Discussion Intel current and future Lakes & Rapids thread

[Dayman1225]

Was this ever confirmed to be real? People have faked slides for years. It's not difficult to do.

I had to get permission to post this slide because I was unsure if it was public or not. I was given the go ahead since it’s so old, it is real whether or not you believe it is up to you

 

[A///]

I had to get permission to post this slide because I was unsure if it was public or not. I was given the go ahead since it’s so old, it is real whether or not you believe it is up to you

Ok. It is very old in that case. The first time I heard of that 8+32 setup was long before the pandemic began or a few months into it. This was before the general public new of the setup and that only came a few months before alderlake launched. My guess is it was shelved due to delays and power use and focusing on dc where margins matter because intel still holds 80% in client including intel based macs.

It might be viable with Intel's tile approach but their tile approach and breadth of ip make it a sticky point imo.

 

[Dayman1225]

Ok. It is very old in that case. The first time I heard of that 8+32 setup was long before the pandemic began or a few months into it. This was before the general public new of the setup and that only came a few months before alderlake launched. My guess is it was shelved due to delays and power use and focusing on dc where margins matter because intel still holds 80% in client including intel based macs.

It might be viable with Intel's tile approach but their tile approach and breadth of ip make it a sticky point imo.

Yeah 8+32 has been dead for a long long time

 

[A///]

Yeah 8+32 has been dead for a long long time

If Intel can pull a 16+32 setup with HT and AVX512 plus AVX10 in future I pray for AMD's engineers.

 

[igor_kavinski]

If Intel can pull a 16+32 setup with HT and AVX512 plus AVX10 in future I pray for AMD's engineers.

Challenge will be to do it ALL in 500W

Come on, armchair engineers, do the math. Can Intel do it in 500W?

 

[A///]

Challenge will be to do it ALL in 500W

Come on, armchair engineers, do the math. Can Intel do it in 500W?

Xeon can. You can bring power use down, you won't get it so low that it'll cause great intrigue. It's how some people bought into the lie of low power ARM servers. Isn't really low power once you crank the frequency up to get the results you want. It's lower than x86, but not by much.

 

[lightisgood]

Core i9-14900KF Beats Ryzen 9 7950X3D, Core i9-13900KS in Geekbench 6

Flagship Raptor Lake Refresh impresses in Geekbench 6 benchmark.

www.tomshardware.com

ASRock Z790 Taichi vs ASUS System Product Name - Geekbench

browser.geekbench.com

 

[SiliconFly]

I’m fairness 8+32 was meant to exist but it got axed

8+32 is very much possible

 

[SiliconFly]

Threadripper CPUs have had 64-cores for a while. Obviously grandma doesn't need one, but professional users can use that kind of processing power.

The most exciting thing about Intel's e-cores is that it gave them a way back into HEDT where they basically gave up trying to compete because they couldn't make a monolithic die that had even a quarter as many cores.

Putting out a die with 64 e-cores is definitely possible even on their older node, but shouldn't be an issue for them going forward. Even though it's a niche market and e-cores may not stand up against a full Zen core, there's so much room for Intel to compete on price and power that I think they could carve out part of that market.

They can release a simple 48 E-cores cpu very easily. It's just that there are no takers. Even a 64 E-cores is very much possible. But frankly, except a few, no one else would buy.

 

[uzzi38]

Was this ever confirmed to be real? People have faked slides for years. It's not difficult to do.

It's real.

 

[SiliconFly]

If Intel can pull a 16+32 setup with HT and AVX512 plus AVX10 in future I pray for AMD's engineers.

16+32 will put it in HEDT territory.

 

[SiliconFly]

Challenge will be to do it ALL in 500W

Come on, armchair engineers, do the math. Can Intel do it in 500W?

With 20A, they can!

 

[SiliconFly]

Core i9-14900KF Beats Ryzen 9 7950X3D, Core i9-13900KS in Geekbench 6

Flagship Raptor Lake Refresh impresses in Geekbench 6 benchmark.

www.tomshardware.com

ASRock Z790 Taichi vs ASUS System Product Name - Geekbench

browser.geekbench.com

It's mighty impressive that Intel managed to outperform 7950X3D with 14900KF 15 - 17%. Awesome!!!

And with no competition in sight, 14900 will hold the performance crown for at least 6 months, if not more.

(But like I always say, I'm not a big fan of the big, old & fat monolithic cpu on an old outdated node. Thank god this is the last of Intel 7 & possibly RWC we'll be seeing. Waiting for 20A & LNC).

 

[mikk]

View attachment 85832
This slide is obviously pretty old considering it mentions MTL-S. but 8+32 ARL certainly existed

It goes back to the old reddit leak from 2021 when he said ARL-S features an updated compute tile with 8/32 config for the high end enthusiast products, it was clearly a real leak back then.

 

[Henry swagger]

Intel Core i5-14600K with 14 cores and 5.3 GHz boost has been tested with Cinebench 2024 - VideoCardz.com

Intel Core i5-14600K makes an informal debut It appears that Chinese media ECSM are now in possession of Intel’s upcoming 14-core Core i5-14600K CPU. We are just under a week away from the official debut of the upcoming desktop processor series, known as the 14th Gen Core, also referred to as...

videocardz.com

Gonna be the best mid range cpu

 

[A///]

16+32 will put it in HEDT territory.

No it would not. HEDT is more than cores. It's more memory channels, it's more memory capacity, it's extensibility via PCI lanes. By virtue you're stating the current 7950X and 13900K/14900K are HEDT when they're not. They're consumer grade processors that have none of the finesse TR or Xeon W have.

16 core TR is HEDT, 16 core Ryzen is not. Completely different platforms. 24 core (8+16) Intel is not HEDT, Xeon W3400 series and similar are "HEDT"

frankly the term HEDT needs to die off. @aigomorla spoke at length about this a while back. If anything once AMD kneedcapped Intel with the 3950X the old concept of HEDT merged with workstation. Where it belonged anyway.

 

[eek2121]

They can release a simple 48 E-cores cpu very easily. It's just that there are no takers. Even a 64 E-cores is very much possible. But frankly, except a few, no one else would buy.

There would be takers with the right price and form factors.

You underestimate the number of folks who have workloads that scale with core counts, not to mention some of the unique projects that could happen with such a system.

Intel could probably release such a chip for under $500 USD.

My last job, for example, had 12 separate docker containers, 4 of which would scale up with core counts. The hit to single core IPC would be more than offset by the increase in core count.

Threadripper is great for those type of workloads, but 32-48 *mont cores would consume less power and produce less heat.

 

[Mopetar]

TR isn't quite client is it? It's watered down Epyc. How many normal consumers are dumping 5 grand for a processor? Find me a 40 core processor in a <$2000 laptop or a <1500 prebuilt.

It's a consumer workstation/HEDT CPU. Most people don't need 16-cores let alone 64, but for those who do it's an option.

Intel could put 40 e-cores into a die that's smaller than their current desktop CPUs if they wanted to and sell it for a lot less than what AMD charges for a 32-core Threadripper.

In what fantasy world do you live in mate?

You're going to have to elaborate. If you're questioning the ability of Intel to build a monolithic CPU with just e-cores, I don't think it's a problem in terms of making a small die that they can price aggressively.

With Alder Lake they could fit slightly less than 4 e-cores in the same area as the p-cores. Swap all of the p-cores to e-cores and they're already at 40. Rip out the graphics and display controller and they can easily fit in another 12 e-cores at least.

 

[lightisgood]

Intel Core i5-14600K with 14 cores and 5.3 GHz boost has been tested with Cinebench 2024 - VideoCardz.com

Intel Core i5-14600K makes an informal debut It appears that Chinese media ECSM are now in possession of Intel’s upcoming 14-core Core i5-14600K CPU. We are just under a week away from the official debut of the upcoming desktop processor series, known as the 14th Gen Core, also referred to as...

videocardz.com

14600K (ES) marks high-voltage, high-clock and low-power-consumption.
Probably, RPL-R is made by new process.

 

[dullard]

Say a big core that has 30% higher IPC than a small core, at same throughput it will work at 0.77x the frequency of the small core and 0.5x its nominal power at same frequency than said small core, so the efficency argument for small cores is moot since a big core will likely have better efficency at any given same throughput.

That argument falls apart when you have more cores and a fixed power budget (I am making the assumption that most computers won't keep allowing more CPU power forever as we add more cores). To add more cores in the same power budget, the big cores no longer are efficient. If there was no power budget, then yes the big cores are better and we should have all big cores. See the right side of the image below.

But, and this is the point that I am trying to emphasize, you cannot add more cores and just ignore the power consumed. To stay within power budgets as you add more cores, you have to start turning down the frequencies. Soon the little cores are actually more efficient (right side of the image below).


I could post more images, this is just a representative one. There is a crossing point. As more cores are added, we cross over to the little cores doing more work per joule.

 

[Abwx]

That argument falls apart when you have more cores and a fixed power budget (I am making the assumption that most computers won't keep allowing more CPU power forever as we add more cores). To add more cores in the same power budget, the big cores no longer are efficient. If there was no power budget, then yes the big cores are better and we should have all big cores. See the right side of the image below.

But, and this is the point that I am trying to emphasize, you cannot add more cores and just ignore the power consumed. To stay within power budgets as you add more cores, you have to start turning down the frequencies. Soon the little cores are actually more efficient (right side of the image below).

View attachment 85842
I could post more images, this is just a representative one. There is a crossing point. As more cores are added, we cross over to the little cores doing more work per joule.

Actually this curve just add more fuel to my point.

It is not relevant once you look at a same frequency but not same throughput.

At 1.8-2Ghz the P core (wich has SMT), has same throughput than the e core clocked at 3.2GHz.

Hence at 1.8-2Ghz the P core will use the same time as the e core to do the task while consuming about 3500 joules to do so, at the other end the e core will consume 4500 joules to do the same work in the same time but at 3.2GHz.

As you can see at same throughput the P core is vastly more efficient.

Now if the e core is a densified core like Zen 4c it would be different since you would have at the same time same throughput and less energy waste at low frequency, wich tell me that Intel s approach of different cores is not the good one technicaly speaking, only cost efficency and marketing wise.

 

[Khato]

Except it's not as cool looking as Ryzen Master.


Seems pretty clear which company's programmers had more fun and freedom in making their tuning tool.

Sorry that your subjective opinion favors Ryzen Master. Mine favors Intel XTU. Don't think there's anything to be gained from 'arguing' over aesthetics when functionality is equivalent between Intel XTU and AMD's copy.

 

[JoeRambo]

when functionality is equivalent between Intel XTU and AMD's copy.

It is not equivalent. XTU has had it's share of bugs, but is way more useful in both CPU OC and undervolting and real time memory tuning.
Ryzen Master is also hilariuosly slow, i guess that's the price of being pretty and barely useful.

 

[A///]

It's a consumer workstation/HEDT CPU. Most people don't need 16-cores let alone 64, but for those who do it's an option.

Intel could put 40 e-cores into a die that's smaller than their current desktop CPUs if they wanted to and sell it for a lot less than what AMD charges for a 32-core Threadripper.

I'm not sure what you're arguing for in the first line of this reply. Are you insisting a 13900K is a HEDT processor? It isn't.

They could, but it's gonna be lousy performance. You're comparing actual full cores to e cores that lack SMT and that cannot run AVX512. Never mind the lack of bandwidth on a consumer level Intel part. Intel chose to release a partly crippled product. That's on them.

 

[A///]

Threadripper is great for those type of workloads, but 32-48 *mont cores would consume less power and produce less heat.

The presumption here being those TR cores would be actively boosting on a consistant basis. The type of work you describe could be done on a much older computer without much trouble.