MSI GTX 1080 Ti Gaming X Review & Thermal Analysis
MSI GTX 1080 Ti Gaming X Review & Thermal Analysis
2017-04-13
ms:i at $750 gtx 980ti gaming ex is one
of the most overbuilt cards we've looked
at lately particularly for its seemingly
excessive thermal solution and use of
16v core vrm dual set packages from
fairchild specifically the FPPC at 5:01
8s geez
controlled by 8 driver ICS for an eight
phase-- BRM with four MOSFETs per phase
as builds are noted in our PC breakdown
of the card the BRM is capable of
sustaining a 250 amp load while
outputting just 15 watts of heat at 125
C even 400 amps would keep us to 40
watts of heat thanks to how much the
phase design spreads heat over a large
service area and high account of that
today we're reviewing the MSI 1080i
gaming exports thermal and gaming
capabilities before that this coverage
is brought to you by the Computex
conference which runs from May 30th to
June 3rd in Taipei Taiwan this year
Computex is the biggest event of the
year for PC hardware and technology
where we preview the newest prototypes
before they come to market
we highly recommend attending or
following this event online for industry
professionals and enthusiasts learn more
at the link in the description below
this generation of GT X 1080 TI cards
has gone big we have the gigabyte 1080i
Xtreme to review next already here and
that's multiple slots and then this one
is fat - it's got not just the normal
twin frozer cooling design on it with
the dual axial 100 millimeter blades but
it also has a much fatter thin stack and
denser array of those fins so the fin
density is greater the fin count
therefore is greater and it's also
spread across the entire card with
partitioned base plates that we'll talk
about momentarily so it's a big card and
that means it can deal with a lot of
heat even though the 1080i doesn't
necessarily output that much even
compared to a 1080 yes it's more yes the
power throughput can be quite high if
you really push the thin but this is a
whole lot of cooling potential so we'll
see how useful that actually is and how
it compares to the EFI model and to our
hybrid model in this review and then the
next review we'll look at the gaming
extreme card from gigabyte MSI ships the
1080 TI gaming X this card at $750 with
3 configurable OC profiles through there
software and those include the usual OSI
profile the gaming mode profile and the
silent profile and respectively
those o'clock at 1683 megahertz boost
with 1569 megahertz base for the OC
profile gaming mode runs ten megahertz
through twelve bit lower net 16:57
megahertz boost and 15 44 megahertz a
base so a bit lower and then silent 1582
and 1480 boost and base our DB LEDs are
present on the card if that's your thing
but we're more interested in the cooling
solution we already have a PCB breakdown
and a teardown of this card in separate
videos though to quickly recap we're
looking at dual 100 milliliter axial
fans as the set up atop the aluminum fin
stack that covers the entire card the
heatsink terminates in two cold plates
one connects the heat pipes and the left
half of the card to the GPU and the
other connects the FETs and the driver
IC to the thin stack via thermal pads
the inductors contact the fins via a
thermal pad and not dissimilar from the
update EVGA did to its FTW series last
year and the rest is mostly open airflow
the left side of the base plate contacts
the vram module directly which then
contact heat pipes this we think is a
noteworthy decision by MSI cooling vram
with an independent base plate and
contact the main heatsink is still
uncommon and something we saw explored
last year with EVGA FTW hybrid card via
the copper vram cooling plate that
connected to the CLC and gigabytes
extreme water cards there's potentially
an increase to GPU core temperature
readings by nature of saturating the
cooling interface and the cold plate
with more heat from neighboring
components but it's not usually a
significant increase and the decrease in
temperatures on the vram is noteworthy
and a more valuable trade-off an extra
couple degrees on the GPU time the back
of the card is enshrouded by a backplate
though no direct contact is made between
the backplate in the back of the PCB
which means it's less for cooling and
more for structural support on this card
that we will have temperature readings
without the backplate in the article
link in the description below and also
in the article below we'll have a few
things one the methodology for testing
as well as a few more charts so there's
a couple extra thermal charts and a
couple extra resolution game tests
charts in the article and secondarily a
couple of other notes on the testing
methods first of all these thermocouples
that are mounted to the card are pretty
good they are the pad that is adhesive
and connects to the card is one one
hundredths of an inch thick it is laser
thin and is designed and engineered
specifically so that doesn't really
interfere with temperature reading so
that's worth noting these are built for
this task that's what they do these
thermocouples the resolution is about
0.1 to C and the response time is about
0.15 seconds on the thermocouples used
here and then we have more information
on the chemical composition of the
actual adhesive pad if you really care
about it in the article below
everything else is defined there as well
including the drivers used for the
testing the settings of the games
although the charts will contains all
that and then the CPU the memory all
that stuff is in the article but let's
roll on to the thermal test first and
then look at some of the FPS starting
with BRM temperatures strictly compared
against the 1080 TI foundered edition
card we see in this chart that both the
gaming x and 1080 TI f e it lands their
vr m component temperatures and PCB
dioxide temperatures at about the same
place the FETs are well under any
actually concerning threshold which you
wouldn't start hitting until something
like 125 plus celsius anyway with the
1080 TI f e card landing just under 70
c4 FET seven represented by the orange
line or right around 65 c4 Fett to
represented by the red line so those
would be the feds one up from the bottom
and about dead center on each of these
cards the gaming X meanwhile shows its
fats both shades of blue at 72 71
Celsius for fat 7 and 65 to 74 fat to
keep in mind of these packages has
multiple sets for highside and low-side
and so they are quite dense and our PCB
analysis and vrm breakdown video has
more information on that if you are
curious moving now to a chart that plots
frequency and core temperature versus
time we get a chart that's fairly
complex and will be staged a few lines
at a time so it's basically three frames
of this chart for the first stage we're
looking at the clock rate of the GTX
1080p I found edition ie reference card
versus its
temperature now that the clock is
enumerated from a power virus scenario
which means that it is lower than you'd
see in a real game but the heat load is
significantly higher especially on the
vrm and that's what we care about here
for this test the clock is a bit spiky
we have a range of about a hundred
megahertz at times caused by continually
reaching the 84c threshold limiter and
dropping clocks to maintain that
temperature the spikes in frequency
often align with spikes in temperature
for this reason as they have an inverse
relationship let's now add the GN hybrid
card here which is just an Fe board with
an EVGA CLC added to it here we see the
hybrid card run the GPU temp at nearly
45 see where the FE card operated around
84 Celsius prior to our mod this could
be resolved of course with a higher fan
speed but that's not really the point of
today's content and we already talked
about that in our hybrid results video
if we look at the hybrids clock rates
though we can see that it is less spiky
than the clock of the reference cooler
version and this gives us the extra 4 to
9 percent performance in games that we
saw in our previous test depending on
the hive o'clock intensive they are
again that's an hour hybrid deep dive
but now let's add the gaming X charts
getting dense here but this is the last
iteration of it the gaming X is pre
overclocked hence the higher frequency
and it is also the flattest frequency
line we have the hybrid card removed the
thermal constraints that's flattening
the line but immediately ran into power
and voltage budget constraints by
limitations on the card itself the
gaming X will have those at some point
as well of course they all do especially
with Nvidia's limitations but they're
pushed back enough to help further
flatten the curve and then the pre OC
boosted a bit in the frequency
Department partly thanks to the massive
erm while keeping GPU temps with a 50%
fan speeds at around 68 Celsius this is
higher than the hybrid mod and lower
than the reference board which is really
all we need we're well below the thermal
limiter point on the clock and the clock
is smoother overall this means we should
expect a generally higher frame rate or
at least better frame times on the
gaming X than the hybrid but we'll see
how that performs in games granted the
FET temperatures aren't really any
different from the 1080i hybrid or the
EFI will at least have some char
showing how this all plays out in a
moment moving on to game tests you find
the rest of thermals in the article
below
we're including overclocked numbers in
some of our 4k or 1440p tests the table
on the screen now shows the overclocked
step in ultimately we have the step down
to seventy-five megahertz from a 100
megahertz core off set and 400 megahertz
memory offset to retain stability and
all games tested so while we were able
to push 100 megahertz core and 5 50
megahertz memory in fire strike those
were not stable in all the games there's
not a significant amount of overclocking
Headroom over the stock card and you'll
see how that plays out in the games
tested starting with Ghost Recon at 4k
the msi 1080i gaming ex places at 58 FPS
average with lows at 51 and 47 this
lands with 1080i gaming X right around
where our 1080i hybrid mod performed
this makes sense given that the hybrid
modern is the thermal limit of the EFI
card and provided an additional 5 to 6%
of clock Headroom prio see compared to
the 1080i reference card we're looking
at a gain of about four point nine
percent on the GX over the F II card the
Gaming Act didn't overclock in any
meaningful way and landed at 60fps
versus 58 for a change of about 3.4
percent we can have frame time plots
there's no meaningful advantage from
overclocking the gaming X there's simply
not enough Headroom on our units when
considering its stock shipping frequency
in fact the 1080i F II it modded with
the hybrid cooler allows for
overclocking that effectively equals the
gaming accent framerate went under
overclocked looking at 1440p Ghost Recon
wildlands at the 1080i gaming X lens at
94 FPS average under stock conditions
with lows at 81 and 78 this is
effectively equivalent to our stock
hybrid mod with the lead over the
reference cooler about three percent
this is again because we're not
encountering a thermal budget limitation
on either the gaming X or the DIY hybrid
cards
overclocked in the MSI GTX 1080 TI
gaming acts a mouthful gets it up to 97
FPS average for a gain of 3.2 percent of
the stock gaming X this is led by the
hybrid that we overclocked though not in
a big way running at 1080p we see the
hierarchy remains mostly the same as gtx
980ti gaming X performs at around 120
FPS average
ahead of the reference 1080i in both F
pS average and the frame times it's
roughly tied with the hybrid stock card
gains from overclocking are largely
insignificant again and introduce
additional frame time variants as the
card struggles to maintain even a
smallest 75 my cards offset you can see
an hour frame time plots on the screen
now that despite higher average FPS and
more frame throughput overall
consistency goes down with the overclock
which begs the question of how much it's
really worth it
with for honor at 4k the MSI GTX 1080i
gaming axe performs places above the
reference card by about 9% in average
FPS with lows mostly comparable the
1080i hybrid sits between the gaming X
and reference 1080i card for honor has
routinely shown issues with frame time
stability when overclocked as
illustrated by the 1080i hybrido see in
these charts we encountered the same
problems with the gaming X and decided
that user experience was better with a
marginally lower average fps in trade
for actually consistent frame times ie
overclocking isn't worth it in this game
at 1440p the story remains the same Emma
size gaming X performs around 138
average with lows at 117 and 102 putting
a head of our hybrid stock mod by barely
1% the lead over the reverence card is
more noteworthy at around six point six
percent frame timescale somewhat
linearly here and that remains true with
1080p where not much has changed the
1080i gaming X is just ahead of our
stock hybrid mod operating that 204 FPS
vs. 202 FPS this lands the gaming ax at
about six point three percent ahead of
the reference gtx 980ti
running a 192 FPS average using a Vulcan
on doom at 4k the GTX 1080 TI GX pushes
98.7 FPS average with lows at 81 1% and
78 FPS 0.1% lows this leads the GTX
1080i hybrid mod was to stock clocks
which performs on 95 FPS average with
our 1% low and a 0.1% low metrics at 78
and 25 the 1080i reference card
meanwhile is closer to 90 making for a
performance lead in the gaming x over
the fe card of about 10% or about 4%
over the hybrid model with no OC
overclocking helps the gaming 2x push
past 100 FPS up to 103 with one
0.1% lo metric that 84 and 81 for dumat
1440p and 1080p check the article below
cyber elite 4 with DirectX 12 and async
places the gaming X at the top of the
current chart where it runs an 86 FPS
average lows around 70 FPS against the
1080i stock cards 84 average it's not a
huge or significant lead but one that
makes sense given the thermal and stock
clock advantages of the msi card and
nothing too exciting overall though
finally Mass Effect Andromeda posts the
1080i gaming act at 69 FPS average with
one percent lows at fifty-six and hour
0.1% low metric at 53 this is ahead of
the 1080i hybrid mod p OC by about 5.6
percent and behind the overclocked
1080 TI GX marginally overclocking the
gaming ax card once again shows very
little significance at 1440p we see
similar results the 1080i gaming X
places at the top of the chart ahead of
a 1080i G on a hybrid card it was time
by about 5 percent that's roughly the
same as the 4k Delta check the website
for synthetic tests and ashes of the
singularity we've got fire striking
times by there as well so overclocking
on this card is really not exciting we
have our PCB and vrm analysis on the
channel already and in that analysis
build Zoid shows where the shunts are
that you can short on this card if you'd
like to try and use liquid metal to get
a bit more out of the power and voltage
budget but short of doing that don't
expect anything special with
overclocking it comes pretty much pushed
to where it's going to be and when
overclocking is useful the gains are
marginal at best it's maybe 3 to 4
percent if you get more than that it's a
very clock intensive game certainly it's
possible but in our testing with our
unit and obviously they very gp2 GPU
overclocking is not exciting on the card
it's pretty much they're out of the box
so that's good news for someone who
wants to buy a card that's already at
the limit and doesn't want to touch it
in terms of overclocking if you're
trying to buy it install it and never
touch it again it's not bad in terms of
the clock rate if you want something to
play with
really sadly Pascal is not going to get
that for you in most cases anyway
especially at the high end where the
running clocks as close as they can
limit point to begin with so that's just
the Pascal thing now this car being prio
seat as height as it as it is does mean
that you have more limitations in terms
of where you can go as an overclocker in
terms of headroom but that's just the
way the game is with these types of
cards in this architecture the $700
cards like the 1080 TI OSI model from
gigabyte and the 1080 t is C model from
EVGA which that one's about 720 I think
those are appealing they have an
interesting price bracket in that 750
for this gets you in terms of thermals a
well-built card in terms of build
quality well-built card in terms of
clock something that's pushed to the
limit but it's $750 and a 1080i OC or
1080i FC will get you pretty close to
the same clock rate if you are willing
to put a few minutes into overclocking
and that's just again how Pascal is they
all stop around the same point in terms
of OC potential so that means that you
could buy a cheaper 700 or 720 dollar
card and get roughly the same
performance you might be plus or minus
3% but if you're also plus or minus 32
$50
it just becomes a question of is that 3%
worth it to you that's something only
you can really answer for your use case
but something to consider if 750 sounds
a bit cheap keep in mind the $700
options this time around are really not
bad as long as it cooled well the FE
card has a damn good PC D and a
well-designed vrm unfortunately it has a
cooler that for our audience
most the time is not a great solution
you could get any axial fan cooling
solution and not run into GPU
limitations while maintaining a lower
noise output whereas the EPI card sure
you can avoid thermal limitations if
you're willing to sacrifice some of the
noise Department so that means that
things that are based on the reference
PCB this time aren't bad you don't need
a custom PCB to have a good card with a
1080i if they're built on reference and
have a better cooler cool grab it
because if you can spend five minutes
overclocking you look at performance
it's about the same anyway but we'll be
looking at more of these soon the the
extreme gaming version from gigabyte is
on the bench all
has been tested a bit mengapa not
entirely to keep an eye out for that
thermals here are good designs good it
just becomes a value question at 7.50
given all of those Pascal considerations
and finally if you're curious about any
of our testing methods the article
explains them we also have a video on
what 1% lows and 0.1% lows are which is
methodology that we defined for the
frame time testing metrics when we
instituted those a while ago so you'd
find that on the challenges type in what
are 1% lows and then we've got our
patreon page patreon.com/lenguin exit on
that for the website store like cameras
nexus not not to buy shirts like these
thank you for watching subscribe for
more I'll see you all next time
you
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.