GPU Throttle Reason Current

blucube

Member
Ryzen 5 5600x
ASUS ROG STRIX X570-I Gaming
Sapphire Technology AMD Radeon RX 6800 XT
TEAMGROUP T-Force Night Hawk RGB 3800CL15
Windows 10 Pro 64-bit
Cooler Master NR200P
Seasonic USA Focus GX Series 850 Watt 80 Plus Gold

Been trying to track down a cause or explanation for the "Current Throttling" taking place on hwinfo. Searching online wasn't yielding too many results. Running OCCT/Furmark for less than 5 minutes this round and it triggered.

1612834417027.png

Thanks!
 

useaname

Active Member
Hi Bluecube, I don't quite get your post - is it a question or a comment?
Is the highlighted row suggesting you think the throttle is triggered by a GPU Hotspot, rather than the GPU PPT Limit, or are you not intending to draw attention to that row?
Kind regards
 

blucube

Member
Highlight is just a highlighted area. Just curious what to go about doing to investigate the Throttle Reason Current as no real thermal reason is listed. Here was another one playing a game for quite a while with the side panel off.

1612845088948.png
 

Martin

HWiNFO Author
Staff member
My understanding is that "Throttling reason - Current" should be tripped when one of the current (Amps) limits is hit (i.e. TDC).
But your TDC limit doesn't seem to be hit, also several other users reported that these throttling reasons don't seem to provide valid data.
Since these are reported by the GPU driver, it will need to be checked with AMD.
 

blucube

Member
My understanding is that "Throttling reason - Current" should be tripped when one of the current (Amps) limits is hit (i.e. TDC).
But your TDC limit doesn't seem to be hit, also several other users reported that these throttling reasons don't seem to provide valid data.
Since these are reported by the GPU driver, it will need to be checked with AMD.

I did end up seeing one you had to replied to along these lines on a 6800. Though - that user had in some way suggested it was due to the way his monitor was phasing in and out of sleep? Not sure... Frankly it's one of those things where I haven't exactly "attributed" any issues to it - but in lieu of troubleshooting some oddities in a very specific scenario it was something I was trying to rule out but admittedly do not have any real data/experience to suggest it's related. Again - just something I was hoping to rule out if possible :)

I also see when the card sits more idle the stats tend to go "grey" then normalize over and over again - which was something I also didn't see happening on my previous card which was a GTX1080SC. Not that this is relative to point of the topic but just more differences with the 6800XT and HWiNFO vs. my previous experience with another GPU.

Just another screen shot where I'm hammering the system pretty good. This time I've undervolted the GPU to 1080mv, set the GPU clock to 2250-2450mhz. Also disabled "Zero RPM" feature on the fan control. That "Hotspot Sensor" gets up there... but I think I read an article where AMD was suggesting you could expect reads up to 110C?

FurMark Vulkan + OCCT Power Stress Test AXV2 + FFXIV Shadowbringers Benchmark loop:

1612898638415.png

1612898726220.png
 

Zach

Well-Known Member
GPU PPT has reached 100% limit though on all of your screenshots.

Personally I find furmark burn test and OCCT stress test to be really brutal and abusive.
If you game or even run a lighter benchmark like UnigineHeaven4.0 does this "Throttle Reason - Current" occurs?

Disabling zero RPM does not change fan curve. You FAN speed is only spinning at 50~55%. You should enable advanced settings in drivers and see what is the default max fan speed setting. I wouldn't like to see me GPU's hotspot reaching 107C. And yes the limit is 110C.
 

Martin

HWiNFO Author
Staff member
I also highly recommend to upgrade to the latest HWiNFO Beta version 6.43 to fix this issue with RX 6xxx series:
 

blucube

Member
I also highly recommend to upgrade to the latest HWiNFO Beta version 6.43 to fix this issue with RX 6xxx series:

I need to read the entire thread but I'll do the update as I'm on 6.42 currently. I've been attempting to do an extended benchmark with Unigineheaven4.0 as suggested above but have been running into some reboots. HOWEVER - I'm testing new undervolt/overclock on my GPU and suspect it may just be unstable.

I will say however - unless you all definitively found some correlation between HWiNFO and these crashes - Since I've made my build I found out of the box AMD was generating WHEA when the infinity fabric was clocked any higher than 1600mhz I believe (I have a whole thread over on overclockers forum on it). Through a series of testing different BIOS while working with ASUS, in my particular case it had much to do, from what I could tell, with AGESA version of the BIOS the Zen 3 CPU's and the X570 boards but not necessarily limited to.

It wasn't just HWiNFO being open where WHEA would occur which is where I first noticed it because it was the very first thing I did once the build was established. After that, I was just testing with OCCT and you would see WHEA generate there as well independent of HWiNFO. I'm not exactly expressing there isn't/wasn't an issue that involved HWiNFO playing a role but did want to express that issue is very well known in other communities and I've been through it myself (with no reason to believe HWiNFO is causing WHEA / Random reboots for me vs. Unstable overclocks while testing or the aforementioned AGESA issues.)

====

This is the current test settings that seem to be relatively stable at this point just running UniengineHeaven4.0:

1612984470896.png
1612984543909.png
1612984581526.png
1612984612758.png
1612984647682.png


Haven't hit 100% PPT as Zach mentioned with this particular load, no throttling detected.


Here's another small capture with HWiNFO playing Stardew Valley which I know is super intensive (about 3hrs worth) ;p

1612984766313.png

Though I will say this read is back from a previous post on this thread and it was taken while playing Jedi Fallen Order which was a bit more intensive where we saw 100% PPT and throttling once again

1612984894278.png

Either way seems like the common cause is 100% PPT = Current Throttling?
 

Zach

Well-Known Member
It seem like its PPT related, although if I’m not wrong...

1. You have another Power limit throttle that was not affirmative? And PPT is about power (PackagePowerTracking)
2. Your PPT 100% value is not the same in every screenshot...(?) what’s that about, I don’t know.
3. Current is about Ampere (A)
 

blucube

Member
It seem like its PPT related, although if I’m not wrong...

1. You have another Power limit throttle that was not affirmative? And PPT is about power (PackagePowerTracking)
2. Your PPT 100% value is not the same in every screenshot...(?) what’s that about, I don’t know.
3. Current is about Ampere (A)

Yes, a little confusing - I wouldn't expect the PPT to affect the current throttling but it's about the only correlation between the reads we can find? There has been some suggestion that perhaps it's a "false flag"? Though at that point it's a good distance out of my understanding of how/why things work.
 

Martin

HWiNFO Author
Staff member
Update: Looks like the reporting of Throttle Reasons on Navi21 isn't correct.
What is currently showing up as the "Current" reason is in fact most likely the HotSpot temperature tripping the limit.
I will update this in the next build of HWiNFO that will also introduce some new reasons.
 
Last edited:

blucube

Member
Thanks. I'll admit in this case cooling is pretty far from optimal on air and the way I have it configured currently. I'm not totally sure at this point how to improve it beyond adding some AIO's or something... Or getting creative with or moving a case fan...

It's pretty interesting, I like it but suppose it's just a caveat of a mini itx build.

*Edit - removed inaccurate content after fully reading article
 
Last edited:

blucube

Member
Here's almost 10hrs of FurMark w/ the backplate off. Didn't make that much of a difference I don't think... maybe a couple deg C here or there. Definitely not 20 ;p lol.

1613523082797.png
 

Martin

HWiNFO Author
Staff member
This doesn't make any sense to me.
According to your results the "Current" reason should be PPT. According to tests done by others, it should be the Thermal/HotSpot. And AMD did tests as well and says it's OK as is now.
I think the best solution for now is to remove these indicators until the situation is fully clarified.
 

blucube

Member
It sounds like you or someone have already been in contact with AMD but I would be curious myself.

Again I'll admit the cooling in my case consists of x2 120mm 2200RPM fans. One on the bottom that is an intake that practically sits just barely below the GPU die. Then the triple fan exhaust of the GPU followed by the double fan scythe fuma 2 as an exhaust and the other 120mm 2200rpm as a top exhaust as well as whatever the psu can provide for top exhaust.

All that said all in all I don't see any real thermals, albeit higher than what one may like to see but, going out of specifications.
 

Martin

HWiNFO Author
Staff member
I was in talks with AMD and they found this to be a bug in drivers and should be fixed in a new driver release.
 
Top