Intel Core i9-14900KS Pricing Confirmed to be $749

[bug]

These are the perfect cpus for those that care about efficiency actually. Better bins = less power at same clockspeeds. This will be significantly more efficient than eg. a 13900k.

Even if a significant proportion of the prospective buyers would care about that, I don't expect efficiency to be 10% than the non-S model. But we'll never have the benches to prove either way, so let's just stop here.

 

[JustBenching]

Even if a significant proportion of the prospective buyers would care about that, I don't expect efficiency to be 10% than the non-S model. But we'll never have the benches to prove either way, so let's just stop here.

The 13900k needs 25% more power than a 14900k for 5.5 ghz (280 to 350). Of course each individual sku is different but both of the cpus I tested were horrible bins.

 

[MaMoo]

I'd love to see some numbers attached to this but considering that idle power is ignored in so many reviews out there I don't have hope to find a deep dive on actual numbers in different settings. Sucks as high idle power was the biggest annoyance when I built my first AMD system, R5 2600.

My Intel i5-8400 idled at 31W total system which dropped to 25W when adding an AMD GPU (same effect as the Dell above). But the R5 2600 used ~45W. 41W with the 5800X3D seems decent in comparison, considering it's increased capability.

My 12700k on a Asus Prime Matx idles at 46 watts with 5600 Mhz RAM with a Gold PSU. I thought I was doing well but your numbers are better by 10% on your X3D. This is with all the ecores disabled and no OC.

 

[bug]

My 12700k on a Asus Prime Matx idles at 46 watts with 5600 Mhz RAM with a Gold PSU. I thought I was doing well but your numbers are better by 10% on your X3D. This is with all the ecores disabled and no OC.

Well, idle power draw is the first casualty of disabling the E-cores, what did you expect?

 

[MaMoo]

Well, idle power draw is the first casualty of disabling the E-cores, what did you expect?

It was the same before. I checked out of the box. After I disabled my ecores, I got 11% better Linpack scores which is what I need my rig for.

 

[bug]

It was the same before. I checked out of the box. After I disabled my ecores, I got 11% better Linpack scores which is what I need my rig for.

Strange, but ok.
I didn't mean to imply you shouldn't disable E-cores, I did disable mine at first.

 

[GhostRyder]

It's very much a perception issue, not a competitive issue in my opinion. Like with this thread. Where people look at the red herring of peak synthetic power draw numbers and conclude intel=inefficient. When the reality is at any given moment an Intel chip is more likely to be using less power than an equivalent AMD one. There's also the fact that they're using an inferior process and what's basically a tweaked three year old architecture, yet still have the IPC advantage and competitive gaming performance.

Besides, while it's true what you've mentioned about the build it yourself community, that's very much the minority in terms of actual chips sold, e.g. Laptops, business, prebuilts etc. Intel also has what seems to be a contemporary fab business, arguably better than Samsung anyway and maybe competitive with TSMC if things pan out. I'd argue this is a better situation than AMD, which has to fight Apple for second dibs on the second tier TSMC capacity, has some mindshare for CPUs amongst youtubers and techheads, but is a massive minority for GPUs and can't even get AI features/upscaling to work better over three generations than Intel did in one generation.

Intel stock is popular in the private communities I'm in for these reasons and more.

I'm also very impressed with Intel drivers and the pace of their improvement within a single generation, open-source development, their Linux flavour "Clear Linux" etc.

Not to mention a stock Intel chip running at almost 400 W under an AIO still runs cooler than a Zen 4 chip which peaks at 95 C even at 100 W.

Unreleased Intel Core i9-14900KS Already De-lidded, 10°C Temperature Drop On Offer

Intel may be launching its enthusiast-segment desktop processor, the Core i9-14900KS Special Edition next week, but the chip is already in the hands of several elite overclockers, such as Pakhtunov from OCN. Once he got his hands on one of these, Pakhtunov wasted little time in de-lidding the...

www.techpowerup.com

(He tested it before and after delidding).

With both backside power delivery and ribbonFET/gate all around transistors coming with 15th gen, besides a more advanced version of the chiplets AMD has (foveros with tiles), I think Intel is going to be coming out of this perception swinging, with the momentum a company of their heritage can muster.

I get and agree with part of what your saying, however I don't agree idle is more important than peak overall at least in cases with at least these higher end chips. I mean we are not buying 14900k's to browse reddit and watch youtube, were gaming and running heavy hitting applications on these chips. A lot of Intel's efficiency is also dependent on which core the application decides to use/the chip decides to let it use because the E-Cores are much better at efficiency. I would be curious to see on a lot of the reviews out there which core it decided to use (An E-Core or a P-Core) because of some of the reviews I have read.

As for the amount of heat on the chips, I was under the impression alot of that has to do with either the X3D chips cache or the temp sensors location on the chip. I don't think the heat output of the 14900K is lower than the 7950X for instance.

As far as market share, its amazing the share they have considering where they started from after the early to mid 2010's. Plus a lot of the market share at least in the business space is going to be based on life cycle of the products/availability of other alternatives. I think in the next couple of years its just going to keep changing as business start more heavily cycling out devices that are 5+ years old.

Still though, in this chips case the power consumption/efficiency only matters in respect to how much more performance you get from this binned chip in the overclocking field. I almost would love to buy it just to have fun pushing it.

 

[dgianstefani]

I get and agree with part of what your saying, however I don't agree idle is more important than peak overall at least in cases with at least these higher end chips. I mean we are not buying 14900k's to browse reddit and watch youtube, were gaming and running heavy hitting applications on these chips. A lot of Intel's efficiency is also dependent on which core the application decides to use/the chip decides to let it use because the E-Cores are much better at efficiency. I would be curious to see on a lot of the reviews out there which core it decided to use (An E-Core or a P-Core) because of some of the reviews I have read.

As for the amount of heat on the chips, I was under the impression alot of that has to do with either the X3D chips cache or the temp sensors location on the chip. I don't think the heat output of the 14900K is lower than the 7950X for instance.

As far as market share, its amazing the share they have considering where they started from after the early to mid 2010's. Plus a lot of the market share at least in the business space is going to be based on life cycle of the products/availability of other alternatives. I think in the next couple of years its just going to keep changing as business start more heavily cycling out devices that are 5+ years old.

Still though, in this chips case the power consumption/efficiency only matters in respect to how much more performance you get from this binned chip in the overclocking field. I almost would love to buy it just to have fun pushing it.

The E cores are in fact not much better in power efficiency, because efficiency is speed/power used, the P cores are so much faster that the increased power draw evens out. The actual peak efficiency point of the E cores is approximately 3.2 GHz, but they are not run at that speed. The E cores are a solution to area optimizing and therefore having enough cores to get rid of hyperthreading, which is a security issue and can lead to slightly worse single threaded performance.

Temperature ≠ heat output. The issue raised is that the Intel CPU/platform design is so much better than the AM5 equivalent, that Intel chips can output 4x the heat, yet still have lower temperatures, because the heat transfer wasn't sabotaged by an engineering design that maintained cooler compatibility with the older generation, rather than just getting end users to install a spacer washer kit, new back plates or something to that end.

 

[GhostRyder]

The E cores are in fact not much better in power efficiency, because efficiency is speed/power used, the P cores are so much faster that the increased power draw evens out. The actual peak efficiency point of the E cores is approximately 3.2 GHz, but they are not run at that speed. The E cores are a solution to area optimizing and therefore having enough cores to get rid of hyperthreading, which is a security issue and can lead to slightly worse single threaded performance.

Temperature ≠ heat output. The issue raised is that the Intel CPU/platform design is so much better than the AM5 equivalent, that Intel chips can output 4x the heat, yet still have lower temperatures, because the heat transfer wasn't sabotaged by an engineering design that maintained cooler compatibility with the older generation, rather than just getting end users to install a spacer washer kit, new back plates or something to that end.

I probably should have worded that better, I meant that they use less power because their boost is limited vs the P-Cores not that if clocked to the same speeds the E-Cores are more efficient. The point I was making is that the Task Scheduler decides which core to use and that could influence results is all (Not saying I have evidence, just based on reviews I have seen out there that could be a factor).

As for the heat, I was mostly just meaning the sensor location can influence those results. However after remembering it was the opposite of what I was thinking I retract that (IE they are lower than the Intel threshold, I thought opposite when I posted but remembered).

 

[dgianstefani]

Even clocked 1-2 GHz higher, the P cores are still more power efficient than E cores clocked above 3.2 GHz.

This can be easily verified by observing CPU power while locking tasks to either P or E core clusters, and comparing benchmark scores with this method.

E cores are great but they're area efficient, only power efficient if clocked conservatively, which they're not on the CPUs we're discussing.

 

[GhostRyder]

Even clocked 1-2 GHz higher, the P cores are still more power efficient than E cores clocked above 3.2 GHz.

This can be easily verified by observing CPU power while locking tasks to either P or E core clusters, and comparing benchmark scores with this method.

E cores are great but they're area efficient, only power efficient if clocked conservatively, which they're not on the CPUs we're discussing.

Ok, but are you saying that at 4.4 max boost on an E-core that it uses more power than the 6.0 on a P-Core? I know these are the ridiculous editions and are clocked higher for more performance, but I have not seen evidence to show that the P-cores at 6.0 are using less power than the E-cores at 4.4 (In a core to core test). That was all I meant by what I was comparing, not which core specifically was more efficient at the same speeds, just at the speeds they are allowed to hit.

 

[dgianstefani]

Ok, but are you saying that at 4.4 max boost on an E-core that it uses more power than the 6.0 on a P-Core? I know these are the ridiculous editions and are clocked higher for more performance, but I have not seen evidence to show that the P-cores at 6.0 are using less power than the E-cores at 4.4 (In a core to core test). That was all I meant by what I was comparing, not which core specifically was more efficient at the same speeds, just at the speeds they are allowed to hit.

Do you understand the meaning of efficiency? A core can use more power but also be significantly faster, so as long as it does more work than a slower core while using the same or less total amount of power, the faster higher power core is more efficient.

The faster core does the work in less time, so it's consuming more energy for a shorter time period, rather than less energy per second, but for enough seconds it's more energy in total.

 

[GhostRyder]

Do you understand the meaning of efficiency? A core can use more power but also be significantly faster, so as long as it does more work than a slower core while using the same or less total amount of power, the faster higher power core is more efficient.

The faster core does the work in less time, so it's consuming more energy for a shorter time period, rather than less energy per second, but for enough seconds it's more energy in total.

Ok, I don't think you and I are on the same page in terms of the points we are trying to say. I am not arguing efficiency of the cores overall(Yes I understand the definition of efficiency), I am saying in the testing showing single core power usage, the core being used matters not because one is going to complete it faster or which core is more efficient, just that it will influence the displayed power consumption of the system and that test at that time. I am just strictly arguing the results showing power usage in single thread doing a specific task may show a system consuming 45 watts (I am making these numbers up, not quoting anything) however it might be 55 watts in another application because it depends on whether its using the E-Core for the task or the P-Core. I am not arguing that the P-Core will complete the task more efficiently, just that the tests results showing power consumption of the system at that moment maybe influenced.

I am not arguing against P-Cores being a more efficient/Binned cores versus E-Cores.

 

[RandallFlagg]

My 12700k on a Asus Prime Matx idles at 46 watts with 5600 Mhz RAM with a Gold PSU. I thought I was doing well but your numbers are better by 10% on your X3D. This is with all the ecores disabled and no OC.

That's at the wall, right? 46W at the wall, for motherboard/CPU/GPU fans etc. at idle, is a good low number.

 

[MaMoo]

Strange, but ok.
I didn't mean to imply you shouldn't disable E-cores, I did disable mine

That's at the wall, right? 46W at the wall, for motherboard/CPU/GPU fans etc. at idle, is a good low number.

Yeah

That's at the wall, right? 46W at the wall, for motherboard/CPU/GPU fans etc. at idle, is a good low number.

Yeah it is. I am fine with it. I only leave the PC on when I use it and off otherwise.

Strange, but ok.
I didn't mean to imply you shouldn't disable E-cores, I did disable mine at first.

No worries at all. My workloads are unfortunately not able to be scaled to hybrid architectures. So I found out that with the ecores on, I lost a lot of performance because processes were late to finish on them while the pcores sat idle. I could not find a workaround so I used the BIOS key option to park the ecores by pressing a key. Later I learned from someone here that the whole CPU shares a power budget and if ecores were disabled, the pcores can boost higher. So I went into the BIOS and hard disabled the ecores. I guess because my workload cannot be asynchronously scheduled, the pcores were all I benefitted from. So that is why I did this and left it this way. I would use the ecores for video encode, which is scalable to hybrid architectures but I do not do that kind of work.

 

[bug]

No worries at all. My workloads are unfortunately not able to be scaled to hybrid architectures. So I found out that with the ecores on, I lost a lot of performance because processes were late to finish on them while the pcores sat idle. I could not find a workaround so I used the BIOS key option to park the ecores by pressing a key. Later I learned from someone here that the whole CPU shares a power budget and if ecores were disabled, the pcores can boost higher. So I went into the BIOS and hard disabled the ecores. I guess because my workload cannot be asynchronously scheduled, the pcores were all I benefitted from. So that is why I did this and left it this way. I would use the ecores for video encode, which is scalable to hybrid architectures but I do not do that kind of work.

Stranger still. What kind of workload is that? I'm asking because the scheduler is supposed to be smart enough to move intensive work from E-cores to available P-cores. (I know, I know, it doesn't always do what it's supposed to.) Is it linpack you mentioned earlier?

 

[donmclarenson]

On the one side there'll be people who are gonna love it, and on the other side there'll be the group that can't afford it. The amount of good silverware I forgot I had will determine which camp I belong to.

 

[below ambient]

You would really need a good reason to buy this over a $593 7950X3D.

AMD Ryzen 9 7950X3D - Ryzen 9 7000 Series 16-Core 4.2 GHz Socket AM5 120W CPU - Newegg.com

Buy AMD Ryzen 9 7950X3D - Ryzen 9 7000 Series Zen 4 16-Core 4.2 GHz - Socket AM5 120W - AMD Radeon Graphics Desktop Processor - 100-100000908WOF with fast shipping and top-rated customer service. Once you know, you Newegg!

www.newegg.com

its faster when tuned

 

[JustBenching]

You would really need a good reason to buy this over a $593 7950X3D.

AMD Ryzen 9 7950X3D - Ryzen 9 7000 Series 16-Core 4.2 GHz Socket AM5 120W CPU - Newegg.com

Buy AMD Ryzen 9 7950X3D - Ryzen 9 7000 Series Zen 4 16-Core 4.2 GHz - Socket AM5 120W - AMD Radeon Graphics Desktop Processor - 100-100000908WOF with fast shipping and top-rated customer service. Once you know, you Newegg!

www.newegg.com

It's a lot faster in both ST and MT performance. What more reason would you need?

 

[bug]

its faster when tuned

It's a lot faster in both ST and MT performance. What more reason would you need?

Just at the power draw while gaming. It looks like the 7950X3D isn't even trying...

 

[JustBenching]

Just at the power draw while gaming. It looks like the 7950X3D isn't even trying...

Yes, because half the cpu is off. That's a tapped out 7950x 3d, you can't really make it any faster. The 14900ks has has 20% headroom just by tuning and ram tuning.

 

[bug]

Yes, because half the cpu is off.

That's like me arguing that one fighter beats another without breaking much sweat and you saying: "of course he didn't break much sweat, he beat him using only one hand".

That's a tapped out 7950x 3d, you can't really make it any faster.

It's fast enough already. It's within 1% in 4k gaming and within 10% for office productivity. I wouldn't pick Intel over this unless I was going for something very, very specific.

The 14900ks has has 20% headroom just by tuning and ram tuning.

Has that been tested and confirmed? TPU's own review says it hits 100C on air ootb, I should think that makes extracting additional 20% rather unlikely. Then again, there's undervolting.

 

[R0H1T]

And the reason you can't tune AMD chips? I assume people think AMD doesn't have much headroom either

 

[bug]

And the reason you can't tune AMD chips? I assume people think AMD doesn't have much headroom either

The 3D cache limits what you can tune on those chips.

 

[JustBenching]

That's like me arguing that one fighter beats another without breaking much sweat and you saying: "of course he didn't break much sweat, he beat him using only one hand".

The point is you can disable half the 14900k in games too with no impact in performance and drop power draw. Just turning HT off drops 50 watts + in something like tlou

It's fast enough already. It's within 1% in 4k gaming and within 10% for office productivity. I wouldn't pick Intel over this unless I was going for something very, very specific.

And how much more power would it need to close that 10% gap? That's the relevant question. Or doing the opossite, how much less power would the 14900ks need to be 10% slower than it currently is?

as that been tested and confirmed? TPU's own review says it hits 100C on air ootb, I should think that makes extracting additional 20% rather unlikely. Then again, there's undervolting.

What do you mean tested? 20% is easily gained just by memory tuning. I've posted some actual footage, a tuned 14900k at 95 watts is faster than it is out of the box at 200 watts in games.

And the reason you can't tune AMD chips? I assume people think AMD doesn't have much headroom either

Locked frequencies, can't disable HT on an 8 core chip and doesn't get much performance from memory tuning cause the 3d cache already covers that.