[Fixed/Workaround] Asus G51J(x) CPU throttling investigation

[unclewebb]

So how safe is it to use the Throttlestop application on this laptop having a 120W adapter and all?

The biggest concern was using ThrottleStop with the woeful 90 watt power adapters that Dell is using on many of their laptops like the XPS-1645 series. A core i7-720QM + ATI 4670 requires more than 90 watts during many games so a person can enjoy their laptop the way Intel and ATI intended.

I recommend that if you are using ThrottleStop you should monitor power consumption with a Kill-a-Watt meter or similar so you can see power usage as it is coming out of your wall. The Dell adapters are about 90% efficient so a 100 watt draw on the Kill-a-Watt translates to about 90 watts DC output from the adapter. Adapters are rated by how much DC power they can deliver to your computer and usually have an efficiency rating on them too.

A 120 watt adapter is going to give you more wiggle room than the Dell users have. Gaming on an XPS 1645 translates to about 100 to 110 watts from the wall. Every computer is unique based on GPU and other hardware so that's why it's best to measure your computer.

[ALLurGroceries]

I am relatively confident, especially after doing the test I ran in my second vid on page 1 again with my cheap hack to unthrottle in Linux, that it's not a power (PSU) issue. It will get to 120W but it never goes above 133, which is the highest I've been able to draw, and that was with copying files from the DVD drive while also running prime95+furmark. It was still stable then and I ran that load for 10+ minutes, more than long enough to "melt" the system.

I'll get back about the gaming results with throttlestop after I have some time to game, and thanks again for the tip!

[Cheeseman]

Let us know how it goes, cause I'm not confident enough to game with Throttlestop enabled.

[ALLurGroceries]

When I use i7z to monitor the CPU, I can see the changes from writing directly to the MSR.

When I yank out all of the ACPI modules, no cpu frequency scaling at all going on, I don't see the 4 second thing at all with my prog. Even on large FFT on mprime (prime95). When I add a gpu benchmark on top of that it starts throttling like crazy and my program is sometimes effective, sometimes it doesn't do anything at all. I check at .002 sec intervals and set it whenever it's not at 13. When this happens I can read the MSR and it says 13. Also directly writing the MSR by hand has no effect, since it's already apparently at the value I am writing.

I read in the intel whitepaper that if the voltage gets to a certain ("factory-configured") level it won't let the cores go higher:

When temperature,
power or current exceed factory configured limits
and you are above the base operating frequency,
the processor automatically steps down core
frequency (-133.33 MHz) in order to reduce
temperature, power and current. The processor
then monitors temperature, power, and current
and continuously re-evaluates.

It's not temperature. It doesn't seem to be power draw from the PSU (where is ryzeki with the 150W psu to test so we know for sure?!) Would that make it a voltage issue? This implies it's hardware, not BIOS, for the throttling. If the CPU needs more juice it would of course have to come from a BIOS update.

[ryzeki]

Sorry, real busy right now.

I didn't have much time but running TMonitor, Furmark and Prime 95 all together gets me 933mhz speed on all cores.

Running Prime95 8 threads custom alone runs all cores at 1733mhz but they show drops to 933mhz (sometimes 1 core, sometimes all 4 simultaneously) for a split second every 4 seconds or so.

Tomorrow I will try more tests though why are the cores running at 933mhz when I try furmakr+ prime?

[ALLurGroceries]

Can you try running furmark+prime95 with throttlestop?
http://www.fileden.com/files/2008/3/...rottleStop.zip

Make sure to check 'set multiplier' and set it to 13... then enable logging and plz post your .txt. If it matches the rest of ours we can finally rule out the PSU for good.

[ryzeki]

Can you try running furmark+prime95 with throttlestop?
http://www.fileden.com/files/2008/3/...rottleStop.zip

Make sure to check 'set multiplier' and set it to 13... then enable logging and plz post your .txt. If it matches the rest of ours we can finally rule out the PSU for good.

I downloaded throttlestop, tried furmark+prime small ftt, furmark with extreme burning test 1440x900 4xmsaa, enabled throttlestop with multiplayer 13....

And the CPU still throttled to multiplier 7.0 as soon as I fired furmark. Soemthing I am not doing right?

[ALLurGroceries]

I downloaded throttlestop, tried furmark+prime small ftt, furmark with extreme burning test 1440x900 4xmsaa, enabled throttlestop with multiplayer 13....

And the CPU still throttled to multiplier 7.0 as soon as I fired furmark. Soemthing I am not doing right?

No... thanks. That is exactly in line with the results we have been getting. Just means the 150W psu is not making a difference. I think the next candidate is voltage. I am no system architect but the intel turboboost paper implies that.

[ahmadmud]

ok, I don't know if this is related or not, but while I was playing RE5 online, I noticed a drop in performance, and then when I check GPU-Z, I saw this:
2010-01-21 19:27:12 , 600.0 , 950.0 , 1500.0 , 61.0 , 87 , 0 , 1 , 0 ,

2010-01-21 19:27:13 , 383.0 , 301.0 , 767.0 , 61.0 , 87 , 0 , 1 , 0 ,

does any body have any idea why it's downclocking itself?

[ALLurGroceries]

ok, I don't know if this is related or not, but while I was playing RE5 online, I noticed a drop in performance, and then when I check GPU-Z, I saw this:
2010-01-21 19:27:12 , 600.0 , 950.0 , 1500.0 , 61.0 , 87 , 0 , 1 , 0 ,

2010-01-21 19:27:13 , 383.0 , 301.0 , 767.0 , 61.0 , 87 , 0 , 1 , 0 ,

does any body have any idea why it's downclocking itself?

That's probably powermizer, you can disable it in the nvidia control panel.