all right so we are back with the
results this is part three of our series
of creating an RX 480 hybrid this thing
I would say I reasonably liked the look
of the 1080 hybrid when we made it this
thing looks legitimately somewhat like a
monster but it worked pretty well and
that's what we're here to talk about so
first of all this coverage is brought to
you by Origin PC and their new origin
Kronos which is customizable upgradable
and ships with the R X 480 video card in
that card is the one we're talking out
here but we've changed it quite a bit so
this is part of the original card this
is part of some garbage this is part of
the original card and and then this is
the faceplate we took all this off quite
obviously and built the hybrid version
using an Arctic Exelero hybrid 3 which
said I just bought a bunch of these and
I hope they would work and that one did
work thankfully so what we've got here
in terms of the final bill just very
quick catching you up there's big mass
on the table that's because we were
doing all kinds of stuff like adding
these individual aluminum vrm heat sinks
we added vram heat sinks and these are
these are pretty cheap I one of you
commented and pointed me to some really
good vrm or a VRAM heat spreaders that
we'll use next time but these did okay
for the job that they had and then we've
got this pivotable fan that's just
positioned over the vrm and that
dissipates heat out of all these these
alloys sinks that are in there and those
are positioned on the MOSFETs chokes are
under there they're obscured they're not
cool they don't need to be the GPU is
cooled by this liquid pump this is an
ASA tech pump it's slightly different
from the EVGA hybrid that we have in the
cold plate only but this still uses a
flat contact cold plate which is what
you want for GPUs because GPU the the
silicon is just basically perfectly flat
as opposed to a cpu see I'll see like
855 which is on the Seahawk where you'll
see reasonably worse performance
actually significantly worse is on with
2x different than than the hybrid but
still good better than the
but you see that they're ones because
it's using a cpu CLC and those aren't
built for it for GPUs and hotspots or
different things like that so we built
this I've made some changes to this
between the second video in this one one
of those we position the additional 120
millimeter fan up here and clamped it to
the expansion slot blow air down and
keeps the the VRM cooler and then I also
after some initial testing I changed
this fan which is this one from being
plugged into the video card directly to
being plugged into the power supply and
that's to improve stability it gives us
a few extra watts for overclocking and
the detection wasn't working anyway so
the sensor thought that the fan was
spinning at 5 million rpm it would go
between two thousand and five million
so obviously the sensor was wrong the
vrm fan was not really adjusting at
speeds as necessary so I just put it on
max blast and let it go so that's that's
what we got for the build and now we're
just gonna fly through the results
somewhat quickly in view more
information in the original rx 40 review
if you want the full analysis of how
things work and why they work the way
they do but we're just gonna go over
thermals and stuff like that here
alright so let's start with the thermal
differences we measure these numbers in
delta T over ambient and that means you
should add in your own ambient
temperature to get the absolute
temperature value for your environment
but we subtract our ambient so you can
do that so with the hybrid three cooler
we've got it's running on out of box
settings for the 480 just with a better
cooler we're able to hit a massively
improved twenty 3.0 to Celsius Delta T
value for the load temperatures the
stock cooler performed at fifty 6.33
Celsius so that's an eighty three point
nine percent difference that's not as
huge as our 102 percent difference on
the GTX 1080 hybrid but that's for a few
reasons one the founders edition cooler
and the GP 104 chip runs a little bit
hotter the founders edition cooler is
better built than this one but the chip
is hotter and larger so that's one
reason we see that and then to the EVGA
hybrid solution does use that extrusion
on the copper the cold plate which also
produces some of the difference we see
this hybrid
three solution keeps the am the Rx for
80 thermals within 40 to 50 Celsius if
you're looking at the absolute value
when running the stock clock rather than
the 80 to 85 Celsius area of the
original cooler absolute values just for
sake of visualizing the data differently
here's a look at the thermals overtime
chart this first chart is a little
complex and includes several video cards
but you can clearly see where the two
hybrid experiments beat out the
competition and if we simplify that
chart you can see this version contains
only 10 80 1070 fe cards are X 480 stock
and the hybrid versions of those cards
the low idle temperature of the our X
480 hybrid is a mix of the new 14
nanometer FinFET process power reduction
methods by AMD to reduce idle
consumption and then the rear-mounted
heatsink that Arctic cooling includes
with it's cooler and also the two fans
we use so this one and the one mounted
above it do help keep idle temperatures
down since they're blowing directly on
the card and this is that VRAM backplate
heatsink we've got testing a liquid
cooled card means that there's one big
note and that's that we should test
based on thermals more than just raw fan
speed percentage because we've hacked
this fan together and we're no longer
using the onboard vrm fan controller and
we're running at such low temperatures
that you can actually reduce your noise
by lowering the fan RPMs on this hybrid
card and that in exchange slightly
increases your thermals but lowers your
DVI output so we'd still be well under
the stock reference cooler in terms of
noise and thermals by doing that so
manually running these fans at their
lowest possible speeds produces the
temperatures that will show on the
screen now and that is what allows us to
drop the noise levels were still
underneath the original cooler thermals
and that's because this thing is really
not that good of a cooler and as part of
why we were hitting the 80 plus range
absolute values so that's that's how you
have to look at this if you're actually
going to build a hybrid card whether
it's this or something else the the main
thing to look at is would I rather have
really crazy low temperatures with a
noise level that's about the same as a
normal card sort of sort of
before pushing to higher fan RPMs or
would I rather increase my temperature
slightly
still keep it below the original card
but have lower noise levels so
overclocking is another big thing with
the liquid cooled cards I was hoping to
get a couple extra megahertz at least
out of this thing and I even changed
around the fan plugged into the board
versus plugged into the motherboard to
see if an extra few watts would really
help and it didn't so what we ended up
with was the overclock stepping table
that's on the screen now previously we
got stuck at 13 40 megahertz core and
2200 megahertz memory and I was forced
to run the fan at around 4000 rpm which
just kept it at 89 Celsius and with the
giana hybrid solution we ended up at 13
90 megahertz and stuck with the same at
2200 megahertz memory clock I tried
pushing the core up to 1400 megahertz
and it did survive the initial 5 minute
run but it failed right around the 5
minute mark so no way did it survive any
kind of endurance this is when unplug
the vrm fan from the card and hope to
get some more stability amped up the RPM
of all the fans in the system to improve
cooling even of the vrm but there is no
luck we cannot unfortunately claim well
we can claim to have hit 1400 megahertz
it was there for 5 minutes but it's not
realistically possible with this
particular card that's not to say you
can't do it but we couldn't do it on
this card now one thing you can hit and
exceed 14 hundred megahertz when testing
with synthetic applications like fir
marker combustor and that's something
we've shown in the past but just because
we're hitting those higher frequencies
in these synthetic applications I don't
count those numbers as valid because it
doesn't survive a real game like The
Witcher 3 or Grand Theft Auto or shadow
of Mordor games which will stress the
GPU in different ways multiple different
ways so you have this boosting and all
this other stuff if it doesn't survive
real gaming then we don't claim that it
worked
so we were at 13 90 megahertz as a 50
megahertz boost over the original core
overclock that was at 13 40 megahertz
and our memory we could not increase
past the 2200 mark so it got stuck at
2200 cannot go to 2250 and
maude just for sort of reference did see
the voltage peaking pretty high I was
hitting 1.15 volts on this thing which
is very high for a GPU and that was as
opposed to the 1.125 and maximally 1.13
volts that we were seeing previously so
it did actually push a little bit higher
now in terms of power drop
according to gpu-z the board power total
was 192 dot 2 watts and that is a good
bit increased over the 170 180 watt
range we've seen on the 13 40 megahertz
overclock I'm not 100% sure what's going
on with the power draw of this card yet
I'm sure many of you have seen the
discussion online we're researching that
talked about that more later all right
so what's the frame rate impact from
this extra 50 mega Hertz and there's
really not much of one in GTA 5 we're
still seeing bad stuttering issues that
we discovered in these 16 not 6.2
pressed drivers for the RX 480
but the FPS is overall marginally
improved we're now at ninety one point
five instead of eighty nine point five
FPS for 1080 ultra over four passes and
that's a gain of two point two three
percent it's not a linear game though
because our original OC had us at a
seven point two six percent FPS gain
from the stock clock the same non linear
gain is mapped across shadow of mordor
where we saw a performance jump of three
point seven percent versus the stock to
13 40 megahertz jump of seven point nine
percent and that's mostly because when I
overclocked this I used watt man and I
forced the frequency to run exactly the
same at all states of operation so we
eliminated boosting that means that in
these games when when going from
something like the stock 1266 boost to
thirteen forty it's more than just a 74
megahertz jump because it's not just
1266 the baseline actually goes much
lower than that when the card sort of D
clocks itself for various circumstances
so that's why we see those big gains in
our original overclock and mirrors edge
also had some nonlinear gains as you can
see on the screen now we had a fall from
10% performance improvements to much
less than that so as far as the hybrid
project for the rx 40 the kind of
takeaways here and
do a part four of this four more that
we've learned takeaways here we are
power limited right now doesn't seem
like we're voltage limit if we have more
voltage than we need and that could be a
bad thing for long-term use but we are
certainly power limited on the six pin
header there's not much I can do about
that other than waiting for a RBE
partner cards and the GP is obviously
cooled significantly better this cooler
I am exceedingly unimpressed with the
card is a $240 card it has a fairly
power efficient in theory architecture
but that's kind of being debated right
now and the reason we see those high
thermals on the stock card is because of
this thing which is not a good cooler so
the AIB partner cards should be much
better now one other thing we could have
actually left this on even if it
involves zip ties but I think we could
have left this on completely but I did
not want to because I wanted to try out
these separate VRAM heat sinks I wanted
to get this on there without any issue
of running out of screw length or
anything like that or just things
bumping into each other and clearance
issues and I want to try it as is so
this arctic cooler the what is the
hybrid three that we bought I'm actually
a pretty happy with how this went
I really enjoyed putting together I'll
probably do a full review on this thing
in the future which I really wasn't
planning to do but I did like working
with it if you saw part two it was
exceedingly difficult to get the pump
and the cold plate and the block
attached to the back plate and I had to
recruit help for that because it's you
need four or five hands to get those
screws through but it was a fun project
performance gains you get an extra 50
megahertz absolutely not worth it for
frame rates and that's pretty much how
it always goes with this overclocking
totally not were there frame rates
definitely worth it for thermals if you
feel like doing something like this I do
not like the reference cooler and this
resolved that issue in terms of noise
were marginally better until you start
getting into the high-end fan speeds so
if you're doing even basic overclocking
the noise does reduce significantly
because we're going from four thousand
five thousand rpm fans beads which are
55 to 60 decibels for toad
the noise down to a much more reasonable
fan speed and relying on liquid instead
to do all the the bulk of the quote so
that's the project as always hit the
link in the description below for more
information subscribe because we're
gonna do a lot more of these including a
part four of this with what we learned
and that will come up shortly but we do
have crossfire testing coming out before
that probably so as always patreon link
post roll video comment below for more
discussion thanks for watching I'll see
you all next time
We are a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for us to earn fees by linking to Amazon.com and affiliated sites.