Results 1 to 15 of 15
  1. #1

    Arrow AMD 7xxx informations ; XDR2 !

    - HD 7990 New Zealand GCN
    - HD 7970 Tahiti XT GCN 1000MHz 32CUs 2048ALUs 128TMUs 64ROPs 256bit XDR2 8.0Gbps 256GB/s 2GB 190W HP
    - HD 7950 Tahiti Pro GCN 900MHz 30CUs 1920ALUs 120TMUs 64ROPs 256bit XDR2 7.2Gbps 230GB/s 2GB 150W HP
    - HD 7870 Thames XT VLIW4 950MHz 24SIMDs 1536ALUs 96TMUs 32ROPs 256bit GDDR5 5.8Gbps 186GB/s 2GB 120W HPL
    - HD 7850 Thames Pro VLIW4 850MHz 22SIMDs 1408ALUs 88TMUs 32ROPs 256bit GDDR5 5.2Gbps 166GB/s 2GB 90W HPL
    - HD 7670 Lombok XT VLIW4 900MHz 12SIMDs 768ALUs 48TMUs 16ROPs 128bit GDDR5 5.0Gbps 80GB/s 1GB 60W HPL
    - HD 7570 Lombok Pro VLIW4 750MHz 12SIMDs 768ALUs 48TMUs 16ROPs 128bit GDDR5 4.0Gbps 64GB/s 1GB 50W HPL

    转[AMD下代28nm显卡规格全露 HD7990旗舰] - GPU/游戏专区 - 超能网玩家俱乐部 - 全新视角全新观点 - Powered by Discuz!

    The question is ; why XDR2 for GCN and not the new GDDR5 at 7ghz (samsung) ?

  2. #2
    Quote Originally Posted by supernewbie View Post
    The question is ; why XDR2 for GCN and not the new GDDR5 at 7ghz (samsung) ?
    May the availability will be better. Additional I guess that the efficiency of GDDR5-Memory won't be that great at higher clock speeds - compared to the capabilities of XDR2.

  3. #3
    XDR™2 memory is the latest generation of the award-winning XDR memory architecture. Capable of unprecedented data rates of up to 12.8Gbps, XDR2 is the world's fastest memory solution. It delivers twice the peak bandwidth per device when compared to GDDR5 DRAM.

    The XDR2 memory solution provides incredible bandwidth performance and does so at far lower power than GDDR5 memory. In fact, XDR2 consumes 30% less power than GDDR5 at equal bandwidth performance. Given equal power budget, XDR2 memory provides 50% more bandwidth than GDDR5. And at full speed, XDR2 DRAM deliver twice the bandwidth of GDDR5 devices.

    XDR2 memory achieves this power and performance leadership through key Rambus innovations, many of which were developed through Rambus' Terabyte Bandwidth Initiative. The first to employ a Fully Differential Memory Architecture (FDMA), XDR2 memory has improved signal integrity and lower voltage swing versus single-ended signaling architectures such as GDDR5. Other innovations include Flexlink™ C/A, for more capacity scalability and flexible access granularity, and 16X Data Rate which allows the XDR2 memory to run at twice the data rate of GDDR5 with a clock running at about half the speed of the GDDR5 clock.

    The Fully Differential Memory Architecture of XDR2 memory enables rock solid signal integrity at blindingly fast speeds. Differential signaling in combination with pin-efficient Flexlink C/A technology, enables XDR2 memory to deliver a breakthrough level of performance with fewer controller and DRAM pins than single-ended GDDR5 memory. Any way you look at it, speed, power, bandwidth, robustness or pin-efficiency, XDR2 memory is the best way to supercharge your next-generation graphics, gaming or multi-core computing design.

    As of 2008 article.

  4. #4
    Name: AMD Radeon HD 7990
    Compute powern single/double precision, TFLOPS: 12.16/3.04
    Technology process: 28nm
    Steam processors: 6400
    TMU: 256
    Texture fill-rate, GTextel/s: 243.2
    ROP: 96
    Pixel fill-rate, GPixel/s: 91.2
    Memory bus width: 384
    Memory type: 6GB GDDR5+
    Memory bandwidth, GB/s: 576
    TDP, full load, W: 300
    TDP, idle, W: 30

    Will 7990 hit this?

  5. #5
    I thought the problem with XDR was that it had incredibly high latencies.

    Issue resolved or does the sheer size make up for it?

  6. #6

  7. #7
    Fake info. Nobody is mass producing it XDR2 memories in any volume. AMD will definitely be sticking to 256bit DDR5 but at a slightly higher clock speed. DDR5 will go up to 7GB/s and the memory controller will be simpler. Also Thames is a mobile chip.

    If AMD does release their high end part at the end of the year, I would expect it to be either
    a) 5-10% faster than the GTX580 and priced just under it or
    b) ~30% faster than the GTX580 and same or just slightly higher price.

  8. #8
    KencanaNET FTW!! wenasong's Avatar
    Join Date
    Jan 2010
    Location
    *Near Resort World Sentosa*
    Posts
    7,071
    Jimmi, I think I can agree with you. No point lowering the price and based on numbers, I can foresee option b) to be the most likely result.
    Bzzzzzzzzzzzzt Blitz.


    Cert : #1212NA195 , #1404TCE008 , #1404WE064 , #1405RE146 , #1405INE010 , #1604UME060

  9. #9
    Nope, the price will be at high side.
    Most probably a hd7970 will more exp than gtx580. Maybe $799-999.
    It is the research break even point, u cannot bse on the 6 generation which used the same architecture becuz 5 series already earned back their research cost.

  10. #10
    It actually very dependant to the power of the card is drawing.
    The more powet draw means is faster.
    If base on te rumours -34% energy means the performance is only on par or sligthly higher than gtx 580

  11. #11
    KencanaNET FTW!! wenasong's Avatar
    Join Date
    Jan 2010
    Location
    *Near Resort World Sentosa*
    Posts
    7,071
    Not necessarily that the performance will be impacted for that much based on energy consumption. Just give you an example of the more cores the better consuming the same amount of energy at some application. 2x Opteron 6166HE, 1.8Ghz getting scores of about 11.8 on Cinebench. The 2 processors combined has a TDP (a figure closely related to energy consumption) vs say a Phenom II x6 which has the same TDP figure (I think scores much much less than half). We are not even talking about difference in manufacturing (both processors are 45nm). Usually the smaller it goes, the power consumption will be reduced by providing the same performance.
    Bzzzzzzzzzzzzt Blitz.


    Cert : #1212NA195 , #1404TCE008 , #1404WE064 , #1405RE146 , #1405INE010 , #1604UME060

  12. #12
    Quote Originally Posted by waikong86 View Post
    It actually very dependant to the power of the card is drawing.
    The more powet draw means is faster.
    If base on te rumours -34% energy means the performance is only on par or sligthly higher than gtx 580
    Makes sense. Similar power draw but in a smaller process will indicate bigger performance increase unless they use the transistor budget for tessellation performance.

  13. #13
    I wonder what the implications regarding the Trinity platform will be.

    Afaik, Trinity apus will feature GCN graphics. Will that mean that the system will have to be equipped with XDR2 as well, or will they use dual controllers? Or just go with fast DDR3 and be done with it?

  14. #14
    Quote Originally Posted by wenasong View Post
    Not necessarily that the performance will be impacted for that much based on energy consumption. Just give you an example of the more cores the better consuming the same amount of energy at some application. 2x Opteron 6166HE, 1.8Ghz getting scores of about 11.8 on Cinebench. The 2 processors combined has a TDP (a figure closely related to energy consumption) vs say a Phenom II x6 which has the same TDP figure (I think scores much much less than half). We are not even talking about difference in manufacturing (both processors are 45nm). Usually the smaller it goes, the power consumption will be reduced by providing the same performance.
    bro, if u read carefully what i am trying to imply is the same as what u say.

  15. #15
    Quote Originally Posted by wenasong View Post
    Not necessarily that the performance will be impacted for that much based on energy consumption. Just give you an example of the more cores the better consuming the same amount of energy at some application. 2x Opteron 6166HE, 1.8Ghz getting scores of about 11.8 on Cinebench. The 2 processors combined has a TDP (a figure closely related to energy consumption) vs say a Phenom II x6 which has the same TDP figure (I think scores much much less than half). We are not even talking about difference in manufacturing (both processors are 45nm). Usually the smaller it goes, the power consumption will be reduced by providing the same performance.
    there are many factors to reduce power consumption not only the lithography process size.
    but i do believe that the gtx580 or hd 6970 are already the best they can do with 40nm by nvidia and amd respectively. So if u take the best they can do for 28nm the power they can save is around 30-40% for same performance. (assumption)

    U can do comparison with gtx 480 vs gtx 285 difference in 40nm and 55nm process. around 30% boost.
    But later they increase the power saving by boosting better performance for 40nm process in gtx 580. However it only 10-15% increament they can go for same lithography size.

Similar Threads

  1. AMD Southern Islands (28nm ?) informations in catalyst 11.4
    By supernewbie in forum News around the web!
    Replies: 0
    Last Post: Mar 10th, 11, 07:18 AM
  2. Replies: 0
    Last Post: Nov 22nd, 10, 08:45 PM
  3. AMD - New informations about K10.5 (releasedate) and 32nm
    By N1truX in forum News around the web!
    Replies: 6
    Last Post: Sep 1st, 08, 08:52 PM
  4. Replies: 7
    Last Post: Jun 1st, 08, 03:33 PM
  5. The Future of XDR2 from RamBus
    By proyb2 in forum Hardware Depot
    Replies: 2
    Last Post: Jul 9th, 05, 12:20 PM