TL;DR: From GPU-Z version 2.17 onward, furmark does not read the GPU core frequency correctly (it only reads as 0 MHz). Using GPU-Z 2.16.0 and earlier fixes this issue. The problem happens regardless of driver version, furmark version, and OS version.
I think this is because in 2.17, GPU-Z changed their name of that sensor from "GPU Core Clock" to "GPU Clock" and furmark can't find "GPU Core Clock" anymore. The same happened with the memory clock stat too.
At some point, Furmark no longer reads the core speed correctly from GPU-Z. It reads them as zero (0). I have some screenshots where it previously worked fine; Driver 418.81, Windows 10 1803, GPU-Z 2.16, Furmark 1.20.0.1. I've been playing around with trying to replicate the problem as well as a non-bugged (show core freq correctly).
I have been able to replicate both bugged and non-bugged on Windows 10 1903.
I have been able to replicate both bugged and non-bugged it on Furmark 1.20.5.0.
I'm currently using Driver 418.81 and trying to see what sequence and version of GPU-Z is causing the issue, because I can run GPU-Z 2.16 and sometimes (I havent fully tested) can get core to show, and sometimes not, but it depends if I have opened gpu-z 2.21 prior or not. Will update...
On Windows 10 1903, Furmark 1.20.5.0, Driver 418.81...
If I open GPU-Z 2.21.0 before I open Furmark, core freq is bugged in Furmark stress test.
I restarted the computer.
If I open GPU-Z 2.21.0 after I open Furmark but before I start stress test, core freq is bugged in Furmark stress test.
I restarted the computer.
If I open GPU-Z 2.21.0 during Furmark stress test, core freq is bugged in Furmark stress test.
I restarted the computer.
If I open GPU-Z 2.16.0 before I open Furmark, core freq is CORRECT in Furmark stress test.
Ending the stress test and restarting stress test = CORRECT freq
Closing furmark and re-opening Furmark and running stres stest = CORRECT freq
Also tried closing GPU-Z 2.16 and furmark, opening furmark, then GPU-Z 2.16, then furmrk stress test = CORRECT freq
Ending the stress test and restarting stress test = CORRECT freq
Closed both programs, started up furmark and started stress test, then opened GPU-Z 2.16 during stress test = CORRECT freq.
So it seems with this driver + GPU-Z version so far, freq detection in furmark is OK
I then tried opening GPU-Z 2.21, then closed it, then opened GPU-Z 2.16, then opened furmark + stress test = CORRECT freq
Double checked nothing was broken: Opened furmark > GPU-Z 2.16 > stress test = CORRECT freq
Closed both again.
I then opened furmark, then opened GPU-Z 2.21, then closed it, then opened GPU-Z 2.16, then started Furmark stress test = BUGGED freq! Aha, we're getting somewhere!
I closed both programs, then opened GPU-Z 2.16, then furmark, then stress test = CORRECT freq! Ok, the bug isn't persistent.
I closed GPU-Z 2.16 and re-opened it during stress test. Still CORRECT freq
I closed GPU-Z 2.16 and then opened GPU-Z 2.21 during stress test. Still CORRECT freq. HMMMMMMM.
I closed stress test and re-started stress test. Still CORRECT freq.
I then closed stress test and furmark and restarted furmark and it's stress test with GPU-Z 2.21 still open. BUGGED freq.
At this point, I tried the current 430.97 driver with 2.16 GPU-Z, Win 10 1903, Furmark 1.20.5.0
I tried all combinations of running order for GPU-Z 2.16 and Furmark = CORRECT freq.
At this point, the issue seems to be GPU-Z specific.
I rebooted, opened GPU-Z 2.16 > opened furmark > started stress test = CORRECT freq. Ok, we already know that.
Rebooted again, opened GPU-Z 2.17 > opened furmark > started stress test = BUGGED freq. AHA! We found the culprit finally.
Quote...I have some screenshots where it previously worked fine...
If you attach these, we would know which GPU you actually have.
EVGA GeForce GTX 1070 SC Gaming ACX3.0, but the kind of video card seems irrelevant since i just went through an arduous process to find out the problem is because in gpu-z 2.17 they changed the names for gpu core and memory frequency in their sensors window.
if you really need screenshots, what do you need screenshots of exactly? gpu-z 2.16? gpu-z 2.17+? furmark without gpu-z? furmark with bugged gpu-z? furmark with correct gpu-z? different video cards? something else?
gpu-z 2.16 with correct freq
(https://i.imgur.com/ppEidFe.png)
gpu-z 2.17 with wrong freq
(https://i.imgur.com/F9RcgLF.png)
gpu-z 2.21 also with wrong freq
(https://i.imgur.com/w2rp0uZ.png)
I wanted to double-check with a similar GPU first.I can confirm the issue with a GTX 1060 mobile.
Now i dig out some older rigs...
FYI instead of rebooting the computer you can reset the GPU by pressing SHIFT+CTRL+WINDOWS+B at the same time.
Further checks -same issue withGeForce 840m / Windows 1803GeForce 750 Ti / Windows Vista / FurMark 1.19
Thanks for the feedbacks. I don't check the compatibility between FurMark and GPU-Z for each new release of FurMark or GPU-Z. That explains the bug. I will fix it as soon as possible!
Thanks for the shortcut. Didn't know. The computer is purely for testing / stressing video cards for resale, so I didn't sweat having to wait for reboots. I wanted to "start fresh" on the earlier tests to remove variables out of the equation (I wasn't sure if GPU-Z inserted a driver / code that persisted after being closed, so I wanted to reboot to make sure that wouldn't happen)
BTW I can confirm the same exact behavior on a Sapphire Radeon 7870 2GB running 19.5.2 drivers. 2.16 GPU-Z = correct frequency in furmark. 2.17 or higher = bugged frequency.