AMD announces laptop-bound Radeon HD 7000M series

Ah, the moment we’ve all been waiting for: next-generation Radeons are finally here! AMD has just announced its Radeon HD 7000M series of mobile graphics processors, which feature 40-nm silicon based on the company’s existing VLIW5 Terascale 2 architecture, and—

…wait, what?

I was just as puzzled when AMD briefed us on these GPUs last month. Despite the model numbers, and despite all of the rumors we’ve heard about 28-nm graphics hardware being on the verge of release, these 7000-series parts are based on the same architecture and 40-nm fab process as AMD’s current, 6000M-series mobile GPUs.

Today’s launch covers only 7400M-, 7500M-, and 7600M-series offerings, so it would appear that AMD has left room in the lineup for real next-gen Radeons. There’s no sign of them yet, though, and the specs for today’s new arrivals aren’t terribly exciting:

  Radeon HD

7400M series

Radeon HD

7500M series

Radeon HD

7600M series

Process 40 nm 40 nm 40 nm
Stream processors 160 480 480
Memory type DDR3/GDDR5 DDR3/GDDR5 DDR3/GDDR5
Memory bus width 64-bit 64-bit 128-bit
Eyefinity support up to 4 displays up to 4 displays up to 6 displays
Target segment “Value” “Mainstream” “Performance-Thin”

(All three GPU families also feature DisplayPort 1.2 and HDMI 1.4a connectivity, as well as support for AMD HD3D and Blu-ray 3D.)

The specs above don’t hint at any great performance improvements over the 6400M and 6500M series. Yes, the 7500M has 80 more stream processors than the 6500M, but the 7400M is also saddled with half the memory interface width of the 6400M. At least all the 7000M-series GPUs have UVD3 video decoding blocks, which support DivX and XviD video acceleration. You might remember that some of the 6000M-series GPUs, including the 6500M, feature silicon repurposed from the Mobility Radeon HD 5000 series and thus have older UVD logic.

AMD’s Product Manager for Discrete Graphics, Ogi Brkic (pronounced birr-kitch), also hinted during the briefing that 7000M-series GPUs fit in smaller packages than their predecessors, allowing them to squeeze into thinner laptops. We weren’t able to coax straight answers or additional specifics out of the company before the launch, though, despite repeated requests. AMD’s layoffs in PR and marketing have hit hard, and they’ve left us with only a vague understanding of what makes these new products, well, new. The best AMD could offer was the promise that, after the launch, more information would become available on its website.

Hopefully, AMD will fill out the 7000M series with truly next-generation GPUs before too long. Launches like this one seem counterproductive, though; they soften the impact of actual next-gen product releases and leave us press guys scratching our heads.

Comments closed
    • swaaye
    • 8 years ago

    I doubt there were ever plans for a full new lineup using bleeding edge manufacturing.

    • NeelyCam
    • 8 years ago

    So, TSMC 28nm is in trouble..

    You know, what this means is that NVidia will rule the ARM roost again with Tegra3. Everybody else is aiming for 28nm, and will now be delayed. People were making fun of Tegra3 (“40nm!!?? Pathetic!”), but I had a feeling that not taking a risk with the 28nm process might end up being a brilliant move for them.

    Turns out it was.

      • Yeats
      • 8 years ago

      I don’t know about that…Tegra 3 has hardly been a quantum leap in power or efficiency compared with A5. And iPad 2 is still the tablet target.

        • NeelyCam
        • 8 years ago

        A5 is special, and only Apple has access to it… Tegra3 is better than anything any of the TSMC customers have or will have for the next six months.

        I did forget about Samsung Exynos, though; that could be the top ARM chip in the near future.. I wonder if it (or its “Apple variant”) will be used in iPad3..?

    • wierdo
    • 8 years ago

    “We expected Southern Island products based on TSMC’s new HKMG 28nm process, but the fact of the matter is that TSMC’s HKMG 28nm process is running late—yields and production capacity just aren’t good enough for the production of high volume retail products for 2011. We may yet see some kind of 28nm product before the year is out so that AMD meets their stated commitment, but a complete 28nm launch in 2011 is off the table. However, AMD is concerned they need to launch new mobile products at the end of this year whether they have new GPUs or not, meaning they need to make do with what they already have”

    [url<]http://www.anandtech.com/show/5200/introducing-amds-radeon-7000m-and-nvidias-geforce-600m-mobile-gpus[/url<] So apparently this isn't using TSMC's 28nm process, just something to refresh OEM products with - for the sake of OEMs claiming they have a brand new shiny product line.

    • HisDivineOrder
    • 8 years ago

    I guess you don’t need PR staff when all your new launches are going to be disappointing and/or catastrophic. You can just reduce PR to, “Visit the website in a few weeks and maybe we’ll give you some more info.”

      • yogibbear
      • 8 years ago

      Yeah but that info will inevitably be wrong… *cough* 2 billion *cough*

        • wierdo
        • 8 years ago

        Just my guess, but AMD decided the enthusiast market is a lost cause – even if they don’t say so outright – and are concentrating their efforts on OEMs, that way they can expand their market in the consumer space by catering to the big boys that buy in bulk, easier to target their needs as well this way.

        They’ll still have some products for enthusiasts of course, just gonna cut expenses where possible, like cutting down on PR and prioritizing OEM targeted products in this case.

    • burntham77
    • 8 years ago

    This is all well and good, but I just want to see a high end mobile AMD GPU paired with their fastest mobile AMD CPUs. I know Intel’s are much faster, but I wants what I wants.

    • Hattig
    • 8 years ago

    So TSMC’s 28nm is not good enough for mainstream products currently – that’s no surprise.

    I guess they had to announce these products because they were ready, whilst the 28nm VLIW4 7700M/7800M/7900M will be coming next month or later.

    Also assuming that 28nm desktop HD7900 will be the only GCN based GPUs from AMD this generation.

    • ModernPrimitive
    • 8 years ago

    On HP’s site, the newly announced Envy’s don’t even list the model of GPU they come with. Just says “1GB discrete Radeon bla bla bla”….Yeah, I can see myself jumping and ordering one of those !

      • indeego
      • 8 years ago

      But it doesn’t bug you that HP exits and then re-enters the PC market? How strange.

        • ModernPrimitive
        • 8 years ago

        Different subject altogether. I don’t recall claming either way, but for the record, yes HP is lost is the woods with the PC business, WebOS and other matters….

    • Chrispy_
    • 8 years ago

    By [i<]anti-marketing[/i<], they're hoping to trick you into [i<]not-caring[/i<] about the actual next-gen GPU's when they arrive, so you [i<]won't[/i<] get angry about the poor availability and paper launch.

    • yogibbear
    • 8 years ago

    I used to be a 6000M series GPU like you, then I took an arrow in the knee.

      • glacius555
      • 8 years ago

      ..and then I started to fly backwards, just like 7000M specs?

        • dashbarron
        • 8 years ago

        Ah, like Skyrim dragons?

    • Silus
    • 8 years ago

    LOL

    That is all!

    • Geistbar
    • 8 years ago

    It was ridiculous when Nvidia re-purposed the 8800 a dozen times in a row, and it’s just as ridiculous now, when AMD re-uses their current architecture and yet gives them a model number bump. Stuff like this only serves to dilute the value of a “new” generation of products, as stated by the model number.

    Also, what kind of target segments are: “Value”, “Mainstream”, and “Performance-Thin”? They could all mean the same thing to anyone that encounters the names in isolation- if not for the first one being “value”- aka “cheap”, I wouldn’t be able to tell which was the better end of the names.

    • Ushio01
    • 8 years ago

    Has the 40nm to 28nm transition been the longest time coming for a die shrink or what? It’s already been over 2 and a half years since 40nm was first introduced and it looks like there are still many months to go.

    (In case anyone’s forgotten the AMD 4770 released 29th april 2009 was the first released 40nm GPU).

      • travbrad
      • 8 years ago

      Yep, and sadly because of this GPU perf/price has barely changed (at least in the lowend/midrange). I got a HD4830 almost 3 years ago for $93 and you can’t really get a faster card in that price range even today, 3 years later. More efficient yes, but not faster.

      The same is true of my GTX460 “SC” I got more than a year ago. It was $190 then, and a GTX560 is the $200 card now (which is only marginally faster).

      I’m really hoping 28nm is going to bring better price/perf, because $200 is sort of my GPU budget “limit” and there’s nothing faster for me to get right now.

        • BestJinjo
        • 8 years ago

        Well Newegg does have HD4870 for $55 and some GTX460 cards dipped below $100. Overall though, there has hardly been any increase in performance from HD5850/5870 which came out in September of 2009. GTX470/480 were mildly faster, GTX570/580 also barely faster, while HD6870/6950/6970 are all in the same ballpark as the 5870 as well. Overall for HD5870 user, there has been 0 reason to upgrade and the same is true for HD4870 user who wants a much faster card for $100 today.

          • Yeats
          • 8 years ago

          [quote<](#10 travbrad), Well Newegg does have HD4870 for $55 and some GTX460 cards dipped below $100. Overall though, there has hardly been any increase in performance from HD5850/5870 which came out in September of 2009. GTX470/480 were mildly faster, GTX570/580 also barely faster, while HD6870/6950/6970 are all in the same ballpark as the 5870 as well. Overall for HD5870 user, there has been 0 reason to upgrade and the same is true for HD4870 user who wants a much faster card for $100 today.[/quote<] What? In some cases the GTX580 is twice as fast as the 5870. [url<]http://www.anandtech.com/bench/Product/294?vs=305[/url<]

      • NeelyCam
      • 8 years ago

      To be fair, 40nm was “introduced” with 4770 when it was nowhere near ready for high volume. In practice, it wasn’t really yielding until maybe December 2009, and even then it was harder than hell to find those 5850s/5870s in stock..

      Meanwhile, Intel seems to be trending towards 2 years 4 months between 32nm and 22nm. Maybe this is the new norm – about 2.5 years between process nodes now that it’s getting increasingly hard.

        • swaaye
        • 8 years ago

        Yeah there were some 40nm 4xxx Mobility Radeons that never came out to my knowledge. The 4830 and 4860. Maybe they came out eventually.

    • ronch
    • 8 years ago

    Reminds me of the time Nvidia renamed the 8000 series to become the 9000 series. Nothing new, just a new badge slapped on. And packaging… yeah, that’s really exciting, I’m spilling my guts just thinking about it.

      • Chrispy_
      • 8 years ago

      And a new price label!

      I bought a 9100 series because it was only half the price of an 8500LE, and the number was bigger, so it [i<]felt[/i<] better. The box it came in felt better, anyway.

    • jjj
    • 8 years ago

    looks like you jumped the gun a bit since AMD hasn’t actually announced anything yet

      • FuturePastNow
      • 8 years ago

      If this is correct, then even after AMD announces something they’ll have announced nothing.

    • JustAnEngineer
    • 8 years ago

    Meh. Krogoth is not impressed.

      • Arclight
      • 8 years ago

      Has it become a internet meme yet?

        • ImSpartacus
        • 8 years ago

        We’re close, but it could be a while. It’s not a big deal.

          • TaBoVilla
          • 8 years ago

          it is a meme for me, I use Krogoth all the time with family and friends

          • Palek
          • 8 years ago

          [quote<]We're close, but it could be a while. It's not a big deal.[/quote<] Surely you meant to say: [quote<]We're close, but it could be a while. [u<]But meh, Krogoth is not impressed.[/u<][/quote<]

        • burntham77
        • 8 years ago

        A reference to Total Annihilation? I don’t see that happening.

      • yogibbear
      • 8 years ago

      I <3 Krogoth too.

    • xeridea
    • 8 years ago

    VLIW5? Isn’t that from the 5000 series and some rebranded 6000 series? Wonder how they made it to the 7000, unless thats a typo. Hopefully some 28nm parts come soon I am interested in seeing the improvement, though I won’t be getting one since I was forced to get a 6850 because my machine with 4850 died a chocolaty death.

      • flip-mode
      • 8 years ago

      VLIW5 was introduced in the Radeon HD 2900 XT.
      [url<]https://techreport.com/articles.x/12458/2[/url<]

        • xeridea
        • 8 years ago

        I know it was 2-5k, I thought it went out mostly with the 5k series though.

          • khands
          • 8 years ago

          Nope, only the 6900 cards had VLIW4, all the rest were still VLIW5.

    • Duck
    • 8 years ago

    Another questionable and disappointing product update from AMD.

    I think I see a pattern forming…

    • BoBzeBuilder
    • 8 years ago

    Well… there goes moore’s law.

      • Buzzard44
      • 8 years ago

      Au contraire. Moore’s law states that every ~1.5 years, model numbers for the same product will double.

      See for yourself. I got that from two totally independent sources – NVidia’s website and AMD value graphics division webpage.

        • Ringofett
        • 8 years ago

        I should dig up my old 9700 Pro. Forget the 7000-series, I’ve already got 2000 on that junk.

        • cynan
        • 8 years ago

        Eh? Moore’s Law, in its original form, applies to transistor density, namely that it will more or less double every 2 years.

        By extension, a doubling of transistor density more or less correlates with a doubling in performance, though the degree by which this correlates is highly dependent on architecture revisions and outright redesigns.

        Anyway, since these mobility chips have been 40nm dies for the last couple of years, and have seemingly similar transistor counts (as they are not taking up more space or have higher TDPs) then the OP is correct. This is not really in conformance with Moore’s Law. This will translate to AMD having similar transistor density, and likely performance ,across similar product segments for a least 3 years (ie, HD 5650 -> HD6550 -> HD7500/7600).

          • OneArmedScissor
          • 8 years ago

          *WHOOOOOOOOOOOOOSH*

Pin It on Pinterest

Share This