*HP dv6z AMD Llano (6XXX series) Owners Lounge*

[Althernai]

But they are supposed to be able exceed their TDP rating, according to AMD, if they aren't in danger of reaching thermal limits.

What? I don't think that's how AMD's Turbo works. It stays within the TDP to avoid being dependent on the environment.

In the demo for the A8-3510MX we saw it occurring. The power consumption was about at 50W the whole time.

Sure, but that 50W is the entire system -- including the display, the motherboard, the RAM, the hard drive and so on and so forth. The display alone is more than 5W.

Let's say, that at 1.5GHz each core is at about 5W consumption, again based on Danube. So that would be 5 x 4 + 10(for the IGP) = 30W. And let's say that a single core at 2.4GHz consumes 8W (5/x = 1.5/2.4, cross-multiply and divide to get x).

This is not how it works. It's true that the power consumption scales linearly with frequency, but it also scales quadratically with voltage. If you look at xxkinetikxx's very helpful post here, you will see that in order to reach that 2.4GHz frequency, the voltage must increase from 1.037V to 1.300V. This means that on top of the 1.60 factor from frequency scaling you also have a (1.3/1.037)^2 ~= 1.57 factor from voltage scaling which leads to a grand total factor of about 2.5. In other words, your 5W becomes not 8W, but instead 12.5W.

All of that said, I kind of agree with your general point despite the math being wrong. Before reading any reviews, I personally expected Llano to be able to keep at least one core at the Turbo frequency, but that's clearly not happening. I don't know why this is the case as it would seem that even ignoring the GPU, the CPU alone using 1 core rather than 4 should be enough to offset that 2.5 jump in power consumption. It should work... but it doesn't. That's an experimental fact demonstrated by both reviewers using machines sent to them by AMD and by users with ordinary laptops from store shelves. I suppose the Turbo Core in MX Llanos can be different, but at this point my expectations are rather low.

[jiggawhat]

I love AMD and I really hope they make a legitimate mainstream part in Trinity, but if I have to be real about it, these are the facts:

1) At the moment the 6750m + llano does not have greater graphics power than the 6770m, though it is possible in the future. Even if it does happen I don't really think it can be *that* significant.

2) It has significantly worse processing power, so much so that I believe even with the most generous crossfire drivers it will still be lacking. Games like Battlefield 3 are known to be heavily multithreaded, and most traditional games require heavy single threaded processing, so Intel wins out.

Get a dv6zqe for these reasons:

1) You want to save money (spend around 700 bucks, no more than 900).

2) You require long battery life and/or cool temperatures.

3) You require OpenGL applications. Until the 6770m gets fixed this is the one to get.

Get the dv6tqe for these reasons:

1) You require heavy processing power for video encoding, compression, things like that.

2) You require superior gaming performance in DirectX games.

3) Your budget is under $1100 but greater than $900.

Get the ASUS G53 for these reasons:

1) Your budget is at or over $1100. $1099 is the cheapest price I have seen for a well-configured G53. Don't pay more than $1300.

2) This is a complete desktop replacement laptop. You do not need portability or battery life.

3) You want to game on the highest settings, but don't want to spend money needlessly. GTX 460m is the best video card you can get without getting into the area where the price/performance ratio skyrockets.

Replace all of these with dv7t or g73 if you want a 17" screen, but its my personal opinion that it is not worth it.

Also make sure you get the 1920 x 1080 screen upgrade for all of these models (comes standard with the ASUS).

[abaddon4180]

What? I don't think that's how AMD's Turbo works. It stays within the TDP to avoid being dependent on the environment.

I remember reading somewhere, don't remember where, that it can exceed the TDP as long as it still remains within temperature limits. This slide would seem to support that. It may be limited based on temperature to keep chip within TDP, according to the slide.

Sure, but that 50W is the entire system -- including the display, the motherboard, the RAM, the hard drive and so on and so forth. The display alone is more than 5W.

I was under the impression that those were just the consumption numbers for the APU? If not, the MX APUs seem to be suffering from the same problem as the M-series.

This is not how it works. It's true that the power consumption scales linearly with frequency, but it also scales quadratically with voltage. If you look at xxkinetikxx's very helpful post here, you will see that in order to reach that 2.4GHz frequency, the voltage must increase from 1.037V to 1.300V. This means that on top of the 1.60 factor from frequency scaling you also have a (1.3/1.037)^2 ~= 1.57 factor from voltage scaling which leads to a grand total factor of about 2.5. In other words, your 5W becomes not 8W, but instead 12.5W.

Even assuming that it is recording the voltages currently, which it might not be considering is gives 76W as the max TDP, at 12.5W per core at 2.4GHz should be enough to keep one core at 2.4GHz even if there was 15W set aside of the IGP. Again, based on Danube and the improved power gating, 3 idle cores should not be consuming more than a total of 7.5GHz. Given the information we have from AMD that the IGP and CPU don't have a separate TDP, it should be able to keep 2 there with the IGP consuming 5W.

All of that said, I kind of agree with your general point despite the math being wrong. Before reading any reviews, I personally expected Llano to be able to keep at least one core at the Turbo frequency, but that's clearly not happening. I don't know why this is the case as it would seem that even ignoring the GPU, the CPU alone using 1 core rather than 4 should be enough to offset that 2.5 jump in power consumption. It should work... but it doesn't. That's an experimental fact demonstrated by both reviewers using machines sent to them by AMD and by users with ordinary laptops from store shelves. I suppose the Turbo Core in MX Llanos can be different, but at this point my expectations are rather low.

My point exactly. Judging by the information that AMD has given us and the math, it should be working better than it it. Thus, AMD deceived us. Like I said, it may be fixable with a BIOS update, though.

[R3d]

You can get a i5 dv6t with 1080p for ~820 with the 30% off coupon, which will still have superior gaming performance.

The benchmarks so far show Llano battery life (not sure about temps) to be about the same as SB battery life.

[kevmanw4301]

You have to realize that this is a new product, and there are bound to be kinks. The Dual Graphics will be sorted out in driver updates, and maybe there will be a BIOS update. Also, whoever said that Crossfire scaling is less than 50% is sadly mistaken.

Conclusion : AMD CrossFire Vs. Nvidia SLI Scaling Analysis

Look at the 5850. 125% to 199% on CrossfireX. Drivers for ATI have only gotten better.

[soguxu]

Conclusion : AMD CrossFire Vs. Nvidia SLI Scaling Analysis

Look at the 5850. 125% to 199% on CrossfireX. Drivers for ATI have only gotten better.

You're reading that chart wrong. Base performance is 5830, which is 100%. 5850 is 122-125%, and crossfire 5850 is 180-199%. So Crossfire provided (180-122)/122 = 47.5% minimum FPS increase and (199-125)/125 = 59% average FPS increase.

So right around 50% for the absolute best case scenario with two identical cards, the benefits will only decrease with hybrid crossfire. It could be even more limited due to the thermal envelopes in a laptop.

[kevmanw4301]

Sorry, I did read it wrong. I've been trying to find a recent Toms Hardware article I read a little while ago that showed XFire scaling in the 80% range since drivers have gotten so much better since that article, but I can't seem to find it.

Edit: This one. http://www.tomshardware.com/reviews/...g,2865-10.html

[soguxu]

Sorry, I did read it wrong. I've been trying to find a recent Toms Hardware article I read a little while ago that showed XFire scaling in the 80% range since drivers have gotten so much better since that article, but I can't seem to find it.

Edit: This one. Multi-GPU Scaling Analysis : Triple-GPU Scaling: AMD CrossFire Vs. Nvidia SLI

Nice, but unfortunately that test doesn't even have a million selling game except the 4 year old crysis. They need some Call of Duty and UE3 games in there to see how it really effects gamers.

[R3d]

You're reading that chart wrong. Base performance is 5830, which is 100%. 5850 is 122-125%, and crossfire 5850 is 180-199%. So Crossfire provided (180-122)/122 = 47.5% minimum FPS increase and (199-125)/125 = 59% average FPS increase.

So right around 50% for the absolute best case scenario with two identical cards, the benefits will only decrease with hybrid crossfire. It could be even more limited due to the thermal envelopes in a laptop.

Yeah I'm guessing that dual graphics while have around 35-40% scaling with improved drivers, which would put it at 6770m level.

Though if we can't find a way to overclock the dual graphics the 6770m would still be at an advantage.

Edit: maybe around 50, since AMD has improved their drivers.

[abaddon4180]

The standard clocked 6770 is only about 15% better than the standard clocked 6750 at best. If dual graphics can increase the 6750 performance by 30-40%, it will be much better than the base 6770. If the 6750 in the dv6z can be overclocked as well as it has been in other notebooks, to about within 5% of the overclocked 6770, dual graphics could possibly beat the 6770 by 30%.