Official Llano overclock/undervolt result

It seems that AMD's new Llano cpu are completely multiplier unlocked and therefore can be overclock without any boundary other than temperature. This thread is dedicated to Llano owners who want to share or learn about the different overclock/undervolt results.

There will be no format, but I recommend 2 criteria for posting results.

1. Include a screenshot of k10stat which includes the info and the model of Llano you have. This can be found here. https://sites.google.com/site/k10stat/

2. Include a screenshot of superpi mod with a successful completion using the 1million digits benchmark. This can be found here. Download Super PI Mod v1.5 | techPowerUp

3. Optional but include information about highest overclock or lowest undervolt.

 

Here is my result.

My CPU is a dual core with base clock of 1.9ghz. I was able to undervolt to 0.8500 as the lowest pstate and overclock to 3.2ghz with the default boost voltage of 1.4125 and temperature was preventing me from going higher. At this clockspeed, performance is similiar to a 3ghz core2duo desktop cpu from the primarily results I've obtained. Not bad for a laptop that cost 379 dollars.

 

Can you give us the specs on your machine? That overclock is very nice!

 

Yeah, if you look in the HP Pavilion DV6z forum, users with the A8-3530MX quad are nearing i5-2410m for two core support. Still not near 2630QM for quad core, but definitely close. So far my DV6z with A8-3530mx, 6750m (6755G2), 6GB DDR3, 1080p matte 60% LG screen, 640GB 5400RPM HDD, Blu-Ray, 9-cell battery is costing me about $730.

 

Yeah but for $800 + delivery you can get a MS-16F2-ID1, pick up a cheap sandy bridge CPU and HDD and you have a platform for the future.

In fact that is what I have just done =)

 

What future is that? Battery life is terrible on that machine, and battery life is part of the reason people buy these HP's. But yes, that is a good option if you're looking for a powerful laptop on a budget. And for not much more you can get an NP8130. They will remove HDD and RAM for your and save you $100-$120 off price, can walk away with a basic NP8130 shipped for < $1100 with i7-2630QM.

 

Dual HDDs was a must for me.

 

This is the one I bought. HP - Pavilion Laptop / AMD A-Series Processor / 14" Display / 4GB Memory / 320GB Hard Drive - Pewter - g4-1117dx

A short review of this unit.

Pro: Cool temperature(on battery the temperature never goes above 45c even after extended flash viewing, without overclock the temperature never exceed 65c even with heavy use), low noise(option in bios to disable fan, it won't ever turn on with slight underclock/undervolt, completely silent even with flash viewing), long battery(I get over 5hours with wifi, 70percent brightness, browsing with some flash videos), cheap price(find another laptop that can overclock to a c2d@3ghz level for this price), overclockable(unlocked multiplier, k10stat reports it capable of 4ghz).

Con: None really. The only negative I can think of is they only gave 4gb of ram in single channel. Using dual channel memory is suppose to increase performance by up to 20percent. I will have to pick up 2 sticks of ram to test this out later.

Here's a cinebench 10 single/multithread score. CPU is at 3ghz, highest temperature is 71c.

Spoiler

 

If you have 4GB of ram, you only need to pick up one more 4GB stick.

 

Here are my settings right now, all fully stable and max temps under heavy load on both CPU and GPU are under 80C. Under heavy CPU load only temps are around 69-70C.



The original voltages were,

P6 800MHz - .9250 stock --> .7500
P5 1000MHz - .9375 stock --> .8250
P4 1200MHz - .9750 stock --> .8500
P3 1400MHz - 1.000 stock --> .8750
P2 1600MHz - 1.0250 stock --> .9250
P1 1700MHz - 1.0500 stock --> .9500
P0 1900Mhz - 1.0750 stock --> 1.075 @2.4GHz (500MHz increase at same voltage)
B0 2600MHz - 1.300 stock --> 1.15 (same turbo, lower voltage)

The highest I have gone is 2.8GHz at 1.2V but I don't keep it at that because just a CPU heavy load pushes temps to 80-82C, and I don't like them that high even if it is stable.

 

Ok, how do those P states work? Is it based on CPU load or something you manually choose?

 

Really impressive! I never thought that AMD´s new platform would be that easy to oc and undervolt!

 

Based on CPU load. You see in the screencap where it says the up and down %? You can change those so it goes into the next p-state when you choose.

 

Nice overclock / undervolt.
Never thought i would see a cool running HP Pavilion notebook.
It uses a single fan cooling system?

If only there was a Clevo system with AMD A8-3530MX for that extra cooling power.
OCing to the max. ^^

 

Yeah it uses a single fan. Dual fans would allow for some great overclocking as using the GPU and CPU together causes a big spike in temps. If they were cooled separately it would probably run at 2.8GHz just fine.

 

I agree. Clevo always seems to lag with innovation. I would greatly like a similar system made by Clevo than HP.

But you can run at 2.8GHz right? Just CPU runs at 80-82C? To be honest that's not that bad. Heck my 2720QM I had was running near at 80C at times.

 

If I run at 2.8GHz and stress the GPU and CPU at the same time, BC2 for example, temps can almost reach 90C.

 

You can overclock the 6750m though right? But not in crossfire mode? MSI Afterburner works for you with that? And still not clear what the %up and %down means. But guess I'll figure it out when I get my laptop.

 

Up% means the % load on the processor that will forced it into the next p-state and down % means the same thing, only it goes down instead of up.

 

Ok, so 60% additional load will push it up to the next P state, but 20% drop in load will drop it down to the lower P-state. Gotcha. My machine has a ship date of Aug 2, I'm hoping it will ship this week though.

 

My i3 380M runs @ stock clocks @ ca. ~ 80C (full load). Sometimes it even reaches 85C but my notebook will shut down @ 90C to prevent damage (BIOS setting).

 

Slightly off-topic, do you think you'd get significantly better OC temps if you managed to re-paste the CPU cooler? (i.e. would it be worth the effort of disassembling/reassembling and voiding the warranty?)

 

Possibly. HP, and most manufacturers, usually use the cheapest thermal pastes around and apply them fairly poorly.

 

Definitely. I wouldn't surprised if you saw a 10 c drop.

 

Good to hear, I ordered a dv6z quad w/ the A8-3510MX (scheduled to ship Aug 9) and I'd like to take a whack at re-applying the thermal paste then. In the mean time, if anyone's feeling brave enough...

 

I had time today to do an extended battery test and here are my results.

I marked the screenshots with arrows indicatng the start time and current time. I started at 6:46am and ended at 10:40am. 4hours has passed. I still have around 25percent battery life with over 1hour of runtime according to windows.

Usage was not simulated. I actually used the laptop and not with some synthetic benchmark. I turned on a streaming internet radio at the start that played audio through the speakers the whole time of the test. Cpu usage was in constant 20-30percent throughout, probably spiking when loading new webpages. Screen time out occasional when I stepped away, but speaker and wifi usage were constantly running. The fan never turned on and temperature never exceeded 42c. Only sound was coming from the harddrive spinning.

Some optimizations were made thanks to the level of control allowed by llano and k10stat. Performance was set to the lowest pstate, running at [email protected] voltage. There is no performance penalty that I noticed while browsing at this pstate.

Overall, battery life is nothing short of superb. Llano has no equal in this price range in my opinion. Intel has nothing to compete in terms of temperature, noise, or battery life for this price. AMD has given user complete control of their hardware and performance is entirely controlled by the user. Llano can be overclocked on ac and undervolt on battery.



Uploaded with ImageShack.us

 

So over an hour remaining after 6:46 to 10:40am? That's four hours + an extra hour. Assuming you have the 6-cell battery? Five hours of real use on a 6-cell is phenomenal.

 

<.< way off topic now guys, lets keep it to the performance of the Llano. Regardless of battery life this threads pretty interesting since the cpu is completely unlocked.

 

Yeah. Posts have been deleted. Take it to PMs if you want to discuss other stuff.

 

I do intend on doing a full comparison of this HP Llano with the Sager NP8130 and see how well it really stacks up. Considering the significant cost savings + battery life, if it can manage the newest games at reasonable framerates it may be a home run for lots of people.

What I'm still struggling with is the performance of the CPU itself. It's a quad core, yet it still struggles to manage Core i5 performance with three or four threads.

 

Have to keep in mind fusions been delayed for how many years now? Can you imagine if this tech had been released even a year ago how much more competitive it would have been.

 

Like I said, SB is way ahead of it in performance-per-clock, which helps a lot, and it has hyper-threading, which helps a little more.

@rorschach - If this had been released with Arrandale then it would really be something. Even with a small overclock it would match the highest Arrandale, and even Clarksfield, with 3+ cores, and 1-2 core performance with my 2.4GHz overclock is right around low-mid Arrandale. Add in graphics performance that makes Arrandale's IGP look like Ion and better battery life than Arrandale at the same price and this would have been a grand slam.

 

In any case I guess it's all relative. Sandy Bridge CPU's seem to be super powered compared to even the previous gen Core i7. Core 2 Quad CPU's are still plenty to manage most tasks well especially gaming.

 

Exactly (10 char).

 

Are you sure it is overclockable?
Numbers changing on K10STAT means nothing.
Even on previous AMD Processor the multiplier is adjustable but no improvement in performance is detected when you set the multiplier higher than the maximum multiplier. Only when you set it Lower then you can see the change in performance.

 

Weinter they are using benchmark tools to confirm the increase in performance.

 

I already went through their screenshots.
No concrete evidence of testing and results.
If you want to test, 2 things to test.
Multiplier @ Turbo Speeds if it works that means multiplier can be changed up to turbo.
Multiplier > Turbo Speeds if it works then the CPU is truly unlocked.

 

Read these threads:

Start at about page 85 here:
http://forum.notebookreview.com/hp-pavilion-notebooks/588705-dv6-now-shipping-amd-llano.html

and at abbadon's review:
http://forum.notebookreview.com/hp-...6755g2-undervolting-overclocking-results.html

 

I think you may be right. Even when the multiplier is adjusted to exceed the turbo speed, and CPU-Z reports it as so, the benchmark scores aren't as high as one would expect. For example, my Cinebench 11.5 scores at 2.8GHz should be around 3.5. 1.88 (score for A8-3500m) / 1.5GHz (A8-3500m clock) * 2.8 = 3.51. Instead it only scores 3.25, which is what you would expect at 2.6GHz.

Same thing is true for there148's scores. At 3GHz you would expect the Cinebench R10 single core to be around 3500. 3390 (Anandtech result for A8-3850) / 2.9 (A8-3850 clock) * 3 = 3506. Instead he scored 2700, which is about what you would expect at 2.5GHz.

 

Is there are linear CPU performance only comparison?
eg.
2.4 GHZ -> CPU Points
2.5 GHZ -> CPU Points
.....
Up to
3.2GHZ -> CPU Points.

 

So in other words you can't overclock past turbo speed? 2.6GHz? Even so fixed 2.6GHz isn't a bad deal either. Maybe make a few runs at different speeds and see if it scales linearly, like 1.9, 2.1, 2.3, 2.5 GHz.

 

Because I was the one who wrote the K10STAT guide.
I had AMD Processor Griffin in the past so I know it was possible to set the multiplier to insane numbers and the processor will still run stably but there would be zero changes in performance(Which means the real multiplier is stuck at the highest value burnt into the chip).
So I believe that AMD still limits the Multiplier to less than the Turbo Value so in that sense it isn't truly an unlocked multiplier.

 

If you want to check if it is overclocking or not just measure the Cinebench render times.

 

Well the question is whether or not it will go beyond the max turbo speed of 2.6GHz for the 3530mx.

 

No need.
Just set the multiplier and run Intel Burn Test.
If there is a real change the timing will still drop not linearly but would still be statistically significant.
Waiting for the results...

A even faster way is to set it to 5GHZ if it doesn't crash it means the multiplier is locked at Turbo, But I wouldn't be responsible for any damages done (if it occurs).

 

There does seem to be a change.





The Cinebench scores reflect a change from 2.6-->2.8GHz as well, though the scores at 2.8GHz are what you would expect if it was running at 2.6GHz and the scores at 2.4GHz are what you would expect at a little slower. Is it possible that HWInfo, K10Stat and CPU-Z are all reporting the actual clock speeds at higher than they really are?

 

Guess the clocks don't scale linearly.

You could try AMD System Monitor to get a reading of clock speeds. It's from AMD and supports fusion so I doubt it'd give a wrong reading.

 

So many incorrect assumptions. First this is not griffin or puma which uses k10 architect. This is llano, which uses k10.5 architect. Second, you may have wrote a guide for an older version of k10stat, which reports false results, but we are using an updated version that just came out a few weeks ago specifically created for llano. Third, I'm one of the few people who truly uses the cpu to its max potential on a regular basis. I can firmly assure you llano is completely unlocked and there is a correlation between performance and overclock speeds.

I've included a sample of my evidence. This emulator is looping an exhibition mode of a character performing their fencing style, measurement of the performance of the emulator is largely stable in this case.

I have included screenshots of the emulator running at 800mhz, 1400mhz, 2000mhz, 2500mhz, and 3000mhz. As you can tell by the fps indicator, the emulator is speeding up according to the overclock. Also bare in mind my cpu has a base clock of 1900mhz and it rarely turbo without intervention from k10stat.

800mhz

1400mhz

2000mhz

2500mhz

3000mhz

 

You need to show an Intel Burn Test result to check for accuracy.
When the clock speed up exceeding turbo the timing will drop.
Intel Burn Test is a more accurate tool because it measures only CPU performance.
There is no need to get fanboish about it when we are just discussing about performance.
Also it is logical for AMD to lock the multipliers because it is not Black Enthusiast Edition, so we are just investigating if that is true.
We need the data from 2.6GHZ onwards on Intel Burn In until the computer crashes to be absolutely sure (The CPU will reset upon restart).

 

You refute real world evidence and prefer synthetic benchmarks? I don't know what type of agenda you have, or what you're trying to disprove. Perhaps you don't believe the overclock is stable? I will amuse your skepticism. However I prefer to use real world programs as a standard for stability. I don't ever intend to worry if my overclock is stable in a synthetic stress test since I don't run my laptop 24/7 at 100percent workload.

Here is a result of the intel burn in test you requested. I don't know what this program measures but it told me my system was stable. Bare in mind running emulators does not produce these types of temperatures, nor games such as wow or witcher2. This program is good at producing heat. But then again my llano was able to maintain its speed, I can't say the same for a sandy bridge which I would assume would throttle.