Comments Locked

47 Comments

Back to Article

  • HollyDOL - Tuesday, March 8, 2016 - link

    Hopefully it's going to be interesting thing to watch.
    Alas, calling the event capsaicin seems quite unfortunate... Especially if you take in consideration the hilarious "second round" after consuming bigger doses of mentioned essence.
  • xthetenth - Tuesday, March 8, 2016 - link

    If it's anything like last generation, their cards have way too much staying power for that comparison.
  • ImSpartacus - Tuesday, March 8, 2016 - link

    Yeah, the naming hilariously bad, but I suppose they are trying...

    Nvidia is just so damn chill in comparison. Jen-Hsun can't beat.
  • atlantico - Wednesday, March 9, 2016 - link

    Polaris is pretty damn cool.
  • Alexvrb - Wednesday, March 9, 2016 - link

    He's so chill he swags around with faked Pascal prototypes that are actually GM204, and nobody bats an eye. Oh well at least this time there were no wood screws.
  • Spunjji - Wednesday, March 9, 2016 - link

    I remember the wood screws! Ah, good times.
  • Flunk - Tuesday, March 8, 2016 - link

    I hope they announce when Polaris is coming out, that's really the only thing people want to know right now.
  • bikal adhikari - Tuesday, March 8, 2016 - link

    Fury x2 maybe?
  • chris200x9 - Tuesday, March 8, 2016 - link

    Nobody cares about Fury x2, it's too late. Polaris is coming soon and when in cross fire it will beat the Fury x2 in price, performance, and power usage. I don't see how they could make a dual GPU solution at 28nm competitive with a dual GPU solution at 14 nm regardless if they are on the same card or not.
  • zoxo - Tuesday, March 8, 2016 - link

    The big gamechanger 14/16nm chips won't be out until 2017 from either AMD or NVIDIA. So while we hopefully get something that's around/slightly above Fury performance (at lower power) this summer/before the holidays, the real heavy hitters, and certainly the dual graphics cards are later down the line.
    The larger of the 2 Polaris gpus they have now is less than half the size of the Fury chip, which means that it will have comparable number of transistors/SPs. I certainly don't expect them to come out with a chip that has a comparable in size to the Fury very soon (which was at the very limit of what TMSC could do), there is certainly room for improvement as the process matures.
  • chris200x9 - Tuesday, March 8, 2016 - link

    Even with flagships launching in 2017 you'll get maybe a year at most longevity out of a $1000+ GPU? Also if we look at benchmarks and assume a 30% increase with polaris a new 390 like GPU will beat Fiji. The only way Fiji x2 is competitive is if AMD gimps the Polaris roll out, which would be extremely stupid. Fiji x2 is DOA.
  • zoxo - Tuesday, March 8, 2016 - link

    Fiji X2 is for VR and D12, not for "regular" gaming, and I think it would work relatively well there. 1 eye per GPU is such a natural fit, that it is probably a lot more useful and lot less problematic as AFR crossfire/SLI.
    From the "middle" Polaris chip (the bigger of the 2), I'm expecting a 390/390X level of performance, within a much lower power envelope. Nothing will challenge the big boys from either camp this year.
  • CiccioB - Wednesday, March 9, 2016 - link

    If the "middle" Polaris has the same size than Hawaii, it will pack double the transistor and will consume less. I wonder how can you say that such a GPU could perform as Hawaii.
    It is logical to think it will perform much near Fiji. And probably even better (seen how bad Fiji scaled with such many shaders which is a think they have probabily fixed with the new architecture). If not, it means that the new architecture will be a completely fail.
    Or the "middle" Polaris won't be that big.
  • Kevin G - Tuesday, March 8, 2016 - link

    Fury was big but it wasn't the biggest chip around. GM200 is 601 mm^2, just a little bit bigger than the Fiji die found in the Fury cards. The largest chip manufactured to my knowledge is the Tukwilia Itanium at a massive 699 mm^2 that dwarfs even the Fury chip. It was manufactured by Intel so there is no indication that what the real limit for TSMC is. Same for IBM's current POWER8 which weighs in at 650 mm^2.
  • extide - Tuesday, March 8, 2016 - link

    Well, it was fairly widely reported that Fiji and GM200 pretty much were right at the Reticle limit for TSMC. We don't know exactly how big they can go, but since both companies chips came in at almost the exact same size, they were probably targeting the max.

    What is pretty amazing, is Intel has been pumping out 650+mm^2 dies on 14nm for quite some time (Knights Landing). TSMC may say they are "beating Intel" in the fab wars, but they really aren't... :)
  • testbug00 - Tuesday, March 8, 2016 - link

    The limit for TSMC is around 600mm^2 due to the tools I believe.

    IBM designs of the past and what Intel do are a bit different. Vertically integrated manufacturing can do some amazing things.
  • eek2121 - Tuesday, March 8, 2016 - link

    That's not what I am hearing.
  • Frenetic Pony - Tuesday, March 8, 2016 - link

    Meh, even if the "big" chips don't come out till next year the price drop for cards from the new node will be worth the wait. $200 less (or more) for a smaller, quieter card that, in AMD's case, has a more future proof AV connection is worth a few months.
  • testbug00 - Tuesday, March 8, 2016 - link

    Game changes for who? For a minuscule portion of the market?
  • CiccioB - Wednesday, March 9, 2016 - link

    If real, the small portion can rapidly grow and change the game.
  • mdriftmeyer - Wednesday, March 9, 2016 - link

    They're coming out in June. Keep dreaming they don't arrive until Jan 2017.
  • SunLord - Tuesday, March 8, 2016 - link

    The first Polaris cards will replace the 370/380 price level and then move up power wise we likely won't see the Fury replacement for 7 or 8 months if not longer so this will probably be the hard launch for the X2 and maybe talk about polaris
  • zoxo - Tuesday, March 8, 2016 - link

    Yeah, I do hope they will reveal the uArch of polaris, and the small/mid polaris chips are already very interesting for notebooks tbh.
  • ImSpartacus - Tuesday, March 8, 2016 - link

    Fury x2 (or Gemini) won't compete with Polaris in traditional sense.

    Amd was clear that they delayed it because vr was delayed. It's attached to vr. I think it'll mostly be used in oem machines like the tiki. In that area, it's really not that bad of a solution. It's definitely not a game changer, but it might do alright for itself.
  • JamJams - Tuesday, March 8, 2016 - link

    Or maybe their Polaris will be there..
  • zoxo - Tuesday, March 8, 2016 - link

    They did demo small polaris before, so that's a possibility.
  • gijames1225 - Tuesday, March 8, 2016 - link

    Hopefully we get Polaris dates and details. Something on Zen would be great as well, but probably not happening.
  • Spunjji - Wednesday, March 9, 2016 - link

    This is RTG, so Zen news is very unlikely.
  • zepi - Tuesday, March 8, 2016 - link

    There are exactly two things in the AMD release pipeline that matter. New CPU microarchitecture (Zen) and 14nm / 16nm GPU's.

    Everything else is just haphazard panic CPR their marketing department comes up all the time to keep the company afloat until they have something worth selling in their hands.
  • SunLord - Tuesday, March 8, 2016 - link

    My money is on a hard launch of the Fury X2 since VR headsets are starting to ship/happen and maybe information on Polaris replacement for the R9 370/380
  • just4U - Tuesday, March 8, 2016 - link

    Their going to unveil their final incarnation of the 380/280/7900 series card... (chuckle) just kidding and to be fair, the 380/380x trades blows with the 970 even though it competes with the 960 which in my opinion it's better than. So it's sort of come a long way.. but they appear to have squeezed all the juice out of that orange that they were able to.
  • xthetenth - Tuesday, March 8, 2016 - link

    380 and 380X are the 285 chip not the 280/280X chip. And with how good it's looking recently it's a real solid card.
  • Alexvrb - Thursday, March 10, 2016 - link

    No doubt, a few seconds with a search engine would reveal that the 380 series is Tonga-based. They can be had at good prices and with modern drivers are all-around good performers. Still, with a newer design and a modern process node, Polaris will be leaps better. Won't be out till around mid-summer at the earliest, though.
  • Danvelopment - Tuesday, March 8, 2016 - link

    Interesting codename.

    All I hear is buy the low end model and it will be just perfect, get the high end model and it will be too hot to use, which is never good for a graphics card.
  • Mr Perfect - Wednesday, March 9, 2016 - link

    To hot to use? Who's spreading such FUD? I haven't even heard trolls try that one out.
  • Danvelopment - Wednesday, March 9, 2016 - link

    Nice reading comprehension. Its the implication of the codename. The most powerful chillies are too hot to eat, whereas the lower end ones are just right.
  • Mr Perfect - Thursday, March 10, 2016 - link

    I didn't realize you where making a joke about peppers, it looked like two separate thoughts. First paragraph was about the code name, but then a line return and whole new paragraph about graphics card being to hot to use.

    Thanks for being understanding though.
  • bill.rookard - Tuesday, March 8, 2016 - link

    I'd love to see them demo a 1Xnm FinFet APU with HBM engineering sample. :D

    Wouldn't that be just sick?
  • Valantar - Tuesday, March 8, 2016 - link

    No, that would be completely awful. Especially if it's a Zen based APU. Teasing amazing products that far down the pipeline is just cruel.
  • boeush - Tuesday, March 8, 2016 - link

    Looks like AMD is about to bring da PAIN!

    Here's hoping it's not in the form of shooting themselves in the foot (again). [Seriously, is there anything left of their feet at this point?]
  • BrokenCrayons - Wednesday, March 9, 2016 - link

    AMD is misguided as usual with their focus on small volume halo products that support relatively obscure technologies that are essentially passing fads like virtual reality. There's just too much expense behind support VR and too little inventive for developers who make programs that use DX12 to invest in it. AMD needs to move forward by pushing for OEM deals in laptops where there are still fairly high sales rates. Focusing their efforts on the handful of gaming desktops is foolish. Yes, there's a little bit of growth in desktop gaming, but those systems are a tiny shred of a declining desktop market. Build better GPUs for laptops and skip the VR until it's practical for VR to work on integrated GPUs where there's only an additional expense of the headset for people who want to deal with that kind of entertainment instead of the headset and multiple GPUs that aren't even going to get next generation game support.
  • Murloc - Wednesday, March 9, 2016 - link

    waiting for VR to work on integrated GPUs is excessive since people have been using discrete GPUs for normal gaming for years without complaining.

    I think AMD gets involved in all sorts of stuff because they just can't beat the competitors through sheer performance and lower power consumption.
  • BrokenCrayons - Wednesday, March 9, 2016 - link

    The number of people using discrete GPUs for gaming is a state of decline. Integrated graphics that were "good enough" in their time have existed for quite a while thanks to Intel's efforts with the GMA and HD series. Due in no small part to the shift from desktop to more mobile platforms, discrete GPUs are ever becoming more of a rarity. At the same time, game sales have increased because integrated graphics cards bundled in inexpensive computers now have the ability to run many games at acceptable (acceptable to average consumers) settings that give them reasonable performance. If VR is ever going to become popular, it needs to have a very low entry cost. In order for developers to see value in supporting VR, there has to be an incentive tied to additional profit otherwise the publisher will mandate the developer skip VR support (the same holds true for multi-GPU support in DirectX 12). So yes, widespread and socially-acceptable VR that has the support of software and hardware must be usable with very low cost computers in order to more than a few ripples in the proverbial computing pond.
  • medi03 - Wednesday, March 9, 2016 - link

    AMD has cheaper GPUs that beat all competitor card, except the 1k$ one.

    And on power consumption front, it's 10-15% more performance for with roughly 20% more power.
  • richardginn - Wednesday, March 9, 2016 - link

    How much Capsaicin are we talking about here???

    Bell Pepper level?
    Jalapeno level?
    Thai Chili level?
    Habanero Level?
    Ghost Chili Level?
    Carolina Reaper level??
  • Danvelopment - Wednesday, March 9, 2016 - link

    These would all make great codenames for each family. Carolina Reaper would be a quad chip GPU and would release at the height of Winter.
  • Ammaross - Monday, March 14, 2016 - link

    Trinidad Scorpion: only available for GPU compute in supercomputers. Melts if not liquid cooled in a LiNO2 bath.

Log in

Don't have an account? Sign up now