In earlier posts I’ve discussed issues of high performance computing as it relates to my engineering work in antennas and microwaves. I include such articles here since our collective security (or lack thereof) is strongly dependent on developments in high performance computing. In this post, I will briefly showcase the Mythlogic Chaos 2012 Notebook (also known as the Clevo W110ER) computer that I purchased this past June. With an 11 inch screen, and weighing under 4 lbs, this machine is only slightly larger than a netbook, yet it packs enough compute power to replace even current desktop and tower sized engineering workstations. Based in Ann Arbor, Michigan, Mythlogic sells a range of customized mobile and desktop computers based on designs sourced from Clevo in Taiwan. Clevo is not a household name but it is my understanding that they source the Dell Alienware computers so they are well-known in the industry.
Mythlogic offers a range of customizations to the basic Clevo designs. You can choose CPU, disk, networking and even custom painted cases. For the CPU, I chose the Intel i7-3612QM chip as the best balance of speed, power and thermal load for my needs. This particular chip is one of Intel’s Ivy Bridge series hyper-threaded quad-core CPUs. It presents as 8 virtual cores to the OS. The i7-3612QM runs at 2.1 GHz but can ramp up to 3.1 GHz in turbo mode when necessary. The lower clock helps conserve battery and keeps idle temperatures cooler.
I maxed out the RAM at 16 GB to get best possible performance on my engineering tools and allow me to run multiple Virtual Machines. I chose only a pedestrian Western Digital 7200 RPM 500 GB drive and promptly replaced it with a 120 GB Intel SSD that I already owned when it arrived. The supplied bare 500 GB went into an external USB 3.0 case and now holds my portable data archive. I run Xubuntu as the host OS on all my machines so I opted to forgo having Mythlogic install an OS. This saved me the cost of an OEM license for Windows 7. Lastly, I opted for the standard dual-band Intel Centrino 802.11 a/g/n wireless for performance and compatibility with Linux.
The CPU, memory, and disk system of this tiny machine are obviously very good and on par with high-end Ultrabooks from the likes of HP, Apples, etc. The thing that sets it apart however is the included nVidia GeForce 650M series graphics with 2 GB of DDR3 video RAM. The 650M offers an amazing 384 CUDA cores that dramatically speed up 3D graphics and/or certain type of calculations.
One of my primary engineering tools is the Remcom X-FDTD 3D electromagnetic field simulator. This program can make use of nVidia CUDA cores to solve antenna and microwave engineering problems many times faster than it can with traditional Intel CPU cores. I do most of my X-FDTD work on a dedicated GPU workstation that uses two nVidia Tesla C2070 cards and an nVidia Quadro 5800. This machine is very powerful but it’s of no use to me when I’m traveling so I was very interested to see the capabilities of the Chaos 2012.
Quick GPU & CPU Benchmark
As a quick benchmark, I ran a simulation of a dielectric filled cavity resonator antenna on the GPU workstation and on the Chaos 2012. I configured the problem so that it fit within the 2 GB of video RAM available on the notebook machine. Runtime on the GPU workstation was 43 seconds versus 215 seconds on the Mythlogic. For this problem, the GPU workstation is a healthy 6.14 times faster than the tiny notebook. This is as it should be since the GPU workstation cost on the order of 10x the price! For reference, the same simulation required 507 seconds using all 8 virtual CPU Intel cores in the Mythlogic machine. I estimated run-time with a single core to be on the order of 30 minutes (1,800 seconds). For simulation work the 650M graphics is 2.36 times faster than all 8 virtual Intel cores. As a practical matter, doing the simulation on the GPU saves even more time! When running on the GPU, the simulator needs very little CPU power so I can continue to work on the machine (running a Virtual Machine, writing e-mails, a report, or configuring the next run for the solver) while I wait for the simulation to finish. When I’m running only on CPUs the machine is pretty much maxed out and I have to do something else like take a nap while the solver is working.
The Chaos 2012 has a lot of horse power and there is no battery technology in existence that will let you run number smashing jobs all day and still fit inside such a small case. That said, you can reasonably expect up to 5 hours or more away from the wall socket if you’re just doing normal office work. If you’re gaming, simulating or trying to crack passwords with the GPU and all 8 virtual cores maxed out you shouldn’t expect more than about an hour of battery life. That’s not a lot, but this computer is more about pure compute power than it is about long battery life. That’s a good trade-off for my needs so I’m not complaining!
While the Chaos 2012 doesn’t replace a dedicated GPU workstation it does provide enough CPU and GPU power to enable me to do real world work with the X-FDTD simulator when I’m not in the office. This is a big advantage for me and well worth the roughly $1300 price.
Where does a machine like the Chaos 2012 fit into the issue of security? Well, it comes down to encryption. Everything we do on the web hinges on the notion that if we encrypt a message that it will take someone a very long time to decrypt it if they don’t have the password. As computers get faster it’s getting easier and cheaper to break encryption. The notion that we now have tiny notebook computers with extremely powerful GPU and CPU processors means that the hacker of the future (today perhaps?) might well be able to do his/her code breaking while they are mobile! This is a sobering thought and should motivate you to start using good passwords that are harder to crack.
The Chaos 2012 along with many other modern notebooks have what is known as Optimus technology. Optimus combines a low power Intel graphics chip with a higher end nVidia chip to get the best of both power saving and powerful 3D graphics. Unfortunately, it seems nVidia is not making available graphics drivers to support this technology under Linux. As a result, the default install of say Xubuntu 12.04 will only use the Intel graphics chip. This is fine for most business and personal applications, but to utilize the nVidia portion of the graphics subsystem you will need to install Bumblebee software. Setup of Bumblebee under Xubuntu was relatively straightforward and you should have no problems following the directions on the Bumblebee site. After Bumblebee is working you can start GPU intensive applications using the optirun command. This gives you the flexibility to selectively run the demanding GPU applications according to whether you’re connected to the grid or on battery.
If you find that your particular GPU application is having trouble with Bumblebee you may need to confirm that the software is written such that it can use the Kepler based GPU. In my case, I had to wait several months for Remcom to play catch up and change their software to use the 650M chip. That’s a big part of the reason that I’m writing this article in December and not July right after I got the machine!
I purchased the above mentioned notebook computer from Mythlogic using my own funds. I have no financial relationship with them.
In this article
- Linux and Open Source
- 3d graphics
- 8 virtual
- 8 virtual cores
- battery life
- chaos 2012
- compute power
- cuda cores
- dedicated gpu
- dedicated gpu workstation
- engineering tools
- gpu workstation
- graphics chip
- high performance computing
- intel cores
- intel graphics
- intel graphics chip
- performance computing
- video ram
- virtual cores