A REAL 64 Core CPU - For SCIENCE!

Linus Tech Tips ·Linus Tech Tips ·2018-05-06 · 1,306 words · ~6 min read
Floatplane YouTube

Transcript

JSON SRT VTT 116
0:00 64 cores
0:04 256 threads running at 1.3 to 1.5 gigahertz with 32
0:12 megabytes of cash meet
0:16 the knights landing zeon phi one of the most insane cpus in the world
0:22 how insane well how does a theoretical two and a
0:26 half teraflop sound got no context for that
0:31 don't stress it's like it's like a lot like
0:34 like like look how big this thing is i can take over the world with this sort
0:38 of power but first i need to pay the bills which
0:42 is why today's video is brought to you by tunnelbear tunnelware makes it easy
0:46 to privately and securely browse a more open internet try telebear for free at
0:50 tunnelbear.com LTT
1:02 so some background information is in order here as this is no ordinary CPU
1:07 knights landing is the current generation of xeon phi Intel's version
1:11 of what's referred to as a mini core processor which is exactly what it
1:16 sounds like it's built around the idea
1:19 of trading off per thread performance
1:22 for an obscene number of threads and maximum throughput
1:28 with the theory being that a simpler scaled down core design can dramatically
1:34 scale up in number while still utilizing existing tools
1:38 so we can actually trace knight's landings lineage back to project larabee
1:44 first unveiled in mid 2008 as a GPU
1:47 designed around the x86 architecture Intel even announced their intent to
1:53 release a consumer graphics card version of it back in 2010
1:57 but since nobody is talking about dedicated team
2:02 blue graphics cards today you should probably realize that that
2:06 never launched so xeon phi is what rose
2:10 from the ashes of that project first appearing in 2012 as a pci express
2:15 add-in card codename knight's corner
2:18 and it actually did end up getting used in what was the world's fastest
2:23 supercomputer until june 2016. the
2:26 newest revision though knight's landing looks a little different it launched
2:30 last year and unlike knight's corner it works exclusively in the lga 3647 socket
2:37 that high-end xeon pearly cpus also slot
2:41 into so adding support for avx-512 and
2:45 its many extensions gives knight's landing a lot of flexibility in its
2:50 target market so then who's the target market i'm glad you
2:54 asked uses for this include protein folding simulation
2:58 weather prediction ai and neural network
3:01 research and development and molecular simulation with libraries like
3:05 tensorflow and the uiuc's namd being
3:09 able to take advantage of it as well as
3:12 GPU acceleration okay then Linus
3:15 so why do you have one
3:20 um i guess because super micro pulled a rookie mistake and
3:25 sent it to me i mean obviously all i was gonna do was game on it and i'm actually
3:30 kidding about that at 1.3 gigahertz i can tell you guys
3:33 right now without firing up a single benchmark what the gaming performance
3:36 will look like like that
3:40 as for productivity though yeah right out of the gate it falls flat
3:45 on its face coming in at under half of
3:48 core i9s and nearly a third of threadripper's score in 7-zip and
3:53 cinebench CPU mark and real bench not looking much better even blender a
3:58 conventionally multi-core friendly benchmark sees terrible performance
4:04 given the 1900 price tag for the chip alone
4:09 so no these results won't do it all
4:13 xeon phi is clearly unlike anything we've ever tried to benchmark before
4:17 you'd need like a phd or something to properly evaluate this thing
4:21 which is where dr kinghorn our friend over at puget systems comes in as a
4:27 chemist and mathematician dr kinghorn is no stranger to high-level computation
4:31 like this and agreed to remote into our night's landing workstation and do some
4:36 of the testing for us while also running his own 14 core xeon 2690 v4 workstation
4:43 through the same suite to give us a point of comparison the results
4:48 actually pretty surprising in linpack the higher core clock of the
4:52 traditional xeon gives it an edge in smaller problem sizes but beyond 5000
4:58 the tables turn dramatically with the molecular dynamics library and
5:03 AMD things seem to scale predictably
5:06 with the number of atoms being simulated and all the tests wind up being far less
5:10 favorable to the xeon phi the greatest gains in this benchmark
5:14 actually come from dedicated compute gpus
5:18 though that is fine for the phi these days
5:21 since it sits in a socket now instead of taking up valuable PCIe slots
5:25 tensorflow neural networks bring things back though into the xeon phi's
5:30 wheelhouse especially in batch sizes larger than 64.
5:34 so xeon phi takes a significant lead over its traditional xeon cousin
5:38 but it should be noted that both of them get creamed by a GPU
5:44 further highlighting the importance of large banks of compute gpus for tasks
5:48 like these though that doesn't mean that xeon phi
5:52 is useless its dramatically better performance than a standard xeon makes
5:57 it great for supplementing GPU compute in workloads
6:00 like these which brings us neatly to the conclusion then
6:04 not quite as good as a GPU for tasks that can be accelerated
6:10 far better than a traditional xeon though in most
6:14 and finally utterly worthless for just about anything on the Windows desktop i
6:19 was actually surprised to see Windows even install we ended up needing a
6:22 server version of Windows to see all the cores but that is sort of irrelevant
6:26 anyway because this is intended for use in super computers where like
6:31 massive nodes of these things can all work
6:34 together to solve complex problems along with the banks of gpus that accompany
6:39 them i mean you're almost certainly not going to be buying anything like this
6:44 personally anytime soon unless you're like a you like to science for funsies
6:48 with that said though you yes you are very likely to benefit from the work
6:53 being done using hardware like this and we're glad we got the opportunity to
6:57 play with it it was fun speaking of things that are fun
7:02 it's fun to tell you guys about squarespace and thank them for
7:06 sponsoring this video squarespace has 24
7:09 7 support via live chat so anyone can create a beautiful
7:13 functional website in just a few minutes
7:17 if all you want is like uh a one-page online presence that's their cover pages
7:22 feature and they've got tons of other great features as well including their
7:25 logo designer the ability to publish in the apple news format and not to mention
7:31 that the whole creation tool is cloud-based so it's always up you can
7:35 access it from anywhere and updating your site is as simple as like dragging
7:38 over some pictures clickity-clacking some text and you are ready to rock
7:42 squarespace starts at just 12 bucks a month and you can start a trial with no
7:46 credit card required by heading to the link in the video description
7:49 squarespace.com forward slash LTT and when you decide to sign up for
7:53 squarespace forever you can use offer code LTT to get 10 off
7:58 on your first purchase so thanks for watching guys if you
8:02 dislike this video you can hit that button but if you liked it hit like get
8:05 subscribed maybe consider checking out where to buy the stuff we featured
8:10 if you like work at a university or something
8:13 it's like i'll know that's what happened if like you know 2 000 xeon 5's show up
8:18 on like our amazon report or whatever anyway the point is also in the
8:22 description is our merch store as well as our community forum which you should
8:25 totally join