doesnt! we're faster!" yet everyone knew this was deceptive to say the least. For one the intel chips running on 65-80W and 4.5ghz out performed the AMD Piledriver at 5.5ghz and 300W. for two, no one liked heating their appt with their cpu.

so pointing to 6.2ghz single core isn't really impressing that many people.

Yeah the only problem with that analogy is that it is completely wrong...
the 14900k runs one core at 6Ghz with 35W while the 7950x needs 41W for 5.75Ghz
The p-cores are intel's advanced node and architecture and those are ahead, in performance, in clocks, and in efficiency.

https://www.techpowerup.com/review/...ke-tested-at-power-limits-down-to-35-w/8.html

 

[rluker5]

no, my point was the design changes from 12th to 13th to 14th gen were so minor they accomplished zero IPC uplift (or close enough) from 12th to 14th gen. it's all just one big refresh.

obviously design changes can result in a new chip, but when the underlaying architecture is the same, and the ipc doesn't change it looks like a refresh to pretty much everyone. from what i recall intel described 13th gen as a refinement of 12th, even intel wouldn't claim it was a new chip, just a refinement, a minor one at that. what performance gain there was for 13th gen was mostly from clock speed improvements. same for 14th gen.


well skylake is a radically different chip design; claiming it's a refresh when it's not made by the same company and doesn't even have the same design or chip performance envelope is sorta silly.


Finally one more point. those arguing intel is "ahead" because of their 6ghz chips you do know how silly that sounds in the wake of piledriver right? I mean the 9000 series for piledrive was basically marketed the same way "look we have 5ghz!, intel doesnt! we're faster!" yet everyone knew this was deceptive to say the least. For one the intel chips running on 65-80W and 4.5ghz out performed the AMD Piledriver at 5.5ghz and 300W. for two, no one liked heating their appt with their cpu.

so pointing to 6.2ghz single core isn't really impressing that many people.

To put the zero IPC uplift to the test, let's do a bit of arithmetic. You are good at that. Taking numbers from this chart:

The 13900k is 206/172, or 1.198x as fast as the 12900k in games at stock boost speeds. At max framerate, for most games all core boost is the effective speed. Stock all core boost for 13900k p-cores is 5.4 ghz. For the 12900k it is 5.1 ghz, or 1.059x as fast (5.4/5.1)

That looks like a 14% IPC boost to me when you equalize for clockspeed. A 14% boost in performance in loads that rely almost exclusively on the P-cores when you subtract the clockspeed gains sure looks like more than a refresh to me. With a refresh, like 13th gen to 14th gen, there is 0% performance boost when you subtract the gain from clockspeed. 14% is different than 0%. It is larger because things were changed going from Alder to Raptor. To make the whole chip work better, similar to when Zen went from all cores having access to half of the L3 on a chiplet to all of the L3 on a chiplet. Just clockspeeds were changed going from Raptor to Raptor refresh.

 

[ingtar33]

To put the zero IPC uplift to the test, let's do a bit of arithmetic. You are good at that. Taking numbers from this chart:

The 13900k is 206/172, or 1.198x as fast as the 12900k in games at stock boost speeds. At max framerate, for most games all core boost is the effective speed. Stock all core boost for 13900k p-cores is 5.4 ghz. For the 12900k it is 5.1 ghz, or 1.059x as fast (5.4/5.1)

That looks like a 14% IPC boost to me when you equalize for clockspeed. A 14% boost in performance in loads that rely almost exclusively on the P-cores when you subtract the clockspeed gains sure looks like more than a refresh to me. With a refresh, like 13th gen to 14th gen, there is 0% performance boost when you subtract the gain from clockspeed. 14% is different than 0%. It is larger because things were changed going from Alder to Raptor. To make the whole chip work better, similar to when Zen went from all cores having access to half of the L3 on a chiplet to all of the L3 on a chiplet. Just clockspeeds were changed going from Raptor to Raptor refresh.

14% uplift because of 15% more cores. seriously tiring arguing with people too blinded by a team shirt color to think for 5 seconds. you talk like intel's marketing dept.

 

[rluker5]

14% uplift because of 15% more cores. seriously tiring arguing with people too blinded by a team shirt color to think for 5 seconds. you talk like intel's marketing dept.

So now e-cores contribute to games? Their contribution is generally a couple % with all on vs all off. Going from 8 to 16 e-cores should make negligible difference. It might even hurt performance. My point stands.

 

[lolvatveo]

Why GPU rtx 4090 can use 400W but not CPU? NVIDIA pause plan release rtx 4090ti 500W beacase it almost melt heatsink, they are upgrading and make new heatsink. you should accept both cpu and gpu can use lot of energy. why happy with rtx 4090 but not cpu

 

[TerryLaze]

14% uplift because of 15% more cores. seriously tiring arguing with people too blinded by a team shirt color to think for 5 seconds. you talk like intel's marketing dept.

Well 15% more cores means that you can run 15% more instruction per cycle...
That's what IPC is. The only way to increase IPC is to have more hardware that executes instructions. If you do that by making one core wider or by adding more cores doesn't really make any difference to us as end users.

Why GPU rtx 4090 can use 400W but not CPU? NVIDIA pause plan release rtx 4090ti 500W beacase it almost melt heatsink, they are upgrading and make new heatsink. you should accept both cpu and gpu can use lot of energy. why happy with rtx 4090 but not cpu

It's about what you get for the power, 400W of GPU will get you more performance, but 400W on the CPU just makes it hot, you do not get any more performance out of it compared to running it at 250W.