New Upcoming ATI/AMD GPU's Thread: Leaks, Hopes & Aftermarket GPU's

Discussion in 'Videocards - AMD Radeon' started by OnnA, Jul 9, 2016.

  1. OnnA

    OnnA Ancient Guru

    Messages:
    17,845
    Likes Received:
    6,739
    GPU:
    TiTan RTX Ampere UV
    -> Source in GER -> http://www.pcgameshardware.de/Vega-...-Edition-Tiled-Based-Rasterizer-Test-1231996/

    30.06.2017 at 14:45 The first US-American hardware magazine has received a Radeon Vega Frontier Edition from the trade. In most game benchmarks, the graphics card is still at the level of a GeForce GTX 1080. There is reason to assume, among other things, that the new Tile-Based-Rasterizer with the current drivers still does not work properly.

    We wrote yesterday that the previous benchmarks for the Radeon Vega Frontier Edition still have little meaning. On the one hand, this was due to the testers, who as a private buyer had no experience with the Benchen, and on the other hand, the available, not optimized games. The first part can now be removed from the equation, because with the US-American website pcper.com, the first experienced hardware online magazine received a Vega-10 graphics card. Randnotiz: In Germany, a Radeon Vega Frontier Edition is still not available.

    Most game benchmarks of pcper.com are still only at the level of a Geforce GTX 1080, only in Fallout 4 from the cooperation partner Bethesda can it be with the GTX 1080 Ti. However, it is interesting to test with the Trianglebin tool, which shows the behavior of the raster power amplifiers.

    The Vega-10 GPU on the FE model behaves exactly like Fiji (R9 Fury [X], R9 Nano): Polygons are sequentially scanned and not grouped, as would be the case with a Tile-Based approach. This prevents the so-called culling, in which the non-visible geometry is discarded - this would increase performance and improve efficiency.

    AMD itself indicates that the Draw Stream Binning Rasterizer does not always have to run in Tile Based Deferred Rendering (TBDR), but can be used as a fallback in the previously used Immediate mode - the driver itself can make the decision. So it is possible that the driver decided - for whatever reason - in the test for the Immediate mode or the grouped processing of geometry still does not work at all and has to be activated in a next driver. AMD says only that the Radeon Vega Frontier Edition uses an older driver, but keeps back with details.

    Raja Koduri, head of the Radeon Technologies Group, emphasized in advance that the RX-Vega cards in games should be faster than the Frontier Editions.
     
  2. Noisiv

    Noisiv Ancient Guru

    Messages:
    8,230
    Likes Received:
    1,494
    GPU:
    2070 Super
    As for measurements, we took a pair of calipers to get some rough measurements. These may be off by 0.25-0.5mm. Here’s what we came up with:
    • 30mm x 30mm total size of interposer + GPU (does not include substrate)
    • 20.25mm x ~26mm GPU die size
    • 10mm x ~12mm HBM2 size (x2)
    • ~4mm package height (this is the one that’s least accurate, but gives a pretty good ballpark)
    • 64mm x 64mm mounting hole spacing (square, center-to-center)
    • PCB ~1mm

    www gamers nexus.net/news-pc/2972-amd-vega-frontier-edition-tear-down-die-size-and-more

    DIE SIZE: (527 ± 23) mm²
     
  3. OnnA

    OnnA Ancient Guru

    Messages:
    17,845
    Likes Received:
    6,739
    GPU:
    TiTan RTX Ampere UV
    I like this one:

    And i have only 1 thing to say:
    ATI is innovative for everyone ! not only overpriced Pro but for Us Gamers + Artists (not Pro) too.

    And this is how it rolls, not mature, sometimes rushed and not Ready Yet ;) etc. but Remember we are Wait™ and i like it :p


    We have HBM + Other usefull Teks + Large Openess and Vega will only Add new Tek to the equation.
    No need to be unmature and laugh at ATI, because this is not fair for them !

    As for this:
    "It is due to the mentality in which they see AMD as tool which keeps nVidia's pricing in reach of their wallets.
    Same things is tradition with intel - AMD CPU 'battle'." -> No comments :look:

    Humanity is build on Proper Prosperity and Progress.

    Don't blame ATI for your short wallet ;)
    Wanna Support nV + Intel ? It's your 'choice' but don't blame ATI/AMD for others that has un-reasonable prices.
     
    Last edited: Jul 1, 2017
  4. There are a few things that really confuse me.

    1. AMD was talking about Draw Stream Binning Rasterizer, that it is supposed to cull out pixels, which are not visible, directly in rasterizer. That in fact is NOT Tile Based Rasterizer. It is closer to Tile Based Deferred Rasterization but AMD officials never said it should be Tile Based. So did AMD fanboys once again made something up or is VEGA really supposed to have Tile Based Deferred Rasterizer?

    2. It is obvious that in the specific test VEGA FE used Immediate Rasterization, just with a bigger L2 cache so the result was a bit different from other GCNs. So why did AMD officialy stated that: "The driver has all the gaming optimizations that the other Radeon drivers would include up until at least the driver branching mentioned above." Obviously it doesn't use some Fiji driver as many would like to hope.

    3. Tiled Rasterization won't help you in most compute tasks. There is no reason to assume that VEGA will do 50mh/s even with enabled TBDR if it is currently doing around 30.

    So IMO there are only a few explanations:
    1. AMD lied about drivers being finished and rasterizer works only in legacy mode. Draw Stream Binning will be enabled later or never.
    2. Rasterizer only enables itself in specific cases not visible in the demo -> it works as it is supposed to already.
    3. Draw Stream Binning is not the same as Tile Based Deferred Rasterization and the demo won't show it properly.

    The demo I am talking about:
    [​IMG]
     
    Last edited by a moderator: Jul 1, 2017

  5. OnnA

    OnnA Ancient Guru

    Messages:
    17,845
    Likes Received:
    6,739
    GPU:
    TiTan RTX Ampere UV
    Features:

    - Next-Gen Compute Units
    - Enhanced Pixel Engine
    - Revolutionary Memory Engine
    - Efficient Geometry Engine
    - and many more...

    Specification:

    - GPU Architecture: “Vega”
    - Next-Gen Compute Units (nCUs): 64 (4096 Stream Processors)
    - Memory Configuration: 16GB High Bandwidth Cache (HBC)
    - Memory Bandwidth: 483 GB/s
    - Pixel Fillrate: 90 Gpixels/s
    - Peak FP32 Compute Performance: 13.1 TFLOPS
    - Peak FP16 Compute Performance: 26.2 TFLOPS
    - Display Output Connectors: 3x DisplayPort™ 1.4 HBR3/HDR Ready, 1x HDMI™ 4K60
    - API Support: DirectX® 12.1, OpenGL® 4.5, OpenCL™ 2.0, Vulkan® 1.0
    - Form Factor: Dual-slot, full length (10.5”) configuration
    - Cooling Solution: Air or liquid-cooled versions available

    -> https://en.wikipedia.org/wiki/DirectX
    -> http://www.genrehow.com/difference-between-directx-12-and-12-1/
     
    Last edited: Jul 1, 2017
  6. No offense but I can't take seriously anything, describing itself only as: Next-Gen-Enhanced-Revolutionary-Efficient. I can write these buzzwords in front of anything and say totally nothing.
     
  7. SHS

    SHS Master Guru

    Messages:
    502
    Likes Received:
    47
    GPU:
    Sapphire Vega 56
    I say it best to hold off in tell the consumer ver is release after all the spec will be diff then Vega Frontier Edition.
     
    Last edited: Jul 1, 2017
  8. What if RX VEGA is a different silicon? :) Or did AMD denied that?
    VEGA FE ~ P100
    RX VEGA ~ P102

    What if RX VEGA doesn't have those complex 4:2:1 ALUs, but just 32bit with some terrible ratio to 16 and 64bit just for compatibility reasons like gaming pascal cards. That would explain a lot, but I'm afraid AMD doesn't have manpower to do multiple GPUs like that.
     
  9. millibyte

    millibyte Maha Guru

    Messages:
    1,282
    Likes Received:
    242
    GPU:
    Smoothbore125mm gun
    The initial Fallout 4 graphs showing Vega FE closer to 1080ti were erroneous, as Ryan admitted in the comments of his own article.

    before:
    http://i.imgur.com/ltw6H8t.jpg

    after:
    [​IMG]
     
  10. Titan29

    Titan29 Master Guru

    Messages:
    306
    Likes Received:
    12
    GPU:
    Vega 64 | 7900XTX
    What is surprising is that vega at 14nm with ~527mm2 die contains the same number of cores/tmus/rops as Fiji at 28nm and 596mm2 die.

    Shouldn't vega at this size contain atleast 1.5x more cores/tmus/rops than Fiji? or is vega still 28nm (highly unlikely)? or is it that vega contains 'extra' hardware compared to Fiji?
     

  11. Loophole35

    Loophole35 Guest

    Messages:
    9,797
    Likes Received:
    1,161
    GPU:
    EVGA 1080ti SC
    Some think it is extra hardware to make up for the pipeline deficiencies of GCN.
     
  12. OnnA

    OnnA Ancient Guru

    Messages:
    17,845
    Likes Received:
    6,739
    GPU:
    TiTan RTX Ampere UV
    What's your next GPU will be?

    Im happy with my Fury for Now in 1440p :)
     
  13. Loophole35

    Loophole35 Guest

    Messages:
    9,797
    Likes Received:
    1,161
    GPU:
    EVGA 1080ti SC
    Was waiting on Vega but latest news pushed me to 1080ti. Maybe when Navi comes around I will switch back to AMD. Haven't had a red GPU since my Caymans. Wish I had bought a 290x over my 780 I had a while back. May have not needed to go to the 1070.

    BTW I sold my 1070 for almost what I paid for it a year ago so that's why I'm on the trusty backup.
     
    Last edited: Jul 1, 2017
  14. OnnA

    OnnA Ancient Guru

    Messages:
    17,845
    Likes Received:
    6,739
    GPU:
    TiTan RTX Ampere UV
    AMD Radeon RX Vega at SIGGRAPH

    In a series of tweets, Radeon RX and AMD employees have confirmed that RTG is launching Radeon RX Vega in about one month at SIGGRAPH. The new tweets do not specifically mention CAPSAICIN, but SIGGRAPH conference in general. CAPSAICIN SIGGRAPH is just a part of SIGGRAPH, so maybe the launch is split, or I’m reading in-between the lines too much?

    When asked about Radeon Vega Frontier reviews and how they reflect gaming performance of Vega GPU in general, Jason Evengelho simply said that it’s premature to worry. The same kind of answer came from ‘the famous’ MSI marketing director, who earlier expressed his concerns about power consumption.

    In the worst case scenario, the RX Vega will simply be the Frontier on steroids, where at least the clock speed is stable. On the other hand, there might still be some secrets hidden in the die that we don’t know yet about. AMD has not really shared ANY new information about Vega to the press after Vega Frontier was released.

    -> https:// videocardz|com/70663/amd-confirms-radeon-rx-vega-is-launching-at-siggraph
     
    Last edited: Jul 1, 2017
  15. Paulo Narciso

    Paulo Narciso Guest

    Messages:
    1,226
    Likes Received:
    36
    GPU:
    ASUS Strix GTX 1080 Ti
    I believe that FE must be downclocking badly, just looking at the crappy stock cooler that AMD used which is equal to the same used on 290x.
    If it was bad for a 250w card, it´s obvious that it´s worse for a 300w card.
    Board partners will play an important role delivering proper cooling to achieve a stable 1600 mhz core.
    That and some driver polishing will help to bring the card to the same performance levels of 1080 or a little bit higher.
    But honestly, Nvidia achieved that kind of performance a year ago with better efficiency.
    I only see this card relevant to freesync user, who were waiting for a better card and continue to take advantage of their monitors.
     

  16. OnnA

    OnnA Ancient Guru

    Messages:
    17,845
    Likes Received:
    6,739
    GPU:
    TiTan RTX Ampere UV
    What’s Inside The Magic Sand of VEGA?


    It appears that Vega, being a new architecture, poses significant challenges for AMD’s driver team. There’s very little doubt that AMD’s driver team is still optimizing for Vega and hasn’t rolled out specific game by game optimizations in the drivers just yet.
    AMD will likely not be doing so until closer to RX Vega’s official release. To what extent this will affect Vega’s current gaming performance ? we’ll have to wait until those drivers are out and we have RX Vega in the labs to say for certain. Don’t expect any miracles though.

    Vega 10 features 256 texture mapping units and 64 next generation Vega compute units arranged in two islets, each housing two compute engines. Every compute engine includes two distinct compute clusters. Each of those clusters features 512 stream processors and 32 texture mapping units. The chip in its entirety has a total of 4096 stream processors and 256 texture mapping units.

    On the front-end side of things there are 64 render output units that make up 16 distinct render back-ends that connect to the 2048-bit HBM2 memory interface. The whole Vega 10 die sits on an interposer and is 2.5D stacked with two HBM2 stacks. Every stack can be configured with up to 8 GB of memory for a total of 16 gigabytes as can be found in the Radeon Pro Vega Frontier Edition.

    For a more detailed look at the Vega 10 GPU specs make sure to check out our in-depth Vega 10 spec break down here.

    (So first things first, the core count. Vega 10 features 64 Next generation compute units, each containing 64 GCN stream processors. The entire chip has a total of 4096 next generation GCN stream processors divided into four divisions, each making up a single Shader Engine. Every 1024 sp shader engine has two Asynchronous Compute Units, one render back-end and 4 texture blocks. Each render back-end is comprised of 16 render output units, for a total of 64 ROPs. Each texture block is comprised of 16 texture mapping units, for a total of 256 TMUs. Vega 10 also supports 8 independent work threads simultaneously.)

    Vega Architecture Key Features


    – 4x Power Efficiency
    – 2x Peak Throughput/Performance Per Clock
    – High Bandwidth Cache
    – 2x Bandwidth per pin
    – 8x Capacity Per stack (2nd Generation High Bandwidth Memory HBM_2)
    – 512TB Virtual Address Space
    – Next Generation Compute Engine
    – Next Generation Pixel Engine
    – Next Generation Compute Unit optimized for higher clock speeds
    – Rapid Packed Math
    – Draw Stream Binning Rasterizer
    – Primitive Shaders
     
  17. pharma

    pharma Ancient Guru

    Messages:
    2,481
    Likes Received:
    1,180
    GPU:
    Asus Strix GTX 1080
  18. Ryu5uzaku

    Ryu5uzaku Ancient Guru

    Messages:
    7,547
    Likes Received:
    608
    GPU:
    6800 XT
  19. pharma

    pharma Ancient Guru

    Messages:
    2,481
    Likes Received:
    1,180
    GPU:
    Asus Strix GTX 1080
    It's interesting in the GamerNexus review one game each was chosen by AMD and Nvidia to benchmark.
     
    Last edited: Jul 2, 2017
  20. Loophole35

    Loophole35 Guest

    Messages:
    9,797
    Likes Received:
    1,161
    GPU:
    EVGA 1080ti SC
    I feel he does a really good job there. He has been blasted by the AMD shills a bit but he is very fair with his conclusions.
     

Share This Page