Insights

Intel have replaced their Core i product line with the new Core and Core Ultra lines.

Intel first introduced Core i as the successor to Core 2 in 2008, where they first introduced the i3, i5 and i7 line of processors. These processors were introduced as the premium line of processors, occupying the 3 star, 4 star and 5 star ratings in Intel’s own processor rating system. At the time, the Pentium line was 2 star, and the Celeron was 1 star. This branding stayed the same until the 7th generation of Core i processors, where due to increased competition from AMD, Intel introduced the Core i9 to be the new flagship processor in their line-up.

Why the change in branding?

As the years went on, the gaps between the now 4 lines of processor started to stretch to the point where with 13th generation, a Core i3 13100 was a 4-core processor, and the i9 13900 was a 24-core processor. It had reached a point where the actual design of the i3 was no longer a cut-down variant of the i9 die, but rather a completely different piece of silicon, just like Pentium and Celeron were when the Core i3 was first introduced.

Now, Intel have severed the Core i3 from its bigger brothers and created a new class: Core 3, 5 and 7. The Core i5, 7 and 9 have now been replaced by Core Ultra 5, 7 and 9, introducing newer technologies, much like Core i did when it was first introduced in 2008.

What is Intel Core and Core Ultra?

Intel Core and Core Ultra are the new brand names for Intel’s Mobile and Desktop processors. Intel announced these changes in December 2023 for the launch of their new Meteor Lake Mobile processors. Alongside Core Ultra, Intel also announced Intel Core, although much more quietly.

Intel Core is the new line of lower-end processors. These currently come as either Core 3, Core 5 or Core 7, and are Raptor Lake-U Refresh, which is basically what would have been 14th Gen Core i. These processors are more likely to be found in cheaper or lower-end devices you might find in a school or for basic home use. I fully expect to find these processors prevalent in the Chromebook market due to their reduced cost compared to Core Ultra.

Intel Core Ultra is the new premium line of processors. These come in Core Ultra 5, Core Ultra 7 and Core Ultra 9. These new Meteor lake processors come with a few new technologies, such as Intel Faveros packaging, Intel Arc graphics, new Low Power Efficiency (LP E) cores and AI acceleration. These new processors come in either U SKU for Thin and Light Ultrabooks and H SKU for powerful workstations and gaming machines.

Components of the new Core Ultra processor

One of the big new technologies that Intel is bringing with Core Ultra is Intel Faveros. This is Intel’s first foray into “Chiplet” or “Tile” based processors, where the processor is literally spit into different dies and manufactured in different fabs with different manufacturing nodes. This is a technology that AMD used to great effect in the late 2010’s to close the gap to Intel with the introduction of their Threadripper and Epyc lines—and even started to utilise in their discrete graphics cards last year.

For Core Ultra Series 1, there are four different dies included on the package, which Intel are referring to as Tiles. A Compute Tile, a Graphics Compute Tile, an IO (Input / Output) Tile and a SOC (System on Chip) Tile.

But why split the processor into separate Tiles? Efficiency and cost mainly. Silicon manufacturing is by nature flawed. When a silicon die is manufactured, there is a high percentage chance that some of the die will not work. This is where i3, i5 and i7 came from. The dies that worked perfectly were i7, those with small errors were i5 and those with multiple errors ended up being i3.

Intel can turn off parts of the silicon to turn the same product into lower end products, this is called binning. By splitting the processor into multiple Tiles, you can mix and match good and bad parts.

Have a Compute Tile with only 4P cores working? Add it with a Graphics Tile that only has 7 EU’s working. Now you have a Core Ultra 5 135H which you can sell rather than literally throwing away.

Compute, Graphics and IO

The Compute Tile contains the Performance (P) and Efficiency (E) cores of the processor along with a large amount of shared Level 3 Cache to help speed up these cores. On the U SKU processors, this Tile is made up of 2 P cores and 8 E cores. On the H SKU processors, this Tile is made up for 6 P cores and 8 E cores.

The Graphics Tile is dedicated to raw graphics compute power with the Media engine and Display engine being moved to the SOC Tile. This means there would not be any form of slowdown on the GPU. These new Tiles are built on Intel’s new Arc Graphics, bringing great improvements in performance and efficiency.

The IO (Input / Output) Tile is in charge of connectivity to outside of the processor. This includes everything from USB to PCIe, and by extension, storage and WiFi. Intel’s new IO Tile is akin to what AMD have been doing for a few generations. These functions don’t typically benefit from smaller manufacturing nodes. In AMD’s case, they make their IO Chiplets (what they call their dies) on older nodes to save cost without impacting performance. Intel have not confirmed if this is the case with Core Ultra Series 1, but I would suspect that the IO Tile is manufactured on Intel’s 14nm process, which it was stuck on for so many years, rather than the new Intel 4 manufacturing process.

The SOC Tile and AI

SOC, or System on Chip, is the term for an integrated circuit that contains most of if not all the components of a computer. You can think of this Tile as the central processor of the central processing unit. An example would be the Apple M series of processors with the CPU and RAM all packaged in one. Intel is kind of skewing this definition for their own processors, as the whole Core Ultra processor could be referred to as a SOC, but I understand what they are trying to do here—as everything else needed for a X86 Processor that doesn’t have its own Tile is located on the SOC Tile.

The SOC Tile is where, for me, two of the most exciting advancements have come. From what we know, the SOC Tile allows all these other Tiles to communicate and act as a single processor.

It includes 2 new Low-Powered (LP E) versions of the same efficiency cores found in the compute Tile. The idea behind these cores is that they are even more efficient for low priority background tasks.

My first thought is for these to be used for small management processes performed when a laptop is in sleep, things like updates. This way the SOC Tile can start up without starting up the big power-hungry Compute Tile.

In everyday use, I do not think these two core that max out at only 2.5GHz in H SKU and 2.1GHz in U SKU are going to make a difference, however I do think it will make a difference in battery life both when Idle and when performing Windows updates in the background.

The other computable part of the SOC Tile and the third type of compute engine in these processors is the new neural processing unit or NPU.

We do not know too much about this processor, other than it has 2 cores running at a max speed of 1.4GHz and a performance of 11 Terra Operations per Second or TOPS. This is very slightly more than the performance of the AMD NPU included in their latest generation mobile Ryzen processors at 10 TOPS.

However, once combined with the CPU cores and GPU cores, the top Intel Core Ultra Series 1 processor can score 34 TOPS, whereas the Ryzen mobile processor, once combined, scores 39 TOPS.

What does this mean? Intel recently said that to run Microsoft Copilot on device, they would need a minimum of 40 TOPS. Ideally, this could be run on the dedicated NPU, so we are probably a generation or two from achieving this.

However, the purpose of the included NPU is not for raw AI performance, but rather for efficiencies in AI workloads.

We are yet to see any actual benchmarks comparing power consumption, while performing AI workloads with and without this NPU, so for now we will have to take Intel’s word for it.

During the writing of this article, Microsoft announced the new Copilot+ PCs. These devices feature an ARM-based Qualcomm Snapdragon X processors, with a built-in NPU and performance of up to 45 TOPS. It seems they beat Intel to the punch on that one.

Dell’s implementation

Chorus is a Dell Gold Partner, giving us insight into Dell’s plans for these future processors. This also enables us to configure fully custom devices ahead of device launch.

So how are Dell implementing these new processors into their line up?

Latitude

At the high-end of Latitude, there is the new 9450 2-in-1 with either a Core Ultra 5 or 7 U SKU, with 16GB, 32GB or 64GB of blazing fast LPDDR5x running at 7467MT/s with either 256GB, 512GB or 1TB of storage. These devices continue with the mini-LED backlit keyboard of its predecessor and premium build quality all round.

The Latitude 7000 series continues to come in 13”, 14” or 16” once again with U SKU Ultra 5 or 7. However, Dell have finally also refreshed their Latitude Detachable with the new 7350 Detachable. This Surface Pro like device comes with either an Ultra 5 or 7, up to 16GB of RAM and a 13” 2880×1920 Anti-Reflect, Anti-Smudge Touch screen with Active Pen Support.

Latitude 5000 series is designed to be the most versatile, although the 13” 5350 does not get the same options as its bigger brothers—with only U SKU Ultra 5 and 7 with the Ultra 5 only coming with a 16GB RAM option. Otherwise, the device is available as a 2-in-1 touch screen or traditional laptop with a 60Hz 1080p display.

The Latitude 5550 and 5450 do get some nice extra options. These devices can be configured with either U SKU Ultra 5 and 7 or a H SKU Ultra 5. I don’t believe this option will come to standard stock shelves, but instead would have to be CTO ordered, which roughly has a 2-week lead time.
These devices retain the upgradeable SODIMM memory, but move to DDR5 running at 5600MT/s. Dell offer up to 32GB or RAM on these, but you could upgrade to 64GB or even 96GB, but at quite an expense.

The Latitude 3000 series seems to be mostly unchanged from last generation, keeping the Intel Core i 13th gen processors to keep costs down, but adopting DDR5 memory for the first time. This means to adopt the newest platform, you would be looking at a minimum of Latitude 5000 for this new generation.

Precision

The precision line starts with the 3000 Series and has 3 models available, the 14” 3490 and the 15” 3590 and 3591. The 3490 and 3590 are available with both U and H SKU Core Ultra 5 and 7 with up to 64GB of DDR5 SODIMM memory, and is available with up to RTX 500 Ada graphics built in. The 3591 is the more powerful brother, with H SKU Core Ultra 5 7 and 9 all running at 45W base (up from 28W as standard on these processors), up to 64GB of DDR5 SODIMM and up to RTX 2000 Ada Graphics.

The Precision 5000 series is the “Thin” and “Light” workstations, and although they are much thinner than the 3000 and 7000 series, they are still workstations, so not that light. The 14” model the 5490 can comes with H SKU Core Ultra 5, 7 or 9, up 64GB of no upgradeable LPDDR5x 7467MT/s Memory and up to RTX 3000 Ada Graphics. It can come with a QHD display and up to a 4TB SSD. The 16” model is also available with H SKU Core Ultra 5, 7 or 9, with the same 64GB of LPDDR5x, but with up to RTX 5000 Ada graphics. There is the option for a 4k OLED display and options for up to two 4TB SSDs.

Finally, the Precision 7000 series is actually unchanged, with Dell not even refreshing the models. This is due to the lack of new HX SKU processors in the Core Ultra line up with the 13th gen Core i9 13950HX still being the most powerful mobile processor that Intel makes, with 24 Core (8P and 16E) and running at up to 5.5GHz on those P cores, and a maximum turbo power of 157W. Both the Precision 7680 16” and 7780 17” are available with up to the Core i9 13950HX, up to 64GB of SODIMM DDR5 or 128GB of DDR5 CAMM memory, up to RTX 5000 Ada Graphics, up to three 4TB SSDs and 4K displays. These machines are big and heavy, but there is no beating their performance.

All images sourced from Intel.