FSB frequency downscales during gaming

I'm having a problem while playing games. The FSB frequency downscales from the moment I start playing 3D games. I noticed the problem when I started to play GTA 4 which is a very CPU intensive game and my result were nowhere near as good as other people with the exact same specs as my notebook and with exact the same settings in GTA 4.

Here are my notebook specs:
Zepto Znote 6224W
Intel Core 2 Duo T7500, 2200 MHz (11 x 200)
nVIDIA GeForce 8600M GT (512 MB DDR2)
Windows Vista Ultimate + Service Pack 1

Here is a snippit of a typical log file (logged with Everest) when I started the benchmark of GTA 4:

Code:

Date		Time		CPU	CPU Multiplier	CPU FSB
				MHz	x		MHz
14/12/08	19:38:08	2162	11		197
14/12/08	19:38:14	2182	11		198
14/12/08	19:38:18	2179	11		198
14/12/08	19:38:21	2179	11		198
14/12/08	19:38:24	1025	11		93
14/12/08	19:38:27	1025	11		93
14/12/08	19:39:28	1016	11		92
14/12/08	19:39:32	1016	11		92
14/12/08	19:39:35	1025	11		93
14/12/08	19:39:38	1025	11		93
14/12/08	19:39:41	1023	11		93
14/12/08	19:39:44	1023	11		93
14/12/08	19:39:47	1029	11		94
14/12/08	19:39:51	1029	11		94
14/12/08	19:39:55	2182	11		198
14/12/08	19:39:59	2182	11		198
14/12/08	19:40:02	2312	11		210
14/12/08	19:40:05	2312	11		210

As you can see in this log, the FSB frequency is halved at a certain moment. When the game is loading, the FSB is still the normal 200 MHZ. The moment the FSB frequency downscales, is exactly the moment the benchmark starts in GTA 4. From the moment the benchmark ends, the FSB frequency jumps back to its normal state.

I've tried everything from putting maximum performance in windows vista, to using RMClock to control everything (even disabling dynamic FSB switching in RMClock), to disabling stuff in the BIOS. But nothing works. The behaviour is always the same.
It's nothing GTA 4 related because I have checked this out with other games and I have exactly the same behaviour.

Is there someone who can help me with this issue? I know there is a dynamic FSB switching feature in this Core 2 Duo processor I have, but the strange thing is that the FSB downscales from the moment it should do the opposite.

 

Have you tried fixing a particular multiplier using RMClock or CrystalCPUID ?

What are the temps like, when the FSB downclocks ?

Try loading the cores using Orthos or Prime95, and see if the FSB downclocks at that moment or not.

 

I think it it becomes too hot and throttles.

 

Thx for the fast feedback.

Indeed, I'm starting to think it has probably something to do with the temps.
The 2 cores can get around 92 degrees when under full load.
I also logged these temps together with the FSB but I do not see a direct connection between the temps and the FSB.

I did a test with wPrime to load the cores. After some period, the temps get to around 90/91/92 degrees and it stays around that temp. At a certain moment, the FSB downclocks for a couple of seconds. The temps can become slightly lower to around 89/90 and the FSB switches back and forth between the 2 states. Maybe there is some switch around the 90 degrees threshold?

I already tested with CrystalCPUID and RMClock to fix it at a particular multiplier but this does not help. The multiplier gets indeed fixed, but the FSB does not.

I also think that the fans do not fully utilize their capacity to cool down the cores. When the temps go from 70 to 90 degrees, the fans keep rotating at the same speed. Is their anyway to control the fans to test the temperatures when I put them at full speed?

 

Event viewer will also confirm this Throttling. 90-92 is a lot. It is not a switch it is the way it is built. It needs to protect itself to avoid "death". It does that by downclocking. In your case the FSB is reduced. The FW tells it do to so.

 

Are you undervolting the CPU or not ?

Try running Orthos Small-FFTs. The multiplier does not remain fixed, and drops by 1 when the FSB downclocks.

If undervolting does not help, you'll have to look to active cooling. Fan Control at the software level is not possible in a Zepto.

Checkout the undervolting guide, it may have some info on how to disable SLFM.

 

Thx for your help everyone!
I will look at the undervolting guide or I will try active cooling for my notebook.

 

Yea there is a reason it throttles down when the temperature is too high. It's to make sure the chip doesn't die of course . Of course users might want to say why they made it not adjustable, but then if they do make it adjustable you know people will make it way higher and fry their chip, probably blaming it on Intel.