Choosing a laptop for machine learning : The Search — Part I
Few months ago, I quit my job to sort out certain mental cobwebs. “What next” is often a questions that bugs many millennials — and I am sure not an exception. Clarity of thought is often inspired from the moments spent out of comfort zone I think.
Now, my work is majorly in the domain of machine learning and natural language processing, building intelligent and production grade systems. Hence, you can’t make do with a underpowered laptop, can you? Out of a job (and a laptop ofcourse, as most work I did was on the office laptop), I needed to buy a new laptop to continue my weekend projects and learning experiments.
A typical machine learning compatible laptop should have the following specs — entirely my preference, don’t hold this against me :)
- 500 GB+ storage, SSD preferable
- i5/i7 processors — quadcore (6XXXHQ/K) preferred over (6XXXU)
- 16 GB RAM
- 4 GB+ NVIDIA Graphics Card (atleast 960M, AMD has limited support, apart from OpenCL)
- 13-14 inch screen
- Ubuntu 14.04
A good laptop with the above configurations will cost atleast 2000 USD in India. I stretched my budget to around 2750 USD. After good few days of research, the following shortlist emerged;
- RazorBlade Stealth
- Alienware R3/R2 (13’/15’)
- Dell Inspiron 7559 ← Highly recommended budget ML laptop
- Lenovo Ideapad Y700/Y900
- Custom build from Kolkata based Azom Systems
While Razorblade is not available in India, Alienware offerings in the country didn’t have a 10 series NVIDIA card. Dell has flimsy hardware at lower price ranges (past experience), and Lenovo has bad customer service in its Indian outlets, or atleast that’s what the web forums concurred.
Azom Systems did give me a extremely competitive price of 2000 USD with the following configuration, with option of extending a warranty for two years with additional payment of 250 USD.
Azom Laptop — Exigo
Clevo Barebone — P750DM2Display — 15.6” IPS Full-HD 1920x1080 16:9 Matte (anti-glare)
CPU — Intel Skylake Desktop Grade i5–6500
GPU — NVIDIA GTX 1060 6GB GDDR5
RAM1–8GB 2133MHz DDR4
RAM2–8GB 2133MHz DDR4
RAM3 — Empty
RAM4 — Empty
Storage1 — SATA 2.5” 1TB 5400rpm HDD
Storage2 — Empty (SATA 2.5" HDD/SSD)
Storage3 — M.2 2280 120GB SATA SSD
Storage4 — Empty (M.2 PCIe/SATA SSD)
Wireless — Intel Wireless-AC 8260, dual band, 2x2 + Bluetooth 4.2
Sound — Sound Blaster X-FI MB5, ESS SABRE Hi-Fi DAC headphone audio, 2 x Speakers
Card Reader — 6-in-1 Push-Push Card reader
Keyboard — Customizable backlit, anti-ghost keys
Camera — 2.0M FHD Video camera
Optical Drive — None
Battery — 8 cell (82Wh)
I didn’t go ahead with them for only one reason — they are a small outfit and required you to ship the machine in case it needed repairs. However, I would recommend them highly if you need a custom build in India.
Now, I rejected all the candidates in my shortlist for one of the following reasons;
- Bad/Limited customer service
- Overpriced with soon-to-be-outdated hardware — I need a laptop to last me atleast 3 years
- Flimsy hardware
- Not available in India
What did I do next? That’s an interesting question to be answered in my next post.
Next: Choosing a laptop for machine learning : The Buy — Part II