WEBVTT

00:00:00.320 --> 00:00:08.559
it all started out innocently enough in late june of this year we asked AMD for

00:00:05.200 --> 00:00:10.160
a sample of the vega frontier edition we

00:00:08.559 --> 00:00:16.080
were hoping that we could compare it against the titan xp and gtx 1080 ti

00:00:13.679 --> 00:00:21.600
that back then cost a bit more and a bit less respectively and also use it to

00:00:19.039 --> 00:00:25.840
guestimate the performance of the then upcoming rx vega

00:00:24.480 --> 00:00:29.279
so why are we finally publishing this video

00:00:28.160 --> 00:00:35.680
today well buckle up my friends because it's kind

00:00:33.360 --> 00:00:40.480
of a long story and that story is brought to you by

00:00:37.520 --> 00:00:44.079
tunnelbear try tunnelbear for free with no credit card required at

00:00:42.440 --> 00:00:56.559
tunnelbear.com LTT it's the easy to use vpn

00:00:56.559 --> 00:01:03.600
so AMD declined our sample request for frontier edition that happens sometimes

00:01:01.920 --> 00:01:08.000
and it's fine now sometimes companies do this because

00:01:05.600 --> 00:01:12.560
they don't want press looking too closely at a product but other times

00:01:10.000 --> 00:01:18.159
it's for perfectly legitimate reasons either way it piqued our curiosity so we

00:01:15.600 --> 00:01:23.920
put on our big boy pants ordered one for ourselves and a week later tore it out

00:01:20.880 --> 00:01:27.040
of its box to bask in its blue brushed

00:01:23.920 --> 00:01:28.560
aluminum and tacky yellow glowing

00:01:27.040 --> 00:01:34.000
glory but what else is there in the box well

00:01:31.439 --> 00:01:38.479
there's some foam and this weird booklet of like kickstartery promotional

00:01:36.240 --> 00:01:42.720
material that's strangely for the product that we just already bought i

00:01:40.880 --> 00:01:46.720
mean look at this stuff oh and

00:01:44.159 --> 00:01:49.840
they are people who boundaries as starting lines did anyone even proofread

00:01:49.119 --> 00:01:56.640
this the box might not seem important but this was actually a pretty pivotal point

00:01:54.240 --> 00:01:59.920
in our experience with the frontier edition

00:01:57.600 --> 00:02:04.159
because this was when we realized just how rushed and poorly thought out this

00:02:02.719 --> 00:02:09.599
product was and it actually went downhill from there

00:02:06.799 --> 00:02:14.720
we set up our closed GPU test bench downloaded the only available driver

00:02:12.640 --> 00:02:18.800
that was on AMD's site up until it finally got an update over three months

00:02:16.800 --> 00:02:23.040
later in october and fired up our blender test suite where vega ended up

00:02:20.800 --> 00:02:27.680
on the bottom in the bmw and classroom tests then all the way on top in the

00:02:25.200 --> 00:02:32.800
fishy cat test making it at best inconsistent and at worst

00:02:31.200 --> 00:02:38.480
oh so at this point we ddu'ed and

00:02:35.440 --> 00:02:40.640
reinstalled the driver but no joy there

00:02:38.480 --> 00:02:45.599
and even tried switching to gaming mode in the dual mode driver which

00:02:42.800 --> 00:02:49.120
also didn't help okay then

00:02:46.879 --> 00:02:54.879
so since we're in gaming mode what about gaming unigin's super position and 3d

00:02:52.160 --> 00:02:59.840
mark score saw it get absolutely spanked by the 700 1080 ti

00:02:57.760 --> 00:03:05.120
and then the butt paddling continued in both deus ex mankind divided and rise of

00:03:02.400 --> 00:03:09.280
the tomb raider at 4k in both directx 11 and directx 12 modes so between the

00:03:07.680 --> 00:03:15.120
performance the stability the thermal throttling and the power

00:03:12.879 --> 00:03:21.280
consumption the whole experience up to that point had frankly been super balls

00:03:19.040 --> 00:03:26.239
at this point we were certain of just two things one was that vega frontier

00:03:23.840 --> 00:03:31.280
edition existed only to cater to hardcore AMD fans

00:03:28.720 --> 00:03:34.480
with what felt like it was an overpriced slap in the face

00:03:32.799 --> 00:03:40.959
and two that rx vega was going to be a power

00:03:37.599 --> 00:03:43.280
hungry dog at the rumored pricing

00:03:40.959 --> 00:03:48.400
now throughout this we hadn't heard back from AMD about our problems so we forged

00:03:45.920 --> 00:03:52.319
ahead and then just as our video was getting ready to be published actually

00:03:50.239 --> 00:03:56.959
while i was down at AMD checking out their vega supercomputer demo they told

00:03:54.879 --> 00:04:02.879
us well hey guys we think it's unfair of you to judge

00:03:59.599 --> 00:04:05.519
vega's gaming performance on frontier

00:04:02.879 --> 00:04:10.560
because raja kaduri posted in a reddit thread somewhere that it wasn't for

00:04:07.439 --> 00:04:13.120
gaming and that our benchmark suite

00:04:10.560 --> 00:04:18.160
wasn't showing frontier strength because workloads like blender didn't benefit

00:04:15.439 --> 00:04:22.000
from its workstation dna and we didn't include enough scientific performance

00:04:20.479 --> 00:04:26.880
analysis so what they offered was to give us some

00:04:24.320 --> 00:04:31.199
help with the second part of it if we held off publishing the video to ensure

00:04:29.040 --> 00:04:36.720
we were doing it properly okay fair enough then i guess but weeks

00:04:34.240 --> 00:04:40.400
and then months passed and that help didn't materialize then while we sat

00:04:39.600 --> 00:04:45.199
there waiting rx vega launched with

00:04:42.720 --> 00:04:50.800
performance that put frontier neck and neck with vega 64 exactly as we'd

00:04:48.400 --> 00:04:56.000
expected and laid out in our original script for this video and it makes

00:04:52.960 --> 00:04:58.880
perfect sense other than color the only

00:04:56.000 --> 00:05:04.400
differences between them are the driver the frame buffer size frontier edition's

00:05:01.520 --> 00:05:08.560
higher bass and boost clocks and a locked maximum fan speed of 40

00:05:07.840 --> 00:05:15.600
so i'm not gonna mince words we felt cheated as a customer remember we bought

00:05:12.960 --> 00:05:21.840
this stupid thing for a thousand dollars and manipulated as press because it

00:05:18.400 --> 00:05:24.639
seemed as though our video got buried to

00:05:21.840 --> 00:05:29.039
prevent the negative pr from raining on rx vegas parade

00:05:26.720 --> 00:05:32.960
and we made these feelings clear to AMD who renewed their commitment to helping

00:05:30.960 --> 00:05:36.320
us benchmark frontier edition properly but

00:05:34.639 --> 00:05:41.199
in the back of our minds what we wondered was if they were just

00:05:38.800 --> 00:05:46.080
hoping that we would just forget about publishing a video about a months old

00:05:43.360 --> 00:05:49.830
and now very irrelevant product and focus on their shiny new wx 9100

00:05:48.720 --> 00:05:56.080
workstation card no such luck

00:05:52.720 --> 00:05:58.400
so then for the first of two AMD

00:05:56.080 --> 00:06:05.199
recommended benchmarks we'll be looking at spec uprf 12.1 a collection of 3ds

00:06:01.919 --> 00:06:08.000
max maya solidworks and other workloads

00:06:05.199 --> 00:06:10.800
for which AMD says frontier edition is optimized

00:06:09.520 --> 00:06:17.360
and actually okay frontier edition does pull out

00:06:13.039 --> 00:06:19.520
ahead of vega 64 and even gtx 1080 ti in

00:06:17.360 --> 00:06:24.800
particular wherever edges are used in the rendering pipeline and in the memory

00:06:22.319 --> 00:06:30.000
shredding scientific energy and medical tests

00:06:26.240 --> 00:06:31.759
the titan xp though blows them all away

00:06:30.000 --> 00:06:36.400
to see how much of this performance comes from the workstation driver path

00:06:34.080 --> 00:06:42.080
we switched the driver mode to gaming where interestingly frontier ends up

00:06:38.880 --> 00:06:44.639
well below vega 64 levels of performance

00:06:42.080 --> 00:06:48.720
so it's clear that unless your loads specifically benefit from

00:06:47.360 --> 00:06:53.759
gigs of vram any advantage to frontier is purely a

00:06:52.080 --> 00:06:58.560
software switch though to be clear we aren't singling

00:06:56.160 --> 00:07:03.039
out AMD for this behavior team green's titan and quadro lineups are

00:07:00.560 --> 00:07:07.120
differentiated from g-force in a very similar manner

00:07:04.800 --> 00:07:12.560
and to AMD's credit frontier's positioning as a sort of workstation

00:07:09.680 --> 00:07:17.759
card like the original titan sparked a response from NVIDIA to give titan xp

00:07:15.199 --> 00:07:23.680
back some of the professional features that they had removed over the years so

00:07:21.599 --> 00:07:29.280
kudos for that the second of our two AMD recommended

00:07:26.160 --> 00:07:33.039
benchmarks monero mining uses the vega

00:07:29.280 --> 00:07:35.520
optimized cast xmr miner and AMD's beta

00:07:33.039 --> 00:07:40.080
blockchain optimized driver and here we are actually greeted with

00:07:37.520 --> 00:07:44.479
excellent results of up to 1.85 kilo hashes per second

00:07:41.759 --> 00:07:48.479
though that quickly drops to 1.5 once thermal throttling kicks in

00:07:46.240 --> 00:07:54.000
i do believe a driver fix unlocking fan speed is coming but until then even a

00:07:51.280 --> 00:07:57.919
vega 56 outperforms frontier edition significantly

00:07:56.319 --> 00:08:01.599
so then is vega frontier edition a legitimate

00:08:00.400 --> 00:08:08.960
product or was it like an early access game with

00:08:05.199 --> 00:08:11.919
day one dlc a beta where you pay AMD for

00:08:08.960 --> 00:08:16.240
the privilege of reporting bugs and third-party software incompatibilities

00:08:14.080 --> 00:08:23.360
like those blender crashes while they polish the real product wx9100

00:08:20.319 --> 00:08:26.160
on paper frontier kind of makes sense it

00:08:23.360 --> 00:08:30.960
gets more RAM and some workstation features compared to a consumer card and

00:08:28.560 --> 00:08:35.039
that's for half of the price of the higher end workstation card that gives

00:08:32.719 --> 00:08:40.399
you ecc memory support frame lock gen lock stereo 3d and far more professional

00:08:38.080 --> 00:08:45.440
application certifications but in the real world wx 9100 street

00:08:43.519 --> 00:08:50.320
pricing has fallen to just fifteen hundred dollars already for the

00:08:47.839 --> 00:08:55.760
full-fledged vega experience and frontier's actual value add over a

00:08:52.480 --> 00:08:57.200
regular gaming card is pretty limited

00:08:55.760 --> 00:09:02.320
leaving us in no man's land unless we're concerned

00:08:59.120 --> 00:09:05.279
with and only with some pretty specific

00:09:02.320 --> 00:09:08.399
use cases i mean even AMD only managed to come up with one

00:09:07.200 --> 00:09:12.560
benchmark that actually demonstrated its value for

00:09:10.800 --> 00:09:16.640
us so then

00:09:13.839 --> 00:09:23.839
if AMD had positioned frontier as a developer kit like the oculus dk1 or dk2

00:09:20.560 --> 00:09:26.720
and even if it was paid and expensive

00:09:23.839 --> 00:09:31.440
but available to developers like through their website we wouldn't be having this

00:09:28.800 --> 00:09:36.000
conversation but as it is whatever they posted on reddit months

00:09:33.279 --> 00:09:42.000
ago their actions were to put this product on sale to consumers in the

00:09:39.360 --> 00:09:46.720
consumer channel even going as far as to bundle quake champions crap with the

00:09:44.160 --> 00:09:50.480
water cooled version get upset when we wanted to talk about why it was a stupid

00:09:48.800 --> 00:09:54.720
product for consumers then come up with only a couple of

00:09:52.640 --> 00:10:00.160
pretty flimsy reasons why we shouldn't be pissed about the performance of our

00:09:57.279 --> 00:10:05.519
castrated wx that we bought instead of a titan xp which incidentally would game

00:10:03.200 --> 00:10:10.720
better in our spare time and consume less power

00:10:08.000 --> 00:10:16.000
smells like early access to me i think the right way to fix this is for

00:10:12.959 --> 00:10:18.320
AMD to just discontinue this thing and

00:10:16.000 --> 00:10:23.360
even if it doesn't help the AMD enthusiast gamers out there who got

00:10:20.160 --> 00:10:27.040
sucked into it turn around and unlock as

00:10:23.360 --> 00:10:30.079
many of the wx 9100 features as they can

00:10:27.040 --> 00:10:32.880
for the people out there who have them

00:10:30.079 --> 00:10:37.279
what do you guys think we really do want to hear from you in the comments

00:10:35.279 --> 00:10:41.040
mass drop is a great place to get quality products at amazing prices and

00:10:39.760 --> 00:10:45.120
this year for the holidays they're making it easier than ever to find the

00:10:43.040 --> 00:10:48.560
perfect gift they've got a collection of over 20

00:10:46.480 --> 00:10:52.240
of their most popular mass drop made products that are available to ship in

00:10:50.399 --> 00:10:58.320
two to three days so they'll make it to you by christmas one of the products is

00:10:54.800 --> 00:11:00.240
the mass drop x akg k7xx audiophile

00:10:58.320 --> 00:11:04.399
headphones they're lightweight they're comfortable they perform way above their

00:11:02.480 --> 00:11:08.480
price point in terms of sound and they are highly recommended check them out

00:11:06.640 --> 00:11:12.399
along with the entire massdrop holiday gift collection at the link in the video

00:11:10.480 --> 00:11:16.000
description so thanks for watching guys if this

00:11:14.240 --> 00:11:19.519
video sucked you know what to do but if it was awesome get subscribed hit that

00:11:18.079 --> 00:11:22.959
like button or check out the link to where to buy the stuff we feature

00:11:23.680 --> 00:11:28.720
in the video description also linked in the description is our merch store which

00:11:26.959 --> 00:11:33.519
has cool shirts like this one and our community forum which you should totally

00:11:30.640 --> 00:11:33.519
join