Further details of Intel’s Kaby Lake-and-Radeon union leak

Intel's Indian web site leaked some tasty details of the company's upcoming CPU with Radeon graphics on package yesterday. One of those CPUs, now officially known as the Core i7-8809G, showed up on the company's overclockable desktop CPU list before being scrubbed. The leak revealed several tantalizing new facts about this union of blue-team and red-team technologies.

Most critically, the info Intel posted revealed that the i7-8809G will have a "target package power" of 100 W. Intel has already said that the CPU on the i7-8809G will be one of its H-series (35 W or 45 W) parts, so that means the graphics portion of the package could have roughly 55 W or 65 W to play with if we presume that power budgets are mostly fixed between this duo of functional units.

The CPU side of the i7-8809G will have four cores, eight threads, a 3.1 GHz base clock and 8 MB of L3 cache. Intel didn't provide boost speeds, but those base-clock and cache figures are dead ringers for the i7-7920HQ that tops Intel's Kaby Lake H-series offerings today. That chip has a 4.1 GHz boost speed, and if the i7-8809G does prove overclockable, builders might be able to push the CPU even further with some tweaks.

Despite the eighth-generation model number, the CPU die on board the i7-8809G bears all the hallmarks of being caffeine-free. The DDR4-2400 maximum memory speed points to this being a Kaby Lake-H quad-core chip. Coffee Lake six-core parts like the Core i5-8400, Core i5-8600K, Core i7-8700, and Core i7-8700K all support DDR4-2666 out of the box, while quad-core chips like the Core i3-8100 stick with the Kaby Lake-standard DDR4-2400.

Intel's leak also confirmed that the i7-8809G boasts Radeon RX Vega graphics power alongside its Intel CPU. Intel didn't reveal any information about the processing resources or graphics memory available from the Radeon RX Vega M GH processor, but the package power figure and our back-of-the-napkin divvying-up of that figure suggest that this could be a Radeon RX 550 or Radeon RX 560-class GPU.

We already know that RX Vega parts tend to scale well down the voltage-and-frequency curve from our experience with Vega desktop cards' power profiles, and we've already seen how well eight Vega compute units perform in a 25 W package aboard the Ryzen 5 2500U APU, so it's possible that drawing conclusions about this GPU's weight class from Polaris chips is pessimistic. We'll really need to wait for further details to peg the position of this chip in the GPU hierarchy.

The Core i7-8809G was already slated for launch in the first quarter of this year, and the 2018 CES kicks off next week. We hope to learn more about Intel's compact powerhouse and the systems it'll inhabit when we hit the ground in Vegas soon. Thanks to TR tipster SH SOTN for catching this leak.

Comments closed
    • stevelarsc
    • 1 year ago

    Well, that was a detailed note on the Intel’s Kaby Lake-and-Radeon union leak. The the i7-8809G boasts Radeon RX Vega graphics power alongside its Intel CPU. Intel didn’t reveal any information about the processing resources or graphics memory available from the Radeon RX Vega M GH processor
    [url=http://www.clippingpathphotoediting.com/clipping-path-service/<]clipping path service[/url<]

    • johnsmith26
    • 2 years ago
    • dodozoid
    • 2 years ago

    I now realised that first consumer oriented implementation of intel’s EMIB technology is not on intel chip at all! Not sure if not first at all – have they already launched any other product bearing this tech?

      • willmore
      • 2 years ago

      FPGAs were formally announced a couple of weeks back using EMIB. Which technically means the first EMIB using chips were ARM.

    • Star Brood
    • 2 years ago

    Can someone please let me know what is unique about this VS just an Intel laptop with integrated Radeon graphics (ie. Mac)?

      • Pancake
      • 2 years ago

      The only functional difference I can think of is that this may have better latency and bandwidth between CPU and GPU for magic Apple stuff.

        • Anonymous Coward
        • 2 years ago

        It should offer pretty mean performance in an unusually small package, probably the GPU silicon is somewhat oversized and undervolted. Basically a luxury product.

      • dodozoid
      • 2 years ago

      This is unique due to small footprint ot the entire package as it can fit full blow workstation on previously unimaginably tiny PCB on one side and due to utilization of HBM via EMIB instead of an interposer.

      • tipoo
      • 2 years ago

      It’s smaller which could lead to more size for batteries (or more likely, more lightness), and regular solutions don’t have HBM2.

    • ronch
    • 2 years ago

    Is TSMC gonna be fabbing the GPU silicon for these or will it be Intel?

    Edit – just a thought, guys. What if they also put the AMD logo on the GPU silicon there? Come on, Intel. That is AMD there. Give them some credit.

    • ronch
    • 2 years ago

    I’d actually consider an APU for my next rig given my use cases. Would it be from Intel or AMD? Yes I’m an AMD supporter but giving away their graphics to Intel, a crown jewel, feels like AMD is letting their fans down. It’s probably how North Koreans would feel about Kimmy if he suddenly agrees to abide by everything the UN says (for the record I’m not for or against NK.. I’m just an observer). Kinda sucks for AMD to do that and in retaliation I just might get an Intel APU like thiis, if you can call it that.

      • freebird
      • 2 years ago

      You are making an invalid assumption that the North Korean populace “support” their current dictator…

      So that was not a good comparison.

        • Rikki-Tikki-Tavi
        • 2 years ago

        Dude, that was a joke.

          • mcarson09
          • 2 years ago

          It was a bad joke.

      • mcarson09
      • 2 years ago

      It’s not an Intel APU. It’s a link between a CPU and GPU. Intel coulds also use Nvidia GPUs instead.

    • mcarson09
    • 2 years ago

    The hype on this thing is killing me! I hope intel makes the video card drivers for this thing. This thing sounds like a great HTPC.

    • WhatMeWorry
    • 2 years ago

    “…figure and our back-of-the-napkin divvying-up of that figure suggest that this could be a Radeon RX 550 or Radeon RX 560-class GPU.”

    I thought the 8809G would be easily faster than the RX 550 or RX 560. Because of the chart below:

    CU = Compute Units or Cluster Units
    SP = Stream Processor or Shaders

    RX 550 8 CU = 512 SP
    RX 560 15 CU = 960 SP
    8809G 24 CU = 1536 SP
    Fenghuang APU 28 CU = 1792 SP
    RX 570 32 CU = 2048 SP
    RX 580 63 CU = 2304 SP
    RX Vega 56 56 nCU 3584 SP
    RX Vega 64 64 nCU 4096 SP

    • Freon
    • 2 years ago

    Desktop? Wondering what the target audience is. Those who care (gamers) will still want something they can upgrade later. Those who don’t will survive on the Intel IGP. The middle ground seems to be thin.

    For mobile this makes plenty of sense. Upgrades are rare and difficult. Intel is ahead of the game on efficiency and speed for CPUs.

    It even might make some sense for consoles, though I wonder if Microsoft and Sony really care about having an Intel vs AMD CPU. Maybe AMD doesn’t want to spend the money to properly scale Zen to an affordable console part.

      • NTMBK
      • 2 years ago

      I imagine it’s for NuC style mini desktops.

        • Freon
        • 2 years ago

        Why not just stick with the IGP for such uses?

          • NTMBK
          • 2 years ago

          …because this plays games a lot faster?

      • derFunkenstein
      • 2 years ago

      Something the size of a Skull Canyon NUC would be cool.

    • derFunkenstein
    • 2 years ago

    I’ll take two, particularly if one is an overclockable desktop part mounted on a mini ITX motherboard.

      • NTMBK
      • 2 years ago

      I’d rather have it on a mini STX board!

        • derFunkenstein
        • 2 years ago

        Either way, but if it’s OC’able I think I’d rather have the room for a decent cooler. I had also pondered low-profile mITX but rejected that for the same reason.

      • tsk
      • 2 years ago

      Or even mini STX would be great.

    • NoOne ButMe
    • 2 years ago

    I swear every site forgets the mobile Polaris11 exists. And it cuts the TPD in under half with like 25% performance loss. 1.7 1.8 Tflop at 35W with GDDR5

    If Intel makes these sanely priced, which Of course they won’t, they should kill in Mobile, everything under a GTX 1060.

    Leaks seem to suggest it’s about 3.5-4 TFLOP peak.

    I think it will fill the market between 1050ti and 470D in performance… it Intel price good (it won’t) it could kill a segment of 1050/1050ti laptops.
    Hopefully they will so Nvidia drops prices on 1060+ laptop parts for buyers 🙂

    Ampere(?) May ‘fix’ that?

      • tipoo
      • 2 years ago

      Forget websites – has every manufacturer outside of Apple forgotten it exists? Maybe they just ordered all they could make?

      Wishes for higher wattage GPUs aside, its perf/watt was indeed pretty good when dialed back to 35W. Particularly the top RP 560, as it has the same wattage as the bottom tier.

        • Goty
        • 2 years ago

        I actually saw a mobile RX580 in one of those several-inch-thick gaming “notebooks” at Best Buy the other day while helping my dad shop for a new machine. I think it was one of the HP Omen models.

        To the chagrin of some around here, I steered him toward a 7700HQ/GTX 1060 MSI model.

          • Chrispy_
          • 2 years ago

          I don’t blame you. The only variants I’ve seen are full-fat 2304 shader parts clocked at 1077MHz.

          Even assuming these are cherry-picked parts and running significantly undervolted compared to their desktop brethren, that’s around 100W for those GPU clocks.

          The 1060 Max-Q, on the other hand, is a drop-in replacement for laptops designed for the 53W of a GTX 1050, and it’s going to offer similar (if not better) performance than an RX580 suffering from a 20% loss of clockspeed.

    • tipoo
    • 2 years ago

    100W target TDP is of course too high for what I was hoping for this in, the 2018 15″ rMBP. But I imagine it could be TDP-down’ed to 60W total and still be a significant upgrade thanks to the HBM2 (every rMBP GPU upgrade is still stuck on 80GB/s memory).

    It would mirror Iris Pro where Apple was the one to ask for it, Intel used it in 84W desktops, but Apple got a 47W part.

      • chuckula
      • 2 years ago

      Given that this listing popped up in the desktop category and specifically cites an overclockable SKU, I’d imagine that mobile variants will of course dial the power consumption down quite a bit. A 65W total package power solution is actually pretty reasonable in the mobile workstation segment.

      • Belldandy
      • 2 years ago

      Well TDP is lower than the combined 7700HQ and GTX1060. Theoritically it shouldn’t be that difficult as that combo is found in many thin and light gaming laptops.

        • Goty
        • 2 years ago

        Are we assuming that’s the Max-Q GTX 1060? What’s the TDP of that part?

        I think the “normal” 1060 laptop part has a TDP of somewhere around 80W, so if we assume the processor burns the same 35W as a 7700HQ (unlikely given the base clock, but maybe with process improvements), this Vega implementation could get as much as 65W to play with. Even given the dedicated HBM stack, I don’t really see it putting up much of a fight against a 1060. The 1050 Ti on the other hand…

          • thecoldanddarkone
          • 2 years ago

          The 7700hq is a 45 watt part typical. It is configurable to 35w but it’s 45w by default.
          [url<]https://ark.intel.com/products/97185/Intel-Core-i7-7700HQ-Processor-6M-Cache-up-to-3_80-GHz[/url<] The Xeon 1535mv6 is also a 45 watt typical 3.1-4.2 [url<]https://ark.intel.com/products/97468/Intel-Xeon-Processor-E3-1535M-v6-8M-Cache-3_10-GHz?wapkw=xeon+1535mv6[/url<] Some notebooks have a higher configured tdp, example my p51 (xeon 1505mv6) has a configured tdp of 55w. In most workloads (including avx workloads) it will stay at 3.6 on all threads. Even if it puts up a fight/is faster than the gtx 1050ti, it's not like NVidia doesn't have options. When NVidia released the 1050/1050ti/1060/1070/1080 they didn't originally release the low power variants. The gtx 1060 max-q parts are 60-70 watts (including dram). If your going for thinness than yes it might be the go to configuration. However if your going with performance/$ it's very unlikely going to be this. I don't think it's going to make a huge wave. If I'm wrong that's awesome. It's one of those, we shall eventually see.

            • Goty
            • 2 years ago

            Right you are regarding the TDP; I just brain-farted a bit when typing after looking at the Ark page.

            I also agree this probably isn’t going to make much of a dent in the laptop market. There are too many good solutions out there now that, especially at the stated TDP, I don’t see much room for it. True SFF systems seem to be a more likely niche for something like this, but I don’t know how big that market is.

      • DavidC1
      • 2 years ago

      Anything lower than 100W will reduce performance disproportionately. The CPU likely can’t go lower than 35W. There’s also a 65W version coming but that seems to have a significant performance difference.

      The rumors are the 100W is VR-capable part with RX 470 like performance and 65W more in the range of RX 560.

      Considering the effort they are putting in, and the amount of memory bandwidth it will have, that makes sense.

    • chuckula
    • 2 years ago

    Raj Kod and Kaby-G gonna [url=https://www.youtube.com/watch?v=aYD3gLCXXuU<]regulate.[/url<] [quote<]Regulators. We regulate any benchmarkin' of his workloads. We’re damn good too. But you can’t be any geek off the street. You gotta be handy with the silicon, if you know what I mean. Earn your keep. Regulators, mount up![/quote<]

    • ronch
    • 2 years ago

    Now I want further details about how Raja Koduri ended up at Intel.

      • K-L-Waster
      • 2 years ago

      Probably by car… outside chance it might have been by helicopter.

        • derFunkenstein
        • 2 years ago

        he drove through the impending Intel CPU security hole.

        • ronch
        • 2 years ago

        With his Ryzen rig in the back seat to show Intel, I’m sure.

    • Anonymous Coward
    • 2 years ago

    Now they just need to put this into a Thinkpad which I can justify at work, and it’ll be xmas all over again.

      • tipoo
      • 2 years ago

      I got bought out by a company that only provides Thinkpads T470S’s as laptops, I’m overdue for a replacement on my big desktop replacement workstation. I’m hoping the refresh to 8th gen quads is soon, a dual core ULV would be pretty brutal with my workload…

    • juzz86
    • 2 years ago

    Wow, what a beast. If that TDP rating turns out close to the truth, this thing should hunt well.

      • HERETIC
      • 2 years ago

      Being a little pedantic-There was no mention of TDP.
      There was TPP at 100Watts………..

        • tipoo
        • 2 years ago

        Is it in effect different? Since there’s three high power consuming chips on one package, TPP just made more sense, but it’s still pretty close to end TDP I assume?

        Of course you already said it’s pedantic

          • HERETIC
          • 2 years ago

          No-I still don’t understand why people are using TDP as Power usage.
          TDP is Thermal design power-For heatsink manufacturers to know how
          much waste heat to design their heatsinks for…………..

            • sleeprae
            • 2 years ago

            Is your complaint that TDP is not reflective of actual power usage for a particular chip or SKU? That’s a completely valid and fair criticism.

            In terms of accounting for energy (as in, the law of conservation of energy), heat is the only meaningful output, so heat is effectively identical to power consumption for processors.

            • Chrispy_
            • 2 years ago

            I think it might be.

            The only difference between TDP(TPP) and actual power consumption is that the TDP is an average when running at full load, so the power delivery circuitry has to be capable of dealing with much higher burst power until the relevant temperatures have been reached.

            Just because the Kill-a-Watt meter might read 160W on a 100W part for half a minute or so doesn’t mean that the average power consumption over a longer period will ever exceed the TDP…

        • juzz86
        • 2 years ago

        Understood – sorry. I saw you post about this in another thread, and I actually smiled because you’re fighting the good fight – and here I’ve gone and done it myself!

        +1

Pin It on Pinterest

Share This