AMD Launching First GPU With HBM Memory At E3, On June 16

At a press conference about Carrizo here at Computex, Lisa Su from AMD introduced the release date for the company's new graphics cards with high bandwidth memory (HBM). AMD will launch it on June 16 at E3, as many of us have speculated.

The HBM memory implementation is a new method of making a memory subsystem for GPUs, and it succeeds GDDR5. We wrote about it here. Rather than placing a handful of GDDR5 memory ICs around the GPU, HBM stacks memory, and uses an interposer to bring the memory much closer to the GPU, placing it on a substrate on the same logic die. The footprint for the memory system is therefore 95 percent smaller, allowing manufacturers to make much smaller graphics cards with much greater power.

For now, that's all that AMD has officially revealed for the new GPUs. Want to know more? Sit tight until June 16th. We'll be there.

Follow Niels Broekhuijsen @NBroekhuijsen. Follow us @tomshardware, on Facebook and on Google+.

Niels Broekhuijsen

Niels Broekhuijsen is a Contributing Writer for Tom's Hardware US. He reviews cases, water cooling and pc builds.

  • jimmysmitty
    That's great. They are announcing something we have known for weeks. I want to see numbers for performance and price. Supposedly there might be two Fury (Fiji XT) based parts, a top of the line one and a lower end one. Current rumored price is $850 but that's a bit much for my blood. As well it is supposed to be only 4GB of HBM until V2 which will have 8GB.

    To top it off, the other rumors are that the R300 series are all rebrands, the R9 390X being a 8GB GDDR5 variant of Hawaii XT with a bit higher clock (about 50MHz).

    So unless Fury X (if that is the name) comes at a good price and offers great performance I might have to wait for a GPU upgrade when I also upgrade my CPU, RAM and such.

    I just hate the wait.
    Reply
  • Larry Litmanen
    Regular people do not care about technology being staked or unstacked, what people want to know is this, if i will go out there now and buy a $200 nVidia card and it gives me 40 FPS on Witcher 3 on max, if i were to buy an AMD card in the same price range would it give me 70 FPS.

    People just want results, not what kind of technology you use, AMD can give us a 1 nano meter chip with hyper threading and turboboost, if it's not outperforming an Intel chip that is priced in the same range it is all meaningless.
    Reply
  • humorific
    I thought this was an Nvidia article. The CEOs look like twins.
    Reply
  • thor220
    That's great. They are announcing something we have known for weeks. I want to see numbers for performance and price. Supposedly there might be two Fury (Fiji XT) based parts, a top of the line one and a lower end one. Current rumored price is $850 but that's a bit much for my blood. As well it is supposed to be only 4GB of HBM until V2 which will have 8GB.

    To top it off, the other rumors are that the R300 series are all rebrands, the R9 390X being a 8GB GDDR5 variant of Hawaii XT with a bit higher clock (about 50MHz).

    So unless Fury X (if that is the name) comes at a good price and offers great performance I might have to wait for a GPU upgrade when I also upgrade my CPU, RAM and such.

    I just hate the wait.

    Let's put it this way - Even if AMD were to only have included HBM (which they didn't) the die space savings alone would give them leeway to increase power by about 30% and reduce power consumption. I'm expecting somewhere between a 50 -80 percent increase over the 290x. Price that bad boy at $600 and you've got a Titan and Ti killer.
    Reply
  • thor220
    Regular people do not care about technology being staked or unstacked, what people want to know is this, if i will go out there now and buy a $200 nVidia card and it gives me 40 FPS on Witcher 3 on max, if i were to buy an AMD card in the same price range would it give me 70 FPS.

    People just want results, not what kind of technology you use, AMD can give us a 1 nano meter chip with hyper threading and turboboost, if it's not outperforming an Intel chip that is priced in the same range it is all meaningless.

    AMD owns the $200 GPU segment right now so yeah, you can get a better card from AMD right now.
    Reply
  • rdc85
    ...People just want results, not what kind of technology you use,....
    I hope u means generally mass..

    because well this is tech site..
    more detailed about tech and such is always welcomed ...
    It's kinda funny u comment that on here...
    Reply
  • crisan_tiberiu
    Regular people do not care about technology being staked or unstacked, what people want to know is this, if i will go out there now and buy a $200 nVidia card and it gives me 40 FPS on Witcher 3 on max, if i were to buy an AMD card in the same price range would it give me 70 FPS.

    People just want results, not what kind of technology you use, AMD can give us a 1 nano meter chip with hyper threading and turboboost, if it's not outperforming an Intel chip that is priced in the same range it is all meaningless.

    AMD owns the $200 GPU segment right now so yeah, you can get a better card from AMD right now.
    Now he is going to tell you that "AMD is hot, Nvidia is not". Like the 980ti having 250W TDP and reaching 85Celsius is OK ^^ But when AMD does that it is the end of the world.
    Reply
  • Kepe
    Regular people do not care about technology being staked or unstacked, what people want to know is this, if i will go out there now and buy a $200 nVidia card and it gives me 40 FPS on Witcher 3 on max, if i were to buy an AMD card in the same price range would it give me 70 FPS.

    People just want results, not what kind of technology you use, AMD can give us a 1 nano meter chip with hyper threading and turboboost, if it's not outperforming an Intel chip that is priced in the same range it is all meaningless.

    Yeah.. This is a site dedicated to computer technology. People who come here want to know the technical details. Tech details of a graphics card define what kind of results you can expect when it hits the market. What are you even doing here if this kind of stuff doesn't interest you? If you want the results, wait until the card is released and read the reviews. No one is forcing you to read this stuff.
    Reply
  • Gaidax
    To be honest, I'm not terribly excited for AMD, seems like they have some major stumbling block there. Making a humongous die size says only one thing to me - they took whatever they had, packed more transistors on it added some frills to help pull it higher and slapped liquid cooling on it and called it a day, simply because they don't really have any actual new architecture yet.

    And even worse yet, all the GPUs they will have to offer to the masses will be rebrands of what they have now and even more ridiculous - rebrands of rebrands. These are like 2-3 year old tech being reused already.

    Meanwhile, Nvidia's Maxwell is in fact a new architecture that is not being pushed to the limit (at least not at the stretch AMD is doing here) and successor is being prepped to that already and their mid-range consumer products also sport that architecture.

    I'm not even going to mention the 980Ti's aggressive pricing, I'm pretty sure AMD can't afford to match this with their latest and greatest, seeing how expensive to make it is.
    Reply
  • hannibal
    Price that bad boy at $600 and you've got a Titan and Ti killer.

    The HBM-memory is completely new technology. It will be expensive. It does not even depend on how fast Fury will, be. The first versions of HBM is going to be expensive. The prediction is a little bit slover than 980ti and priced between 980ti and TitanX. But that is just a rumour, but quite possible. I expect that Fury will be a beast in 4K and other situations where you need a lot of memory bandwidth, but at smaller resolutions like 1080p and 1440p the Nvidia may have the edge. The Fury is a concept of proof for HBM, more like limited edition prototype. It can be really fast in some situations, but very fast memory does not make GPU any faster in most common situations.

    ... simply because they don't really have any actual new architecture yet.

    And even worse yet, all the GPUs they will have to offer to the masses will be rebrands of what they have now and even more ridiculous - rebrands of rebrands. These are like 2-3 year old tech being reused already.

    Some things to consider: GCN is not bad architecture, AMD does not have too much money for development at this moment (that is a fact), HBM is new technology.
    I think that AMD is trying out HBM in niche product, so they can test how it does work, also they have been waiting for smaller production node for ages (same as Nvidia).
    So there is not any sensible reason for AMD to sell Fury cheaply, nor to make whole new series of GPUs. If they would have a lot of money, they could do that, but now they have to think very carefully, where to put their money. 2016 is the year of new production node (14nm or 16nm), it can also be the year of new memory type HBM in larger scale (depending on how expensive it will be to produce HBM memory and how useful it will be in graphic cards and APUs and how much there will HBM memory available in the market, when also Nvidia is moving toward that direction)
    So AMD is saving money for 14nm upgrade and new HBM based products. They have to. If they would make whole new products now, they would not have money for upgrades next year, when really big things are happening. It is pity, but AMD is hampering in GPU part, because CPU part has not been very lucrative lately.
    We will see much better balanced competition in the next year, because AMD will move to 14nm production node, but still AMD need a little bit of luck, so that they wil get decent amounts of working GPUs and CPUs and APUs from production line. Even Intel did have some hiccups when starting 14nm production, ans they have all the money that development department can hope for...
    Reply