1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Intel Atom C3955 – 16 Core Denverton Benchmarks Leaked

Discussion in 'STH Main Site Posts' started by Patrick Kennedy, Feb 16, 2017 at 9:32 AM.

  1. Geran

    Geran Member

    Joined:
    Oct 25, 2016
    Messages:
    84
    Likes Received:
    4
    I wonder what TDP this chip will be at compared to the Xeon D-1500 series equivalent.

    I am trying to hold out as long as possible for my pfSense build for these chips to get released but the C2000 bug put a damper on things.
     
    #2
  2. Evan

    Evan Active Member

    Joined:
    Jan 6, 2016
    Messages:
    474
    Likes Received:
    56
    Sounds like TDP not super less than Xeon-D at least in compatible CPU power (although AES-NI on c3000 may be stronger)
    They are both 14nm and have 10g etc.

    Thing is maybe the 2 or 4 core boxes will make awesome pfSense boxes due to low power and good performance.

    I still love the idea we can use ddr4 rdimm which is a big money saver.

    SM have an N4200 box that has 2 x intel nic and good AES-NI performance , i.e. Over 1G per core
     
    #3
  3. mstone

    mstone Active Member

    Joined:
    Mar 11, 2015
    Messages:
    252
    Likes Received:
    57
    The goldmont aes-ni is still a lesser implementation than broadwell, it's just much better than silvermont. I don't have benchmarks handy, but I'd guess goldmont won't do better than half the number of bytes per clock cycle--which is still enough to make possible things which were out of reach for silvermont.

    Can you cite a source for this? That's significantly higher than I'd expect given its clock rate, and even higher than I'd expect it to hit (but not sustain) in turbo. Unless the benchmark @Patrick published wasn't for a 2.4ghz chip?
     
    #4
    Geran likes this.
  4. Evan

    Evan Active Member

    Joined:
    Jan 6, 2016
    Messages:
    474
    Likes Received:
    56
    #5
  5. Evan

    Evan Active Member

    Joined:
    Jan 6, 2016
    Messages:
    474
    Likes Received:
    56
    C3958 (16 core) 31w TDP
    C3850 (12 core) 25w TDP
    C3758 (8 core) 25w TDP
    C3558 (4 core) 16w TDP
    C3338 (2 core) 9w TDP

    Can't config ghz or the sata or network port setup, high end chips do support 4 x 10g it looks like, and a heap of sata.
     
    #6
  6. mstone

    mstone Active Member

    Joined:
    Mar 11, 2015
    Messages:
    252
    Likes Received:
    57
    Ok, yeah, geekbench results. I've found them to be pretty bogus, no idea why. If you look at results for the same CPU they very widely, and are often significantly higher for AES in particular than I've seen more consistently using other benchmarks. There are also discrepancies between geekbench 3 & 4 results, but I haven't figured out if they're within the huge error bars of results from a single version. I'm also not sure whether the two versions are even running the same "AES" benchmark, because it's so vague about what it's running. (What's the key size? Block size? Cipher mode? Is there an authentication scheme?) I would not trust geekbench at all for estimating crypto performance.
     
    #7
  7. MiniKnight

    MiniKnight Well-Known Member

    Joined:
    Mar 30, 2012
    Messages:
    1,852
    Likes Received:
    454
    Where'd that come from?
     
    #8
  8. Geran

    Geran Member

    Joined:
    Oct 25, 2016
    Messages:
    84
    Likes Received:
    4
    That C3850 looks like a potential winner
     
    #9
  9. Evan

    Evan Active Member

    Joined:
    Jan 6, 2016
    Messages:
    474
    Likes Received:
    56
    Fishing around info resellers have on supermicro a2sdi motherboards. Think a2sdi-4c-hln4f, -ln4f, -tp8f, -h-tf (substitute 4c for different and different combo's interfaces etc)
    From 4 to 12 sata3 and network from 4x1g right up to 2x10g copper + 2x10g sfp + 4x1g copper.
    All seem mini-itx size, some use full size ddr4 ecc rdimm some ddr4 ecc sodimm
    All seem to have 4x pcie slot, some m.2
     
    #10
  10. Evan

    Evan Active Member

    Joined:
    Jan 6, 2016
    Messages:
    474
    Likes Received:
    56
    I guess there is also a c3858 , the 0 being general purpose and the 8 being network comms style chip like in c2000 series.
     
    #11
  11. Evan

    Evan Active Member

    Joined:
    Jan 6, 2016
    Messages:
    474
    Likes Received:
    56
    @mstone i also don't read much into geekbench results either but I did presume that the same test on different cpu showing big differences in performance may have some basis in fact. Anyway for sure the new generations have better crypto performance, and the full cores better than atom.
     
    #12
  12. Geran

    Geran Member

    Joined:
    Oct 25, 2016
    Messages:
    84
    Likes Received:
    4
    I think the whole line up is going to be interesting...can't wait for the new solutions to start rolling out.
     
    #13
Similar Threads: Intel Atom
Forum Title Date
STH Main Site Posts New Official Intel Atom C3000 “Denverton” Details Yesterday at 6:32 AM
STH Main Site Posts Intel Atom C3000 Denverton – First benchmarks and what we can expect when it is finally... Feb 14, 2017
STH Main Site Posts The Intel Atom C2000 Series Bug – Why it is so quiet Feb 7, 2017
STH Main Site Posts Intel Apollo Lake Launched – Intel Atom E3900 Series Oct 25, 2016
STH Main Site Posts New Intel Xeon D Networking Chips Are Coming Yesterday at 6:32 AM

Share This Page