Discussion Intel current and future Lakes & Rapids thread

[Asterox]

Huh, that is some fat CPU frequency drop when TDP is fixed.

https://twitter.com/i/web/status/1371932560826195970

 

[Hulk]

Huh, that is some fat CPU frequency drop when TDP is fixed.

https://twitter.com/i/web/status/1371932560826195970

View attachment 41270

That's the nuclear fusion result from the "Sun-ny" cove reactor core.

On a more serious note now we can see why for the past 30+ years whenever transistors were added there was also a process shrink.

 

[dullard]

And to be honest, i do not see the point of a desktop CPU supporting big.LITTLE at all in general.

That is why I was speculating on the possible benefits (you know the things you dismissed as being already solved or just rumors). The main benefit is obvious: keeping the base load of work out of the big cores lets the big cores go much higher in frequency for much longer. But I was simply speculating about a lot of possible other use cases where little cores could help.

And if none of those are what you need, then Alder Lake is not for you. Wait for Golden Cove.

 

[Hulk]

That is why I was speculating on the possible benefits (you know the things you dismissed as being already solved or just rumors). The main benefit is obvious: keeping the base load of work out of the big cores lets the big cores go much higher in frequency for much longer. But I was simply speculating about a lot of possible other use cases where little cores could help.

We keep going around and around on this. Why Big/Little for the desktop?
I have two semi-educated guesses at this point:

If the Intel Big/Little technology is primarily designed for mobile and can't easily be separated in fabrication to "Big only" for the desktop, owing to either technical or economic factors, then this could be the reason or a contributing factor.

Perhaps there is an enormous amount of context switching going on in modern software applications/operating systems. These context switches are more detrimental to the Big cores because you lose more compute when they are essential shut down for an equal period of time as the Little ones. So, the Little cores would handle the little chores and allow the Big cores to focus on the heavy lifting without having to constantly flush out for a light lifting job.

Just more unfounded guesses here.

 

[firewolfsm]

The benefits of big.little are more apparent on laptops, where there will be a very real increase in battery life. But there should be benefits on desktop too. Hypothetically, if you just wanted to maximize multicore performance on desktop, you would ONLY use little cores. They're more efficient both in terms of energy/operation and die space/operation, since you can fit more cores in the same die space. That fact alone makes the use case for big.little on desktop pretty obvious, as it leverages some of that, while still maintaining good single core performance.

 

[trexfromouterspace]

And if none of those are what you need, then Alder Lake is not for you. Wait for Golden Cove.

Alder Lake is Golden Cove + Gracemont.

Realistically, I could see a desktop variant shipping with the Gracemont cores disabled. Would be a good way to check the impact of adding the little cores, if nothing else.

Edit: disabled not enabled, I'm a dumdum

 

[Racan]

Some very bold claims by CapFrameX

https://twitter.com/i/web/status/1371772118715797504

Any chance this could be possible? Maybe in a few titles that favor this architecture but on average?

 

[lobz]

Wow just goes to show how different perspectives can be. I was thinking just the opposite. AMD looks to equal or beat Intel's best with their #3 (5800X). From my point of view Intel needs to move the 11900K price to $399 and drop all 18 of the other binned Rocket parts below that price. I think AMD is more than competitive right now without lifting a finger.

umm... okay?

 

[lobz]

Some very bold claims by CapFrameX

https://twitter.com/i/web/status/1371772118715797504

Any chance this could be possible? Maybe in a few titles that favor this architecture but on average?

No amount of RAM OC will make a 4.4 GHz CPU beat the entire Ryzen 5000 lineup in gaming. Not to mention that for the price premium of a consistently (i.e. not just a single golden sample kit from a reddit post) OC'able RAM kit you can buy a better CPU and / or a better mobo, therefore have a well-balanced, really strong PC instead of a niche benchmark hunter hinging on the mood swings of its RAM timings. He's either having really abstract delusions or just farming clicks

 

[lobz]

Alder Lake is Golden Cove + Gracemont.

Realistically, I could see a desktop variant shipping with the Gracemont cores enabled. Would be a good way to check the impact of adding the little cores, if nothing else.

Speaking of realistically, there are going to be multiple desktop SKUs with Gracemont cores enabled, that is not real speculation at this point, only officially unannounced.

 

[Panino Manino]

Some very bold claims by CapFrameX

..

Any chance this could be possible? Maybe in a few titles that favor this architecture but on average?

Intel is now the absolute budget option?
Nice.

Talking about Intel, this is also amazing!

There's more on Intel channel.

 

[coercitiv]

Some very bold claims by CapFrameX

Just watch as it gets spinned from "beat in gaming" to "beat in gaming value". The latter will be true, although if you think about it... Comet Lake CPU are already doing that.

 

[lobz]

Just watch as it gets spinned from "beat in gaming" to "beat in gaming value". The latter will be true, although if you think about it... Comet Lake CPU are already doing that.

inb4

 

[Ajay]

So, time to ask my ~ bi-monthly question: any updates on the status of Intel 7nm production? Anything on Meteor Lake?

 

[coercitiv]

Talking about Intel, this is also amazing!

It's about real-world performance!

It's about the benefits and impacts of the technology we create!

 

[Hulk]

umm... okay?

I mean based on the fact that the 5800X beats the 11700K pretty easily I would expect the 11900K to perhaps match the 5800X and be price accordingly. But more than likely based on Ian's 11700K review the the 5800K will still overall be ahead of the 11900K performance-wise.

CPU 2021 Benchmarks - Compare Products on AnandTech

CPU 2021 benchmarks: Compare two products side-by-side or see a cascading list of product ratings along with our annotations.

www.anandtech.com

 

[trexfromouterspace]

Speaking of realistically, there are going to be multiple desktop SKUs with Gracemont cores enabled, that is not real speculation at this point, only officially unannounced.

rip I meant to say disabled

 

[jpiniero]

rip I meant to say disabled

The small die is rumored to be 6+0 so there definitely will be.

 

[DrMrLordX]

On a more serious note now we can see why for the past 30+ years whenever transistors were added there was also a process shrink.

Well a certain other company just respan their entire core on the same process with a net increase in transistor count and hit paydirt doing it. The real problem is that Cypress Cove wasn't purpose-built for 14nm. It seems like a hackjob.

 

[Hulk]

Well a certain other company just respan their entire core on the same process with a net increase in transistor count and hit paydirt doing it. The real problem is that Cypress Cove wasn't purpose-built for 14nm. It seems like a hackjob.

Yes that's true and the "other" company also redesigned the architecture for that same node from the start. Imagine what that "other" company will do when the move to a smaller node.

I remark should have been written, "Now we know what happens when you design your next architecture for a smaller node and have to backport. It's not good."

 

[gdansk]

Some very bold claims by CapFrameX

https://twitter.com/i/web/status/1371772118715797504

Any chance this could be possible? Maybe in a few titles that favor this architecture but on average?

If you cherry pick the games I'm confident one could construct a graph which supports his broad statement.

 

[ondma]

Intel is now the absolute budget option?
Nice.

Talking about Intel, this is also amazing!

There's more on Intel channel.

If they can beat the entire Ryzen line up, then they certainly will not be a "budget" option. I doubt this will happen though, except maybe a few selected titles.

 

[JoeRambo]

No amount of RAM OC will make a 4.4 GHz CPU beat the entire Ryzen 5000 lineup in gaming. Not to mention that for the price premium of a consistently (i.e. not just a single golden sample kit from a reddit post) OC'able RAM kit you can buy a better CPU and / or a better mobo, therefore have a well-balanced, really strong PC instead of a niche benchmark hunter hinging on the mood swings of its RAM timings. He's either having really abstract delusions or just farming clicks

What if Intel went berzerk and brought back bus speed overclocking? then 44 multiplier with some MCE applied is more than enough to beat anything AMD and Intel has. One can dream

Memory tuning benefits both camps. When AMD had horrible latency with ZEN they used to benefit from RAM tuning more, nowadays they have so much L3 that they simply gain less. Sites that run JEDEC timings are their best friends. 3200CL20? Takes actual effort to buy memory this bad. DYI system will get 3200CL16 stuff and will get XMP profile loaded. That is in no way optimal and leaves performance on the table, but no JEDEC timings in gaming systems.

Intel was at last forced to stop shooting themselves in the foot, allowing RAM tuning is huge boost for their cache starved CPUs.

 

[Panino Manino]

If they can beat the entire Ryzen line up, then they certainly will not be a "budget" option. I doubt this will happen though, except maybe a few selected titles.

They can't, that's why he is bragging about the cheaper Intel chip.
Now Ryzens are too expensive to be worth buying.

 

[lobz]

What if Intel went berzerk and brought back bus speed overclocking? then 44 multiplier with some MCE applied is more than enough to beat anything AMD and Intel has. One can dream

Memory tuning benefits both camps. When AMD had horrible latency with ZEN they used to benefit from RAM tuning more, nowadays they have so much L3 that they simply gain less. Sites that run JEDEC timings are their best friends. 3200CL20? Takes actual effort to buy memory this bad. DYI system will get 3200CL16 stuff and will get XMP profile loaded. That is in no way optimal and leaves performance on the table, but no JEDEC timings in gaming systems.

Intel was at last forced to stop shooting themselves in the foot, allowing RAM tuning is huge boost for their cache starved CPUs.

Intel will never do such an AMD move again, when their cheapest SKU can surpass the need for their K CPUs and Z mobos.