AMD ZeroCore broken

Hello,

This is to see if anyone else is noticing this issue. AMD ZeroCore is being marketed for mobile users hard. And why not? Shuts down the GPU and only uses 3W of power. It's great idea, except it doesn't work... But you know what's great about ZeroCore, it works for desktop users, yeah... seriously.

Anyway, if your 7970M fan is always on, even if it's not loud, just always on, that's because your 7970M is always active. Sure it may say GPU utilization is at 0%, but it's still consuming a lot of power and if you are battery, draining your battery. How nice.

It should say 0mhz and 0mhz for both core and memory. You can use HWInfo or AMD's own system monitor to verify. Mine is broken. It works on the rare occasion, after I put the machine to sleep. On wakeup, sometimes ZeroCore will be working. Other than that, I'm screwed.



Yes it is important to have it working for desktop, but screw those mobile users. Seems to be AMD's management's priority.

 

That's very odd, have you checked policies for background processes like steam? For some reason optimus is set to use the nvidia GPU for steam so if you don't set it to integrated every time my system would start it would run the 680M.

I managed to get it working pretty well with the enduro beta drivers with the GX60.

 

I have nothing but the Chrome browser. It's stupid. I'm on battery! I'm on power saver profile. Even put it to sleep and turn it back on. Really really really stupid.

 

What happens if you select the option to use the integrated graphics on all programs and restart?

 

That's how I have it now, it's on forced power saving option. AMD is always on. Heck the fan is louder than on CPU when it's supposed to be doing nothing.

 

So you also have tried running without any monitoring programs and just listening for the fan?

 

Yes. Using 13.1 WHQL.

 

If HD7970M was captured and being monitored by HWINFO64 or any GPU monitoring program, it'll keep the HD7970M turned on, this is the same for GTX 680M users. GPU-Z, GPU temp monitoring in ThrottleStop, desktop GPU monitoring gadgets, any sorts of tools that are capable of monitoring / requesting temperature value from the dGPU, will prevent the dGPU from being shutdown.

By the way the fan EC takes the last value of the temperature reported by the dGPU when the dGPU is shutdown, it's the reason why the fan keeps on running. What I normally do is let it run for few min and open up GPU-Z (temperature is usually around 29~33C) and wait until the fan stops (at around 40C) and close GPU-Z right after.

 

I've had it monitoring properly with GPU shut down before. As for fan and EC, I've updated the ME firmware and drivers and that is that.

 

Catalyst 13.1 / 13.2 broke ZeroCore, I can confirm this.

Before 13.1 (e.g. 12.11 betas): GPU will power down (core/memory clock drops to 0 MHz) if no game is running. If a game is running, the GPU will power back up, and show core/memory clock as the normal values (for me, 815/1300).
After 13.1 (including 13.1, 13.2): GPU will stay on; clock speeds do not drop to 0 MHz, even when NO game is running. Even when I have "Force power saving graphics" in CCC. The option does "work" in that if you run a game with the option enabled, the performance will be very poor since the game IS being rendered on IGP. But the dedicated GPU is still "on".

Serious QA issue that AMD missed for us mobile users. It's funny when 12.11 betas worked perfectly fine in this regard... and 13.1 is WHQL'd.

However, game performance in 13.2 is better than the 12.11 betas.

 

Same for me. The 12.11 Performance driver had zero core working, the 13.1 is not. But even then, it didn't always work, seemed random when it would and wouldn't, other than putting it to sleep.

 

Sorry for the newbie question, but how do you know if it's working or not? How can you be 100% certain the AMD card is at 0% usage, not consuming power? Is there any software to view your laptop's current energy consuption?

 

Not HWMonitor. It did not power up the 680M for me but only used HD4000. I had to start up Steam to see my GPU

 

Easiest would be to put your hand at the vent. If the air being pushed out is hot/warm, that would indicate the GPU is being utilized and on. Before I fixed my fan issue with proper ME 8 firmware and drivers, the fan would blow out cold air. Now the air is always warm, indicating to me the GPU is not powered down.

The other would be to use AMD System Monitor and HWinfo. They will tell you whether the 7970M is running at 0mhz or 150mhz.

 

If you launch an app that uses the dGPU and then launch HWINFO64, HWINFO64 will capture 680M and start requesting temperature value, and the dGPU won't turn off unless the app and also the HWINFO64 are closed.

I guess AMD's zero core is slightly different to Optimus, able to shutdown when being requested to report the temp value by monitoring tools is pretty good. For NVIDIA Optimus everything has to be closed completely.

 

I can only see the 7970M in HWINFO64 when a High Performance application is open. In that case, the core clock stays at 300 MHz and the memory clock stays at 149.5MHz, and that is at % usage. I am guessing that when I open HWINFO64 without have a High Performance program open, the 7970M is not visible because it is not active and in a complete 0% usage, 0MHz clock state. At the moment, the 7970M side is pushing out cool air, so I must be lucky.

It is quite cool how the system switches between either card without a single visual discrepancy. I guess that is one of the main advantages of Enduro.

If HWINFO64 is set for High Performance that is likely the issue. If not, and you cannot find a program that was mistakenly marked for High Performance, the Clevo 9.10 driver isn't that much older.

 

Well it's only ever the intel chip outputting an image to the display, the AMD card is sent graphics information to render and puts it in the buffer for the integrated graphics to render so there is never a hand over for the output.

 

Facing similar issue but not in a gcn card but its a southern island 600x card, usualy it should downclock to 0/27 mhz when idlling, happens with older 11.12 driver but anything new 12.10 or 13.1 the clock always runs at 100/150 mhz crippling battery life.
forum.notebookreview.com/asus/705493-amd-gpu-zero-core-issue-asus-k53ta.html#post9049886