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

[nenforcer]

First 10 P Core only Bartlett Lake has a name the Core 7 253PE with an appearance in Passmark.

Intel Bartlett Lake Core 7 253PE Benchmarked On PassMark For The First Time; 20% Faster Than Core i5 14400 In Multi-Threaded Test

We have the first-ever PassMark score for the Intel Bartlett Lake processor called Core 7 253PE, which shows drastic MT performance boost over i5 14400F.

wccftech.com

"Core 7 253PE" CPU Gets PassMarked, 10 P-Core "Bartlett Lake-S" Model Nearing Launch?

Intel's "Bartlett Lake-S" (Performance) P-Core-only processor design has made a surprise appearance, after a long period of developmental silence. Earlier, today X86 is dead&back put a spotlight on the intriguing entry of Team Blue's unreleased "Core 7 253PE" SKU within PassMark's CPU benchmark...

www.techpowerup.com

Looks promising I would like to see the 12 P core only benchmarks still.

I can see these getting a consumer launch due to their support for DDR4 memory. I dont see them ever seeing the light of day in the current DDR5 memory situation we see ourselves in.

 

[ToTTenTranz]

12P Bartlett Lake might actually be my next CPU..

I'm looking to upgrade my 5900X, for which I own 128GB DDR4 and I don't want to go down to 16GB DDR5.

 

[Wolverine2349]

If this is released it would be awesome. Seeing light of day in DDR5 memory situation, well pair it with that and all 10-12 P cores with modern IPC low latency on monolithic die high memory speeds and we got a winner. DDR5 costs more now, but still!! So did stuff in 2020-2021.

Hopefully it is unlocked. The perfect future proof CPU while maintaining backwards compatibility set and forget with no big.little arch nor dual CCD.

 

[511]

Well sorry to shoot your hopes guys but it's a edge product

 

[Hulk]

Generally, when an application can scale to 8 cores, like rendering or video encoding, it can scale further. Since 16E's are more performant than 4 P's when it comes to MT, I don't see many scenarios where 12P's would be more performant than 8+16, and Intel seems to agree with me because that has been their flagship configuration for the past 5 years. I'm sure Intel did the testing with a variety of applications and that is how they came to go with 8+16.

That being said I'm extremely curious for the release of these parts so that they can be tested and we can finally have a definitive answer.

 

[ToTTenTranz]

The original rumor from February said these chips were meant for Intel's Networking and Edge group (NEX). They were not primarily intended for the consumer platform.

Those plans were made before the current RAM crisis, though.

If AMD is looking at reviving some Zen3 models, Intel could be making some changes too.


Regardless, has anyone ever tried to put one of the older hybrid Bartlett Lake into a Z790 and see if it boots up, for example?

 

[511]

They should boot up

 

[DrMrLordX]

Generally, when an application can scale to 8 cores, like rendering or video encoding, it can scale further.

Not necessarily, as has been pointed out in other threads where we looked at (for example) video encoding where core scaling is pretty limited.

 

[coercitiv]

Those plans were made before the current RAM crisis, though.

If AMD is looking at reviving some Zen3 models, Intel could be making some changes too.


Regardless, has anyone ever tried to put one of the older hybrid Bartlett Lake into a Z790 and see if it boots up, for example?

My post was a bit before the RAM crisis too, almost 1.5 years ago to be exact Damn, time flies in this dystopian future.

It would be very nice to see these chips hit the consumer market in the current context of RAM shortage, but I'm not holding my breath. Intel recently had issues meeting demand for their Intel 7 chips and even increased prices for 12th-14th gen products, to the point where Arrow Lake Ultra 5 chips are cheaper than Alder Lake i5.

That being said, I acknowledge we are heading for some very unpredictable times in the consumer market. I assume computer parts sales for everything DDR5 related will fall off a cliff, probably the entire ecosystem will suffer even with renewed demand for DDR4 platforms. Hardware manufacturers are going to do very weird things in 2026.

 

[ToTTenTranz]

They should boot up

Well the hardware is 100% compliant, but the microcode isn't listed in the bios. I don't know what happens in this situation.

 

[Hulk]

What is the app

Not necessarily, as has been pointed out in other threads where we looked at (for example) video encoding where core scaling is pretty limited.

We shall see.

 

[LightningZ71]

Heck, just produce a low end Xeon with 12P with AVX-512 re-enabled and you'd have an instant market. It would be very niche, but there would be a market for additional units for a low cost of R&D.

 

[511]

Heck, just produce a low end Xeon with 12P with AVX-512 re-enabled and you'd have an instant market. It would be very niche, but there would be a market for additional units for a low cost of R&D.

Intel can't do that cause they don't have Empty Intel 7 fabs all Intel 7 silicon is getting sold as fasat as they can make it

 

[Hulk]

Heck, just produce a low end Xeon with 12P with AVX-512 re-enabled and you'd have an instant market. It would be very niche, but there would be a market for additional units for a low cost of R&D.

Or even easier just get a 9900X.

 

[DrMrLordX]

What is the app

We shall see.

Handbrake. x264 definitely, but x265 also. Example:

AMD Ryzen 9 9950X & Ryzen 9 9900X Deliver Excellent Linux Performance Review - Phoronix

www.phoronix.com

First test on this page: 9950x with a 33% increase in core count only scores 12.8% higher than the 9900x. 4k shows a better performance delta for the 9950x (predictably) but you're still getting diminishing returns. If you normalize for power and/or clockspeed you can change the view a little bit, but the point still stands.

 

[Hulk]

Handbrake. x264 definitely, but x265 also. Example:

AMD Ryzen 9 9950X & Ryzen 9 9900X Deliver Excellent Linux Performance Review - Phoronix

www.phoronix.com

First test on this page: 9950x with a 33% increase in core count only scores 12.8% higher than the 9900x. 4k shows a better performance delta for the 9950x (predictably) but you're still getting diminishing returns. If you normalize for power and/or clockspeed you can change the view a little bit, but the point still stands.

Some codecs scale better than others in Handbrake. But handbrake is still scaling past 12 cores.
x264 scales the worst with more threads (oldest). x265 is better and AV1 is better yet.

 

[DrMrLordX]

Some codecs scale better than others in Handbrake. But handbrake is still scaling past 12 cores.
x264 scales the worst with more threads (oldest). x265 is better and AV1 is better yet.

Right but everything has its limits. That's just one example. People shouldn't be in a hurry to embrace e-core spam (per se) if a smaller-but-still-substantial number of P cores could be had instead. Which, generally-speaking, is not really the case with modern Intel consumer products. Bartlett Lake is obviously not a great solution for anyone looking at a brand new desktop PC provided they can stomach DDR5 prices, but I would not be at all surprised if a 12P Bartlett Lake managed to outperform an 8P+16e competitor in some MT tasks.

 

[Hulk]

Right but everything has its limits. That's just one example. People shouldn't be in a hurry to embrace e-core spam (per se) if a smaller-but-still-substantial number of P cores could be had instead. Which, generally-speaking, is not really the case with modern Intel consumer products. Bartlett Lake is obviously not a great solution for anyone looking at a brand new desktop PC provided they can stomach DDR5 prices, but I would not be at all surprised if a 12P Bartlett Lake managed to outperform an 8P+16e competitor in some MT tasks.

Agreed. There is nuance. This is not a "one size fits all" situation.

 

[nenforcer]

Core 9 273PTE 5.5Ghz 36Mb cache
Core 9 273PE 5.7GHz 36Mb cache
Core 9 273PQE 5.9GHz 36Mb cache

These are all being listed on some European electronics wholesales websites and should be drop-in LGA1700 compatible. However, unless there is consumer motherboard BIOS updates to support them no one will probably ever be able to test them unless its in some industrial medical motherboard perhaps.

 

[DavidC1]

First test on this page: 9950x with a 33% increase in core count only scores 12.8% higher than the 9900x. 4k shows a better performance delta for the 9950x (predictably) but you're still getting diminishing returns. If you normalize for power and/or clockspeed you can change the view a little bit, but the point still stands.

Those two chips are also 24 vs 32 threads though, not 12 vs 16. 12 vs 16 might show a bigger gain.

 

[DrMrLordX]

Those two chips are also 24 vs 32 threads though, not 12 vs 16. 12 vs 16 might show a bigger gain.

You can look at the 9700x and 9900x for a similar comparison. 16t->24t shows ~26% increase in performance from a 50% increase in thread count/core count. And the 9600x almost matches the 9700x which is certainly a product of the 9700x hitting a power limit. Make of the data what you will.

 

[Aeonsim]

AMD Ryzen 9 9950X & Ryzen 9 9900X Deliver Excellent Linux Performance Review - Phoronix

www.phoronix.com

First test on this page: 9950x with a 33% increase in core count only scores 12.8% higher than the 9900x. 4k shows a better performance delta for the 9950x (predictably) but you're still getting diminishing returns. If you normalize for power and/or clockspeed you can change the view a little bit, but the point still stands.

33% more cores but zero extra memory bandwidth both for CCD to IO or IO to Memory. Some workloads that don't need the bandwidth really love the extra cores but others are very much bandwidth limited.

If you look at the final page where performance is broken down by category the Rendering tasks really like the extra cores going from 9900x to 9950x ~29% improvement. While for the other categories it's typically 10-20% improvement. Some of the audio workstation tests also like the extra compute and don't seem to need more bandwidth sitting around the ~30% mark.

 

[DrMrLordX]

33% more cores but zero extra memory bandwidth both for CCD to IO or IO to Memory. Some workloads that don't need the bandwidth really love the extra cores but others are very much bandwidth limited.

Yup, stuff like Vray (or Cinebench) scales almost perfectly with core count and is also insensitive to memory bandwidth. Though that would imply that the 9950X is starved for bandwidth which isn't necessarily the case.

 

[DavidC1]

X264/X265 also has two different workloads, one of which is less threaded than the other and the very threaded one.