Titan RTX Frozen at 1350MHz: VBIOS Bug on Retail Card
Titan RTX Frozen at 1350MHz: VBIOS Bug on Retail Card
2018-12-20
we have to tighten our TX's and today
we're doing a shot here in the test lab
because one of them is not working
it's stuck at 13 50 megahertz you might
remember this issue that would be that
one right there might remember this
issue from when we got some user cards
and some cards from you all and one of
the EVGA cards was stuck at 1350
megahertz that was a twenty atti this is
a Titan RT X so quite a bit different in
terms of the issue it's about two times
the difference in price and this is
something we are now actively working
with Nvidia they actually have a local
office to us with an FA e there and
they're sending someone out to swap
cards with us so we're planning to give
them this card because it is stuck at 13
to 30 megahertz for perspective it
should be over 1900 megahertz and
they're gonna give us a different one
that way we can get back to testing so
we have this card has gone through all
the game tests that one we need for SLI
and that's gonna put a bit of a hold on
things but today I'm going to
demonstrate how that is stuck at 1350
and what exactly we're going through
before that this video is brought to you
by us and the gamers access store you
can go to store documents nexus net to
pick up one of our ceramic mugs
critically-acclaimed mod mats or
educational video card tear down and PCB
Anatomy posters that teach the names and
placements of all the key PCB components
learn more at store des cameras XS net
or click the link below here's the issue
here's the recap originally with the
EVGA 20 80 TI that we had from a viewer
that was stuck at 1350 megahertz it
raised some alarm bells at in-video
which is a good thing but this was the
only instance that they knew of that the
card was stuck at 1350 same for EVGA and
so the problem is we were able to fix it
and that's a problem because that is
hard to diagnose it because it's been
fixed so we fixed it by we tried
flashing it once with the original bios
that was on there and didn't do anything
I'm still stuck at 1350 and the power
target was showing 0 which is something
another symptom we're seeing here so we
zoom in on the screen you'll see a 0%
right there and that's the same exact
thing we saw in the EVGA card so to fix
this
we just reflashed it and we flashed it
twice the first flash was that the
original bios did not fix anything
the second flash was with a slightly
altered v bios but there are no
significant changes that we knew of if
any changes at all and that fixed it
so problem solved we sent it back to
EVGA anyway on behalf of the user and
they I mean you can't really do anything
with it because now it's fixed
just for BIOS versions if you're curious
this messed up card is ninety point oh
two point two three point zero zero
point zero one and I saved previously
the V BIOS from this card and the other
one and on the original ROM for the good
Titan
it's 9002 2300 one which is the same
exact ID and if you do a comparison of
the two binaries the same and if we open
up NV flash we can also if you've never
done this you can download it on
techpowerup or similar site any flash 64
is what you want you just do like - -
protect off and that will disable the
right protection and then you can do a
save name the BIOS ROM and that exports
it which is what we've done there now if
we wanted to load the good BIOS what we
do is - - 6 I'm not going to type the
rest I don't want to accidentally hit
enter and then mess up and videos
ability to troubleshoot this and then he
type in the good BIOS name and flash it
and that would be the end of that so we
do think this is probably still an
uncommon issue but if you did run into
it and you don't want to go through the
RMA process this tool and me flash is
how you would probably fix it you could
potentially just export your own and
BIOS from the card to enemy flash 64 DXE
- - you do protect it off and then the
next thing you would do is save and give
it like original BIOS ROM we can show
how that works actually and then after
that you would do the dash dash 6 like
the number 6 and and target the ROM and
then that would write it to the card
write it back to it you have to reboot
in between
and and then it should probably fix it
for you but we're not going to do that
today so we want to help try and
diagnose this issue at this point even
though I really want to do sli testing
now and I know I can flash that with the
V bios from the one behind me or even
probably just reflash it with its own
and fix it because when we extracted the
V bios from both of these cards the
binaries are the same it's the same v
bios so there's really no reason it
should be messed up but I'm 99%
confident it's a V bios issue in flash
and we'll fix it so anyway I'll show you
what's going on what's happening is yes
the card is stuck at 1350 megahertz but
some key identifiers to what a potential
problem might be would be to look at the
the readouts for hardware monitor or
through GPU C and we've shown these in
the past and gbz especially where you've
got a perf cap perf cap reason
performance cap reason this is something
we've used for overclocking you'll see
it's display an idle right now and
ideally under a load like fire strike
ultra on the background it should be
showing something like maybe power PWR
or thermal which would be THM and and
this is neither of those another common
one is V rel for voltage limits so those
are kind of the limits you hit typically
and you will hit a limit at by design
you'll hit something it's just a matter
of which one idle should not be where
we're sitting because it's not so
there's a problem and 13 50 megahertz
also just happens to be the base clock
of the card and by coincidence the 20
ATT I was also stuck at 1350 which makes
me kind of curious about what's going on
there because that wasn't its base clock
so it might just be coincidence but
other than that we can look at a lack of
settings here so gpu-z typically for
this card on on the other one we have
that works it will have a readout for
the current power the power consumption
and that should be like - 80 watts +
under stock conditions so we don't even
have that it's not appearing which means
something's wrong and other things not
appearing under hardware monitor on
precision x1 with this red number one we
should have another option above GPU
clock
says total power or something to that
effect and that is not appearing so
we've enabled the other ones that we
have power limit voltage limit temp
limit no load limit and GPD usage all
enabled and if we scroll through all of
this stuff you'll see GPU clock 1350
memory clocks actually functioning
because GPZ shows 1750 point 2 megahertz
then you multiply it out and that gives
you the correct clock number so that the
memory is fine GPU temperature is at 63
which is fine power limit this is a
binary so it's either 0 or 1 and we are
not hitting a power limit so it's
showing zero voltage limit binary it's
showing 0 and the limit
currently temp limit 0 no load limit is
showing 1 so we are at no load limit
which just means it's not really being
treated like it's under low it's not
being treated like it's doing anything
GPU usage is still showing up about a
hundred percent though and then memory
use it's not really relevant but it is
showing to about two gigabytes but it
doesn't really matter and then frame
rate
I don't know how perfectly accurate this
is but it is consistent with the frame
rate number we're getting in fire strike
right now you probably can't even read
it is too blurry I was about 33 so it is
actually putting out frames clearly just
the wrong everything else in terms of
the core clock there's also this really
cool bug so other than the power being 0
right now if we change the fan speeds
I've had issues here I don't know if
it's gonna repro right now yes
it it did do that change the fan speed
the fan speed did not go up you'd
probably hear it on my mic if it did and
what we're getting is everything slowing
to a crawl so it's still rendering in
one frame so it's we're at about
probably fractions of a frame per second
right now or maybe maybe out about 0.25
frames per second so this is this is
clearly a bug it should be increasing
the fan speed not sure what's causing
that the last thing to show here is we
have thermocouples hooked up to this so
we have a teardown that will go live
after this video goes live and we did
the teardown to stick thermocouples on
here to try and diagnose things like we
did
the 20 series 20 atti so just like with
the 2080 TI the thermals are completely
within reason they're well within spec
and even when the when fire strike is
running at its 36 fps or whatever there
was fire strike ultra so 53 degrees
Celsius that is for the that is for the
memory vram and that's for the hottest
memory module 52 point 5 degrees for the
hottest MOSFET that we could pinpoint
both of those are way within spec
memories under 90 95 that's fine the
MOSFETs under 120 550 by a lot so that's
fine then ambient is 23 just for
reference so yeah took it apart to stick
thermocouples on it just to confirm just
like with the 20 series cards that you
all were having problems with for the
2080 TI is that everyone sent in one of
the things we looked at was is it
thermal is it some component that's not
reported in GP z like mosfet memory the
answer was no and so we redid that test
here stock thermocouples on mosfet and
memory and the answer is again no it is
not thermal so tearing it down
unfortunately in putting thermocouples
on it doesn't help us here because it
didn't really teach us anything and the
board itself you know it's it's creating
this 13 50 megahertz lock you look at
the board all the board components look
fine so this is a probably a either a
lower-level issue that we can't see like
in silicon or something but that seems
unlikely and or AV bios issue and that
seems likely because last time it was a
V bios issue and flashing this would
almost certainly fix it but we are going
to leave this one defective so that we
can help Nvidia try and troubleshoot it
because I'm very curious what the
problem is I know they are as well the
engineers there you know we can respect
the position they're in because you make
millions of video cards and then there
are two that I happen to get that gets
stuck at 1350 and one of them wasn't
even it wasn't even mine to begin with
it was a viewers we just ended up with
it so it's not an issue with our test
setup either because like there was a
third party involved that sent it to us
and yeah we can we can respect the
difficulty of making that many
millions of cards and not being able to
encounter one in your own controlled
environment or labs so we're gonna try
and help them out maybe they can solve
it
does seem to be a low-frequency issue
but if you also encounter this let us
know because it's it's hard to know
exactly how many people are running into
this so first account below if you've
encountered this on yours otherwise
we're still gonna be doing SLI testing
we're just waiting from point of this
video about maybe 16 hours 12 16 hours
to get a swab from Nvidia locally and
then we can continue SLI testing we
already have the review for the the
single card done and that's using this
one which is working without issue so
we're good on that but anyway just
wanted to show you the problem and show
that it's come back 13 50 megahertz lock
was not apparently entirely unique to
that one EVGA card and videos very
interested in it they want to solve it
and so that's good and we're gonna try
and help them out with it because it's
it's a pretty serious problem I mean the
card is are they useless otherwise so
you could RMA it but hopefully just
doesn't happen again so keep us posted
if you run into this with any of your
cards and we'll do the same for you as
always subscribe for more go to that's
lovely go to stored on cameras accessed
tot net to pick up a shirt not like the
one I'm wearing because it's out of
stock you bought too many of the
disappointment shirts we're gonna
refresh this design into a front only
design at some point or go to
patreon.com/scishow and access and check
back for the review of the Titan r-tx
thank you 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.