Official Llano overclock/undervolt result

I am trying to test if it is really an unlocked multiplier.
Intel Burn In forces the CPU to calculate a large number, it is a CPU bound application which is used to test for pure CPU arithmetical.
If you haven't heard of it you don't really know much.
Apparently your CPU throttled from the results displayed.
It is used to test if setting above the max turbo multiplier works.

 

there148 I think he's asking you to prove him wrong by pushing the cpu further than it should be able to go. So far it seems like a 3ghz clock is very possible for this cpu. The next step would be to push it past stable and start getting errors or a blue screen showing the overclocks real.

 

There is a huge difference between the 32bit and 64 bit version of cinebench 10. I used the 32bit to allow others to compare with the list on notebookcheck which uses the 32bit version. My 64bit single core benchmark is actually 3.3k. Most review sites such as anandtech include a cinebench 10 result but do not indicate if it's 32bit or 64bit. This can lead to rather ambiguous interpretations.

 

3.3ghz @ 1.3750vcore

300point increase from 3ghz.

 

Man these Llano chips are awesome.

 

A couple posts above yours I posted screenshots of IntelBurn at 2.6GHz and at 2.8GHz and there was a difference.

@there148 - I didn't even notice the 32-bit. That proves it is unlocked past the turbo speed. Like I said in the dv6 Llano thread, it looks like the only thing holding it back is the cooling system. With a separate fan for the CPU and GPU and some better thermal paste the quad-cores could likely run constantly at 2.8-3GHz, maybe more, with no heat problems.

 

If I end up keeping my HP, I will do a repaste with IC Diamond.

 

If you do that...could you take pictures and upload? Maybe even do a little how-to? I'd happily donate some BTC for a good guide on re-pasting

 

Absolutely will do a how-to. It's just unfortunate the OEM's can use a better paste and do a better job. It's all automated for crying out loud! I've always had significant improvements by repasting the CPU and GPU myself.

 

A new version of k10stat just came out today. The version is now at 1.53. I decided to test this version out, no new feature, probably a bug fix release.

I managed to increased my old overclock by another 100mhz. Seems to pass super pi without a hitch. Currently testing to see if it's world of warcraft stable.

 

Hey There (sorry, couldn't resist) Didn't your machine have only the 1 4gig stick in it? did you ever pick up another to enable dual channel memory? Reason I ask is that memory bandwidth is everything to these llano chips.

Seer

 

what about games? does it affect gaming is there a significant boost on fps?? jsut by looking at the performance and price i am definetely ditching my lenovo t61 for tha 379dls hp

 

sorry for double post but has anyone tried that program with a zacate APU? maybe i could work and get better performance out of those APU as well

 

Does anyone know the difference between the AMD thermals and power reading on CPUID hardware monitor vs Intel ?

When looking at the "powers" section of Hardware Monitor, I notice that sandy bridge has some super low wattage 3w during idle, whereas our llano shows 16w..... all the way up to 64W.
There's no way this thing goes up to 64W right? the TDP is 35w ?

I'm gonna steal someone else's link here for reference-
http://img683.imageshack.us/img683/8836/cinebench24ghz1075v.png


I tested the core i3 sandy for battery life during idle, and it lasted 7 hours on a 48whr battery,
whereas my llano A6 lasted only 6 hours on a 56whr battery during idle.
both with the same power settings. I presume it's because the core i3 has a super low power state during idle.

I thought llano was supposed to have better battery mgmt results than sandy bridge ?
But I got less time out of a bigger battery.

 

By default voltages are set way high on the A6/A8 CPU's. If you adjust them to lowest stable voltages you will probably be able to increase your battery life by a decent amount. Plus the A6 is a quad core vs. i3 dual core.

Also, did the i3 system have one or two RAM chips? With low voltage laptops having one RAM chip vs. two can reduce power consumption by maybe 1W which if system normally only consumes 6-8W on idle can make a significant improvement, as can the type of HDD and SSD.

 

both my i3 and a6 have only 1 stick of ram.

http://img683.imageshack.us/img683/8836/cinebench24ghz1075v.png

this guy has lower voltage setting, but hardware monitor reports higher wattage than mine.
my default p6 state is .9375v.

Does hardware monitor report powers incorrectly ?

 

Yes it does. Use HWInfo, BatteryMon or BatteryBar to figure out consumption. My whole system consumes about 10-14W idle and 80-85W under load, that is with the 6750m.

 

is that at the wall socket?
or is that what is reported by HWINFo ?


I did lower voltage of the lowest p states by 0.03v, and i squeezed out over 1 hour of extra battery life.
7 hours with 10% battery left, on a 56wh battery

 

That is what HWiNFO and BatteryMon report the total consumption on battery when idle as.

 

what do i need to overclock the a4-3300m in my hp pavilion g4-1117dx? thanks.

 

nevermind. i found the "enable clock control" option in the k10stat sys tray.

 

You don't even need "enable clock control". Just set your P-states and voltages and let it go.

 

Picked up a g4-1117dx for a friend and I thought I would post the results of the overclock, though slight, here. I also switched the 1x4GB of RAM to two 4GB sticks and it improved the graphics performance a lot.

3dMark11, 1 x 4GB, 1.9GHz

3dMarkVantage, 1 x 4GB, 1.9GHz

3dMark06, 1 x 4GB, 1.9GHz

3dMark11, 2 x 4GB, 2.4GHz

3dMarkVantage, 2 x 4GB, 2.4GHz

3dMark06, 2 x 4GB, 2.4GHz

 

Here's all my Asus K53TA results. I've since been able to run 2.3GHz stable at 1.0500V and temps reasonable. I also can run at 2.53GHz at 1.1500V but temps reach 90C:

Spoiler

 

Damn my K10 M520 Caspian can't overclock.

 

I can't make the clocks stick anymore when I choose High Performance power options in the control panel.

before, it would automatically jack up to P0 state, and stay there, but now it just stays at P6 unless I close k10stat.

Then, somehow, with k10stat closed, it will then stay pinned at p0 state.

basically, it works AFTER I close k10stat.

anyone know what's going on here ?

 

@ link626

(1) Double check your advanced power profile in Windows and see if min CPU is still at 100%

(2) Is K10stat in "control" of your clocks or is control disabled?

(3) Try clicking boost check box, then click again, and click apply and see if it works. Sometimes K10Stat isn't doing anything unless you apply the settings.

 

Can someone please to explain me how to use K10Stat for undervolting? Actually it seems work well, but something is wrong. I get error massage only below 0.9V what is unreal...
Here is screen-shot:

 

AMD CPU Undervolting tutorial - ?ást 1. - YouTube

 

Thanks, I do correctly but seems K10Stat do not support well AMD LLano. It changes voltages, but do not change MHz. Might change MHz as well, but won't after a certain MHz. I'm still confused with this undervolting / overclocking. Is there any program which will show me the real MHz of my CPU?

 

Regards to the newest version Aida 64 my highest CPU clock is 1730MHz. Whatever I set with K10Stat it won't be more. Maybe the Vostro 3555 cannot handle well the 3530mx APU? What can be the problem?

 

The only issue I'm having with K10STAT is, when running it as a system process at startup, I can't get the taskbar icon to show up at all. I have the -StayOnTray argument enabled, but the only way to get the icon there is to manually launch the program. When I boot up, K10STAT works perfectly. It sets the profile and clock frequencies appropriately (verified with CPU-Z), and it even retains those settings after resuming from sleep.

I followed the step-by-step guide for launching the program as a system startup process, which is here, but no joy in getting the taskbar icon to appear.

If there's anything I'm missing, I'd appreciate someone pointing it out.

 

It won't be on the tray when running as a system process at startup. You have to set it under your account. Like it says in the guide,

 

I am a complete idiot. +rep for you, self-inflicted head slap for me.

 

NM. <ten chars>

 

There is no way it will be able to run at 2.6GHz @ .9V. With .9V the best you can hope for is about 1800, judging by what others have achieved. For 2.6GHz you are going to need to set it to 1.1V minimum, likely higher.

 

If 0.9V for only 1.8GHz, than AIDA64 probably right with the 1730MHz. I think something is wrong my system or K10Stat, because it should run at least on 1.9GHz...
How much Gflops do you get on 1.8GHz when you run IntelBurn?

 

So far my best result is in my sig: 2.8 Ghz @ 1.235v.

I usually run @ 2.6 Ghz with 1.15v for everyday tasks.

 

2.6GHz P0? Wow. I run 2.2GHz. I can boost it up to 2.6GHz but even with my cooling mod gets quite toasty.

 

Yeah. It really doesn't run that hot, usually under 80c unless it's something super intensive. 2.6 Ghz is the sweet spot.

 

I'm finding 2.2 is the sweet spot for me as well. Beyond that, it gets into diminishing returns on performance versus temp. My voltages have to be on the higher side in order to maintain stability. 2.3 at 1.1v is doable with load temps below 80, but there's little real world benefit for my usage. 2.2 at 1.0875 has load temps in the 73-75 range with Prime95. I can play Portal 2 and other Source-engine games smoothly at that speed. Even Crysis in DX10 with some tweaking via the Triple-C mod pack is above 40 fps the vast majority of the time. No point in putting even more strain on the CPU at that point.

 

I've been lurking around, reading the HP dv6z owner's lounge thread for a few days now, only half way through. Interesting stuff, though.
Might as well make my first post in this forum useful and show what my new Llano and I have been up to


Anyway, I got me a €660 ProBook 6465b from HP with the following specs:

A6-3410MX (1.6/2.3 GHz) with Radeon HD 6520G
1366x768 anti-glare screen
1x4GB of DDR3-1333
Western Digital Scorpio Black 320GB
CD/DVD-RW
Bluetooth v2.1, Gigabit Ethernet, b/g/n Wireless, USB2, 1xUSB/E-Sata combined port, etc...
Windows 7 Professional x64
Recovery discs for both x86 and x64 versions of the OS, as well as for applications and drivers, was included.


No discrete graphics, so it probably won't be all that great for newer games. Good thing I mostly play older games

Anyway, on to the K10stat results

These are the stock frequencies and voltages:


This is what I've saved as my overclocking profile:

Temperatures max out around 68 degrees Celsius in IntelBurnTest with these settings.

And these downclocked/undervolted settings will be for light use, train rides and such:


I haven't been at this for more than more than ~10 hours, so it's possible that there's still a little more to get out of this thing, but it's looking pretty good for my needs so far

 

Made a post in the HP Pavilion forum. Basically I can get 2.8GHz without turbo with 1.1750 with temps of 89C with intelburn test. That's with an HP DV6z-6135 with an AMD A8.

 

Overclocking the mobile Llano CPU using K10stat is one thing, but has anyone tried increasing the frequency of the FSB?

I suspect this would give a better boost to overall performance since the GPU and RAM-bandwidth will increase too (while beeing a more complex overclocking process though).

(A6-3410MX - HP Probook 6465b)

 

Just a note to those of you overclocking. IntelBurnTest does not stress the CPU that much. Try using Prime95. And for ultimate stability, run Prime95 (large fft) + MSI Kombustor (be sure to select DirectX 9,10, or 11 as it defaults to OpenGL). This will put an ultimate stress and temperature on your machine.

 

Here are my results of A6-3400M APU:


Tested with Prime95

The overclocking really brings great performance increase for CPU intensive programs, but nothing for games. At least not for new games, might for some older CPU intensive games.

Base clock;
IntelCPUBurn: 15 Gflops
3DMark2001: 16493
3DMark2003: 14325
3DMark2011: 666

2.4GHz Overclock;
IntelCPUBurn:25,5 Gflops
3DMark2001: 19706
3DMark2003: 14610
3DMark2011: 670

I think the 35W TDP is the problem, the CPU cannot keep run on 2.4GHz when the GPU need to work as well.

 

Hi,

It's very interesting thread. I have also APU in my notebook and thanks to this tread I can fully control its voltage and frequency. I have only one question. Is it possible that the same APU in the same type of notebook has different default voltages in each P-state? Furthermore, significantly higher in one of the same notebooks?

 

Games like Bad Company 2 and Battlefield 3 are quite CPU intensive that the faster clock helps a lot. But most other games will be more limited by the GPU than anything else.

But good overclocks/volts there.

 

I would like to see (somehow?) which P states using the CPU during gaming or 3Dmark11. Probably when the GPU is working the CPU cannot run P0 or P1 State, but rather switch to P5-P6 to keep the 35W limit. I'm not sure, but this could the explanation why I do not get any performance increase in 3DMarks when I'm overclocking. Perhaps those who has discrete graphics card, the CPU overclock matters a lot in gaming as well.

 

It will run at P0, only resources from one core is dedicated to GPU, but not the full core. Run HWInfo64 and it will show you real-time statistics of your cores. But the Llano CPU has a strange self throttling "feature" where it will clock all cores lower when certain conditions are met. I haven't 100% figured it out, but as you set your P0 closer to the designed turbo speed, it is a lot more likely to happen, even with reasonable temperatures.