WEBVTT

00:00:00.080 --> 00:00:06.799
we've had high-end desktop or hedt processors for quite a while now you

00:00:04.319 --> 00:00:11.280
know those cpus that sit at the very top of the food chain that straddle the line

00:00:08.800 --> 00:00:15.440
between home pc and workstation with a high price tag to match

00:00:13.040 --> 00:00:20.080
and for a long time Intel dominated that market segment with its extreme edition

00:00:17.440 --> 00:00:26.880
and x-series processors but AMD has since fired back in a big way literally

00:00:24.240 --> 00:00:30.320
this is the AMD threadripper and i mean look at the size of it it's practically

00:00:28.640 --> 00:00:33.520
a coaster you can put your drink on this thing and save your table but other than

00:00:32.000 --> 00:00:38.879
leaving your neighbor's desktop with a feeling of inadequacy is it right for

00:00:36.640 --> 00:00:42.800
you and what should you know about it let's start out by looking at something

00:00:40.480 --> 00:00:48.160
straightforward core count although typical ryzen desktop chips have between

00:00:44.960 --> 00:00:51.360
4 and 16 cores threadripper cpus start

00:00:48.160 --> 00:00:53.199
at 8 cores and go all the way up to 64

00:00:51.360 --> 00:00:57.520
cores meaning that the chips are heavily marketed toward content creators who

00:00:55.120 --> 00:01:03.120
need fast rendering times and they all have simultaneous multi-threading so

00:00:59.760 --> 00:01:05.040
we're talking 128 threads on just one

00:01:03.120 --> 00:01:10.159
socket speaking of sockets it's pretty clear that a chip this chunky won't fit

00:01:08.080 --> 00:01:16.640
into your standard am4 motherboard instead threadripper uses the tr4 and

00:01:12.720 --> 00:01:19.119
str4 sockets which each have over 4 000

00:01:16.640 --> 00:01:24.159
contacts in case you're wondering AMD switched to str4 with the threadripper

00:01:21.600 --> 00:01:29.200
3000 series and despite looking the same as the older tr4 this two sockets are

00:01:27.280 --> 00:01:33.119
not compatible but aside from having more cores and a bigger footprint what's

00:01:31.360 --> 00:01:37.600
going on under the hood that really separates it from a normal ryzen chip

00:01:35.680 --> 00:01:42.560
well because threadripper is an enthusiast level lineup it's not just

00:01:39.920 --> 00:01:46.399
core count that's important it's also i o Threadripper trips have a lot more

00:01:44.640 --> 00:01:51.920
pci express lanes than the ryzen counterparts the first two generations

00:01:48.479 --> 00:01:54.399
had a full 60 PCIe 3.0 lanes coming

00:01:51.920 --> 00:01:59.119
directly off the chip with 48 linked to the slots on your motherboard enabling

00:01:56.439 --> 00:02:03.040
multi-GPU and the other 12 reserved for storage if you're using a 3000 series

00:02:01.600 --> 00:02:09.039
threadripper the number of lanes is bumped up to 64. and you get the new pci

00:02:06.240 --> 00:02:12.959
express 4.0 revision as well even if you're not into multi-GPU and hey we're

00:02:11.440 --> 00:02:17.599
not going to blame you if you're not this can still be useful if you're into

00:02:15.200 --> 00:02:21.599
virtualizing you need tons of storage or you're using something like fpga for

00:02:19.680 --> 00:02:25.840
crypto mining threadripper also features a larger level three cache than

00:02:23.840 --> 00:02:30.160
mainstream ryzen chips to keep all those cores fed and reduce cache misses during

00:02:28.400 --> 00:02:34.319
demanding workloads and if you're not sure what a cache miss is you can learn

00:02:32.319 --> 00:02:38.720
all about the importance of CPU cache in this video additionally all threadripper

00:02:36.959 --> 00:02:42.560
chips support quad channel RAM for performance boost as well as ecc or

00:02:40.720 --> 00:02:46.879
error correcting memory which can prevent crashes helpful if you're doing

00:02:44.640 --> 00:02:50.560
really critical work on your pc or using it to host a server we'd also be remiss

00:02:48.800 --> 00:02:56.040
if we didn't mention the threadripper pro lineup for actual workstations which

00:02:53.360 --> 00:03:01.200
uses yet another socket called swrx8 and will give you support for

00:02:58.480 --> 00:03:05.840
octa-channel memory but hold on a sec surely chips this large and powerful

00:03:03.440 --> 00:03:10.560
have to have some kind of downside right predictably threadripper cpus cost a lot

00:03:08.560 --> 00:03:16.640
more money when threadripper debuted back in 2017 the lower cost skew was 549

00:03:14.879 --> 00:03:20.720
and the current lineup at the time we shot this video starts at 1

00:03:19.159 --> 00:03:25.920
399 for a 24 core model and it goes all the

00:03:23.280 --> 00:03:29.599
way up into 4 grand if you want a 64 core model however although this is

00:03:28.000 --> 00:03:33.519
obviously a lot of money threadripper has had success beating Intel setups

00:03:31.599 --> 00:03:37.840
that cost around five times as much in benchmarks of course high cost isn't the

00:03:35.840 --> 00:03:41.519
only thing to look out for threadripper cpus are a little trickier to install on

00:03:40.000 --> 00:03:46.319
a motherboard than you might be used to and their higher tdps up to 280 watts

00:03:44.480 --> 00:03:51.440
mean you'll need a cooler that's up to task but we wouldn't recommend setting a

00:03:48.640 --> 00:03:55.280
glass of ice water on it even if it is the size of a coaster this episode is

00:03:53.680 --> 00:03:58.239
brought to you by curiositystream curiosity stream is a subscription

00:03:56.959 --> 00:04:02.640
streaming service that features thousands of streamable documentaries

00:04:00.400 --> 00:04:07.040
and non-fiction tv shows they offer award-winning exclusives and originals

00:04:04.640 --> 00:04:10.560
on topics like history nature science food technology travel and more one of

00:04:09.200 --> 00:04:14.720
the documentaries that's popular around here is the joy of ai where professor

00:04:12.560 --> 00:04:18.560
jim al-khalili tells the story of the pursuit of ai the emergence of machine

00:04:16.799 --> 00:04:22.000
learning and the recent breakthroughs brought about by artificial neural

00:04:20.079 --> 00:04:25.520
networks curiosity stream is available on many platforms including Android and

00:04:23.840 --> 00:04:30.080
iOS you can get unlimited access starting at just 2.99 a month or 19.99 a

00:04:27.919 --> 00:04:33.759
year and for our audience the first 30 days are completely free if you sign up

00:04:32.000 --> 00:04:37.520
using the link in the description and use the promo code techwiki during sign

00:04:36.000 --> 00:04:41.919
up so thanks for watching guys if you like this video like it subscribe it and

00:04:40.080 --> 00:04:46.560
hit us up in the comments section with your suggestions for videos that we

00:04:43.360 --> 00:04:46.560
should cover in the future