Page 93 of 115 FirstFirst ... 43839192939495103 ... LastLast
Results 921 to 930 of 1143

Thread: RYZEN (including Pinnacle Ridge/Peak Ridge? )

  1. #921
    Senior Member
    Join Date
    May 2014
    Posts
    1,462
    For some reasons ($$$?) Computerbase still insist with an outdated version of a bench designed by AT s Ian Cutress,
    one has to wonder why they dont do the same with Cinebench and use R10 instead of R15...

    Computerbase "test", whose impact in their average is huge :

    https://www.computerbase.de/2017-08/...ment-benchmark


    And AT s updated software, Ian Cutress still use the old version but he relegated it to his legacy software page tests :




    Also, an other discrepancy, at Hardware.fr this time with about 0% scaling for Ryzen from 12 to 16C in 3DS Max :




    http://www.hardware.fr/articles/967-...ray-v-ray.html
    Last edited by Abwx; 11th August 2017 at 07:18 AM. Reason: typo

  2. #922
    Senior Member
    Join Date
    Aug 2017
    Location
    GA/GB
    Posts
    158
    Quote Originally Posted by Bluefoot View Post
    IPC? What do you mean? Single threaded? Ok. Who's buying these processors for those workloads and running nothing else in the background? Virtually nobody. 'IPC' across multiple threads - it's blatantly obvious the AMD offerings scale much better.
    Intel's 7900X trades blows with Threadripper in a few benches. IPC does matter, even more so in multi-core situations as you don't typically have the clock frequency of lower-core-count chips.


    Quote Originally Posted by Relayer View Post
    Who is going to buy these CPU's a GTX 1080ti and game at 1080? Or run single thread workloads? Intels IPC advantage is worthless in the market these CPU's compete in. Anyone who points at that is really grasping at straws.
    No one will. And whoever does point that out, is simply being ridiculous. You see it quite a lot in the comments section of WCCF, for example. But, I've also seen it across forums as well.
    CPU Grade [WiP]

  3. #923
    Senior Member
    Join Date
    Nov 2011
    Posts
    2,632
    Quote Originally Posted by Relayer View Post
    Who is going to buy these CPU's a GTX 1080ti and game at 1080? Or run single thread workloads? Intels IPC advantage is worthless in the market these CPU's compete in. Anyone who points at that is really grasping at straws.

    Edit: And let's look up price? https://www.newegg.com/Product/Produ...82E16819117637
    You have way too much faith in the average computer builder.
    Within 27% margin of error and 3mm off my estimate of something 4mm, good enough to predict the future

  4. #924
    Senior Member
    Join Date
    May 2012
    Posts
    541
    Quote Originally Posted by DylRicho View Post
    Intel's 7900X trades blows with Threadripper in a few benches.

    IPC does matter, even more so in multi-core situations as you don't typically have the clock frequency of lower-core-count chips.
    Barely, in a small minority, with the 1920X. Calling it competitive with the 1950X is laughable. The 7900X is at massively higher clocks both stock and OC achievable than Intel higher core count chips will be. It already goes miles beyond socket spec even at 10 cores ... see what happens when you add up to 8 more cores ...

    Um, right ... higher core count Intel chips are going to be severely limited on clocks by thermals and power draw, even with exotic cooling, and their / the mobo's life expectancy is rather open to question, too. The AMD chips are much cooler, more energy efficient, and look like they will slightly exceed the clocks of the lower binned, lower core count Ryzen chips.

    Then there are the cost and price factors. You can bet that the 1950X probably costs AMD less than a quarter of what it costs Intel to make a 7900X, once yields are accounted for ... the costs are only going to get worse for the bigger monolithic dies.

    Ordinarily you would think Intel, the monopoly (soon to be former monopoly), would be in a position to price AMD out of the market. But they can't win a price war with huge monolithic designs against 'glued' chiplets. The same situation may loom for NVIDIA if NAVI isn't hugely delayed.
    Last edited by Bluefoot; 11th August 2017 at 09:08 AM.

  5. #925
    Senior Member
    Join Date
    Feb 2010
    Location
    Planet Earth, Solar System, Milky Way
    Posts
    3,533
    Quote Originally Posted by Bluefoot View Post
    Barely, in a small minority, with the 1920X. Calling it competitive with the 1950X is laughable. The 7900X is at massively higher clocks both stock and OC achievable than Intel higher core count chips will be. It already goes miles beyond socket spec even at 10 cores ... see what happens when you add up to 8 more cores ...

    Um, right ... higher core count Intel chips are going to be severely limited on clocks by thermals and power draw, even with exotic cooling, and their / the mobo's life expectancy is rather open to question, too. The AMD chips are much cooler, more energy efficient, and look like they will slightly exceed the clocks of the lower binned, lower core count Ryzen chips.

    Then there are the cost and price factors. You can bet that the 1950X probably costs AMD less than a quarter of what it costs Intel to make a 7900X, once yields are accounted for ... the costs are only going to get worse for the bigger monolithic dies.

    Ordinarily you would think Intel, the monopoly (soon to be former monopoly), would be in a position to price AMD out of the market. But they can't win a price war with huge monolithic designs against 'glued' chiplets. The same situation may loom for NVIDIA if NAVI isn't hugely delayed.
    Ah, and the AMD Ryzen processors are not equipped with a special chewing gum under the lid...

  6. #926
    Senior Member
    Join Date
    May 2012
    Posts
    541
    Quote Originally Posted by GeorgiD View Post
    Ah, and the AMD Ryzen processors are not equipped with a special chewing gum under the lid...
    Has it been confirmed whether the 18 core 7980XFE is soldered or toothpasted?

  7. #927
    Senior Member
    Join Date
    Jan 2010
    Location
    London
    Posts
    3,255
    Quote Originally Posted by Bluefoot View Post
    Has it been confirmed whether the 18 core 7980XFE is soldered or toothpasted?
    All of them are toothpastes as far as I remember

  8. #928
    Junior Member
    Join Date
    Aug 2017
    Posts
    8
    It will be hilarious to see what happens when people try to overclock the 7980XE or any other chip above the 7900x.

    If the toothpaste doesn't burst into flames first, then you can count on a few blown motherboard VRMs under stress testing considering how far the 10C 7900X pushes things already:







    (P95 27.1 uses AVX1/Sandy stuff.... 28.3 uses AVX2+FMA3, 29.2 uses AVX512, I can't imagine how worse it could get under those conditions, it's already throttling in every way possible in that video)

    Maybe v2 motherboards with actual heatsinks on the VRM instead of RGB heat traps could do better.

    ----------------------------------------------------------------------------------

    Has anyone tested TR with decent 3200-3600 memory + tight timings so far? I'd like to see how far it scales when fed properly.

  9. #929
    Junior Member
    Join Date
    Aug 2017
    Posts
    10
    Quote Originally Posted by muziqaz View Post
    All of them are toothpastes as far as I remember
    It's a mistery, to me at least, how is it possible to repurpose xeons as hedt, apply some different binning rules on that new product line-up, but somehow bother to differentiate the packaging.
    Would it be less complicated to just pick some xeons from their packaging and throw them in a new verification line?

    Ok, it's not like making a new arch and going back costs several billions and another 2-3 years, but this is just a paste problem. Send it back and repaste it.
    It seems to me that the toothpaste decision was made on purpose.

  10. #930
    Senior Member
    Join Date
    Feb 2010
    Location
    Planet Earth, Solar System, Milky Way
    Posts
    3,533
    Quote Originally Posted by Yorgos View Post
    It's a mistery, to me at least, how is it possible to repurpose xeons as hedt, apply some different binning rules on that new product line-up, but somehow bother to differentiate the packaging.
    Would it be less complicated to just pick some xeons from their packaging and throw them in a new verification line?

    Ok, it's not like making a new arch and going back costs several billions and another 2-3 years, but this is just a paste problem. Send it back and repaste it.
    It seems to me that the toothpaste decision was made on purpose.
    Yes, the purpose is to bleed more money..

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