• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

FM2 observations and questions

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Ruiner

Member
Joined
Jan 23, 2001
I've got an A8K Trinity in my HTPC with the budget Biostar A75 matx board.
I'm using a Thermaltake TMG A3 http://www.thermaltakeusa.com/Product.aspx?S=1138&ID=1580 for quiet operation. Since it's only rated for 65W, I've undervolted (1.3) and underclocked (3.4GHz) to ~65W A8 settings. Temps (via Coretemp) still drift into the low 60C range under Prime. FWIW I have a better cooler on the way. I can't fathom running the dinky stock cooler. I can't imagine that it could keep the hotter A10 from throttling even at stock.

Some things I've noticed playing around and want to know if they are typical for this platform:
a) I tried OCing the IGP to 800 Mhz(default A10 clock). It runs stable, but noticed that the gpu clock will not downclock anymore when idle....it stays at 800 regardless of load. Stock it would downclock to ~200 at idle.

b) With the cpu multi and Vcore modified, it doesn't appear that turbo works. Furmark is single threaded, but all 4 cores stay pinned at 3.4 (load is 100 for core 1, and low single digits for the other 3).

c) I flashed the new beta bios, but didn't reset to 'optimized defaults' before flashing, and now the 'auto' settings for the cpu default to the underclocked/undervolted ones I had before the flash (even after using 'load optimized defaults'). I'm planning on resetting the CMOS, but will wait til the cooler arrives to take it apart (the jumper is less than easy to access.
 
It seems that the thermal sensors for FM2 socket are not working properly whatever APU you're using, so don't really consider the temps you may see. HWmonitor seems also to be one of the most accurate in this field for FM2 but still no miracle :)
 
I'm getting some realistic readings with HW monitor and the biostar utility. The reported temps are consistent with those of the HS body itself. Temps are also realistic during normal use. The high numbers are only observed during prime or prime+furmark.

I'm more curious about the behavior of turbo-core and the change of 'optimized default'.
 
Here is a scheme from AMD about their turbo technology:

IMG0036606.jpg

Each benchmark will make the turbo run at a different frequency.

You can have a look to the FID/VID tables with CPU-Z, here is one for an A10 5800K

FID 0x1A - VID 0x0E - IDD 20 (21.00x - 1.462 V)
FID 0x16 - VID 0x28 - IDD 15 (19.00x - 1.300 V)
FID 0x12 - VID 0x34 - IDD 12 (17.00x - 1.225 V)
FID 0xD - VID 0x42 - IDD 10 (14.50x - 1.137 V)
FID 0x8 - VID 0x52 - IDD 7 (12.00x - 1.037 V)
FID 0x3 - VID 0x64 - IDD 5 (9.50x - 0.925 V)
FID 0x10C - VID 0x6C - IDD 3 (7.00x - 0.875 V)

A x19 multiplier provide a CPU frequency of 3.8GHz and the voltage is about 1.30V.
The turbo technology is switching between those different multipliers/ voltages in order to be the most adequate.
But honestly, it is EXTREMELY rare that the chip goes over 4GHz through turbo.

So if you fix a multiplier/ voltage, the turbo obviously won't work since it has to change those settings... that you've just fixed and freezed! :)

Here is a table of the consumption with Prime 95+ Furmark for different APU so that you can have an order of magnitude (credits to hardware.fr) :thup:
IMG0040078.png

If you plan to undervolt, disable the turbo, if you don't, the potential is going to be really low for undervolting.
On the table above, you can see that disabling the turbo has a big impact on the consumption when only prime 95 (CPU) is running. It is logical since the CPU tends to use its TDP at maximum. So disabling the turbo gives you a nice margin.
When you run Furmark (GPU), both parts CPU and GPU will have to share the TDP, so the potential decreases, even with the turbo off.

For your "optimized defaults" question, I have to admit I didn't understand where the problem is ^^'
If you plan to underclock/ undervolt, it is a good thing, isn't it?
Btw, 60°C under Prime 95 isn't that bad (if temperatures are accurate, see first post), since it is about the max the CPU will ever do under load... ;)

EDIT: pictures fixed ;)
 

Attachments

  • IMG0036606_1.jpg
    IMG0036606_1.jpg
    77.1 KB · Views: 211
  • IMG0040078.png
    IMG0040078.png
    6.2 KB · Views: 172
Last edited:
Thanks. So fixing the multi will automatically disable turbo, as opposed to shifting both speeds up or down proportionately. The bios does not have a specific setting to disable turbo. your links are broken btw.

60C under prime is at 3.4GHz. It hit 70C at 3.6.

Re. my other question, I'd assume that once I set cpu settings back to auto/default in bios, that multipliers would return to stock (3.6/3.9T) with turbo functioning again. They aren't. It may take a cmos reset, but I haven't dismantled the system yet to try it.
 
your links are broken btw

What links? :eh?:
You mean you can't see the pictures? (I didn't put any link at all :D)

Here is a link to the article (in french, google translate ;) ) where the pictures comes from :)
http://www.hardware.fr/articles/878-1/amd-a10-5800k-5700-a8-5600k-5500-apu-desktop-deuxieme.html

It sounds weird you can't disable/unable turbo, your MOBO is the TA75MH2 right?
And you CPU is the A8 5600K? (I found it rated at 100W and not 65W as you said)

In the manual of the BIOS for the MOBO, I found it:

Performance ->APU FID/VID Configuration -> Custom P-states [disabled] (default)

It should enable turbo :)
 
Yeah, no pics. That article linked is helpful, though.

I'm temporarily running multiplier settings closer to the locked, 65W, 5500 (3.2/3.7T) chip until I get my new cooler.

Since the bios flash, disabling custom p-states in bios doesn't result in stock speeds anymore. Ii may have to reset cmos.
 
Running a Shurikien cooler now at default speeds and vcore. CMOS reset fixed the multi issues I described.

I'm playing with the ram now. My Corsair 1600 1.5v 9/9/9/24/41 2T sticks default to 9/13/14/34/48/2T 1.6v when set to 1866. Pretty loose, but seems memtest friendly so far. Will try 2133 next. I am using the IGP btw.
 
Back