TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

New P2 Instance Type for Amazon EC2 – Up to 16 GPUs

223 pointsby jeffbarrover 8 years ago

19 comments

Smerityover 8 years ago
$0.9 per K80 GPU per hour, while expensive, opens up so many opportunities - especially when you can get a properly connected machine.<p>Just as an example of the change this entails for deep learning, the recent &quot;Exploring the Limits of Language Modeling&quot;[1] paper from Google used 32 K40 GPUs. While the K40 &#x2F; K80 are not the most recent generation of GPU, they&#x27;re still powerful beasts, and finding a large number of them set up well is a challenge for most.<p>In only 2 hours, their model beat previous state of the art results. Their new best result was achieved after three weeks of compute.<p>With two assumptions, that a K80 is approximately 2 x K40 and that you could run the model with similar efficiency, that means you can beat previous state of the art for ~$28.8 and could replicate that paper&#x27;s state of the art for ~$7257.6 - all using a single P2 instance.<p>While the latter number is extreme, the former isn&#x27;t. It&#x27;s expensive but still opens up so many opportunities. Everything from hobbyists competing on Kaggle competitions to that tiny division inside a big company that would never be able to provision GPU access otherwise - and of course the startup inbetween.<p>* I&#x27;m not even going to try to compare the old Amazon GPU instances to the new one as they&#x27;re not even in the same ballpark. They have far less memory and don&#x27;t support many of the features required for efficient use of modern deep learning frameworks.<p>[1]: <a href="https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;1602.02410" rel="nofollow">https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;1602.02410</a>
评论 #12611643 未加载
评论 #12611311 未加载
评论 #12611349 未加载
评论 #12611123 未加载
__mpover 8 years ago
MeteoSwiss uses a 12 node cluster with 16 GPUs each to compute its weather forecasts (Piz Escha&#x2F;Kesch). The machine is in operation since more than a year. We were able to disable the old machine (Piz Albis&#x2F;Lema) last week.<p>The 1.1km forecast runs on 144 GPUs, the 2.2km probabilistic ensemble forecast is computed on 168 GPUs (8 GPUs or 1&#x2F;2 node per ensemble member). The 7km EU forecast is run on 8 GPUs as well.
评论 #12612312 未加载
评论 #12613121 未加载
paukiatweeover 8 years ago
For anyone who interested in ML&#x2F;DL on cloud.<p>Google Cloud Platform just released Cloud ML beta with different pricing model, see <a href="https:&#x2F;&#x2F;cloud.google.com&#x2F;ml&#x2F;" rel="nofollow">https:&#x2F;&#x2F;cloud.google.com&#x2F;ml&#x2F;</a><p>Cloud ML costing $0.49&#x2F;hour to $36.75&#x2F;hour, compared to AWS $0.900&#x2F;hour to $14.400&#x2F;hour<p>The huge different of $36.75&#x2F;hour (Google) compared to $14.400&#x2F;hour (AWS) make me wonder what Cloud ML are using, they mentioned GPU (TPU?) but not exact GPU model.
评论 #12611235 未加载
评论 #12611484 未加载
评论 #12611182 未加载
topbananaover 8 years ago
<p><pre><code> All of the instances are powered by an AWS-Specific version of Intel’s Broadwell processor, running at 2.7 GHz. </code></pre> Does anyone have any more information about this? Are the chips fabricated separately or is it microcode differences?
评论 #12611486 未加载
mrbover 8 years ago
Keep in mind the Nvidia K80 is a 2 years old Kepler GPU. Nvidia launched 2 newer microarchitectures since then: Maxwell, Pascal. I would expect to see some P100 Pascal GPU &quot;soon&quot; on AWS. Maybe 6 months? (Maxwell&#x27;s severely handicapped double precision performance reduces its utility for many workloads.)
评论 #12611455 未加载
评论 #12612332 未加载
sysexitover 8 years ago
Pretty classless how Jeff describes TensorFlow as an &quot;Open Source library,&quot; without atributing it to Google.
评论 #12612124 未加载
raverbashingover 8 years ago
One thing I discovered recently is that for GPU machines your initial limit on AWS is 0 (meaning you have to ask support before you start one for yourself)<p>(This might be an issue of my account though - having had only small bills so far)
评论 #12611173 未加载
评论 #12611372 未加载
评论 #12611696 未加载
phs318uover 8 years ago
My first thought: &quot;I wonder what the economics are like, re: cryptocurrency mining?&quot;<p>My second thought: &quot;I wonder if Amazon use their &#x27;idle&#x27; capacity to mine cryptocurrency?&quot;<p>With respect to my second thought, at their scale, and at the cost they&#x27;d be paying for electricity, it could quite possibly be a good hedge.
评论 #12611477 未加载
评论 #12612374 未加载
评论 #12611387 未加载
评论 #12611677 未加载
评论 #12611836 未加载
chrisconleyover 8 years ago
This is great - we&#x27;ll try to get our Tensorflow and Caffe AMI repo updated soon: <a href="https:&#x2F;&#x2F;github.com&#x2F;RealScout&#x2F;deep-learning-images" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;RealScout&#x2F;deep-learning-images</a>
seanwilsonover 8 years ago
How do Amazon (or any other cloud provider) make sure they have enough of these machines to cope with the demand for them without getting too many?
评论 #12612002 未加载
评论 #12612936 未加载
评论 #12611223 未加载
评论 #12612050 未加载
评论 #12611878 未加载
ivan_ahover 8 years ago
According to [1], the K80 GPUs have the following specs:<p><pre><code> Chips: 2× GK210 Thread processors: 4992 (total) Base clock: 560 MHz Max Boost: 875 MHz Memory Size: 2× 12288 Clock: 5000 Bus type: GDDR5 Bus width: 2× 384 Bandwidth: 2× 240 GB&#x2F;s Single precision: 5591–8736 GFLOPS (MAD or FMA) Double precision: 1864–2912 GFLOPS (FMA) CUDA compute ability: 3.7 </code></pre> Is that a good deal for $1&#x2F;hour? (I&#x27;m not sure if a p2.large instance corresponds to use of one K80 or half of it)<p>How much would it cost to &quot;train&quot; ImageNet using such instances? Or perhaps another standard DDN task for which the data is openly available?<p>______<p>[1] <a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Nvidia_Tesla#cite_ref-19" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Nvidia_Tesla#cite_ref-19</a>
ajaimkover 8 years ago
Priced this config (or close enough) on <a href="http:&#x2F;&#x2F;www.thinkmate.com&#x2F;system&#x2F;gpx-xt24-2460v3-8gpu" rel="nofollow">http:&#x2F;&#x2F;www.thinkmate.com&#x2F;system&#x2F;gpx-xt24-2460v3-8gpu</a><p>Comes to just under $50,000 for the server or roughly 4.5 months @ $14.40
评论 #12613546 未加载
ravenstineover 8 years ago
Sounds like a great way to build a custom render farm. My home comouter has a dirt cheap GPU but it works well enough for basic modeling &amp; animation. Terrible for rendering, though. I&#x27;ve been thinking of using ECS to build a cluster of renderers for Maya that I can spin up when needed and scale to the appropriate size. I don&#x27;t know for certain if it&#x27;s cheaper than going with a service, but it sounds like it is(render farm subscriptions cost hundreds), and I would get complete control over the software being used. I am glad to hear that Amazon is doing this. Granted, I&#x27;m more of a hobbyist in this arena, so maybe it wouldn&#x27;t work for someone more serious about creating graphics.
spullaraover 8 years ago
It is interesting to compare this to NVidia&#x27;s DGX-1 system. That server is based on the new Tesla P100 and uses NVLink rather than PCIe (about 10x faster). It boasts about 170 Tflops vs the p2.16xlarge&#x27;s 64 Tflops. If you run the p2.16xlarge full time for a year it would cost about the same as buying a DGX-1. Presumably Amazon releases their GPU instances on older hardware for cost savings.<p><a href="http:&#x2F;&#x2F;www.nvidia.com&#x2F;object&#x2F;deep-learning-system.html" rel="nofollow">http:&#x2F;&#x2F;www.nvidia.com&#x2F;object&#x2F;deep-learning-system.html</a>
cm2187over 8 years ago
Stupid question: are GPU safe to be shared by two tenants in a datacenter? I read previously that there are very few security mechanisms in GPU, in particular that the memory is full of interesting garbage. So I would assume no hardware enforced separation between the VMs too.
评论 #12611099 未加载
评论 #12614117 未加载
clishemover 8 years ago
Pricing?
评论 #12611087 未加载
epberryover 8 years ago
Yess thank you, thank you, thank you. I was just signing up for the Azure N Series preview but we&#x27;re good to go now :).
nik736over 8 years ago
Anyone knows if video transcoding on GPUs (with FFMPEG) is viable nowadays? If yes, what are the gains?
评论 #12611264 未加载
评论 #12612251 未加载
评论 #12611412 未加载
asendraover 8 years ago
Damn, I would love to have an excuse to play around with this.