Intel's upcoming Nova Lake desktop SKU to require 65W of separate power delivery for its iGPU, leaker claims — beefy integrated graphics could require two VCCGT phases for 12 Xe3P cores

News
By
Published

One VCCGT phase can usually sustain up to 30W safely.

Raptor Lake CPU
Raptor Lake CPU (Image credit: Intel)

Intel's next-gen Nova Lake family of desktop CPUs is shaping up to be the company's most exciting launch in years, with a major performance leap expected across the board. A new leak from Jaykihn now claims that one of the SKUs from this lineup will have an iGPU so strong that it will require 65W of power delivery to achieve full performance. It will also apparently need two VCCGT phases to handle said power.

The SKU in question is a 16-core part composed of 8 P-cores, 4 E-cores, and 4 LP-E cores, while the iGPU is said to have 12 Xe3P cores. Just as a reminder, Xe3, also known as "Celestial," is an enhanced, optimized version of the Xe3 (Battlemage) graphics architecture that already debuted on Panther Lake. Similarly, Nova Lake is expected to use Coyote Cove P-cores and Arctic Wolf E-cores (and LP-E cores).

Top-end Panther Lake SKUs come equipped with the Arc B390 iGPU, which also has 12 Xe3 cores, so this Nova Lake SKU with 12 Xe3P cores should perform even better. Especially when you consider the thermal and power headroom at its disposal. AMD's desktop APUs, the Ryzen G-series, are usually 65W parts as well, but that's the TDP for the entire chip, not just the integrated graphics.

Latest Videos FromTom's Hardware

Requiring 65W of dedicated power delivery via two VCCGT phases would constitute a kind of top-end iGPU performance we haven't seen before. In fact, this rumored Nova Lake chip is arguably veering into Strix Halo territory where the up to 40 Compute Units on flagship SKUs can sip around 70W-80W of power. However, that's still a mobile part with an integrated SMU that can dynamically allocate power between the CCD and iGPU.

Swipe to scroll horizontally
Nova Lake-S Rumored SKUs

SKU

Core Config (P+E+LP-E)

bLLC

TDP (Unlocked/Locked)

52 Cores (dual-tile)

(8+16)+(8+16)+4

288MB

175W

44 Cores (dual-tile)

(8+12)+(8+12)+4

264MB

175W

28 Cores

8+16+4

144MB

125W

28 Cores

8+16+4

-

125W / 65W

24 Cores

8+12+4

132MB

125W

24 Cores

8+12+4

-

125W / 65W

22 Cores

6+12+4

108MB

125W / 65W

22 Cores

6+12+4

-

125W / 65W

16 Cores

4+8+4

-

65W / 35W

12 Cores

4+4+4

-

65W / 35W

8 Cores

4+0+4

-

65W / 35W

6 Cores

2+0+4

-

65W / 35W

Nova Lake-S would likely use a traditional desktop rail split where the motherboard's VRMs must deliver up to 65W via two separate dedicated VCCGT phases (since one wouldn't be sufficient) to power the integrated graphics tile independently from the VCCCore CPU phases. That's unprecedented territory, and it serves as just one of many rumors from the Nova Lake launch that have us excited for the competition that's brewing for next year.

Current reports pin both AMD and Intel's next-gen releases to show up at CES 2027 instead of later this year. Therefore, take everything with a grain of salt; a lot could change between now and then, and we don't have confirmation on any SKU from Intel. The last time we saw Intel put out a strong desktop APU was 2015's Core i7-5775C, so a product like this has been a long time coming.

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.

10 Comments Comment from the forums
  • usertests
    People (me) have been asking for a socketed Intel desktop APU for years, and now we're finally gonna get it.
    Reply
  • thestryker
    The prospect of this is genuinely exciting to me if the SKU is priced right and there at least some affordable motherboards with appropriate VRM. Entry level discrete graphics is effectively dead since the only parts that make sense under the B570/5050 are used purchases. By allowing the GPU to have free reign power wise it should maintain boost clocks extremely well. Intel should be sticking with a large cache (for low core count) design which will mitigate slower desktop memory speeds if there's no reasonably priced high speed DDR5 available. This all combined with an improved architecture I really wouldn't be surprised to see somewhere around desktop RTX 3050 performance. This would be a great place to be as either a starter gaming system or small size/low powered system.

    The biggest problem with all better integrated graphics on desktop to date has been that they're tied to upper CPU SKUs. This should sit in the lower/middle of Intel's desktop stack.

    The biggest problem with AMD APUs has been that they're compromised should you upgrade to discrete. With less cache and PCIe connectivity it means not only are you no longer using the integrated graphics you paid for but the CPU you're left with is worse than the regular core count equivalent. This for sure won't have a compromised cache setup and there's no reason it should have compromised PCIe connectivity either.
    Reply
  • usertests
    thestryker said:
    The biggest problem with all better integrated graphics on desktop to date has been that they're tied to upper CPU SKUs. This should sit in the lower/middle of Intel's desktop stack.

    The biggest problem with AMD APUs has been that they're compromised should you upgrade to discrete. With less cache and PCIe connectivity it means not only are you no longer using the integrated graphics you paid for but the CPU you're left with is worse than the regular core count equivalent. This for sure won't have a compromised cache setup and there's no reason it should have compromised PCIe connectivity either.
    It's an afterthought. AMD puts the mobile die onto the socket (although desktop APUs could be made with chiplets soon), you get whatever the compromises were. Hopefully the user understood the compromises going in, or got it at a budget price.

    Competition is key. AMD phoned it in with high MSRPs, skipped/late generations, and now an OEM-only Krackan Point lineup (no Strix models with 880M/890M). The desktop APU is often derided for poor price/performance, and that's been 100% tied to AMD's decisions. Now we could see a different way of doing things from Intel.

    If AMD wants to challenge 12 Xe3P graphics performance, they're going to have to use Medusa Halo Mini (24 CUs RDNA5, 128-bit). CPU performance should technically be in the lower/middle of AMD's desktop stack, if it is packing 4+8+2 cores (no optional compute chiplets added), which should be slower than an Olympic Ridge CPU with a single 12-core CCD. Probably with 10 MiB graphics L2 cache to challenge Intel's ~16 MiB L2 cache, which was a big part of Panther Lake's secret sauce.
    Reply
  • Nota ReAlperson
    Intel has made some powerful igpus since the i7 5775c. The iris pro p580 and the Kaby lake g with amd vega graphics come to mind.
    Reply
  • Gururu
    This is a great solution for the casual gamer who can get by without a dGPU. I survived about a year on UHD 770, however this could set someone up for years.
    Reply
  • usertests
    Nota ReAlperson said:
    Intel has made some powerful igpus since the i7 5775c. The iris pro p580 and the Kaby lake g with amd vega graphics come to mind.
    The i7-5775C was an experiment for Intel, with eDRAM + large iGPU only carried forward to a handful of mobile processors and BGA desktop packages instead of socketed desktops. Kaby Lake G was another weird experiment, obviously not using Intel's own graphics.

    They have the expertise, they have the technology. We'll see if their new desktop APUs survive in the face of Nvidia deal, or if they simply use Nvidia tiles later on.
    Reply
  • samopa
    My question is: how far can this iGPU help increase the performance of a discrete GPU?
    I'd always opt for a processor without an iGPU if possible, so it can run at lower power and/or heat. But recently, an article in Tom's throws a possibility to add a second GPU, a lesser one than your main GPU, to increase performance by handling other tasks (such as AI, physics, etc).
    I wonder if such a case is possible with an iGPU, and if it is worth it ?
    Reply
  • Gururu
    samopa said:
    My question is: how far can this iGPU help increase the performance of a discrete GPU?
    I'd always opt for a processor without an iGPU if possible, so it can run at lower power and/or heat. But recently, an article in Tom's throws a possibility to add a second GPU, a lesser one than your main GPU, to increase performance by handling other tasks (such as AI, physics, etc).
    I wonder if such a case is possible with an iGPU, and if it is worth it ?
    It is possible to devote certain applications to an iGPU and others to the dedicated GPU in Win11 settings. I have tried this and it was more trouble for me than what it was worth. I don't know if two dGPUs have a better outcome.
    Though it is always advisable to get the CPU with integrated GPU for troubleshooting purposes, a very powerful iGPU is not needed. The iGPUs here are very powerful and would mostly be useful for casual gamers allowing one to save a few hundred bucks by not needing a low-mid tier dGPU. It will be interesting to see if this iGPU setup can be maximized for tasking like you suggest in combination with a dGPU.
    Reply
  • TerryLaze
    samopa said:
    My question is: how far can this iGPU help increase the performance of a discrete GPU?
    I'd always opt for a processor without an iGPU if possible, so it can run at lower power and/or heat. But recently, an article in Tom's throws a possibility to add a second GPU, a lesser one than your main GPU, to increase performance by handling other tasks (such as AI, physics, etc).
    I wonder if such a case is possible with an iGPU, and if it is worth it ?
    Since something like that would need the game devs to do work I wouldn't hold my breath for it happening.
    In theory yes the igpu could handle upscaling/frame generation or general filters but the game would have to be aware of the igpu and actually give these tasks to it.
    Reply
  • usertests
    TerryLaze said:
    Since something like that would need the game devs to do work I wouldn't hold my breath for it happening.
    In theory yes the igpu could handle upscaling/frame generation or general filters but the game would have to be aware of the igpu and actually give these tasks to it.
    I don't think game devs necessarily need to do any work, if the drivers can handle some of these dual-GPU applications that don't actually involve a game rendering on more than one GPU.

    Here are some I've seen:

    1. One GPU handles game, the other GPU handles Lossless Scaling.
    2. One GPU handles game, the other GPU handles PhysX. This became necessary for a handful of games after Nvidia deprecated 32-bit PhysX/CUDA support on 50-series, until they reversed it.
    3. When Nvidia first showed off DLSS 5, it was running on two 5090s.

    GDvfIbRIb3U

    ClmKqKrrJGA

    h4w_aObRzCc

    xfv6g7DIYf0

    Whether any of this is a good or practical is up for debate, but dramatically faster iGPUs could open up more options, while working with boards with only one x16 slot.
    Reply