Yesterday we wrote that in the field of desktop processors, Intel will still only have Arrow Lake with 8 large and 16 efficient cores for the next two years, because it canceled the desktop version of the next 1.8nm generation of Panther Lake as well as the refresh of Arrow Lake, which should have at least significantly increased the number nuclei. However, it seems that the reality is more complicated. In parallel, Intel will keep the LGA 1700 socket alive and release completely new processors for it.
You may recall that there was a media report some time ago that Intel might be preparing some kind of 15th generation Core on the LGA 1700 platform, which would follow the Raptor Lake Refresh chips and be based on a chip called Bartlett Lake (or maybe just Bartlett). This chip was supposed to be an alternative to the new Core Ultra processors, which will use the more expensive new LGA 1851 platform. Parallel to it, Bartlett processors would be sold, which would use the cheaper LGA 1700 boards, and therefore could also be used with DDR4 memory.
However, it later turned out that Bartlett (Lake) is actually a product of Intel’s networking division, so it should be a CPU intended for example for NAS and similar purposes, and it stinks a lot that it is perhaps just an adaptation of Raptor Lake for this market segment ( so they could come out as Xeon D), and not a new product that could also look into the desktop. But this has changed now.
Leaker Jaykihn, who brought various information about Arrow Lake processors, has now said that Intel is indeed planning Bartlett Lake processors, and more importantly: although they are developed by the networking division, they will also be released for desktop PCs, that is, for gaming computers. Indeed, they will form a kind of fourth generation of processors on the LGA 1700 platform, which will be sold in parallel with Arrow Lake’s “Core Ultra”. And what is even more important – this will not be a refresh.
Another rumor has also been confirmed – Bartlett Lake will bring a completely new chip that will have 12 large P-Core cores without efficient E-Core cores. So it could be a very interesting gaming CPU or a processor for those who don’t like the big.LITTLE concept with E-Core. Only consumption will probably not be a strong point.
Bartlett Lake: Two different processor families under the same name?
According to Jaykihn, the Bartlett Lake processors will actually be made up of two different families. One part of the processors will still be hybrid and should be based on the old Alder Lake and Raptor Lake chips, so these models will still only be a refresh, of course they will still have Golden Cove (or Raptor Cove) and Gracemont cores. Their offer will not include a Core 9 model, only a lower one. These processors do not have an NPU and are therefore not Core Ultra, but at the same time they will probably switch to the new marking, so they will be a Core “number” without the letter i. For example, the CPU could be called Core 7 260.
According to this source, the top of the range will be the processor Core 7, which will have 8 Goldmont cores and 16 Gracemont cores, for a total of 32 threads. Processors will also be on the menu Core 5 with 6 large and 8 small cores (20 threads). These two configurations are supposed to have B0 silicon in them, which is a Raptor Lake chip with larger L2 caches (2MB for P-Core, 4MB per E-Core cluster). In terms of performance and capabilities, these models can probably be seen as a refresh of the Core i9-13900 and i9-14900, respectively Core i5-13500/13600 and i5-14500/14600.
Cheaper version Core 5 but they will be based on the C0 chip, which is Alder Lake generation silicon with 1.25MB L2 cache on the P-Core and 2MB L2 cache per E-Core cluster. These processors will only have 6 large cores and 4 E-Core cores. So it will be a refresh of models such as i5-13400 and i5-14400.
The menu will also include a quad core Core 3 and dual-core “Core only” (so maybe Intel will quickly drop the strange “Intel Processor” designation, but it’s possible that the leaker is wrong and the dual-core won’t be called Core). These lower models will be based on the Alder Lake generation H0 chip, which is silicon with only six large cores and no E-Cores.
The important thing is that these processors will not have 125W versions, only 65W and 45W models (the TDP of the 45W version could therefore be higher than the current 35W T series, but of course the question is more what the maximum turbo consumption will be). This probably means they won’t be selling the “K” version unlocked for overclocking. Users with such gradients will therefore continue to have to buy 125W Core 14th generation models.
These processors, which as you can see won’t be quite what’s really interesting about Bartlett Lake, are due out sometime around the turn of the year. Jaykihn first listed a planned release in December or January, later corrected it and now only lists January. So they could probably be officially announced and released during CES 2025, in parallel with the mainstream 65W Arrow Lake processors, which are also expected to be released around that time. This perhaps explains why there are only 45W and 65W models. It will be similar to the “Core non-Ultra 100” refresh for laptops, which Intel released this January as a complement to Meteor Lake processors, only this time it will be in a desktop.
True Bartlett: Up to 12-core non-hybrid alternative to Core Ultra
What will interest you about the Bartlett Lake generation is the second family, which is based on a separate and apparently completely new chip. The latter is said to be non-hybrid and has more P-Core cores instead of E-Core. So there are a total of 12 P-Cores with HT in it, giving 24 threads. These cores will have the Golden Cove architecture (respectively Raptor Cove with a larger L2 cache), i.e. without changes compared to Raptor Lake. It can be said that it will be a new version of Raptor Lake chips, only with P-Core only. It should still use the LGA 1700 socket.
So the number of stops on the ring bus is the same as an 8+16-core hybrid CPU, as E-Core clusters only need one stop per four cores, allowing them to scale to higher thread counts. It will be interesting to compare how these two alternative configurations (which are a direct match, as big.LITTLE according to Intel always replaces one P-Core with four E-Cores) will perform in multi-threaded software and which one will have better performance and energy efficiency. This will probably be the best opportunity yet to evaluate the pros and cons of the Alder Lake and Raptor Lake hybrid architecture.
This branch of Bartlett Lake will offer the following configurations: The most powerful 12-core version will be sold as Core 9. A cheaper alternative will be the ten-core version, which will be marked Core 7. And the last model or models will be octa-core, they will bear the designation Core 5. Lower models based on this silicon will no longer be available, so no Core 3 and so on. These will be provided by the hybrid variant.
From these non-hybrid Bartlett Lakes, there will be not only a 45W and 65W locked version, but also enthusiast models with a 125W TDP. So these should be unlocked for overclocking and will again provide an alternative for enthusiasts. We do not yet have any information about the real consumption – the so-called PL2 or maximum turbo consumption – which is usually much higher than the TDP for Intel processors.
Still without AVX-512?
Jaykihn went on to say on Twitter that he has no information that these processors have AVX-512 instructions enabled again (but it’s also not confirmed that they should still be disabled). Golden Cove and Raptor Cove core can theoretically provide them, their support is still accessible in silicon. And the fact that Bartlett Lake’s non-hybrid processors have no E-Cores theoretically makes it affordable.
So there is some chance that AVX-512 might show up on these non-hybrid Bartlett Lakes, but on the other hand, Intel has left it disabled on the E-2400 Xeons, which also don’t have E-Core, and the fact that support isn’t advertised makes me very hopeful doesn’t sound So don’t wait for the AVX-512, although if Intel feels a lot of pressure from the competition, it might change its mind…
Bartlett processors should use the same LGA 1700 socket and design as Alder Lake and Raptor Lake and work in today’s boards
But the release is far away, in a year?
The worse news is that these processors are still a long way off. While the hybrid version is due to be released in January, and will therefore be an alternative to the first generation of Arrow Lake desktop processors (Core Ultra 200), this non-hybrid version of Bartlett Lake will reportedly come to the market only in the third quarter of 2025 (that is, in a year in July, August, or September).
From this, it seems that it should be an alternative for the next generation and annual cycle of desktop processors. At that time, Intel is supposed to release an Arrow Lake refresh, differing by integrating a more powerful NPU (but the number of CPU cores will remain the same), which will probably be called Core Ultra 300. These 12-core Bartlett Lake models could probably only be released as a parallel alternative to nim, so perhaps under the name Core (non-Ultra) 300.
New “Ultra” processors with chiplets only a minority, still 7nm raptors for the mass market?
So it seems a bit that the arrival of Core Ultra processors using new modern manufacturing processes (3nm TSMC technology, 2nm Intel process) and advanced chiplet encapsulation will not be some kind of wave that would spread throughout the market and these new technologies would take over Intel’s portfolio.
At least part of the market will probably be covered by Intel for a long time with older technologies – monolithic processors without chiplets, which will probably still use the old 7nm production process, the roots of which go back to the years 2015-2019, when it was still the troubled 10nm process, which was later developed into successful versions of the 10nm SuperFin (2020) and Intel 7 (2021) production processes. It is possible that this will be a necessity for Intel, as new technologies (“Ultra”) with advanced manufacturing processes, chiplets and 3D encapsulation will be expensive and have limited production capacity.
Intel Core Ultra Meteor Lake processor composed of chiplets with 3D encapsulation
In the desktop CPU space, perhaps even the 7nm “alternative” Bartlett Lake processors and the LGA 1700 platform could do the majority of sales for the next two years, while the advanced Core Ultra “Arrow Lake” on the new LGA 1851 platform could be in the minority. According to various speculations, LGA 1851 will have much more expensive boards. At the same time, there is the problem that Core Ultra processors manufactured with new manufacturing processes (3nm and 2nm Arrow Lake, 1.8nm Panther Lake) may suffer from lower frequencies, so they will not even be able to beat Bartlett Lake based on the old architecture and old manufacturing process in single-threaded and game performance. Or maybe Bartlett and Raptor Lake will beat it, but not by much.
However, the continued use of the 7nm process and architectures from the Raptor Lake processors will probably be a disadvantage in terms of consumption, which will probably remain quite higher than the new Arrow Lake processors and also compared to the competing Ryzen processors from AMD. So this will be the dark side of that long life of the LGA 1700 platform and Bartlett Lake.
Will the instability and degradation issues be resolved?
In addition, there is another uncertainty. Currently, both the 13th Gen and 14th Gen Raptor Lake processors suffer from widespread issues, with perhaps tens of percent of the 125W Core i9-13900K and Core i9-14900K processors affected by instability, with other lower models not being safe either.
After almost half a year, Intel still hasn’t come up with a solution, and it seems more and more that these processors might actually be subject to a gradual degradation (failure) of the silicon, so that even the ones that are stable so far can gradually stop working. And this could not be solved retroactively even with any repair.
So the current versions of the Raptor Lake processors or their 7nm manufacturing process (or both) very likely have an extremely serious problem. Which also raises doubts about the Bartlett Lake processors, which will be a refresh of these processors, or in the non-hybrid ones, a new chip, but based on their IP.
Hopefully, by the time Bartlett Lake is released, it will be clear what the situation with that Raptor Lake issue is, and somehow satisfactorily resolved. Hopefully by that time it will be possible to rely on the new Bartlett Lake processors and we will know that the (relatively fast) silicon degradation (if this problem is confirmed) will no longer affect them.
Source: www.cnews.cz
