Intel's 7 new CPUs exposed, Nova Lake desktop CPU lineup leaked. The lineup for Intel's next-generation desktop CPUs is basically complete. 7 models, ranging from the top-tier Core Ultra 9 4970K BFC all the way down to the entry-level Core Ultra 5 4650KF, covering three tiers: Ultra 9, Ultra 7, and Ultra 5.

💡 What You Will Learn

Intel's 7 new CPUs exposed, Nova Lake desktop CPU lineup leaked. The lineup for Intel's next-generation desktop CPUs is basically complete. 7 models, ranging from the top-tier Core Ultra 9 4970K BFC a

📜 Table of Contents

• Seven Models, Two Approaches
• BFC Suffix: Intel Copies AMD for the First Time
• Back to Four Digits, Because Model Numbers Ran Out
• 2027: Head-On Collision with AMD
• Final Thoughts

Intel's 7 New CPUs Leaked, Nova Lake Desktop CPU Lineup Revealed

Intel's next-gen desktop CPU lineup is basically complete.

Seven models, ranging from the top-tier Core Ultra 9 4970K BFC all the way down to the entry-level Core Ultra 5 4650KF, covering the Ultra 9, Ultra 7, and Ultra 5 tiers, all switching to the new LGA 1954 socket.

First, let's be clear about the source: this list is not officially announced by Intel. It was released by the X platform leak account LC Tech Leaks on October 1st. The official specs won't be confirmed until Intel's CES event in early 2027.

But these two signals are already clear: Intel is preparing to go back to four-digit naming, and it's finally starting to stack cache — this is the first time in a decade it's directly taking on AMD.

Seven Models, Two Approaches

According to the leaked list, these seven chips can be cleanly split into two groups based on "whether they have a large last-level cache." Models with cache carry a BFC suffix, with cache reaching up to 144MB:

Within the same tier, the BFC variant always has 4 more low-power efficiency cores (LPE) plus a block of large cache. Take the Ultra 9 tier, for example: the 4970K BFC has 4 more cores and an extra cache block compared to the 4950K, yet the TDP is the same 125W — this is Intel's tiered strategy, making you pay extra for the large cache.

By the way, only the 4650KF among the seven has its integrated graphics cut; all others come with 32EU iGPUs. The vPro commercial feature is only given to the three BFC models, making the enterprise positioning very clear.

BFC Suffix: Intel Copies AMD for the First Time

AMD's 3D V-Cache has been the most formidable card in desktop CPUs over the past couple of years, going from the 5800X3D all the way to being standard on today's flagships, beating the competition in real-world gaming performance. From 12th gen all the way to Panther Lake, Intel's weakness has always been in process and packaging — it never went head-to-head on cache.

This time, bLLC single-compute tops out at 144MB, dual-compute at 288MB, directly targeting AMD's top X3D's 288MB 3D V-Cache tier. What's more noteworthy is the tiering approach — Intel didn't make large cache standard across the board, but turned the suffix into a SKU differentiator, letting users decide whether to pay extra.

What BFC stands for exactly, Intel hasn't said. The current speculation is something like "Big Cache Special Edition," but that's just outside guesswork — it won't count until officially announced.

Back to Four Digits, Because Model Numbers Ran Out

Arrow Lake uses the Ultra 200 series, Panther Lake uses the Ultra 300 series, and Nova Lake will return to the 400 series.

The reason is SKU explosion: single-compute plus dual-compute, two cache versions at 144MB and 288MB, with K, KF, and no-suffix variants — three-digit numbering is no longer enough to differentiate. The leak also mentions that mobile HX models will use four-digit naming, while H and no-suffix models continue with three digits.

The socket is also brand new — LGA 1954, launching with Z990 and Z970 motherboards. According to rumors, this socket will be compatible with at least three generations of processors: after Nova Lake comes Razer Lake, Titan Lake, and Hammer Lake.

2027: Head-On Collision with AMD

The timeline is even more interesting. Nova Lake-S is expected to debut in early 2027, launching in Q1, while AMD's Zen 6 desktop version, Olympic Ridge, is also squeezed into 2027 (around mid-year), with up to 24 cores, also on a 2nm process.

The last time the two collided in the same year goes back to late 2024, with Zen 5 versus Arrow Lake. What's more subtle is the platform strategy: Intel is switching to a new socket, while AMD stays on AM5 — users who buy an AM5 motherboard now won't need to change boards to upgrade to Zen 6 in 2027.

Intel also has a card up its sleeve: the 52-core dual-compute version isn't expected until Computex 2027, with the first batch of 28-core models just being an appetizer. Rumor has it that this dual-compute flagship has a very high power wall, with PL2 reaching 500W.

The product line is mostly certified already — Nova Lake-HX passed USB-IF's USB4 80Gbps certification, the desktop 900 series chipset passed USB 3.2 Gen2x2, processor samples with native PCIe 5.0 x16 have appeared in PCI-SIG records, and certified samples can already be seen in HWiNFO. All that's missing is an official announcement.

Final Thoughts

For regular users, the most useful takeaway from this list is: if you're planning to build a PC in 2027, you can start planning your budget now. Intel and AMD both launching new products the same year, one switching sockets and one not, with both flagships colliding head-on just months apart — this could be the most worthwhile round of desktop CPUs in recent years.

As for whether Intel's 28-core flagship really has 28 cores, we'll have to wait for the early 2027 launch event, or for the next round of leaks to unify the details in this table.