Amazon supercharges GPU power, spits out Nvidia-backed G3

Get your office benchmarking Crysi- *cough* I mean, working

By Andrew Silver

Posted in HPC, 14th July 2017 15:09 GMT

Amazon has rolled out its latest GPU computing box instance line, G3.

It comes in three flavours: g3.4xlarge (1 GPU), g3.8xlarge (2 GPUs), and g3.16xlarge (4 GPUs). The line is meant for 3D modelling, visualisation, video encoding and other graphics-intensive apps.

Amazon's G2 line first came out in 2013. The high-horse, introduced later on in 2015, was the g2.8xlarge, which came with four Nvidia Grid GPUs with 4GB of video memory and 1,536 CUDA Cores each. You could encode four 1080p video streams or eight real-time 720p video streams, Amazon said in a blog post.

The GPUs in the new G3 family come with 8GB of GPU memory, 2,048 parallel processing cores, multi-monitor support, enhanced graphics-rendering effects, Nvidia GRID Virtual Workstation features, and enhanced networking. You could record 10 H.265 (HEVC) 1080p30 streams and up to 18 H.264 1080p30 streams, Amazon said in another blog post.

To compare to the previous g2.8xlarge with 60GiB of memory and 32 vCPUs (because why not compare the best to the best), the new g3.16xlarge instance comes with 488 GiB of main memory and 64 vCPUs.

The g2.8xlarge cost $2.808 hourly on demand with Linux in Ireland, whereas g3.16xlarge costs $4.84, according to Amazon's current price list.

So if you want the grunt, it'll cost you. ®

Sign up to our NewsletterGet IT in your inbox daily

2 Comments

More from The Register

Nvidia reports record revenues in latest fiscal quarter

But cryptocurrency not growing for the GPU giants

Nvidia: Using cheap GeForce, Titan GPUs in servers? Haha, nope!

Nice try, but no, you're gonna have to cough up for these expensive data center chips

Nvidia says Google's TPU benchmark compared wrong kit

You're faster than the Kepler, but what about the newer and better Pascal?

Nvidia's profits so far this year are GPU-ge

The anti-AMD racks up 48 per cent revenue jump

Nvidia: Pssst... farmers. Need to get some weeds whacked?

HPC blog Jetson-powered sprayer turns GPU-flinger into weed-whacker

Google ramping up AI in China, Nvidia's Titan V, Intel's hip-hop misstep

Roundup And more in your machine-learning news summary

'Self learning' Intel chips glimpsed, Nvidia emits blueprints, AMD and Tesla rumors, and more

AI roundup A quick guide to this week's reveals

AMD, Intel hate Nvidia so much they're building a laptop chip to spite it

Analysis Just months after Chipzilla trashed its new best friend as an 'unreliable supplier'

Hey, Nvidia – who loves you, and who do you love? A. Cloud giants

Cryptocurrency mining is another story, though

Facebook AI guru alt-tabs out, Nvidia EULA audits, Baidu changes, chip tricks, and more

Roundup Machine-learning news and code to pore over