Computer specs

1,088 views
Skip to first unread message

Mikec

unread,
May 26, 2016, 3:09:10 PM5/26/16
to LAStools - efficient tools for LiDAR processing
I am looking into purchasing a dedicated lidar processing computer. What are peoples' current recommendations in terms of system specs (RAM, CPU, # of cores, etc.)? Thanks.

Jeff Romais

unread,
May 26, 2016, 3:38:20 PM5/26/16
to last...@googlegroups.com
I suggest something of the Wintel persuasion because of lower costs. For me it was important to have redundant drives for internal backup and a cpu that can work at peak power without heating up. The station can recall in one or two seconds a Arcmap 10.2 file containing 100 high resolution DEM extracted from LiDAR tiles without a single hiccup.  A couple years ago I bought a custom built computer with the following specs:

Cooler Master HAF XB w/ USB 3.0, HDD Hot Swap & X-Dock, Removable Motherboard Tray
CD: LG 14X Internal Blu-ray Burner, BD-RE, DVD+RW Combo Drive (Black Color)
CPU: Intel(R) Core™ i7-4790 3.60 GHz 8MB LGA1150
CS_FAN: Maximum 120MM Case Cooling Fans for your selected case 
FAN: Cooler Master Seidon 120M Liquid Cooling CPU Cooler - Enhance Cooling Performance Single Standard 120MM Fan
COOLANT: Standard Coolant
SSD 500Gb (System Drive)
SSD 250Gb (LiDAR and Mapping drive)
HDD:2x  2TB (2TBx1) SATA-III 6.0Gb/s 64MB Cache 7200RPM HDD (for general file storage and back up)
IUSB: Built-in USB 2.0 Ports
MEMORY: 16GB (8GBx2) DDR3/2133MHz Dual Channel Memory [+100] (ADATA XPG V3)
MIR_VCSSD: * $35 MIR for AMD R9 270 2GB or R9 270X 2GB + ADATA 512GB SP610 SSD Combo
MOTHERBOARD: ASUS Z97-K ATX w/ Intel GbLAN, 2 PCIe x16, 2 PCIe x1, 2 PCI, 1 x M.2, 6x SATA 6Gb/s 
MOUSE: AZZA Optical 1600dpi Gaming Mouse with Weight Adjustable Cartridge
NETWORK: Onboard Gigabit LAN Network
POWERSUPPLY: * 750 Watts - EVGA SuperNOVA 750B1 80 Plus Bronze Modular Power Supply
VIDEO: AMD Radeon R9 270 2GB GDDR5 PCIe 3.0 x16 Video Card (Single Card)
 

Good luck.

On Thu, May 26, 2016 at 1:52 PM, Mikec <mike....@gmail.com> wrote:
I am looking into purchasing a dedicated lidar processing computer. What are peoples' current recommendations in terms of system specs (RAM, CPU, # of cores, etc.)? Thanks.

Nicolas Cadieux

unread,
May 26, 2016, 3:40:01 PM5/26/16
to last...@googlegroups.com

Hello
If you have the cash as Xeon CPU would be better as you can put more memory.

I have a Intel extreme CPU with 6 cores an 12 thread.  64GB of memory (the max memory on a Intel i7).  The biggest boost comes from the pciE SSD.  I would get an x99 MB. 

Nicolas

On May 26, 2016 13:52, Mikec <mike....@gmail.com> wrote:
I am looking into purchasing a dedicated lidar processing computer. What are peoples' current recommendations in terms of system specs (RAM, CPU, # of cores, etc.)? Thanks.

carlisle haworth

unread,
May 27, 2016, 2:56:50 AM5/27/16
to last...@googlegroups.com

If cost is no concern, the new MacPro cylinder running windows boot camp is most impressive. These run anywhere from $3k to 10k fully stacked. 

sent from a phone™

On May 26, 2016 12:09 PM, "Mikec" <mike....@gmail.com> wrote:
I am looking into purchasing a dedicated lidar processing computer. What are peoples' current recommendations in terms of system specs (RAM, CPU, # of cores, etc.)? Thanks.

Newcomb, Doug

unread,
May 27, 2016, 8:39:22 AM5/27/16
to last...@googlegroups.com
It depends on the extent and density of LiDAR data you will be processing and the methods and software you use.  I do most of my LiDAR processing on Linux and occasionally have single processes that exceed the 32 GB of RAM available on my workstation.

I have a home system with an 8 core AMD Vishera CPU and 32 GB of RAM that I bought a couple of years ago does pretty well on processing large LAZ files.  I like more cores over raw speed per core.  

Agree with SSD for the OS Drive.  The prices of SSD's are coming down, but depending on the amount of LiDAR data you have to process a  "working" SSD combined with Larger spinning drives( 4TB +)  for redundancy/slower storage might be better  .


Doug  

 
--
Doug Newcomb
USFWS
Raleigh, NC
---------------------------------------------------------------------------------------------------------
The opinions I express are my own and are not representative of the official policy of the U.S.Fish and Wildlife Service or Dept. of the Interior.   Life is too short for undocumented, proprietary data formats. As a federal employee, my email may be subject to FOIA request.

Jameson Harrington

unread,
May 27, 2016, 8:49:32 AM5/27/16
to LAStools - efficient tools for LiDAR processing
Has any one found a workstation graphics card (Quadro) to be necessary or will a high end gaming card work? I know some of the Riegl software will leverage GPU processing but are there others where you will see a benefit? 

Thanks,
Jameson

Michael Treglia

unread,
May 27, 2016, 9:09:25 AM5/27/16
to last...@googlegroups.com
I have actually spec'd out the last few PCs w/ a good GPU thinking some good applications for it would be coming down the pipe, but not so much. R has some packages that will let you leverage good GPUs; some bioinformatics software has GPU capabilities, and the only GIS package I know that really takes advantage of it is Manifold (http://manifold.net/), for which they're beta-ing a new version now (I think with additional GPU capabilities).

Re: Computer specs - I agree with what everyone else is suggesting - check out the specs for the Motherboard/processor, but Xeon typically let you add more RAM (which is a cheap upgrade), and multicore processing is becoming more common it seems, so more cores can be advantageous.  SSD for OS (and you can get a good 1TB SSD for ~$300, so maybe even for data if that's within your budget and needs), and plenty of RAM.  

And depending on what you're doing, might be able to avoid Windows needs. (I do most of my work on Windows, but using Free/Open Source which runs great on various linux distros)

mike


batt...@eaglemapping.com

unread,
May 27, 2016, 11:47:42 AM5/27/16
to LAStools - efficient tools for LiDAR processing
We built a workstation back in late 2014 running a Nvidia Titan Black GPU, which was around $1200 at the time. Noticed a minor boost running Riegl RiAnalyze waveform extraction, but not enough to justify the price over a mid-range GPU. The money is far better spent on CPU and SSDs. Many, many SSDs... 

Our newest workstation is running a GeForce GTX 970, 2 x Xeon E5-2690 v3 @ 2.6GHz, 64GB DDR4-2133 RAM, and 24TB of Samsung 850 EVO SSDs running in RAID 50. Granted, this was extremely expensive, but worthwhile given the production boost we've seen.

All depends on the purpose, and how much processing you intend on doing. Our latest workstation is running data almost non-stop, so the ~50% performance increase was well worth the investment. If it is for more sporadic use, I'd recommend a mid-range GPU, nothing fancy, a moderate amount of fast RAM, great hard drives in RAID, and the best CPU you can afford. For the record, that Intel 4790 is great, but for an extra $40 you can get the 4790K @ 4.0GHz which I have found to be fantastic for LiDAR acquisition and processing, with probably the best single thread performance on the market at the moment.

Cheers,
Bryan

Nicolas Cadieux

unread,
May 27, 2016, 12:22:43 PM5/27/16
to last...@googlegroups.com

Hi,

I agree with Brian on the CPU and SSD being more important than the GPU. The EVO is a nice drive but if you have the cash, the OCZ SSD that plugs into the PCIe video card bus is very fast and big (1GB).

I you don't want to invest in a Xeon now, you can buy an X99 MB that will take the Intel extreme.  Later, you can change it for a 32core Xeon. The memory will have to be changed also when you swop the CPU.   Going second hand in a few years will save you lots of cash on those big Xeons...

Nicolas

Inuk Simard (GMTX)

unread,
May 27, 2016, 2:59:51 PM5/27/16
to last...@googlegroups.com
Hi,
 
Am still not a huge lidar data processor, but I use processed lidar data in my projects along with a lot of imagery (aerial, satellite, UAV) and other point cloud data sources, arcgis, pix4dmapper and other suites.
 
Am running on a Dell Precision T7600, 64-bit WIN10 (upgraded from 8), RAM: 256MB, 2x16-core Xeon E5-2687W @ 3.10GHz, GPU: Nvidia Quadro 5000 with Tesla C2075 (in CUDA config), HD1 (OS): 1TB, RAID1: 5.5TB & RAID2: 3TB for project data and backups.
 
Most applications do not cringe on heavy data set visualization and that's the positive of hacing high RAM, but if your mostly crunching numbers in the background, less ram, more CPU and fast drives would be better. I rarely get frustrated with this system, I only wish I would have installed SSD drives instead of regular HDs, for less RAM. Even at full 32 logical processing capacity (in multithread applications) i rarely consume more than 30% RAM, which is already a lot.
 
Inuk

Evon Silvia

unread,
May 31, 2016, 8:22:46 PM5/31/16
to last...@googlegroups.com
I'll add another vote for the SSDs. We've stuck with the Samsung EVO/PRO SSDs because we have had noteworthy reliability issues with other brands when they're subjected to heavy processing loads.

GPU processing has shown a lot of promise in theory, but as others have said there isn't much out there that's leveraging it for LiDAR processing. Yet. The main issues seem to be that most time-consuming procedures either are IO-bound and therefore processing speed is more or less irrelevant (thus the emphasis on SSD arrays), or they require access to data structures and neighbors, which GPU processing makes extremely complicated, wasteful, and therefore slow, negating its advantages.

Evon

David Brubacher

unread,
Jun 1, 2016, 11:03:46 AM6/1/16
to LAStools - efficient tools for LiDAR processing
The Samsung SSDs on the SATA bus are great (e.g. 850 Pro), but if you have an M.2 slot, the 950 Pro will give you 2,500 MB/s read, 1,500 MB/s write vs both numbers in the 500-600 range due to the limitations of the SATA interface. If you don't have a recent motherboard with the M.2 slot, you can get an adapter which should be nearly as fast.

I've also heard tell of an adapter card that will take multiple M.2 format drives and goes in a secondary video card slot, but haven't found a source.

If you are looking at an SSD RAID solution purely for speed as opposed to data safety, this solution is better and cheaper.
Reply all
Reply to author
Forward
0 new messages