AMD ZeroCore observations

They make noise, too.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
sneaker
Posts: 133
Joined: Tue Dec 24, 2002 8:37 pm
Location: Australia

AMD ZeroCore observations

Post by sneaker » Thu Apr 18, 2013 3:33 pm

So I picked up a just-released ASUS 7790-DC2OC-1GD5 the other day, and one of the features that's long appealed to me about the 7000-series Radeons is "ZeroCore". Since I switch off my monitor overnight while leaving my PC on (to run automatic backups, routine scans, etc.), it makes sense for the video card to shut itself down as much as possible in that situation.

But there's not much feedback here on ZeroCore, so I thought I'd kick off a topic on it with a few of my experiences so far. To start with, here's my system power consumption (not including monitor) in various idle states, measured at the power socket with my Belkin energy meter.

Desktop a few mins after boot: 49W
Notepad (maximised, empty window): 48W
Firefox (maximised, displaying static web page): 49W
Display off, ZeroCore activated: 39W


And here are the "annoyances" I've encountered so far - I titled them in bold to more easily distinguish them, not to suggest they're definitively true:


Not fully compatible with GPU monitoring apps

Running GPU-Z minimised with 'continue refreshing in the background' checked would apparently cause ZeroCore to cycle on and off, or something along those lines (I observed the video card fans initially cycling off and on under this condition). Power consumption would be reduced, but only by a few watts - e.g. from 49W to 46W instead of 39W.

In the case of SpeedFan, I needed to remove the tray icon using Windows' notification area settings to allow ZeroCore to function. The GPU temperature and fan speed readings would also freeze after ZeroCore had been activated (they'd display -999C and 0 RPM respectively), so I had to remove those.


It's dependent on display power management

Switching off the display via its power button wouldn't activate ZeroCore - it only kicked in when Windows power management turned off the display. If I manually switch off the display first, ZeroCore will still subsequently activate based on the power management timer.


The display had to be static (watch out for those blinking cursors)

A notepad window or a login prompt with a blinking cursor are some forms of screen activity that prevented ZeroCore from activating. Scheduled tasks that involve display activity can deactivate it, and I've had to modify one of my backup scripts which displayed results a notepad window (with blinking cursor) and prevented ZeroCore from activating again after the task was done.


Power consumption increased after ZeroCore

This is a really strange anomaly I encountered - most of the time, after coming out of ZeroCore, idle power consumption increased in some cases, with the most dramatic being the extra 12W consumed while displaying the desktop (including when the desktop is partially visible behind non-maximised foreground windows):

Desktop a few mins after resuming from ZeroCore: 61W
Notepad (maximised, empty window): 51W
Firefox (maximised, displaying static web page): 47W

Weird, isn't it? According to GPU-Z, the GPU gets stuck at 0.988V when displaying the desktop, instead of 0.825V pre-ZeroCore. This is reproducible and persists indefinitely until the system is cycled with a reboot or suspend, undoing the efficiency gains provided by ZeroCore. Note that I have a static desktop - there are no widgets or anything, just a static wallpaper and shortcuts.


EDIT: This issue appears to be fixed in more recent drivers I just found and tested (the 7790's drivers are separate from the main Catalyst downloads at the moment, which I didn't realise before so I used ASUS' ones).


May not work over HDMI or DisplayPort

This isn't a personal observation, but a claim I repeatedly encountered while troubleshooting - there were many reports that ZeroCore won't function over HDMI, and a couple that it didn't function over DisplayPort (with none confirming it did). I'm using a single display connected with a HDMI (video card) to DVI (monitor) cable, and it works in this configuration.


I'd appreciate any observations from others about ZeroCore.


P.S. I suspect a few of you might ask about the noise levels of the ASUS 7790, but I can't offer anything more than a subjective "it's really damn quiet at idle". Techpowerup has reviewed the card and they seem better than most sites in the noise measurement department.

Here is the fan speed range as reported by SpeedFan:

0%-10%: 1084 RPM
20%: 1110 RPM (this is the card's default idle speed)
100%: 3281 RPM

The fans are 0 RPM when ZeroCore is active :)

edh
Posts: 1621
Joined: Mon Mar 17, 2008 1:49 pm
Location: UK

Re: AMD ZeroCore observations

Post by edh » Thu Apr 18, 2013 11:52 pm

That looks like quite a few problems!
sneaker wrote:Since I switch off my monitor overnight while leaving my PC on (to run automatic backups, routine scans, etc.)
Wouldn't it be better instead to somehow script things such that when youleave the system at night, it then runs all of the tasks before shutting itself down when it's finished?

I would guess that the increase in power when coming back up is that it comes on in full 3D clocks. Many graphics cards do this on startup too when the video driver is loaded.

mkk
Posts: 687
Joined: Sun Sep 05, 2004 1:51 pm
Location: Gefle, Sweden
Contact:

Re: AMD ZeroCore observations

Post by mkk » Fri Apr 19, 2013 10:08 am

Very interesting, thanks for the investigation. I'll set the Windows display timeout a little lower then, otherwise I've been in habit of starting the "blank" screensaver when walking away from the system. Will have to hook up a power meter to the system some day and compare. Most of the limitations sound pretty reasonable to me, but it's good to know which mechanics it may depend on.

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: AMD ZeroCore observations

Post by CA_Steve » Fri Apr 19, 2013 10:32 am

My habits are contrary to ZeroCore's benefit -> before I walk away from the PC, I put it to sleep.

The one thing I wish ZeroCore did that it doesn't do is work in tandem with Virtu. It would be all kind of awesome if the discrete GPU completely powered down when the iGP was in charge and only powered up for games.

sneaker
Posts: 133
Joined: Tue Dec 24, 2002 8:37 pm
Location: Australia

Re: AMD ZeroCore observations

Post by sneaker » Fri Apr 19, 2013 10:38 am

edh wrote:Wouldn't it be better instead to somehow script things such that when youleave the system at night, it then runs all of the tasks before shutting itself down when it's finished?
That can be convenient for a consistent task or two, but if there is an unpredictable array of tasks in a wide variety of programs that one wishes to be completed in that time (backups, scans, updates, browser downloads, torrents, ftp uploads, video encodes, etc.) it can be too much work trying to get everything to play nicely with sleep, and keep it that way.


Back to the topic at hand - I woke up this morning expecting to see my power meter reading about 39W before waking my monitor. Instead, it was reading 64W, and it's still at that level as I type this - the GPU voltage and memory speed have become stuck at 0.988V and 1600 MHz respectively instead of the 0.825V and 300 MHz they should be during idle. So, it seems the power consumption increased after ZeroCore issue is not fully resolved - it's evolved and is now occuring even when the desktop is not visible, but it's not clear how reproducible this variation will be.

I've just rebooted and idle readings are back to where they should be - about 47W at the wall socket, GPU voltage 0.825V, GDDR clock 300 MHz.

Further testing has indicated the Windows login screen, when displaying a password prompt with blinking cursor, prevents ZeroCore activating. What if I activate the "blank screen" screensaver at 1 minute in conjunction with "on resume, display logon screen", and monitor standby at 2 minutes? ZeroCore was having none of it.

static desktop > monitor standby = zerocore
windows login screen > monitor standby = no zerocore
windows login screen > blank screen screensaver > monitor standby = no zerocore
static desktop > blank screen screensaver > monitor standby = no zerocore
windows login screen (password prompt manually defocused) > monitor standby = zerocore

Cistron
Posts: 618
Joined: Fri Mar 14, 2008 5:18 am
Location: London, UK

Re: AMD ZeroCore observations

Post by Cistron » Sat Apr 20, 2013 4:01 am

CA_Steve wrote:The one thing I wish ZeroCore did that it doesn't do is work in tandem with Virtu. It would be all kind of awesome if the discrete GPU completely powered down when the iGP was in charge and only powered up for games.
Wasn't there big press about nVidia optimus and AMD powerplay implementing this in ~2008? A shame it only manifested on some mobile platforms.

sneaker
Posts: 133
Joined: Tue Dec 24, 2002 8:37 pm
Location: Australia

Re: AMD ZeroCore observations

Post by sneaker » Mon Apr 22, 2013 10:25 pm

I remember that Optimus-on-desktop announcement years ago and searching in vain for updates about it in the months that followed.

An update on the power consumption increased after ZeroCore issue - it seems there is no longer a link between this and ZeroCore with the newer set of drivers, because since my last post, there have been multiple occasions on which the GPU has become stuck at 0.988V and GDDR at 1600MHz (when it should be 0.825V and 300 MHz) during idle, without the monitor having gone into sleep mode since the last reboot. It appears something else is causing the card to remain in this more power hungry idle state, but as yet, I don't know what - seems I'm going to have to place the power meter next to the monitor, point a video camera at it to capture what's happening on screen at the time the idle power consumption jumps, and hope that reveals what's triggering it.

As mentioned before, the issue increases idle power consumption of the system from about 50W to above 60W, for no apparent purpose.

Cistron
Posts: 618
Joined: Fri Mar 14, 2008 5:18 am
Location: London, UK

Re: AMD ZeroCore observations

Post by Cistron » Tue Apr 23, 2013 3:10 am

sneaker wrote:I remember that Optimus-on-desktop announcement years ago and searching in vain for updates about it in the months that followed.

An update on the power consumption increased after ZeroCore issue - it seems there is no longer a link between this and ZeroCore with the newer set of drivers, because since my last post, there have been multiple occasions on which the GPU has become stuck at 0.988V and GDDR at 1600MHz (when it should be 0.825V and 300 MHz) during idle, without the monitor having gone into sleep mode since the last reboot. It appears something else is causing the card to remain in this more power hungry idle state, but as yet, I don't know what - seems I'm going to have to place the power meter next to the monitor, point a video camera at it to capture what's happening on screen at the time the idle power consumption jumps, and hope that reveals what's triggering it.

As mentioned before, the issue increases idle power consumption of the system from about 50W to above 60W, for no apparent purpose.
I remember my HD3870 power-play acting up at some point as well. After a power-hungry application the voltages and clocks wouldn't drop back to idle settings. I believe the culprit was some driver setting, but despite all my best google efforts I can't find it now.

mkk
Posts: 687
Joined: Sun Sep 05, 2004 1:51 pm
Location: Gefle, Sweden
Contact:

Re: AMD ZeroCore observations

Post by mkk » Tue Apr 23, 2013 11:06 am

Perhaps a card BIOS bug even, unless a future driver solves it. The HD7790 is a new chip after all. My HD7950 has never had its clock speeds locked up over the couple of months I've had it. Haven't monitored the voltages though.

sneaker
Posts: 133
Joined: Tue Dec 24, 2002 8:37 pm
Location: Australia

Re: AMD ZeroCore observations

Post by sneaker » Tue Apr 30, 2013 4:03 pm

Couple of updates:

- It's six days since I installed the Catalyst 13.4 drivers released last week, and I'm pleased to report that in that time, the card hasn't become 'stuck' at the higher GPU voltage and memory speed, whereas previously it was happening once or several times during the course of a day.

- The workaround I'm currently using for the Windows 7 login screen (aka Win-L lock screen) preventing ZeroCore from activating is to have at least two user accounts, and to click "switch user" on the login screen, which results in a screen without a blinking cursor. As yet, I don't know of a workaround that will allow ZeroCore to activate without manual intervention after the login screen appears (i.e. to go ZeroCore after the screen has locked and gone to sleep automatically after a set idle time, per the "on resume, display logon screen" setting in screen saver properties).

QUIET!
Posts: 238
Joined: Sat May 11, 2013 8:33 am

Re: AMD ZeroCore observations

Post by QUIET! » Sat May 11, 2013 6:49 pm

I'm expecting my 7790 to arrive on Monday.

Do you have any updates?

I haven't investigated the driver issue but I guess I should start by looking for 13.4.

Thanks.

sneaker
Posts: 133
Joined: Tue Dec 24, 2002 8:37 pm
Location: Australia

Re: AMD ZeroCore observations

Post by sneaker » Sun May 12, 2013 2:23 pm

Nothing new to report re the card becoming stuck at higher GPU voltage/memory speed - hasn't reoccured since the 13.4 drivers were installed.

It's likely issues with ZeroCore will tend to fly under the radar, since it's a bit of a niche interest (the broader community of PC hardware enthusiasts will scoff at someone concerned over 10 watts) and few people keep a power meter hooked up to their PC - even those attracted by the feature would be largely unaware if it weren't functioning as they'd expected. I don't think AMD has done much to describe the practicalities of it. Most of it is marketing type descriptions along these lines:

https://www.youtube.com/watch?v=FNJ_Ak3zYtQ

"We know that there are millions of PC users who leave their systems on overnight for convenience..."

...many of whom presumably lock their PC at the Windows login screen with its ZeroCore-thwarting cursor. To complicate things, display activity doesn't consistently present an issue. Consider this video:

https://www.youtube.com/watch?v=Sl-59jy65bQ

Notice how the user closed Afterburner but left a bunch of desktop widgets active on the right. This is display activity, right? Yet ZeroCore apparently works in that video, and I can confirm that if I stick Windows 7's clock (with seconds hand enabled) and CPU monitor gadgets on the desktop, these don't impede ZeroCore. It appears, when sending the monitor into sleep mode, Windows disables certain types of display activity (such as the taskbar clock and desktop widgets) but not others (such as a blinking cursor in a Notepad window).

QUIET!
Posts: 238
Joined: Sat May 11, 2013 8:33 am

Re: AMD ZeroCore observations

Post by QUIET! » Sun May 12, 2013 11:01 pm

Thanks, its good to know that the big issue seems to be resolved.

I expect AMD will eventually fix the remaining bug too.

QUIET!
Posts: 238
Joined: Sat May 11, 2013 8:33 am

Re: AMD ZeroCore observations

Post by QUIET! » Sun Jun 16, 2013 7:51 am

I was getting spectacular (to me) idle power with my i3-3220 and 7790 combo at 23 watts before I entered 3d mode, then it jumped to about 36 watts after that. It looks like I need a driver upgrade but at least I'll know what to look for.

Thanks.

wanky
Posts: 27
Joined: Mon Jan 24, 2005 1:58 am

Re: AMD ZeroCore observations

Post by wanky » Tue Jun 25, 2013 11:50 pm

My observations with zerocore are pretty much the same as OP. Im using gigabyte 7870, Win8, 13.6 drivers.

-Did not work with HDMI so had to use DVI
-Dropped 10w when operating
-If active window has a blinking cursor then zerocore did not operate. (Can disable blinking cursor in keyboard settings to get zerocore to work)
-Login screen did not prevent zerocore from operating correctly. (even though it has a blinking cursor)

james
Posts: 22
Joined: Tue Jun 11, 2013 9:46 am

Re: AMD ZeroCore observations

Post by james » Fri Aug 09, 2013 1:07 pm

I ahven't done any power measurements yet, but I've found a lovely bug...

If I'm running a fullscreen app (e.g. game), and Alt+tab to the desktop, and I leave it to idle (and screen to turn off), then when I return to 'wake' the PC again, the screen turns on, but shows black. Only way to get it back (short of rebooting) is to unplug the monitor from the card, into the integrated graphics port, clos the game down using task manage, then put the cable back to the gfx card. :(

Only happens on fullscreen apps though...

Once my build is done, I'll get some power measurements...

fnukyguy
Posts: 14
Joined: Sat Sep 26, 2009 5:34 pm
Location: norway

Re: AMD ZeroCore observations

Post by fnukyguy » Sun Aug 25, 2013 8:29 am

The windows taskbar clock in the bottom right corner wakes up the card from zerocore every minute.
In other words, when time goes from say 10:13 to 10:14 the card wakes up for about 4 seconds, then goes back to zerocore.

Anyone else getting this?

I have to remove the windowsclock for zerocore to work like it should.
Im using a desktop gadget clock as a workaround.


Using catalyst 13.4 on windows 7 64-bit with HIS radeon 7870.

Vicotnik
*Lifetime Patron*
Posts: 1831
Joined: Thu Feb 13, 2003 6:53 am
Location: Sweden

Re: AMD ZeroCore observations

Post by Vicotnik » Mon Aug 26, 2013 3:36 am

Power efficiency is great, but when they make me jump though hoops to make it work I tend to not care. A non-changing screen might just as well be black, as in computer is in deep sleep. :roll:

fnukyguy
Posts: 14
Joined: Sat Sep 26, 2009 5:34 pm
Location: norway

Re: AMD ZeroCore observations

Post by fnukyguy » Sat Oct 05, 2013 12:14 pm

Found another issue with zerocore..

Accessing uTorrent via uTorrent WebUI from another computer brings GPU out of zerocore (which is kind of weird since nothing is happening on the screen).
Once utorrent is accessed via webui, zerocore is not enabled again, even after closing webui-tab in browser.

Using "uTorrent Remote" instead is a good workaround, it does not interfere with zerocore at all.

wanky
Posts: 27
Joined: Mon Jan 24, 2005 1:58 am

Re: AMD ZeroCore observations

Post by wanky » Tue Oct 08, 2013 3:41 am

another issue I had was the GPU voltage will reset to stock when waking from zerocore. annoying if running a custom overvolt/undervolt. I disabled zerocore a while ago because of that

Post Reply