Page 11 of 21 FirstFirst ... 910111213 ... LastLast
Results 101 to 110 of 205

Thread: Piledriver-based Vishera, Next AMD FX

  1. #101
    640k who needs more?
    Join Date
    Mar 2010
    Posts
    995
    There is a chance DH misunderstood some of the NDA documents.It would be nice to see 15+ "IPC" improvement but so far AMD only stated "10% higher x86 performance" for FX Next (which is Vishera).

    That being said, Charlie mentions 10+% higher IPC than Bulldozer version1 in his latest Trinity article and we know that Trinity has no L3 cache and has memory controller which is a bit different than the one in Vishera. So there is possibility (at least) to see >10% IPC versus original FX. AMD needs 10% IPC and 10% clock uplift to make FX a chip it was meant to be in the first place: a worthy successor to K10 which is close to its IPC(~90% instead of now closer to 80%) while clocking 30+% higher and supporting higher thread count.

    Bulldozer in FX8150 variant is on average just ~9% faster than 1100T,stock vs stock. This is average from a mix of desktop workloads,some of which are mostly serial while others are MTed. If Vishera is clocking 10% higher while having around 10% higher "IPC" in most workloads,then FX Next will be around 30-35% faster than 1100T. This is really an uplift which can justify the "next gen. design" name that Bulldozer carries.

  2. #102
    Senior Member
    Join Date
    Nov 2011
    Posts
    2,593
    Quote Originally Posted by wuttz View Post
    just to make it clear this time, its not amd overpromising now and "underdelivering" later..
    Well, AMD fired all the marketing people responsible for that, so I don't know if we will see a repeat.

    Wasn't desktop Piledriver supposed to be different than the Piledriver in Trinity (beyond just L3 and memory controller) or is it just me?

  3. #103
    Yes, Vishera has a revised Piledriver core, model 2h. Not sure what the differences are exactly. Some of them are listed in that optimization guide, but let's wait and see.

    The days of AMD being way too late and too slow than promised are gone... AMD are executing very well nowadays.

  4. #104
    Quote Originally Posted by sdlvx View Post
    Wasn't desktop Piledriver supposed to be different than the Piledriver in Trinity (beyond just L3 and memory controller) or is it just me?
    Only the Piledriver in Kommodo, which has been canceled. Vishera will probably be the same as Trinity, because some AMD exec. said after the last Financial A. Day, that Vishera 2013 will be different from Vishera 2012.

  5. #105
    Quote Originally Posted by inf64 View Post
    There is a chance DH misunderstood some of the NDA documents.It would be nice to see 15+ "IPC" improvement but so far AMD only stated "10% higher x86 performance" for FX Next (which is Vishera).

    That being said, Charlie mentions 10+% higher IPC than Bulldozer version1 in his latest Trinity article and we know that Trinity has no L3 cache and has memory controller which is a bit different than the one in Vishera. So there is possibility (at least) to see >10% IPC versus original FX. AMD needs 10% IPC and 10% clock uplift to make FX a chip it was meant to be in the first place: a worthy successor to K10 which is close to its IPC(~90% instead of now closer to 80%) while clocking 30+% higher and supporting higher thread count.

    Bulldozer in FX8150 variant is on average just ~9% faster than 1100T,stock vs stock. This is average from a mix of desktop workloads,some of which are mostly serial while others are MTed. If Vishera is clocking 10% higher while having around 10% higher "IPC" in most workloads,then FX Next will be around 30-35% faster than 1100T. This is really an uplift which can justify the "next gen. design" name that Bulldozer carries.
    10% more per core performance, not just 10% more x86 performance. That is what the official slide says, so I'm expecting 110%*8 more x86 performance here.

  6. #106
    ???? Each core is 10% faster = chip is 10% faster.
    OK, if scaling gets worsened, chip is faster by less than that amount, if scaling improves, it gets faster by more. But I dont expect huge gains or losses here.

  7. #107
    640k who needs more?
    Join Date
    Mar 2010
    Posts
    995
    Zizy is correct. 10% better x86 performance automatically means that each core is 10% faster than each BD core. The whole chip is 10% faster at the same clock,not 8x110% or anything similar to that nonsense

  8. #108
    It was early, I wasn't thinking straight :P

  9. #109
    640k who needs more?
    Join Date
    Aug 2010
    Posts
    783
    Quote Originally Posted by inf64 View Post
    There is a chance DH misunderstood some of the NDA documents.It would be nice to see 15+ "IPC" improvement but so far AMD only stated "10% higher x86 performance" for FX Next (which is Vishera).

    That being said, Charlie mentions 10+% higher IPC than Bulldozer version1 in his latest Trinity article and we know that Trinity has no L3 cache and has memory controller which is a bit different than the one in Vishera. So there is possibility (at least) to see >10% IPC versus original FX. AMD needs 10% IPC and 10% clock uplift to make FX a chip it was meant to be in the first place: a worthy successor to K10 which is close to its IPC(~90% instead of now closer to 80%) while clocking 30+% higher and supporting higher thread count.

    Bulldozer in FX8150 variant is on average just ~9% faster than 1100T,stock vs stock. This is average from a mix of desktop workloads,some of which are mostly serial while others are MTed. If Vishera is clocking 10% higher while having around 10% higher "IPC" in most workloads,then FX Next will be around 30-35% faster than 1100T. This is really an uplift which can justify the "next gen. design" name that Bulldozer carries.
    so 20% vs current bd? that would be a leap forward.

    I don't buy the marketing though:

    * 4 extra cores on the 8xxx parts are a feature, not a given.

    * 4xxx and 6xxx are crippled parts and should be treated/priced as such.

    * overclocking is also a feature not a given - not everyone wants to overclock or is willing to spend extra $$$ for fans, power supplies and such. stop the bluffing and give me the stock benchmarks, thank you very much.

    * matching mb's needs to be addressed - nforce 630a a 6 year old chipset is listed as supporting FX on at least one website???

    from all the posts i've read on various boards, AMD has almost become a non-factor in desktop and that needs to get fixed asap.

    What I'm saying is AMD should price their desktop parts a lot cheaper. 8xxx Vishera should be $149, IMO.

  10. #110
    10% to 15% higher CPU performance
    15% to 20% higher NB/HT performance(*)
    16% to 28% lower CPU/Mobo power consumption(*)(**)

    (*)-Assumption
    (**)-Only if RCM only touches the cores for the CPU portion/If Hypertransport 8GT/s is also used

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  
WordPress Appliance - Powered by TurnKey Linux