Intel Silvermont(Next-gen OOE Atom)

Discussion in 'PC Industry' started by DSC, May 6, 2013.

Tags:
  1. cal_guy

    Newcomer

    Joined:
    Jun 27, 2008
    Messages:
    216
    Likes Received:
    2
    Jaguar's numbers are really surprising
    AESENC/DEC 3 cycles latency (5 according to Agner Fog) 1 cycle throughput. Double instruction.
    AESIMC/KEYGENASSIT 2 cycles latency 1 cycle throughput. Single instruction
     
  2. mczak

    Veteran

    Joined:
    Oct 24, 2002
    Messages:
    3,012
    Likes Received:
    112
    Yes that's quite nice. And the PCLMULQDQ too has (like ordinary simd muls on jaguar) ultra-low latency and throughput of 1! AESKEYGENASSIST would be like 10 times faster than on haswell according to Agner's data (not sure why intel has published so vastly different figures for that one). But even without considering this one, the encryption instructions on jaguar actually look better than on haswell even, and of course much better than on silvermont.
     
  3. liolio

    liolio Aquoiboniste
    Legend

    Joined:
    Jun 28, 2005
    Messages:
    5,723
    Likes Received:
    193
    Location:
    Stateless
  4. DSC

    DSC
    Banned

    Joined:
    Jul 12, 2003
    Messages:
    689
    Likes Received:
    3
  5. ToTTenTranz

    Legend Veteran Subscriber

    Joined:
    Jul 7, 2008
    Messages:
    9,853
    Likes Received:
    4,463
    GPU performance of their top solution is about the same as the Adreno 320 in the Snapdragon 600.

    Good enough for Android games, useless for Windows games. Unless someone wants to play 10 year-old games in a tablet.

    Speaking of tablets, there are at least two tablets officially announced with Silvermont: a 8" Dell and a 10" Asus.
     
  6. mczak

    Veteran

    Joined:
    Oct 24, 2002
    Messages:
    3,012
    Likes Received:
    112
    As far as I can tell from these benches, Jaguar would be quite a match for this chip in performance if it would support turbo, IPC indeed seems a bit higher on Jaguar (based on the assumption that most likely silvermont is running closer to 2.4Ghz rather than 1.4Ghz in most of the tests). Granted it still wouldn't be a match in perf/w but at least if AMD would have implemented some turbo (say 1.5Ghz up to 2Ghz) it wouldn't look like it drew much more power while still being slower...
     
  7. epicstruggle

    epicstruggle Passenger on Serenity
    Veteran

    Joined:
    Jul 24, 2002
    Messages:
    1,903
    Likes Received:
    45
    Location:
    Object in Space
    Sorry for the late reply. We moved mountains to make it happen.
     
  8. Thorburn

    Regular

    Joined:
    Oct 8, 2006
    Messages:
    323
    Likes Received:
    19
    Location:
    UK
    Been working with Baytrail chips for the most of the year and was actually sufficiently impressed that I went out and bought an ASUS Transformer Book T100 last week - my first ever tablet and the first time I've bought a laptop in about 7 years!

    Spent the weekend and this week doing some testing with it, mostly in games, and have posted up some videos here: http://www.youtube.com/user/BrueComputing

    Graphics performance is somewhere in-line with the 1st-gen Core i3/i5 stuff, which isn't too bad considering this is a 2W SDP part. Done some testing against Temash chips and they do edge it in graphics benchmarks, but are 8W TDP parts.

    In the T100 it's the first Windows tablet I've felt it's worth owning in terms of price/performance/usefulness.
     
  9. swaaye

    swaaye Entirely Suboptimal
    Legend

    Joined:
    Mar 15, 2003
    Messages:
    8,456
    Likes Received:
    578
    Location:
    WI, USA
    Truecrypt results from my Dell Venue 8 Pro with Atom Z3740D. Seemed like some here might be interested in these results. I can't find an OpenSSL build for Windows with AES-NI unfortunately.
    [​IMG]

    edit - I suppose this info is readily available now that I think more about it.... TR has Truecrypt results.
     
  10. monstercameron

    Newcomer

    Joined:
    Jan 9, 2013
    Messages:
    127
    Likes Received:
    101
    the 2W SDP vs 8W TDP is a skewed comparison, the smaller tablet can draw ~11W under load http://www.notebookcheck.net/Review-Asus-Transformer-Book-T100TA-C1-GR-Convertible.106219.0.html

    while the larger notebook with amd temash uses only ~13W under load http://www.notebookcheck.net/Review-Samsung-ATIV-Book-9-Lite-905S3G-K01DE-Subnotebook.106346.0.html

    just a few W more and there are many variables that could factor in on power consumption, however it seems that atom stays closer to idle under normal usage while amd is within the upperbound hence lower idle and usage numbers for temash compared to baytrail.
     
  11. swaaye

    swaaye Entirely Suboptimal
    Legend

    Joined:
    Mar 15, 2003
    Messages:
    8,456
    Likes Received:
    578
    Location:
    WI, USA
    I thought I'd post and say that it looks like OpenSSL for Windows does have AES-NI support.
     
  12. homerdog

    homerdog donator of the year
    Legend Veteran Subscriber

    Joined:
    Jul 25, 2008
    Messages:
    6,134
    Likes Received:
    906
    Location:
    still camping with a mauler
  13. swaaye

    swaaye Entirely Suboptimal
    Legend

    Joined:
    Mar 15, 2003
    Messages:
    8,456
    Likes Received:
    578
    Location:
    WI, USA
    It does support it. It's just that I was confused by some documentation about using it on Linux that didn't work with the Windows build I have. My Venue 8 Pro scores:

    openssl speed -evp aes128
    zoom.
    Code:
    OpenSSL 1.0.1e 11 Feb 2013
    built on: Mon Feb 11 16:33:34 2013
    options:bn(64,32) rc4(8x,mmx) des(idx,cisc,2,long) aes(partial) idea(int) blowfish(idx) 
    compiler: cl -D_USE_32BIT_TIME_T  /MD /Ox /O2 /Ob2 -DOPENSSL_THREADS  -DDSO_WIN32 -D_USE_32BIT_TIME_T -W3 -Gs0 -GF -Gy -nologo -DOPENSSL_SYSNAME_WIN32 -DWIN32_LEAN_AND_MEAN -DL_ENDIAN -D_CRT_SECURE_NO_DEPRECATE -DOPENSSL_BN_ASM_PART_WORDS -DOPENSSL_IA32_SSE2 -DOPENSSL_BN_ASM_MONT -DOPENSSL_BN_ASM_GF2m -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DMD5_ASM -DRMD160_ASM -DAES_ASM -DVPAES_ASM -DWHIRLPOOL_ASM -DGHASH_ASM -DOPENSSL_USE_APPLINK -I. -DOPENSSL_NO_RC5 -DOPENSSL_NO_MD2 -DOPENSSL_NO_KRB5 -DOPENSSL_NO_JPAKE -DOPENSSL_NO_STATIC_ENGINE    
    The 'numbers' are in 1000s of bytes per second processed.
    type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
    aes-128-cbc     164791.16k   249379.82k   294641.11k   311612.47k   317188.81k
    
    openssl speed aes-128-cbc
    not using AESNI
    Code:
    OpenSSL 1.0.1e 11 Feb 2013
    built on: Mon Feb 11 16:33:34 2013
    options:bn(64,32) rc4(8x,mmx) des(idx,cisc,2,long) aes(partial) idea(int) blowfish(idx) 
    compiler: cl -D_USE_32BIT_TIME_T  /MD /Ox /O2 /Ob2 -DOPENSSL_THREADS  -DDSO_WIN32 -D_USE_32BIT_TIME_T -W3 -Gs0 -GF -Gy -nologo -DOPENSSL_SYSNAME_WIN32 -DWIN32_LEAN_AND_MEAN -DL_ENDIAN -D_CRT_SECURE_NO_DEPRECATE -DOPENSSL_BN_ASM_PART_WORDS -DOPENSSL_IA32_SSE2 -DOPENSSL_BN_ASM_MONT -DOPENSSL_BN_ASM_GF2m -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DMD5_ASM -DRMD160_ASM -DAES_ASM -DVPAES_ASM -DWHIRLPOOL_ASM -DGHASH_ASM -DOPENSSL_USE_APPLINK -I. -DOPENSSL_NO_RC5 -DOPENSSL_NO_MD2 -DOPENSSL_NO_KRB5 -DOPENSSL_NO_JPAKE -DOPENSSL_NO_STATIC_ENGINE    
    The 'numbers' are in 1000s of bytes per second processed.
    type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
    aes-128 cbc      32189.81k    36132.41k    37676.07k    72772.13k    74997.63k
    
     
  14. homerdog

    homerdog donator of the year
    Legend Veteran Subscriber

    Joined:
    Jul 25, 2008
    Messages:
    6,134
    Likes Received:
    906
    Location:
    still camping with a mauler
    Huh, guess I misread your post. Weird.
     
  15. Thorburn

    Regular

    Joined:
    Oct 8, 2006
    Messages:
    323
    Likes Received:
    19
    Location:
    UK
    Had a play with a Venue 11 Pro today and compared it to the T100 in 3DMark to see how the Z3740 stacks up vs. Z3770.



    The Z3740 actually comes out 11% faster in graphics performance, Z3770 29% faster in the CPU Physics test.

    I'd wager that it is down to how the power usage is balanced on the SoC - the upper power and GPU clock limits are the same for both chips but the higher base clock on the Z3770 is presumably leaving less power headroom for the GPU to utilise...
     
  16. Albuquerque

    Albuquerque Red-headed step child
    Veteran

    Joined:
    Jun 17, 2004
    Messages:
    3,845
    Likes Received:
    329
    Location:
    35.1415,-90.056
    Wow, great test! I'd rather have the 11% higher GPU, and I'm sure a lot of other people probably feel the same. Thanks for posting the results.
     
  17. swaaye

    swaaye Entirely Suboptimal
    Legend

    Joined:
    Mar 15, 2003
    Messages:
    8,456
    Likes Received:
    578
    Location:
    WI, USA
    The Venue 8 Pro BIOS has a stunning load of settings if you access it through the Win8 advanced boot, UEFI settings menu. Lots of IGP settings are available including a graphics boost option that is disabled by default.

    I haven't bothered with benchmarks but have played some games. I have found it to run some quite well. Guild Wars and Defense Grid for example. SupCom2 is too slow though. It has interesting performance characteristics. Low bandwidth and fillrate but the ALU capability isn't bad for what it is. I find it totally amazing for a 2W SOC.
     
  18. Silent_Buddha

    Legend

    Joined:
    Mar 13, 2007
    Messages:
    16,072
    Likes Received:
    5,023
    Interesting. I was actually considering the Dell Venue 11 Pro, but from pretty much everything I've read the Synaptics digitizer pen solution is significant worse than the Wacom pen digitizer. /sigh.

    Too bad this doesn't come with a Wacom digitizer.

    Regards,
    SB
     
  19. mczak

    Veteran

    Joined:
    Oct 24, 2002
    Messages:
    3,012
    Likes Received:
    112
    I always consider it like a ~5W chip or so (just get away with that "scenario design power" please...).
    I dunno though I actually think it's the opposite, the graphics has high bandwidth for that class but low ALU capacity (the bandwidth exceeds that of half of AMDs discrete mobile gpu lineup (which says a lot about that lineup...) whereas those 4 EUs aren't a whole lot. A Haswell GT3 (without edram) part will have 1.5 times the bandwidth but 10 times the EUs...
    I thought that intel improved the power distribution scheme some time ago to give more preference to the gpu part a while ago (Ivy Bridge?), but it's interesting that it seems the nominally slower part can still be faster due to that. Though if you compare different systems, you need to be really careful, since just like their big brothers these chips all have configurable TDP (plus configurable turbo).
     
  20. Thorburn

    Regular

    Joined:
    Oct 8, 2006
    Messages:
    323
    Likes Received:
    19
    Location:
    UK
    Took a look on the Venue 11 Pro this morning, no IGP settings that I can see other than Screen Brightness.

    There is an option to "Enable Turboboost\x99" which is set to Enabled by default - "Allows the Intel Turbo driver to increase the performance of the CPU OR graphics processor."
     
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...