226
edits
Changes
Jump to navigation
Jump to search
==Motherboard/CPU/Fan==
*Not a huge deal, but used for data preparation
*If using multiple GPUs, at least one core (two threads) per GPU
*Should have enough PCIe slots
*Motherboards: ASUS Z10PE-D16 [http://www.newegg.com/Product/Product.aspx?Item=N82E16813132257&Tpk=N82E16813132257], Dual LGA 2011 R3, DDR4 - Up to 32GB RDIMM, 16 slots
*Chips: Intel Haswell Xeon e5-2620v3, 6 core @ 2.4ghz, 6x256k level 1 cache, 15mb level 2 cache, socket LGA 2011-v3 [https://www.amazon.com/gp/product/B00M1BUUMO/ref=oh_aui_detailpage_o04_s00?ie=UTF8&psc=1]
*CPU Fans: Intel Thermal Solution Cooling Fan for E5-2600 Processors BXSTS200C [https://www.amazon.com/gp/product/B007HJAM50/ref=oh_aui_detailpage_o03_s00?ie=UTF8&psc=1]
Using multiple GPUs adds a lot of complexity. It has a few benefits: possible speed ups if the network can be split up (and is big enough), able to train multiple networks at once (either copies of the same network or modified networks), more memory for huge batches. Tensorflow doesn't scale well for multiple GPUs, but could be used better (probably) to train multiple networks at once.
==RAM==
*At least as much RAM as GPUs, probably twice that much (2 * Number of cards * 11 GB [GTX 1080 Ti size])
*RAM: Crucial DDR4 RDIMM [http://www.newegg.com/Product/Product.aspx?Item=9SIA0ZX39C3002], 2133Mhz , Registered (buffered) and ECC, comes in packs of 4 x 32GB
==PSU==Pros of multiple GPUs: *Some say 1Adds a lot of complexity.5x-2x wattage Cons of GPU+CPUmultiple GPUs:*Possible speed ups if the network can be split up (and is big enough), some say GPU+CPU+100Wbut tensorflow is not great for this*PSUs: Corsair RM Series 850 Watt ATX/EPS 80PLUS Gold-Certified Power Supply - CP-9020056-NA RM850 [https://wwwAble to train multiple networks at once (either copies of the same network or modified networks).amazon.com/gp/product/B00EB7UIXM/ref=oh_aui_detailpage_o03_s00?ie=UTF8&psc=1] Allows for running long experiments while running new ones*More memory for huge batches (not sure if necessary)
==Storage==
*M.2 Drives: Samsung 950 PRO -Series 512GB PCIe NVMe - M.2 Internal SSD 2-Inch MZ-V5P512BW [https://www.amazon.com/gp/product/B01639694M/ref=oh_aui_detailpage_o03_s01?ie=UTF8&psc=1]
*Solid State Drives: Intel Solid-State Drive 750 Series SSDPEDMW400G4R5 PCI-Express 3.0 MLC - 400GB [https://www.amazon.com/gp/product/B00UHJJQAY/ref=oh_aui_detailpage_o07_s00?ie=UTF8&psc=1] or 800GB [https://www.amazon.com/gp/product/B013QP8XUE/ref=oh_aui_detailpage_o00_s00?ie=UTF8&psc=1]
*Regular Hard drives: WD Red 3TB NAS Hard Disk Drive [https://www.amazon.com/gp/product/B008JJLW4M/ref=oh_aui_detailpage_o00_s00?ie=UTF8&psc=1] - 5400 RPM Class SATA 6 Gb/s 64MB Cache 3.5 Inch
no edit summary
}}
==Single vs. Multi GPU==
*[https://www.nvidia.com/en-us/geforce/products/10series/geforce-gtx-1080-ti/ GTX 1080 Ti Specs]
* Since we are using Tensorflow, it doesn't scale well to multiple GPUs for a single model
# If the network doesn't fit in the memory of one GPU (11 GB),
* [https://devtalk.nvidia.com/default/topic/743814/cuda-setup-and-installation/advice-on-single-vs-multi-gpu-system/ Advice on single vs multi-GPU system]
# Might want Want to get two graphics cards, one for development, one (crappy card) for operating system [https://stackoverflow.com/questions/21911560/how-can-i-set-one-nvidia-graphics-card-for-display-and-other-for-computingin-li]
*[https://stackoverflow.com/questions/37732196/tensorflow-difference-between-multi-gpus-and-distributed-tensorflow Different uses of multiple GPUs]
# Intra-model parallelism: If a model has long, independent computation paths, then you can split the model across multiple GPUs and have each compute a part of it. This requires careful understanding of the model and the computational dependencies.
====TL;DR====
==Misc.Parts==
*Cases: Rosewill 1.0 mm Thickness 4U Rackmount Server Chassis, Black Metal/Steel RSV-L4000[https://www.amazon.com/gp/product/B0056OUTBK/ref=oh_aui_detailpage_o04_s00?ie=UTF8&psc=1]
*DVDRW (Needed?): Asus 24x DVD-RW Serial-ATA Internal OEM Optical Drive DRW-24B1ST [http://www.amazon.com/Asus-Serial-ATA-Internal-Optical-DRW-24B1ST/dp/B0033Z2BAQ/ref=sr_1_2?s=pc&ie=UTF8&qid=1452399113&sr=1-2&keywords=dvdrw]
==Single GPU Build==
==Double GPU Build==
[https://pcpartpicker.com/list/ZQjKf8 PC Partpicker build] missing a lot of pieces
===Motherboard===
*Should have enough PCIe slots
*Motherboards: ASUS Z10PE-D16 [http://www.newegg.com/Product/Product.aspx?Item=N82E16813132257&Tpk=N82E16813132257], Dual LGA 2011 R3, DDR4 - Up to 32GB RDIMM, 16 slots
===CPU/Fan===
*Not a huge deal, but used for data preparation
*If using multiple GPUs, at least one core (two threads) per GPU
*Chips: Intel Haswell Xeon e5-2620v3, 6 core @ 2.4ghz, 6x256k level 1 cache, 15mb level 2 cache, socket LGA 2011-v3 [https://www.amazon.com/gp/product/B00M1BUUMO/ref=oh_aui_detailpage_o04_s00?ie=UTF8&psc=1]
*CPU Fans: Intel Thermal Solution Cooling Fan for E5-2600 Processors BXSTS200C [https://www.amazon.com/gp/product/B007HJAM50/ref=oh_aui_detailpage_o03_s00?ie=UTF8&psc=1]
===GPU===
* 2x GTX 1080 Ti for computation
* Video card for monitor
===RAM===
*At least as much RAM as GPUs, probably twice that much (2 * Number of cards * 11 GB [GTX 1080 Ti size])
*RAM: Crucial DDR4 RDIMM [http://www.newegg.com/Product/Product.aspx?Item=9SIA0ZX39C3002], 2133Mhz , Registered (buffered) and ECC, comes in packs of 4 x 32GB
===PSU===
*Some say 1.5x-2x wattage of GPU+CPU, some say GPU+CPU+100W
*PSUs: Corsair RM Series 850 Watt ATX/EPS 80PLUS Gold-Certified Power Supply - CP-9020056-NA RM850 [https://www.amazon.com/gp/product/B00EB7UIXM/ref=oh_aui_detailpage_o03_s00?ie=UTF8&psc=1]
===Storage===
*M.2 Drives: Samsung 950 PRO -Series 512GB PCIe NVMe - M.2 Internal SSD 2-Inch MZ-V5P512BW [https://www.amazon.com/gp/product/B01639694M/ref=oh_aui_detailpage_o03_s01?ie=UTF8&psc=1]
*Solid State Drives: Intel Solid-State Drive 750 Series SSDPEDMW400G4R5 PCI-Express 3.0 MLC - 400GB [https://www.amazon.com/gp/product/B00UHJJQAY/ref=oh_aui_detailpage_o07_s00?ie=UTF8&psc=1] or 800GB [https://www.amazon.com/gp/product/B013QP8XUE/ref=oh_aui_detailpage_o00_s00?ie=UTF8&psc=1]
*Regular Hard drives: WD Red 3TB NAS Hard Disk Drive [https://www.amazon.com/gp/product/B008JJLW4M/ref=oh_aui_detailpage_o00_s00?ie=UTF8&psc=1] - 5400 RPM Class SATA 6 Gb/s 64MB Cache 3.5 Inch
