Custom Polaris GPUs (RX 480) incorrect parameters

JustEner

New Member
Hello!

I have tried multiple versions of HWInfo64 but none of them could show me correct values for my Radeon Rx 480. I have tried with Sapphire's and MSI's Custom versions so I do not know if the problem exists with Reference versions.

It looks like this:
 
polaris_HWinfo_false.jpg



I hope you can solve it. :)
 

Attachments

  • HWiNFO64.DBG
    1.4 MB · Views: 3
  • hwinfo_polaris.HTM
    160.2 KB · Views: 4
Thanks for the report.
This looks like a bug in AMD drivers which return totally invalid values. Do other tools report such values too ?
But don't worry, I have a workaround for this, which will be included in the next HWiNFO build ;)
 
Thank you!
Yes, GPU-Z could not show correct fan speed while idle either. Additional to this it can not measure VRM temps.
On the other hand previous, non-beta versions of Sapphire TriXX and MSI AB did not recognise the VGA which results in not showing any parameters at all.
If it helps. :)
 
Thanks.
I noticed that GPU Core Voltage (VDDC) and GPU Core Power are invalid, which I can fix. Any other invalids seen in HWiNFO ?
Do other tools show correct VDDC ?
 
Every other seems fine.
GPU-Z shows correct VDDC values.
MSI AB v4.3.0 beta 14 shows it correctly also.

Edit.: Actually GPU Fan Speed is not correct. It shows a random number in RPM while the fans are actually not rotating.
Maybe it shows the fan speed which was measured right before driver turned it off.
 
When the fans are spinning, does it show the correct speed?
And what exact GPU model is that? Seems to be from MSI.
 
Then I believe the fan tachometer doesn't properly report 0 RPM and just holds the last value. I don't know how to fix this problem yet.
 
JustEnter

Well, if you're a frequency (amiss Mustech M320 mode "Frequency measurement") on the fan board (normally the red wire) with respect to its ground (black wire) will see the pulses on its tachometer with a frequency of 8 - 10 Hz and above (there on fan basis is a Hall sensor and is fixed on the impeller magnet) - questions to AMD because there a board failure, or rude driver error (probably the latter, for the driver from ATI / AMD always contain a lot of errors, even from the time of ATI Mach), no - change possible fan.
 
Hi, everybody!

I've got a couple of Sapphire RX 470's 4GB (Reference build).
HWINFO64 ver.5.35.2950, Win 7/64.

Everything is OK except:

1. Above mentioned GPU VDDC/Power/Current. (GPU-Z 1.10 displays them correctly.)
2. GPU [#1] is connected to a monitor and there's no "GPU VRM" monitoring for such GPU.
You can see the "GPU VRM" monitoring at the bottom of the screen for the non-monitor GPU [#2] but not for the connected GPU [#1]:
(the same situation is when there's only one RX 470 in the system - no VRM-monitoring at all)

TAxS5HS.png


Thanks in advance

Cheers!

P.S. There's Integrated GPU [#0] in the system also but it is disabled under Windows Device Manager - there's no option for disabling it in my mobo Bios.
 
For the first issue (GPU VDDC/Power) please try the latest HWiNFO Beta build (v5.35-2955), which should fix this. Please let me know if it's OK now.

Please attach the HWiNFO Debug File so I can have a detailed look at the second issue. Is it possible that GPU is switched off via ULPS ? What happens if you put load on both GPUs, does the VRM sensor appear? Also, does GPU-Z report VRM value for both GPUs ?
 
Martin said:
For the first issue (GPU VDDC/Power) please try the latest HWiNFO Beta build (v5.35-2955), which should fix this. Please let me know if it's OK now.

Please attach the HWiNFO Debug File so I can have a detailed look at the second issue. Is it possible that GPU is switched off via ULPS ? What happens if you put load on both GPUs, does the VRM sensor appear? Also, does GPU-Z report VRM value for both GPUs ?

2955 - I'll try and report.

Load: As you can see from the above screen both cards are at 100% load ("GPU Utilization").

This is screen from GPU-Z (both 470 cards) along the HWINFO at 100% load:

VhoVGqc.png


Debug File - how to make it?
 
Good news!

I removed old INI file and clean Registry, download 2955 version, run it and almost everything I wrote about is sorted out:
VDDC is normal now and VRM's are nicely monitoring :) ...

Just the "GPU Core Power" is lower than that of the GPU-Z:
(both apps are at 1 sec interval !!!)

3Q9aY5o.png


At GPU-Z, "GPU only Power Draw" is at 53W and there was a lot of forties.
But when you look at new HWINFO the setting is 37W max  - even not in forties!

This is my last thing to not use the GPU-Z ;)

TIA
 
cez4r

The measurement method in GPU-Z and informs the driver that the data HWiNFO different, and that the most important and where and what they measure? To say exactly who and what measures and whose results are right and whose wrong is necessary to measure the power consumption of the calibrated instruments and compare their performance programs, data, or else we will customize the reality under the pre-assigned us a good result.
 
Well, I'm not sure how GPU-Z measures GPU power and what the "GPU only Power Draw" really means. Are you sure it's correct ?
HWiNFO calculates the power as VDDC * GPU Core Current, which is a true measured value. Note, that this value covers only the GPU Core rail (not GPU memory and other rails). And the VRM power for VDDC seems to be quite close to GPU Core Power.
On the screenshot it looks like the value in GPU-Z is almost twice as in HWiNFO. Do you maybe see some relation between both tools when trying different GPU load scenarios ?
Perhaps if you could provide the HWiNFO Debug File, I could see something there...
 
Martin said:
Well, I'm not sure how GPU-Z measures GPU power and what the "GPU only Power Draw" really means. Are you sure it's correct ?
HWiNFO calculates the power as VDDC * GPU Core Current, which is a true measured value. Note, that this value covers only the GPU Core rail (not GPU memory and other rails). And the VRM power for VDDC seems to be quite close to GPU Core Power.
On the screenshot it looks like the value in GPU-Z is almost twice as in HWiNFO. Do you maybe see some relation between both tools when trying different GPU load scenarios ?
Perhaps if you could provide the HWiNFO Debug File, I could see something there...

Yes, it's correct and it should be that high. As the name suggests it's for power draw of the GPU Core only.
In GPU-Z it's like this (when you look at it): ...-46W-43W-46W-46W-53W-46W-53W-46W-43W-46W-53W-...
So it's closer to 50W on avg - and it should be at those clocks/voltages. BTW the screen is taken at the Luxmark running (100% utilization).
I can run the HWINFO only (without the GPU-Z) and the GPU Core Power still is too small.

I will provide the Debug File later. Thanks!
 
Can you please try to compare actual power values between GPU-Z and HWiNFO under various loads and write them down?
I want to see whether there's some relation, like for example GPU-Z power is *always* two times higher, etc.
 
Martin said:
Can you please try to compare actual power values between GPU-Z and HWiNFO under various loads and write them down?
I want to see whether there's some relation, like for example GPU-Z power is *always* two times higher, etc.

Yep, GPU-Z power is *always* higher!

I'm using only OpenCL apps - GPU load always at 100%:
Note 1: GPU-Z or HWINFO always run separately w/ tested apps.
Note 2: Scores in Watts are in format Average-Peak.

             GPUZ    HWINFO
App 1    46-53    28-37
App 2    49-55    32-38
App 3    59-61    43-45


Hope this helps :)
 
Back
Top