R600 has 512-bit external memory bus, Orton promises it takes no prisoners

Discussion in 'Beyond3D News' started by Geo, Dec 15, 2006.

  1. Geo

    Geo Mostly Harmless
    Legend

    Joined:
    Apr 22, 2002
    Messages:
    9,116
    Likes Received:
    213
    Location:
    Uffda-land
    Speaking at yesterday's AMD Financial Analysts Day, Executive Vice President of Visual and Media Businesses, Dave Orton, appeared to throw down the performance gauntlet in favor of AMD's upcoming R600 GPU. Having had over a month to study NVIDIA's G80, Orton did not seem the least bit intimidated. In a slide entitled "R600: Why we lead in graphics", Orton promised that even if the name of the company had changed, that the commitment to GPU performance leadership had not. He promised a "take no prisoners" approach to performance leadership for AMD's new GPU.

    More interestingly, in his verbal remarks Orton reported (at roughly the 1:22:30 mark of the <a href="http://www.amd.com/gb-uk/Corporate/InvestorRelations/0,,51_306_14668,00.html" target=_b3dout>webcast</a>) that one of R600's key advantages would be "new levels of memory bandwidth to the graphics subsystem, and bandwidth is critical to graphics performance." As all graphics geeks know, AMD pioneered the move to GDDR4 memory with the Radeon X1950 XTX, which gave them a temporary advantage in bandwidth. However, in the period since NVIDIA has released the 384-bit GeForce 8800 GTX, whose memory bandwidth crushes the X1950 XTX by 86.4GB/s to 64.0GB/s. It is impossible that AMD could regain a significant enough advantage in bandwidth to be cited by Orton as a major competitive advantage without following NVIDIA north of the 256-bit bus that has been a mainstay of the ATI/AMD high-end products since 2002's Radeon 9700 Pro.

    R600 will therefore feature a 512-bit external memory bus, likely using 1.2GHz GDDR4 for ~153GB/sec from memory pool to chip, to back up the smack AMD's Executive Vice President laid down.

    In other tidbits, Orton also vowed to be first to the 65nm technology process, but did not disclose which product he had in mind for the honor, nor even product type, GPU or chipset. Our graphics-oriented notes (and a few selected slides) on the rest of the conference are included inside, if you dare to take the red pill.
     
    digitalwanderer likes this.
  2. Geo

    Geo Mostly Harmless
    Legend

    Joined:
    Apr 22, 2002
    Messages:
    9,116
    Likes Received:
    213
    Location:
    Uffda-land
    AMD Analyst Day 2006, Dec 14, 2006

    Usual caveat --What geo found interesting in my own idiosyncratic way as best I heard it. No warranty express or implied to be a complete and accurate record.

    Dirk Meyer, COO
    • "Maniacal" focus on Moore's Law single threaded is over. Then multi-core homogenous cores. Now integration of heterogenous cores.
    • Ex-ATI was 4,000 employees, 2,500 of which were engineers.
    • Will continue to be focused on creating world-class stand alone CPU and GPU.
    • Procurement cost synergies showing up. Ultimately new manufacturing strategy.
    • Both partnering with others and doing own integrated going forward.
    • "AMD" will be "master brand", but ATI will still be visible with GPUs.
    • Support structure has changed to integrated as well.
    • CPU, GPU, Consumer is the business groupings. Chipsets straddle CPU and GPU. Own organizations for R&D, marketing, etc.
    • Small group will ensure that leverage and synergy across all three happens.
    • Sales forces integration is complete world-wide.
    • Single supply chain organization as well.
    • Leadership goals:
      • Engergy efficient computing. Ultimate Visual experience. Affordable internet access for 50% of world population.
    Henri Richard, Sales & Marketing
    • Working now to convert brand customers to portfolio customers.
    • Never had the opportunity to create a new catagory and drive it. "Accelerated Processing Units" creates that opportunity.
    • Slide: "Exclusive: Dynamic Video Decoding" Hmm, what's that? Is that supposed to be a proprietary name?
    Dave Orton, Visual and Media Businesses
    • UVD (Universal Video Decode) aimed at BRD/HD DVD. "Critical" for value platforms.
    • "About half" of the 2,500 engineers of former ATI are software engineers. On the consumer side its a little higher.
    • Vista driver work with Microsoft began in 2003.
    • "Take no prisoners" performance approach key at the top. Slide title "R600: Why we lead in graphics". Key advantage is second generation of Unified shaders.
    • "New levels of memory bandwidth to the graphics subystem, and bandwidth is critical to graphics performance."
    • Buzz phrase alert! "Dynamic Game Computing" will provide new levels of visual realism.
    • Process technology leadership will continue. "We will be the first in the market to bring high performance 65nm as well"
    • R600 technology will be driven down market.
    • "There have been questions about what's our commitment to the GPU market as a company?". They consider the gaming market and technologies to be critical so that high-end tech can be leveraged down market.
    • Vista can grow GPU market, including business market. (yeah, yeah, heard it all before --time to prove it)
    • Expects the AMD sales team to help significantly in selling workstation GPUs in enterprise market.
    • Unified software stack with AMD will help in business/notebook markets.
    • GPGPU opportunity just beginning.
    • Fusion in 2009: Mainstream (both notebook and desktop), Portables, Emerging markets.
    • 3D technology "waterfalls" from high-end to low-end, but must be reengineered on the way to a degree to get die size and power reqs down. Former ATI has good experience with this from the Consumer business.
    • Multimedia: UVD in integrated and low-end provides value.
    • H.264 coming to AMD cell phones.
    • Fusion helps make new convergence segments possible. Fundamental for performance, cost and power.
    • GPUs are the core of the business (presumably of his division), and "the opportunity to put that business back on track in the gross margins area".
    • Shared vision of convergence of PC and consumer area in 2009+ time-frame was the real driver of the AMD/ATI merger.
    Phil Hester, CTO
    • The industry "made a mistake" getting into the megahertz war. What really mattered was system performance.
    • Doesn't [Fusion] mean that discrete GPU goes away? Not at all --designed to provide good base capability. Never do high-end enthusiast-class graphics that way.
    • Software transition is "non-disruptive", contrary to popular conception.
    Slides

     
    _xxx_ likes this.
  3. Skrying

    Skrying S K R Y I N G
    Veteran

    Joined:
    Jul 8, 2005
    Messages:
    4,815
    Likes Received:
    61
    Very interesting, I guess this semi-confirms 512-bit. Interesting indeed.
     
  4. Cypher

    Newcomer

    Joined:
    Jun 28, 2005
    Messages:
    85
    Likes Received:
    1
    Yeah, what he said. I'm tempted to say that this kind of smacktalk is almost a confirmation of this. At the very least, certainly above 384-bit (although, any takers on 448-bit?)
     
  5. Bouncing Zabaglione Bros.

    Legend

    Joined:
    Jun 24, 2003
    Messages:
    6,363
    Likes Received:
    82
    It's worth noting that for the DX9 inflection point, ATI jumped to the 256 bit bus even though Nvidia said it was not needed and commentators predicted it was too expensive and difficult to produce. Possibly for the DX10 inflection point, AMD/ATI are doing the same for 512 bit in order to invest their flagship DX10 product with the memory technology they intend to use for this generation of product.
     
  6. Sound_Card

    Regular

    Joined:
    Nov 24, 2006
    Messages:
    936
    Likes Received:
    4
    Location:
    San Antonio, TX
    Very good news Geo!!!! I loving it already. Do you mind if I post this in the ***The official R600 thread*** @ R3D? :wink:
     
  7. Geo

    Geo Mostly Harmless
    Legend

    Joined:
    Apr 22, 2002
    Messages:
    9,116
    Likes Received:
    213
    Location:
    Uffda-land
    Well, when we put it on the front page it means we're willing to tell the world, so sure. :)
     
  8. AlBran

    AlBran Just Monika
    Moderator Legend

    Joined:
    Feb 29, 2004
    Messages:
    20,002
    Likes Received:
    4,931
    Location:
    ಠ_ಠ
    < Joe > BRIIiing It On!

    When he says "second gen unified shaders" does he mean Xenos is 1st gen and R600 is second?
     
  9. Dave Baumann

    Dave Baumann Gamerscore Wh...
    Moderator Legend

    Joined:
    Jan 29, 2002
    Messages:
    14,079
    Likes Received:
    647
    Location:
    O Canada!
    Yes.
     
    Acert93 likes this.
  10. pax

    pax
    Newcomer

    Joined:
    Mar 9, 2002
    Messages:
    64
    Likes Received:
    5
    Probly just stock picts in that R600 slide but the ati chip wasnt at an angle there...
     
  11. Sobek

    Sobek Locally Operating
    Veteran

    Joined:
    Dec 17, 2004
    Messages:
    1,774
    Likes Received:
    18
    Location:
    QLD, Australia
    Second generation unified shaders? What's the difference...I would have thought unified shaders would be a fairly straight-forward thing that shouldn't require any modifications. Or would it just be some small changes in the way it handles data (yes, I know nothing on unified shaders).

    See I love this news, this is what gives me that good old feeling down inside. Either that or my lunch has taken a turn for the worse..then again, i'm not on the Nvidia website so that can't be it :twisted:
     
  12. Mark

    Mark aka Ratchet
    Regular

    Joined:
    Apr 12, 2002
    Messages:
    604
    Likes Received:
    33
    Location:
    Newfoundland, Canada
    Classic!

    (sort of)
     
  13. Rangers

    Legend

    Joined:
    Aug 4, 2006
    Messages:
    12,295
    Likes Received:
    1,107
    Yep. Kind of hints at some sort of improvement as well. Otherwise they'd just be re-used first gen.

    Guess all the 512-bit doubters and "ATI to exit graphics" crowd have been proved wrong.
     
  14. Sobek

    Sobek Locally Operating
    Veteran

    Joined:
    Dec 17, 2004
    Messages:
    1,774
    Likes Received:
    18
    Location:
    QLD, Australia
    Right right...but what is the difference between First/Second generation Unified shaders? :?:
     
  15. Anarchist4000

    Veteran Regular

    Joined:
    May 8, 2004
    Messages:
    1,437
    Likes Received:
    359
    1 Generation?

    But the typical improvements from one generation to the next I'd imagine. Increased precision in the ALUs, improved scheduling/caching, maybe different ALU:TEX ratios. If Hexus's comments on 2GHz are to be believed I'd say different clock domains as well.

    The biggest bet will likely be scheduling and caching seeing as how Nvidia went the ultra-efficient scalar approach.
     
  16. ZoinKs!

    Regular

    Joined:
    Nov 23, 2002
    Messages:
    782
    Likes Received:
    13
    Location:
    Waiting for Oblivion
    Because I'm in a contrarian mood

    How about using an eDRAM scheme similar to Xenos?
     
  17. ZoinKs!

    Regular

    Joined:
    Nov 23, 2002
    Messages:
    782
    Likes Received:
    13
    Location:
    Waiting for Oblivion
    It's not Xenos but they are kinfolk.

    One area that may change is dynamic allocation of the shader alu's. Xenos uses 48 alu's divided into 3 groups. So you have a few options: all 48 running pixel shaders, all 48 running vertex shaders, or 16 running one type and the other 32 running the other. (See here for more detail.)

    Perhaps r600 will allow more then 3 banks of alu's. Perhaps it'll even allocate the precise number needed at the moment. Former seems more likely then the later, imho, because I doubt you really need to allocate each individual alu and choosing how to allocate them doesn't sound like an easy task. :razz:


    edit: doh! just realized Anarchist said pretty much the same thing, just with a lot less words. :lol:
     
    #17 ZoinKs!, Dec 15, 2006
    Last edited by a moderator: Dec 15, 2006
  18. satein

    Regular

    Joined:
    Aug 17, 2005
    Messages:
    483
    Likes Received:
    21
    Location:
    Sheffield, UK.
    Sound interesting!! But the leaked die shot of the R600 has no sight of additional eDRAM. Even the die size itself is almost full the chip top area and the die was also placed in an incline angle (about 45 degree).

    I think you have got another interesting point and you can also heat up discussion on the R600 rumour theard here...
    The ATI R600 Rumours & Speculation Centrum

    Edit: typo
     
  19. Geo

    Geo Mostly Harmless
    Legend

    Joined:
    Apr 22, 2002
    Messages:
    9,116
    Likes Received:
    213
    Location:
    Uffda-land
    Huddy ruled that out some time back. I'd have to give him the "David Kirk LLPOF* Award" if they did that.







    *Liar, Liar Pants on Fire.
     
  20. Rys

    Rys AMD RTG
    Moderator Veteran Alpha

    Joined:
    Oct 9, 2003
    Messages:
    4,138
    Likes Received:
    1,337
    Location:
    Beyond3D HQ
    I've just edited the original news item to put our flag in the ground and confirm 512-bit, since we're certain it's definitely so.
     
Loading...

Share This Page

  • About Us

    Beyond3D has been around for over a decade and prides itself on being the best place on the web for in-depth, technically-driven discussion and analysis of 3D graphics hardware. If you love pixels and transistors, you've come to the right place!

    Beyond3D is proudly published by GPU Tools Ltd.
Loading...