Intel's upcoming 'Wildcat Lake' low-power series breaks cover in Geekbench listing — 'Core 3 304' is twice as fast in single-core performance versus last-gen

News
By published

And about 50% more powerful in multi-core tests.

Intel
(Image credit: Intel)

Intel has been working on its Wildcat Lake family of budget CPUs for a while now; we've had leaks since at least late 2024 teasing a proper next-gen successor to Twin Lake. The launch of Panther Lake has recently reignited the Wildcat Lake rumor mill. As such, leaker @harukaze5719 has just spotted a new Geekbench listing for an Intel "Core 3 304," and it's quite impressive for its class.

Latest Videos From

Intel Core 3 304 (Wildcat Lake)

(Image credit: Future)

As for the specs of the CPU, we're looking at a six-core (2P + 4LPE) config, which is the only one Wildcat Lake has, but the Geekbench listing is reporting 1P + 4LPE across the two clusters, meaning one of the performance cores is disabled. It's an early engineering sample after all, so take all these numbers with a grain of salt; with an extra P-core in the mix, the thermal demands will change the performance equation.

Speaking of which, these Wildcat Lake processors are aimed at the ultra-budget market where efficiency is key. Think Chromebooks, mini-PCs, NAS, etc. For years, Intel has stayed on its Alder Lake platform (Twin Lake is just a refresh) for this category, so they've only had Gracemont E-cores for the past two generations. Wildcat Lake is expected to be the first to introduce Cougar Cove P-cores and pair them with Darkmont Low Power E-cores to provide a significant performance bump.

Intel Core 3 304 (Wildcat Lake)

(Image credit: Future)

That performance bump will come with power demands, however, as Wildcat Lake is apparently rated for only 15W, whereas Alder Lake-N and Twin Lake were between 9-15W. The TDP is what separates these bottom-tier Core Ultra 300 series chips — the Core 3 310 and Core 5 320 that were leaked a few days ago, too — as they seem to have identical specs to the "Ultra" branded Core Ultra 5 322 and 332 CPUs.

The Geekbench listing for the Core 3 304 confirmed its base frequency is 1.5 GHz, and it can boost up to 4.3 GHz, which is just 100 MHz lower than the lowest-end Core Ultra 5 SKU in Panther Lake. We don't see it on the Geekbench page, but we know that Wildcat Lake is expected to feature 2 Xe3 cores, which means basic integrated graphics support but still a huge leap over "Intel UHD" graphics in prior generations.

All in all, Wildcat Lake is shaping up to be a potent release for the value-oriented consumer, especially during a global component crisis. The exact release date for these chips is unknown at the moment, but we should be seeing them soon, given how much they've been popping up lately. Initial leaks from back in the day actually pointed to a 2025 release, but that was when Panther Lake was also expected in 2025.

Google Preferred Source

Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.

Hassam Nasir
Hassam Nasir
Contributing Writer

Hassam Nasir is a die-hard hardware enthusiast with years of experience as a tech editor and writer, focusing on detailed CPU comparisons and general hardware news. When he’s not working, you’ll find him bending tubes for his ever-evolving custom water-loop gaming rig or benchmarking the latest CPUs and GPUs just for fun.

7 Comments Comment from the forums
  • DS426
    Here we go again with Geekbench. Do the results include iBOT?

    I don't doubt that Wildcat Lake is a huge improvement over Twin Lake, but I'm not going to get excited over a single synthetic benchmark result on an engineering sample product.
    Reply
  • usertests
    As for the specs of the CPU, we're looking at a six-core (2P + 4LPE) config, which is the only one Wildcat Lake has, but the Geekbench listing is reporting 1P + 4LPE across the two clusters, meaning one of the performance cores is disabled.
    If the leaked 320 and 310 from a few days ago have the full 2P + 4LPE config, it wouldn't surprise me if the 304 has only 1P + 4LPE as indicated here. Maybe we'll see 1P + 3LPE or 0P + 4LPE under that.

    Wildcat Lake SKUs will bring a big ST and decent MT uplift over Alder Lake-N and Twin Lake, but the big thing to look out for is the graphics. Maybe in the range of +100-200% is possible.

    All in all, Wildcat Lake is shaping up to be a potent release for the value-oriented consumer, especially during a global component crisis.
    It needs to be quite cheap to matter. $300 and under laptops. They may feel like a sidegrade from i3-1215U/i3-1315U/100U (which can be found under $300) because of their single-channel memory and LPE-cores likely lacking L3 cache access. Graphics will also be closer than Alder Lake-N.

    Those Alder Lake-based cheap laptops are often coming with 8 GB soldered LPDDR4-3200 and an empty DIMM slot, so another DDR4-3200 stick will about match the bandwidth of a typical Wildcat Lake LPDDR5X system, with higher capacity at a relatively low cost.
    Reply
  • ezst036
    Please stop with Geekbench.

    Please blacklist Geekbench.
    Reply
  • IntelUser2000
    DS426 said:
    Here we go again with Geekbench. Do the results include iBOT?

    I don't doubt that Wildcat Lake is a huge improvement over Twin Lake, but I'm not going to get excited over a single synthetic benchmark result on an engineering sample product.
    It's a 4.6GHz Cougar Cove core, the same one in Pantherlake. The top Pantherlake core is only at 5.1GHz, and that's for the highest 388H. So in ST it won't be much behind.
    usertests said:
    Those Alder Lake-based cheap laptops are often coming with 8 GB soldered LPDDR4-3200 and an empty DIMM slot, so another DDR4-3200 stick will about match the bandwidth of a typical Wildcat Lake LPDDR5X system, with higher capacity at a relatively low cost.
    Pantherlake has memory compression that makes its throughput 20-30% better than it would be otherwise. LPDDR5-7466 single channel with that is then roughly equal to DDR-4500 dual channel. Bandwidth doesn't really matter for CPU workloads, but it will for GPU.

    CPU-wise, it will be better than 1315U, as that's only 2+4 on older gen. Wildcatlake is 2+4, but the E's are essentially like the P cores in 1315U.

    But, the Xe3 in Pantherlake isn't too memory bandwidth bound. It's less than half the bandwidth, but it has 1/4th the shader count. It's weakness is going to be the absolute GPU performance due to having such a cut down config.

    The older Alderlake-N and Twinlake actually did decent in gaming because the CPU was much better than the predecessor, and it was DDR5 so it was faster in 3D than the predecessor by a noticeable amount.

    You also get the battery life advantages Alderlake and Raptorlake won't have, plus latest media standard. I wouldn't get Alder/Raptor over this just because of the battery life advantages and media alone.
    Reply
  • usertests
    IntelUser2000 said:
    The older Alderlake-N and Twinlake actually did decent in gaming because the CPU was much better than the predecessor, and it was DDR5 so it was faster in 3D than the predecessor by a noticeable amount.

    You also get the battery life advantages Alderlake and Raptorlake won't have, plus latest media standard. I wouldn't get Alder/Raptor over this just because of the battery life advantages and media alone.
    It's going to be a close call for me. I'll be heavily scrutinizing Wildcat Lake. But I did not know about the memory compression, that's interesting.

    One weakness of Alder Lake-N was that the cheapest models would be paired with DDR4 memory. That's impossible with Wildcat Lake, so there will be a big jump in bandwidth in every unit you can buy. The (LP)DDR5 support of Alder Lake-N was also capped at 4800 MT/s.

    If Wildcat ends up with up to 2-3x the GPU performance of Alder Lake-N 32 EUs, then bandwidth improvements could be helpful. Maybe the higher bandwidth is also relevant to Wildcat Lake's LPE-cores, which aren't likely to have any L3 cache access and would be hitting memory more often than E-cores.

    I'm not sure I care about the media engine improvement. Is it better than Lunar Lake? Compared to Alder Lake-N, you'll get VVC (H.266) decode, which could be a niche codec for years, and 12-bit AV1 decode. Also AV1 encode, which matters to some users.
    Reply
  • IntelUser2000
    usertests said:
    It's going to be a close call for me. I'll be heavily scrutinizing Wildcat Lake. But I did not know about the memory compression, that's interesting.

    One weakness of Alder Lake-N was that the cheapest models would be paired with DDR4 memory. That's impossible with Wildcat Lake, so there will be a big jump in bandwidth in every unit you can buy. The (LP)DDR5 support of Alder Lake-N was also capped at 4800 MT/s.

    If Wildcat ends up with up to 2-3x the GPU performance of Alder Lake-N 32 EUs, then bandwidth improvements could be helpful. Maybe the higher bandwidth is also relevant to Wildcat Lake's LPE-cores, which aren't likely to have any L3 cache access and would be hitting memory more often than E-cores.
    They did per clock testing and while on Lunarlake the LPE cores were significantly slower than Arrowlake's E cores, despite the same uarch, on Pantherlake, they were only 5% slower than the regular E cores within the same chip. The LPE in Pantherlake is ~20% faster per clock than LPE in Lunar. Because Darkmont is few % faster than Skymont, it makes LPE in Pantherlake essentially equally to Arrowlake's Skymont, or roughly in line with Raptorlake P cores.

    Whatever they did with the memory controller(improved both bandwidth and latency), and I suspect even the dog-slow memory side cache on Lunarlake was improved on Pantherlake. The memory side cache on Lunar was useless in terms of performance, it was only little bit faster than system memory. Single thread CPU improvements are mostly latency bound, meaning Pantherlake improves there substantially.

    GPU improvements are harder to guess. But the 4 Xe core version is 3-5x the performance of the one in N305. So it depends on how it scales, but graphics-wise half the cores and half the bandwidth is at the worst half the performance. If anything, I don't believe the 4 Xe core version is needs the 7466 memory at all, and Wildcatlake might do better than 1/2.
    usertests said:
    I'm not sure I care about the media engine improvement. Is it better than Lunar Lake? Compared to Alder Lake-N, you'll get VVC (H.266) decode, which could be a niche codec for years, and 12-bit AV1 decode. Also AV1 encode, which matters to some users.
    I don't know much about those codecs, but back when Intel introduced VP9 hardware support on Broadwell, they talked about significant CPU usage reductions(thus power reduction) on video playback. Later that came to be important as it became used for Youtube.

    I tested bunch of Haswell laptops and it takes 3-3.5W on a U-class CPU to do playback on Youtube. A Y-class Kabylake one I tested does it at 1.5-2W. While a little may have to do with U vs Y, the testing across multiple laptops were roughly consistent with reviews, meaning the successful improved support had to do with the reduction.
    Reply
  • usertests
    By the way, there was a rumor that we'll eventually get Wildcat Lake Refresh with 4P + 4LPE, with no changes to GPU, NPU, or PCIe. I assume this reuses the Wildcat Lake I/O tile, while bringing in the smallest Panther Lake CPU tile.

    For all I know they could be using that tile in Wildcat Lake already but disabling 2+ P-cores and some cache. I have no idea, and Intel is not rushing to talk about Wildcat.
    Reply