AMD's RX 7900 XTX Reference Cooler Can Measure Ambient Intake Temps

News
By Aaron Klotz
published

AMD's flagship GPU can apparently check for overheating PCB components and cool them with a 2nd fan curve

AMD RX 7000 Series Reference Card
(Image credit: AMD)

According to a video by Gamers Nexus, AMD's reference RX 7900 XTX features a unique ambient thermal sensor that allows the GPU to measure air-inlet temperatures coming directly into the cooler. We don't know why AMD added this sensor to its cards, but it has the potential to be very useful.

There are several benefits to having an ambient thermal sensor inside your GPU cooler. Gamers Nexus notes that one of the most significant advantages is to monitor ambient temperatures while the fans are off (0db mode), to keep PCB components on the card from overheating.

Gamers Nexus continues, saying this exact issue was a problem on EVGA's GTX 10 series graphics cards before the ICX coolers were introduced. The 0db mode of the cards accidentally caused some of the power delivery components to overheat and die prematurely. As a result, EVGA recalled those cards and eventually replaced them with ICX coolers featuring additional thermal sensors.

However, it's worth mentioning that this was a one-time ordeal, and almost every card on the market today has a zero RPM fan mode, with coolers designed to take advantage of that feature. 

In the case of the 7900 XTX, AMD probably added this sensor as an added security measure, just in case some of the PCB components get too hot while the GPU core is still cool enough to keep the fans off. Once temps get too warm, the card can initiate a secondary fan curve to keep the PCB components cool while operating the fans at a very low RPM.

Hopefully, AMD will provide end-users access to this sensor inside the Adrenalin control panel (and 3rd party applications) since it could be helpful in several scenarios - not just for keeping the card cool. For example, the ambient sensor can also be used to measure the actual ambient temperature inside the entire system to monitor and troubleshoot high temperatures on other components and/or fault chassis fans.

Unfortunately, AMD neglected to share why it added this sensor to its reference cards, leaving us to speculate. But if we get an official comment from AMD, we’ll update you here on the details.

Aaron Klotz
Aaron Klotz
Freelance News Writer

Aaron Klotz is a freelance writer for Tom’s Hardware US, covering news topics related to computer hardware such as CPUs, and graphics cards.

See more GPUs News
8 Comments Comment from the forums
  • InvalidError
    An inlet temperature sensor doesn't exactly tell the board much about board components overheating. If you want to know that board components are overheating, you put your temperature sensor on the PCB near those components. With so many VRM designs using integrated smart power stages, you can get the individual power stages' internal temperature by reading their temperature output pin.
    Reply
  • -Fran-
    If my thermodynamics memories are not that bad, it could be to calculate the temp delta for the optimal fan pressure needed so you use the optimal fan rotation speed (RPM) for the optimal temp of the air flow in order to move the hotness out. This would give way way finer grained control over fan speed and noise.

    ...Or something along those lines.

    Regards.
    Reply
  • InvalidError
    -Fran- said:
    If my thermodynamics memories are not that bad, it could be to calculate the temp delta for the optimal fan pressure needed so you use the optimal fan rotation speed (RPM) for the optimal temp of the air flow in order to move the hotness out. This would give way way finer grained control over fan speed and noise.
    You don't need an ambient temperature sensor for that: the VRM tells you the board power being pumped into the HSF, the HSF engineers should have characterized the HSF's thermal resistance as a function of fan speed, so you can already derive ambient temperature from characterized junction-heatsink-air thermal resistance as a function of fan speed, junction temperature and board power.

    I suppose stuffing a sensor in there is the lazy work-around for not characterizing the HSF beyond making sure it can cope with max load under worst-case supported conditions.
    Reply
  • -Fran-
    InvalidError said:
    You don't need an ambient temperature sensor for that: the VRM tells you the board power being pumped into the HSF, the HSF engineers should have characterized the HSF's thermal resistance as a function of fan speed, so you can already derive ambient temperature from characterized junction-heatsink-air thermal resistance as a function of fan speed, junction temperature and board power.

    I suppose stuffing a sensor in there is the lazy work-around for not characterizing the HSF beyond making sure it can cope with max load under worst-case supported conditions.
    No, it just adds more information for an overkill solution to ensure the optimal fan speeds in all thermal conditions.

    I guess that's the best way to characterize that external sensor: overkill XD

    Regards.
    Reply
  • helper800
    -Fran- said:
    No, it just adds more information for an overkill solution to ensure the optimal fan speeds in all thermal conditions.

    I guess that's the best way to characterize that external sensor: overkill XD

    Regards.
    Its also an very useful way to sanity check what's actually going on with the fluid dynamics of your case and fan setups.
    Reply
  • DavidLejdar
    It would sure seem nice if there would be end-user access to that sensor. One use for it could perhaps be even to link it to PWM chassis intake fan/s at the bottom of the inner chassis (when one has that option).

    Not that I may necessarily need it, having thermal sensor on MB, according to which there isn't any heat building up, with so far sufficient air flow. But at least getting the read-out would sure seem nice.
    Reply
  • Alvar "Miles" Udell
    It seems to me that this would be a feature much more suited to enterprise level cards with multiple to a case rather than consumer level with...one. Perhaps they're testing and refining the feature on these Radeon cards so they are implemented correctly on FirePro.
    Reply
  • InvalidError
    Alvar Miles Udell said:
    It seems to me that this would be a feature much more suited to enterprise level cards with multiple to a case rather than consumer level with...one. Perhaps they're testing and refining the feature on these Radeon cards so they are implemented correctly on FirePro.
    In an enterprise server rack, you'd want a fan-less heatsink with a front-to-back straight-through flow relying on the server's double row of "blowiematrons" to keep everything cool. Putting fans on CPUs and GPUs in those would only lead to their premature failure from over-speeding from the amount of airflow being forced through by the blowies.
    Reply