Differences

This shows you the differences between two versions of the page.

Link to this comparison view

HPC:Village [2017/09/21 20:22]
solar [Hardware] updated Wikipedia link for Radeon HD 7900 series
HPC:Village [2020/12/10 12:55] (current)
mator [Related third-party projects/offers] GRID5000
Line 11: Line 11:
   * [[http://​www.supermicro.com/​products/​system/​4u/​7047/​sys-7047gr-tprf.cfm|Supermicro GPU SuperWorkstation 7047GR-TPRF]] workstation/​server platform with MCP-290-00059-0B rackmount rail set   * [[http://​www.supermicro.com/​products/​system/​4u/​7047/​sys-7047gr-tprf.cfm|Supermicro GPU SuperWorkstation 7047GR-TPRF]] workstation/​server platform with MCP-290-00059-0B rackmount rail set
     * 4U chassis     * 4U chassis
-    * Two 1620W PSUs ((The system'​s AC power consumption at idle is around 360W. At full load on all components, it increases to almost 1600W. These are totals for the two PSUs, which are normally ​sharing the load.))+    * Two 1620W PSUs (normally both are active and are sharing the load)
     * Dual socket 2011 motherboard with IPMI, 16 memory sockets, four PCIe 3.0 x16 slots for full-length dual-width PCIe cards and a fifth slot for a shorter card     * Dual socket 2011 motherboard with IPMI, 16 memory sockets, four PCIe 3.0 x16 slots for full-length dual-width PCIe cards and a fifth slot for a shorter card
     * A full set of cooling fans, including those pulling hot air out of passively-cooled accelerator cards     * A full set of cooling fans, including those pulling hot air out of passively-cooled accelerator cards
Line 25: Line 25:
     * 60 cores (x86-ish with 512-bit SIMD units) seen as 240 logical CPUs (four hardware threads per core), 1053 MHz, 8 GB GDDR5 ECC RAM on a 512-bit bus, 320 GB/s     * 60 cores (x86-ish with 512-bit SIMD units) seen as 240 logical CPUs (four hardware threads per core), 1053 MHz, 8 GB GDDR5 ECC RAM on a 512-bit bus, 320 GB/s
     * Peak performance of about 2 TFLOPS single-precision,​ 1 TFLOPS double-precision     * Peak performance of about 2 TFLOPS single-precision,​ 1 TFLOPS double-precision
 +  * [[wp>​GeForce 1000 series|NVIDIA GTX 1080]] gaming graphics card (short form factor, manufactured by Gigabyte)
 +    * NVIDIA Pascal architecture
 +    * One GP104 GPU with 2560 SPs typically at 1607 MHz to 1771 MHz, 8 GB GDDR5X RAM on a 256-bit bus, 320 GB/s
 +    * Peak performance of over 8 TFLOPS single-precision
   * [[wp>​GeForce 900 series|NVIDIA GTX Titan X]] gaming graphics card (reference design, manufactured by Gigabyte)   * [[wp>​GeForce 900 series|NVIDIA GTX Titan X]] gaming graphics card (reference design, manufactured by Gigabyte)
     * NVIDIA Maxwell architecture     * NVIDIA Maxwell architecture
-    * One GM200 GPU with 3072 SPs at 1000 MHz to 1076 MHz, 12 GB GDDR5 RAM on a 384-bit bus, 336 GB/s +    * One GM200 GPU with 3072 SPs at 1000 MHz to 1177 MHz, 12 GB GDDR5 RAM on a 384-bit bus, 336 GB/s 
-    * Peak performance of over 6 TFLOPS single-precision, 0.2 TFLOPS double-precision+    * Peak performance of over 6 TFLOPS single-precision
   * [[wp>​GeForce 700 Series|NVIDIA GTX TITAN]] gaming graphics card (Zotac GeForce GTX TITAN AMP! Edition)   * [[wp>​GeForce 700 Series|NVIDIA GTX TITAN]] gaming graphics card (Zotac GeForce GTX TITAN AMP! Edition)
     * NVIDIA Kepler architecture     * NVIDIA Kepler architecture
-    * One GK110 GPU with 2688 SPs at 902 MHz to 954 MHz in single-precision mode, 6 GB GDDR5 RAM on a 384-bit bus, 317.2 GB/s+    * One GK110 GPU with 2688 SPs at 902 MHz to 1045 MHz in single-precision mode, 6 GB GDDR5 RAM on a 384-bit bus, 317.2 GB/s
     * Peak performance of over 5 TFLOPS single-precision,​ from 1.3 to 1.5 TFLOPS double-precision in the corresponding mode     * Peak performance of over 5 TFLOPS single-precision,​ from 1.3 to 1.5 TFLOPS double-precision in the corresponding mode
     * This is a budget replacement for the TESLA K20X GPU card intended for workstations and servers (which would cost at least 3 times more and would run considerably slower at single-precision and integer code, but would offer ECC RAM)     * This is a budget replacement for the TESLA K20X GPU card intended for workstations and servers (which would cost at least 3 times more and would run considerably slower at single-precision and integer code, but would offer ECC RAM)
-  * [[wp>Radeon HD 7000 Series#​Radeon_HD_7900|AMD Radeon ​HD 7990]] gaming graphics card +  * [[wp>AMD RX Vega series|AMD Radeon ​RX Vega 64]] gaming graphics card (reference designmanufactured by MSI with slight overclocking
-    * AMD GCN architecture +    * AMD GCN 5th gen architecture 
-    * Two “Tahiti” GPUs, which provides 2x2048 SPs, 6 GB GDDR5 RAM on two 384-bit buses, 576 GB/s +    * One Vega10 XT GPU with 4096 SPs typically ​at 1401 or 1576 MHz, GB HBM2 RAM on a "2048-bit" ​bus, 483.GB/s 
-    * Custom core clock rate: 501 MHz for GPU0 (heavily underclocked)997.5 MHz to 1050 MHz for GPU1 (almost same as HD 7970 GE) ((Normally, HD 7990 runs at 950 MHz to 1000 MHz - however, when we added a GTX Titan X to this machine in 2015, we had to underclock HD 7990's GPU0 as we no longer had a free slot near this dual-GPU card, so it was no longer able to keep its GPU0 cool when running at stock clocks. Going down all the way to 501 MHz is overkill, but it is the highest where the standard firmware would use a lower core voltage of 950mV instead of 1200mV, and this lower voltage is needed to prevent this GPU from overheating in our current setup.)+    * Peak performance of over 10 TFLOPS single-precision
-    * Peak performance of over 6 TFLOPS single-precision,​ about 1.5 TFLOPS double-precision +
-    * This is a budget replacement for the FirePro S10000 GPU card intended for servers (which would cost at least 3 times more, but would offer ECC RAM) +
-  * [[wp>​Radeon HD 6000 Series#​Radeon_HD_6700|AMD Radeon HD 5750/6750]] gaming graphics card marketed as "​PowerColor Radeon HD 6770 Green Edition (AX6770 1GBD5-HV4)",​ one half of a HD 5850 +
-    * AMD TeraScale 2 (VLIW5) ​architecture +
-    * One Juniper PRO GPU with 720 SPs at 700 MHz, GB GDDR5 RAM on a 128-bit bus, 73.GB/s +
-    * A short card that fits into this motherboard'​s 5th dual-width PCIe slot +
-    * Not a high performance card, but usable for testing/​benchmarking on the old VLIW5 architecture,​ such as to avoid performance regressions for users with older cards like this (HD 5000 and 6000 series up to and including 6870) +
-    * Peak performance of over TFLOPS single-precision +
- +
-Total peak performance is over 20 TFLOPS single-precision,​ about 4 TFLOPS double-precision.+
  
 +Total peak performance is over 31 TFLOPS single-precision.
 ===== Pictures ===== ===== Pictures =====
  
 Here's what the server looks like (click on the thumbnails for higher resolution pictures). Here's what the server looks like (click on the thumbnails for higher resolution pictures).
 +
 +2019 upgrade (added Vega 64 and GTX 1080, removed HD 7990 and HD 6770 Green Edition):
 +
 +{{super2019-uncovered1.jpg?​direct&​160}}
 +{{super2019-uncovered2.jpg?​direct&​160}}
  
 2015 upgrade (added GTX Titan X, as well as HD 6770 Green Edition into the short slot): 2015 upgrade (added GTX Titan X, as well as HD 6770 Green Edition into the short slot):
Line 71: Line 71:
 ===== Software ===== ===== Software =====
  
-The operating system is Scientific Linux 6.(with several devtoolsets installed, such as providing a variety of newer GCC versions), since this is a common free option to run Intel MPSS as needed to access the Xeon Phi card (which, in turn, runs its own copy of Linux, coming from Intel MPSS).+The operating system is Scientific Linux 6.10 (with several devtoolsets installed, such as providing a variety of newer GCC versions), since this is a common free option to run Intel MPSS as needed to access the Xeon Phi card (which, in turn, runs its own copy of Linux, coming from Intel MPSS). We also have CUDA 10.1 with its driver version 418.39, and AMD AMDGPU-PRO 18.50.
  
 Here's what this looks like via OpenCL: Here's what this looks like via OpenCL:
  
 <​code>​ <​code>​
-[solar@super ~]$ clinfo | egrep '​^ ​ (Platform |)Name:' ​| tail -n +4 +[solar@super ~]$ clinfo | fgrep Name: | tail -n +4 
-  Platform Name:                                 ​AMD Accelerated Parallel Processing +  Platform Name:  AMD Accelerated Parallel Processing 
-  Name:                                          ​Juniper +  Name: gfx900 
-  Name:                                          Tahiti +  Platform Name:  Intel(R) OpenCL 
-  Name:                                          Tahiti +  Name:         ​Intel(R) Xeon(R) CPU E5-2670 0 @ 2.60GHz 
-  Name:                                          Intel(R) Xeon(R) CPU E5-2670 0 @ 2.60GHz +  Name:  Intel(R) Many Integrated Core Acceleration Card 
-  Platform Name:                                 ​Intel(R) OpenCL +  Platform Name:  NVIDIA CUDA 
-  Name:                                                 ​Intel(R) Xeon(R) CPU E5-2670 0 @ 2.60GHz +  Name: GeForce GTX 1080 
-  Name:                                          Intel(R) Many Integrated Core Acceleration Card +  Name:​  ​GeForce GTX TITAN X 
-  Platform Name:                                 ​NVIDIA CUDA +  Name:  GeForce GTX TITAN
-  Name:                                          GeForce GTX TITAN X +
-  Name:                                          GeForce GTX TITAN+
 </​code>​ </​code>​
- 
-Curiously, "​Tahiti"​ appears twice because there are two such GPUs (they'​re device 0 and 1, respectively),​ whereas the CPUs appear twice because they'​re available via both AMD's and Intel'​s OpenCL SDK, and either SDK will use all cores of both CPUs. 
- 
-===== Additional resources ===== 
- 
-We also host a changing number of other development boxes, as of this writing including with Intel AVX2, Intel HD Graphics 4600 (with a configured and working OpenCL "​driver"​),​ AMD XOP, AMD GCN 1.1, NVIDIA Fermi, some non-x86 architectures (ARM, MIPS64, Epiphany), some FPGAs (ZedBoard with Xilinx Zynq 7020, ZTEX 1.15y with quad Spartan-6 LX150). ​ Please feel free to inquire about availability of these or/and other resources if relevant to your project. 
  
 ===== Who is eligible ===== ===== Who is eligible =====
Line 112: Line 104:
 ===== Credits ===== ===== Credits =====
  
-The HPC Village project is provided by [[http://​www.openwall.com|Openwall]] (idea, most computer hardware parts, software configuration,​ system administration) and [[http://​www.df.ru|DataForce]] (assembly and hosting of servers, Internet connectivity). ​ NVIDIA GTX Titan X purchase was fully sponsored by [[https://​sagitta.pw|Sagitta HPC, a subsidiary of Stricture Group LLC]]. ​ AMD Radeon HD 7990 was [[:​john/​contests|team john-users]]'​ prize in Hash Runner 2013 organized by [[http://​www.ptsecurity.com|Positive Technologies]]. +The HPC Village project is provided by [[https://​www.openwall.com|Openwall]] (idea, most computer hardware parts, software configuration,​ system administration) and [[http://​www.df.ru|DataForce]] (assembly and hosting of servers, Internet connectivity). ​NVIDIA GTX 1080 and AMD Vega 64 purchase was sponsored by a grant from [[https://​www.zfnd.org|Zcash Foundation]]. ​NVIDIA GTX Titan X purchase was sponsored by [[https://​sagitta.pw|Sagitta HPC, a subsidiary of Stricture Group LLC]]. AMD Radeon HD 7990 (available in this machine until January 2019, then replaced with the Vega 64) was [[:​john/​contests|team john-users]]'​ prize in Hash Runner 2013 organized by [[https://​www.ptsecurity.com|Positive Technologies]].
 ===== Related third-party projects/​offers ===== ===== Related third-party projects/​offers =====
  
Line 133: Line 124:
  
   * [[http://​research.microsoft.com/​en-us/​projects/​catapult/​|Project Catapult]] free access as [[https://​www.tacc.utexas.edu/​-/​tacc-to-launch-new-catapult-system-to-researchers-worldwide|announced by TACC (Texas Advanced Computing Center at The University of Texas at Austin)]] and [[http://​blogs.msdn.com/​b/​msr_er/​archive/​2015/​11/​12/​project-catapult-servers-available-to-academic-researchers.aspx|Microsoft Research]]   * [[http://​research.microsoft.com/​en-us/​projects/​catapult/​|Project Catapult]] free access as [[https://​www.tacc.utexas.edu/​-/​tacc-to-launch-new-catapult-system-to-researchers-worldwide|announced by TACC (Texas Advanced Computing Center at The University of Texas at Austin)]] and [[http://​blogs.msdn.com/​b/​msr_er/​archive/​2015/​11/​12/​project-catapult-servers-available-to-academic-researchers.aspx|Microsoft Research]]
 +
 +OSUOSL hosting and OSS services, as well access to ARM and POWER machines:
 +  * [[https://​osuosl.org/​services/​]]
 +
 +GRID5000 - A large-scale testbed for distributed computing, used by CS researchers in HPC, Clouds, Big Data, Networking, AI:
 +  * [[https://​www.grid5000.fr/​w/​Grid5000:​Home]]
HPC/Village.1506018158.txt · Last modified: 2017/09/21 20:22 by solar
 
Except where otherwise noted, content on this wiki is licensed under the following license: CC Attribution-Noncommercial-Share Alike 3.0 Unported
Recent changes RSS feed Donate to DokuWiki Powered by PHP Valid XHTML 1.0 Valid CSS Driven by DokuWiki Powered by OpenVZ Powered by Openwall GNU/*/Linux