News Intel Bartlett Lake-S: up to 12P-Core or up to 8P-Core +16E-core

[KompuKare]

They would have to prove that this happens on a CPU that never ran outside of intel specs, so it will never happen.
You can't claim damages for something the company never promised.
A class action could be had against mobo makers that didn't clearly state that they where using out of spec values.

So Supermicro W680 boards run out of spec?!

Pull they other one, it has bells on it!

About the dozen P cores only model, the area ratio beneath P and E is meant to be 1:4 so "about"the same size as 12P/16E (40T) part.
But I'd be far more worried about power; would they not have to clock these far lower than RPL?

 

[TheELF]

So Supermicro W680 boards run out of spec?!

Pull they other one, it has bells on it!

About the dozen P cores only model, the area ratio beneath P and E is meant to be 1:4 so "about"the same size as 12P/16E (40T) part.
But I'd be far more worried about power; would they not have to clock these far lower than RPL?

The chipset does support both cpu and ram overclocking.
L1tech did not share any settings of how they where run from the start.
The fix he shows for fixing instability is to overclock all cores to 5.3 and to overclock the ram to 4.2, so overclocking is a part of it.

Anyway, it doesn't matter what you or I believe, they would have to prove in court that it happens on unmodified CPUs and just saying well servers do it as well doesn't cut it.

 

[Wolverine2349]

Broadwell-E had 10 cores on a single ring back in 2016:

But beyond 10 cores they switched to dual ring, presumably because a ringbus scales poorly beyond that point. And Skylake-E moved to a mesh.

Yes they did. What was referring to wanting an option of more more than 8 Big cores of homogenous arch cores on a single node, I mean on a modern arch with modern IPC and PCIe. Latest arch is Skylake based Comet Lake 10900K and 10850K which had 10 cores on a single ring.

Nothing since and a 12 P core on a ring bus would be welcome. I am sure they could do 12 P cores on a single ring bus since they have 16 e-cores and 4 e-cores or even slightly less like maybe 3.5 e-cores fits in 1 P core so 12 P cores on one ring can be done.

But stability please. Though is Raptor Lake has stability problems, just do Alder Lake die then maybe they need to go to 10 P core only model. Or maybe they can do 12 since 4 e-cores is a little bigger than 1 P core even on Alder Lake die?

 

[Wolverine2349]

Sounds interesting until i read:

Yeah a litte too little too late. Given that Alder and Raptor Lake will be at least 3.75 years and 2.75 years respectively at that time cause no sooner than July 2025.

But IPC is advancing so much more slowly than the jump from Skylake to Zen 4 and Golden Cove. Even Zen 5 and Lion Cove has much less jump and future gens at least through 2026 or 2027 probably will not have any better than 10-12% jumps as a wall has been hit form both Intel and AMD it seem in IPC and clocks. SO maybe Golden Cove and a stability fixed Raptor Cove will age great and not be too behind even 3-4 years old?

 

[Thunder 57]

Yeah a litte too little too late. Given that Alder and Raptor Lake will be at least 3.75 years and 2.75 years respectively at that time cause no sooner than July 2025.

But IPC is advancing so much more slowly than the jump from Skylake to Zen 4 and Golden Cove. Even Zen 5 and Lion Cove has much less jump and future gens at least through 2026 or 2027 probably will not have any better than 10-12% jumps as a wall has been hit form both Intel and AMD it seem in IPC and clocks. SO maybe Golden Cove and a stability fixed Raptor Cove will age great and not be too behind even 3-4 years old?

IMHO ARL and Zen 5 will age better. Also if it is still just RPL-R I wouldn't trust it.

 

[dttprofessor]

12E skymont Intel 3 LGA 1700 ?

 

[SteinFG]

12E skymont Intel 3 LGA 1700 ?

Core i3 N300 is probably the closest you'll get to that. It's not releasing on desktop though, too low-power

 

[Wolverine2349]

IMHO ARL and Zen 5 will age better. Also if it is still just RPL-R I wouldn't trust it.

Will they ever have 12 P cores on Arrow Lake single tile or will AMD have more than a 8 cores on a single CCD within a single CCX?

If not I am not so sure. Certainly not for 2017-2020 games and even some 2022 games like Elden Ring.

Some games even modern ones or 2023-2024 ones are not designed for hybrid arch and have issues cause game devs these days are lazy.

And you say RPL-R refresh would not trust it? You mean stability issues? Though is it really a refresh if it is a new 12+0 die? Would you still not trust it then?

Cause really Raptor Lake has only had one die thus far being 8+16 right? And that is what has had the problems. Its sold with parts disabled that are defective. And even lower end non K SKUs some use different stepping's/dies which are really just Alder Lake dies sold under RPL brand.

So basically a 12 + 0 die if they are Raptor Cove cores, you would not trust it given the stability and degradation issues Raptor Lake which has Raptor cove cores have had?

I may not trust it either. But i so badly want more than 8 modern IPC cores with PCIe Gen 4 or newer on a single node/die. It appeared there would not be any option coming soon if ever. Now there is hope. But the catch is, can it be trusted to be rock stable and not degrade too easily to make the CPU I really wanted a reality in the not too distant future?

Or please Intel just release a 12 + 0 Arrow Lake die. Please you have a buyer waiting right here.

I want a set and forget it CPU best for all games the last 5-10 years and future games as well with the even better 15% IPC uplift and lower power and oh yeah guaranteed stability fixes.

 

[jpiniero]

Intel doing this for more than just embedded is kind of an admission that OEMs must be unhappy with Arrow Lake's costs.

 

[AcrosTinus]

This might be it.
If Intel really runs this strategy, the core arch will stay one level behind the current one.

And in Q3 2025 I might buy it if it is on a newer node and at least RWC or LNC.

 

[Wolverine2349]

IMHO ARL and Zen 5 will age better. Also if it is still just RPL-R I wouldn't trust it.

Zen 5 and ARL aging better. Well for Zen 5 no way for sure after the testing and benchmarks.

I mean look at the gaming and general workload benchmarks. Underwhelming 5% uplift in gaming and general workloads.

If Bartlett Lake 12 P core is stable and does not degrade, it will age much better. 12 P cores on one die with similar IPC to Zen 5 given Zen 5 5% IPC uplift in gaming and general workloads. It would slaughter vanilla Zen 5 in gaming and only be slightly behind or trade blows behind 7800X3D and 9800X3D in gaming. And age better than the 8 core X3D parts as games get more thread heavy while trading blows in lower threaded games.

 

[KompuKare]

To get decent clocks, how much would 12 P cores consume?

Any guesses?

 

[Wolverine2349]

To get decent clocks, how much would 12 P cores consume?

Any guesses?

Who knows. Though 16 e-cores take up a little more power I think than 4 extra P cores. Or could be similar? Not sure?

 

[Tech Junky]

Any guesses?

If you go to PCpartpicker.com and select the CPU and then click on the power box it will give you the min/max for the CPU.

 

[Thunder 57]

Zen 5 and ARL aging better. Well for Zen 5 no way for sure after the testing and benchmarks.

I mean look at the gaming and general workload benchmarks. Underwhelming 5% uplift in gaming and general workloads.

If Bartlett Lake 12 P core is stable and does not degrade, it will age much better. 12 P cores on one die with similar IPC to Zen 5 given Zen 5 5% IPC uplift in gaming and general workloads. It would slaughter vanilla Zen 5 in gaming and only be slightly behind or trade blows behind 7800X3D and 9800X3D in gaming. And age better than the 8 core X3D parts as games get more thread heavy while trading blows in lower threaded games.

Maybe you will get lucky and Intel decides to release them to the public. Then maybe we can find out who was right. My money is still on Zen 5 and ARL.

 

[dullard]

If you go to PCpartpicker.com and select the CPU and then click on the power box it will give you the min/max for the CPU.

Do you mind telling us which Bartlett Lake-S CPU is in PCPartPicker?

 

[Wolverine2349]

Maybe you will get lucky and Intel decides to release them to the public. Then maybe we can find out who was right. My money is still on Zen 5 and ARL.

Arrow Lake is really TBD until its released, but have you seen Zen 5 benchmarks. Vanilla Zen 5 in gaming and most consumer workloads is barley if at all any better than vanilla Zen 4.

A 12 P core Raptor Lake all cores being on one die with no cross core latency penalty will age much better than Zen 5 for thread heavy games.

Now Zen 6 and Nova Lake will be incoming eventually. But it appears Intel and AMD are nboth struggling to improve IPC much and have hit a wall. So I think a 12 P core Bartlett Lake on a single ring bus as long as its stable and does not degrade and can clock at least 5GHz all core will age very well and give many what we want as no scheduling quirks set and forget it solution.

Eventually IPC will increase a lot and leave the 12 P core CPU in dust but given Intel and AMD's recent struggles to improve IPC that much from today's levels, I think we are looking at 2027-2028 or longer before we see a CPU with a big IPC uplift (20% or more) from either company. Clock speeds have hit a wall too.

 

[dullard]

To get decent clocks, how much would 12 P cores consume?

Any guesses?

At base, the whole 12 P-core chip gets 125 W, so each P-core has ~10 W. At turbo ~253 W, each P-core gets ~21 W. That is a drop from a Raptor Lake chip with the E-cores turned off (~15 W at base, and ~31 W at turbo). So, the 12-core chip will have to run at lower frequencies.

I'd say look at the Raptor Lake lineup and subtract ~400 MHz due to the lower power allowed per core. Maybe subtract less (only ~300 MHz) if Intel optimized it for these lower power levels.

 

[ondma]

Arrow Lake is really TBD until its released, but have you seen Zen 5 benchmarks. Vanilla Zen 5 in gaming and most consumer workloads is barley if at all any better than vanilla Zen 4.

A 12 P core Raptor Lake all cores being on one die with no cross core latency penalty will age much better than Zen 5 for thread heavy games.

Now Zen 6 and Nova Lake will be incoming eventually. But it appears Intel and AMD are nboth struggling to improve IPC much and have hit a wall. So I think a 12 P core Bartlett Lake on a single ring bus as long as its stable and does not degrade and can clock at least 5GHz all core will age very well and give many what we want as no scheduling quirks set and forget it solution.

Eventually IPC will increase a lot and leave the 12 P core CPU in dust but given Intel and AMD's recent struggles to improve IPC that much from today's levels, I think we are looking at 2027-2028 or longer before we see a CPU with a big IPC uplift (20% or more) from either company. Clock speeds have hit a wall too.

Nova Lake was supposed to be a big IPC uplift for Intel, but I am losing hope for anything they promise. Granting the slim chance still that ARL will be a hit, nothing Intel has done on the desktop has been a winner since Alder Lake. (RL looked like it was good until the stability issues appeared.)
As far a Zen 6, cant imagine AMD will be a dud 2 releases in a row, but who knows. Maybe they have just shifted their focus too far toward servers to really care about consumer desktop.

 

[jdubs03]

Nova Lake was supposed to be a big IPC uplift for Intel, but I am losing hope for anything they promise. Granting the slim chance still that ARL will be a hit, nothing Intel has done on the desktop has been a winner since Alder Lake. (RL looked like it was good until the stability issues appeared.)
As far a Zen 6, cant imagine AMD will be a dud 2 releases in a row, but who knows. Maybe they have just shifted their focus too far toward servers to really care about consumer desktop.

It’s 2 years away. Got lots of time to sort out where NVL will stand compared to ARL-R and Zen 6.

Surely they’ve realized with their recent mess that they won’t be able to play fast and loose with their performance/efficiency claims.

Also ARL could still surprise on the upside. The most recent leak even though it’s just one GB6 and GB5 result shows it’s competitive. We could see some slight increases on those results by the time products are in reviewers hands in two months. All in all not as bad as feared by some, and hoped among others.

 

[ondma]

It’s 2 years away. Got lots of time to sort out where NVL will stand compared to ARL-R and Zen 6.

Surely they’ve realized with their recent mess that they won’t be able to play fast and loose with their performance/efficiency claims.

Also ARL could still surprise on the upside. The most recent leak even though it’s just one GB6 and GB5 result shows it’s competitive. We could see some slight increases on those results by the time products are in reviewers hands in two months. All in all not as bad as feared by some, and hoped among others.

Actually, I was hoping for a ground up redesign from intel for their P cores. (Royal Core, Nova Lake???) Sadly I dont think it is going to happen now, with their financial woes. I think they are going to continue to struggle as long as they simply continue to add more resources to try to improve IPC on a P core that is already too big and inefficient.

 

[Wolverine2349]

Actually, I was hoping for a ground up redesign from intel for their P cores. (Royal Core, Nova Lake???) Sadly I dont think it is going to happen now, with their financial woes. I think they are going to continue to struggle as long as they simply continue to add more resources to try to improve IPC on a P core that is already too big and inefficient.

Why not make e cores new p cores for homogeneous design.

If Skymont rumors of it truly having Raptor Cove IPC are to be believed. And it can get good latency, it could be something smaller and more efficient to replace the big p cores permanently?

 

[ondma]

Why not make e cores new p cores for homogeneous design.

If Skymont rumors of it truly having Raptor Cove IPC are to be believed. And it can get good latency, it could be something smaller and more efficient to replace the big p cores permanently?

They still have a clock speed deficit and lack hyperthreading (of course P cores do now too). I dont think they are there yet, but it is a possibility.

 

[Wolverine2349]

They still have a clock speed deficit and lack hyperthreading (of course P cores do now too). I dont think they are there yet, but it is a possibility.

Though could intel create a Skymont only die with say 12 to 16 of them and sell it cost effective. Or is a new die even with more efficient less expensive cores still super expensive.

Like I mean they clock at 4.6 to 4.7GHz and so do p cores of 12700k

So intel could sell a higher core count more power efficient better 8% better IPC golden cove substitute. All 12 to 16 on a single ring/tile. Just no ht but so what. No need for ht with that many cores. And 4.6 to 4.7GHz still very good. Would crush Zen 3 and trade blows with Zen 4 as it's better IPC would begate slight clock disadvantage to Zen 4.

Some would say well just disable p cores and you have that. Though not so sure. The Skymont cores are in clusters of 4 so they are designed to be used with Lion Covr cores that have their own node each on the ring bus. It would need a new die with the Skymont cores dedicated nodes on a ring or better mesh to work as the 12700k more than 8 p core ht off perfect substitute.

 

[Thunder 57]

Though could intel create a Skymont only die with say 12 to 16 of them and sell it cost effective. Or is a new die even with more efficient less expensive cores still super expensive.

Like I mean they clock at 4.6 to 4.7GHz and so do p cores of 12700k

So intel could sell a higher core count more power efficient better 8% better IPC golden cove substitute. All 12 to 16 on a single ring/tile. Just no ht but so what. No need for ht with that many cores. And 4.6 to 4.7GHz still very good. Would crush Zen 3 and trade blows with Zen 4 as it's better IPC would begate slight clock disadvantage to Zen 4.

Some would say well just disable p cores and you have that. Though not so sure. The Skymont cores are in clusters of 4 so they are designed to be used with Lion Covr cores that have their own node each on the ring bus. It would need a new die with the Skymont cores dedicated nodes on a ring or better mesh to work as the 12700k more than 8 p core ht off perfect substitute.

You are quickly becoming similar to a former member who kept saying AMD should have made a "3600 x 2". If it was all that obvious I'm sure Intel would be working on it. Nobody is going to put money into developing a "gaming CPU". It's all server first for Intel and AMD. Intel decided to target Apple with LNL. Well see how that works out.

There is no need for a gaming CPU. Developers will target for what is available. I don't understand your fetish for a 12-16 core homogeneous CPU. As @dullard posted about it would have to sacrifice frequency.

I don't know what else to say. No one else seems to understand why you want what you do. Should Bartlett Lake come out in a 12+0 form, I would love to see it compared to a 9900X. Too bad we have a year to wait.