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

Huh...I'm late seeing these. Not badly priced.

So it looks like the fastest CPU includes a 400 core slow clocked GPU, and then for $100 more you can add a second GPU?

Do those work in Crossfire? It doesn't really tell you what the second one is, or how that works. If it's two 400 core GPUs, a solid CPU, and Blu Ray for $900...that ain't too bad.

Do AMD's normal drivers work on these? I assume so. I won't buy AMD + Intel because of the driver issues.

 

Ordered mine on 6/28, was supposed to ship today, no luck on the tracking info and after reading this thread I don't think it's going to be shipped anytime soon.

 

Wow that's quite a reversal from when I was ordering.

10 days ago the DV6t was expected to ship on the 29th due to screen shortages and the DV6z was expected to ship on the 14th...

 

The crossfire is with the built-in AMD APU (the whole chip is a hybrid CPU and GPU) working together with the 6750M.

 

Okay, cool, so it IS doing Crossfire then. I just wasn't sure if you were getting a 6750 and then it was just shutting off the internal 400 core thing...

Hmm...so that leaves you with a dedicated 480 core part, and an integrated 400 core part with no video RAM? How do those work together? That still speeds things up?

And this all works with AMD's reference drivers, right?

With maxed out GPU(s)/CPU, 8GB, and Blu Ray, it's like $900, which really isn't bad, and would finally let me try AMD again.

Theoretically I guess this should even get you up to high end GPU performance...theoretically...if Crossfire does it's job well enough and the normal drivers work and no microstuttering or whatever.

 

To some extent, yes, but not nearly as much as the hardware suggests. It's not like ordinary CrossFire where you can get almost the entire performance of the cards. Precisely because of the different memory bandwidths, it's asymmetric CrossFire which gives you much less than that. Also, because the CPU part of Llano is so pitiful, you end up being CPU-bound in some games and then the extra graphics doesn't help you.

Eventually, yes. At the moment, it doesn't work with DX9 games (you actually get lower frame rates). I don't think it works with OpenGL either (AMD's site suggests that OpenGL always runs on the integrated GPU, regardless whether it's Llano or Sandy Bridge).

 

Oh, okay...but the reference drivers still at least install, right? (No hacks or weirdness?)

I skipped the Intel + AMD systems after learning normal drivers don't work.

 

is this Memory KHX1600C9S3P1K2/8G good for the new hp dv6zqe [config: AMD Quad-Core A8-3530MX Accelerated Processor (2.6GHz/1.9GHz, 4MB L2 Cache)]

 

You are very ignorant and your "grey box" example reflect your lack of knowledge. In this era of multicore systems, unresponsive program windows is not indicative of poor CPU performance because the operating system would split the work load across multiple cores. Web browsers would most likely only consume one core, leaving the other cores alone. Only in extreme cases where a program utilizes all cores would your assumption be correct. However, in your example, there may be a myriad of possibilities at fault such as an unrecoverable crash, memory leak, or software conflict. I could code a simple console shell which does an infinite loop that will crash the system like you described but that does not in anyway indicate poor cpu performance. Programs with poorly design exception handling will always tax any processor. Please stop trying to spread propaganda using your limited experience.

And for your definition, average users would be consumers who use their laptops for web services. If you believe average users today buy laptops for offline activities such as mission critical encoding, rendering, or compiling, then you are beyond out of touch.

 

You're contradicting yourself. My OS is never unresponsive, but Firefox does become unresponsive at times. Therefore, single core performance is extremely important in getting rid of such things, which is precisely my point and where Llano is way behind on Intel.

Nope, this happens with Firefox and IE9 on some sites. It's not an "unrecoverable crash" when you can just wait a bit and it's back to normal, is it?

Here is what happens:
1. I'm on a webpage with AJAX, this happens sometimes (rarely) even with a simple site such as twitter or google maps. I might have some other tabs open, such as some newspaper sites. Those don't tax my computer since I got adblock, but they tax the C2D cpu of my friend's macbook pro because he doesn't have adblock, and all those (non-video) flash becomes a bit too much to handle, spooling up the fans on the mac. Average user won't have adblock and flash heavy sites will be slow on a slower CPU.

2. Page becomes unresponsive and I get the grayed out/faded window when I try to close it and windows tell me the program is taking too long to respond. In the mean time, I know CPU usage is high because the fans are ramping up. Bringing up the task manager confirms this.

3. Windows gives me an option to close the program or wait for it to respond. I choose to wait and it comes back, and is working normally after 10 seconds.

Now, this is purely CPU single-thread performance, because when the same happened with my C2D, the wait would be 20-30 seconds and most times I'd just choose to close the browser instead of waiting.

Stop with this condescending tone, you don't know anything about how much experience I have. I see cases like this every day and I don't do mission critical compiling and encoding, etc. Just Firefox+adblock plus is enough to stress my CPU at times, depending on the website. A faster single thread CPU makes for a faster user experience and you don't need specialized programs to see it. A faster CPU helps with even browsing the web, because I guarantee a lot of the sites won't be optimized or have flash or other features. Since software is a lot harder to get right and bug-free than hardware, I'd rather have a fast CPU that can plow through anything than to count on well optimized programs and websites.

 

I can honestly say I don't experience this, even with my E-350. I will sometimes experience a slight slow down, though even that is rare, but Firefox has never just frozen like that for me no matter what I am doing. And I am not using AdBlock, either. With my previous Danube CPU, though, I never even noticed a slow down. If your SB laptop is doing that then there is something else wrong.

 

Happens to me sometimes on my 1.73 ghz C2D laptop using chrome. Though I'm not sure whether that's because I need to reinstall Windows or because of the processor itself.

 

It depends on what other sites are being visited. A poorly written Java script can overload even a high-end desktop CPU. I haven't seen ones that bad in a while as presumably people have gotten better at writing them over time, but it is certainly possible. One of the nice things about having decent hardware is that it serves as a buffer against random, incompetently written software.

 

I guess I don't visit sites with poorly written code, then. I would imagine the average user doesn't either. Back on topic, Notebookcheck has a review of the dv6 with A6-3410mx/6750m

Google Translate

According to them, the CPU stayed between 1.6 and 1.7 no matter the benchmark.

 

Hm, interesting.

It seems like the Llano based DV6s run cooler and quieter and have slightly more battery life (except when watching movies).

But the i7+6770m combo performs better in both the CPU and GPU department.

I'm surprised that it couldn't use turbo more often, even with the extra 10w tdp.

 

Me too. The Crossfire, though, I think will get at least to the 6770m level soon enough.

 

Yeah, I hope AMD gets their stuff together and fixes the driver problems with all of the DV6 models.

 

I guess it's different expectations. I expect 1-second response time from any task I do that's not internet-speed limited and I get impatient and start clicking around. If I don't touch anything the tasks would probably finish in 5-10 seconds but I don't like waiting for that. Having a fast CPU and a SSD spoiled me. There is nothing wrong with my system and I have a clean windows install, etc.

I also have a Windows XP VM, Outlook, and at least one other Office application open. Also, miranda64 sometimes messes up and starts hogging the CPU...I'm glad I have the horsepower reserve to handle such situations.

If you could get a 300hp car for just a little bit more than a 150hp car and it wouldn't burn more fuel, wouldn't you go for the 300hp car if all else was equal? That's how I look at CPU's. Sure I don't need all that power all the time, but it's good to have it when you need it.

 

A little surprising, but not that much. Check out the voltage on that thing: it requires 1.1V just to hit 1.6GHz. I shudder to think of what you'd need to get to 2.3GHz (remember, voltage hurts your power draw quadratically). For comparison, the 2720QM needs only needs 1.176V to run at 3GHz.

 

Again, even providing the CPUZ is correctly reporting the voltage on Llano, which might not be the case given that it lists the max TDP as 76W, there should still be enough to run one core at the max turbo speed on the 35W APUs, two on the 45W APUs. And if 76W is the max TDP like CPUZ says, very unlikely, then that leaves enough room to turbo all 4 cores. It is more than a little surprising that Turbo Core is not working.

 

It's reading the TDP from a table which is wrong. The voltage is reported dynamically.

In any case, there is definitely thermal headroom for at least one core -- the GPU on the 3410MX is only 320 cores and they're clocked at only 400MHz (i.e. a 20% and 10% reduction from the 3500M respectively). Even with that, the TDP is somehow higher (45W rather than 35W). It's not clear what is constraining AMD, but it's not thermal headroom.

 

The solution is simple, and it can be deduced from AMD's own blog.
CPU's until recently have been tested very conservatively (all transistors switching simultaneously to determine TDP, which will never happen in real life) and Turbo Core is a method to unlock more performance when there's more TDP headroom.

Intel does this by measuring the heat dissipation directly and overclocking the CPU accordingly, while AMD does it by estimating how many transistors would switch based on workload and calculating heat dissipation. This is the first generation and AMD's calculations are very conservative, thus not much overclocking is going on, that's the first theory.


The second theory is that the weak Turbo might be on purpose since we all know K10 is not very efficient at higher clocks so the chip was optimized (transistor sizing, layout, etc) for low power at sub 2 Ghz clocks and it'd be very inefficient above that so the Turbo Core was designed to keep that at a minimum.

 

I think it is the first theory, or something similiar.
Look:
CPU Overclocking vs. Power Consumption. Page 17 - X-bit labs

As long as you don't up the voltage, the efficiency actually increases (the total W * h).
Excluding games, flash and similar workloads because they don't actually end, unlike, for example, rendering, where once it finishes rendering, it idles.

 

You have to up the voltage to run at higher clocks though, that's at least what my SB does:




I find it very impressive that a chip can increase its clock by more than 4x when needed...What's the lowest clock speed of Llano when it's idle?

 

I think you mean that it estimates how many transistors are active and calculates the power consumption. Heat dissipation and power consumption are not the same thing. Still, even given incredibly conservative estimates of consumption, using the consumption numbers from Danube and early Llano, show that there should be room within the TDP for the 45W APU's to overclock at least one core. Even if they were estimating that half of the transistors on each core active, 45W should be enough for one core to be at 2.3GHz.

Another problem is that AMD's slides lead us to believe that Llano can exceed the TDP as long as it doesn't exceed temperature limits. It is running at 75C tops in Notebookcheck's most strenuous tests. Unless the temperature limit is set absurdly low, that is not what is stopping it from turbo.

 

Nvidia's current gen GPUs clock down to 50mhz! That's awesome/insane IMO!

 

So has anyone received a shipping notification yet?

 

They are pretty much the same thing when computing how much turbo to use, and heat dissipation is easier to measure than power consumption though.

I don't disagree with that at all, so either AMD is ultra-conservative in its boost, so they might be calculating for the worst case that you're in Death Valley, or they don't want to use up too much power by boosting up the clocks and having low battery life, but then why not turbo when on AC power? Therefore the first case is more likely, they're just very conservative and assume all their chips are used in Death Valley in laptops with poor cooling designs.

 

You can get away with that in GPU's since it's not much work drawing a desktop on the screen, your computer would be very unresponsive if you clocked the CPU down much further.

 

Lmao, I ran the 5870 in my M15x at 25 core, 40 mem by the AMD GPU Clock Tool, for battery life, so thats not that impressive. Anyways, I think its safe to say I'm holding off on buying a Llano chip until turbo improves, if it does.

 

Shouldn't be any different. It can't actually do that at 50mhz, but it can quickly jump up as needed, if you move a window or whatever even. Could be similar on future CPUs.

 

I don't think that you understand how AMD's turbo is supposed to work. AMD's turbo works, or is at least supposed to work, like this,

1. Measure the load on each core individually
2. Estimate power being drawn, temperature/heat dissipation does not come into play as long as the chip doesn't exceed limits that would force throttling to lower than P0, based on load on each of the cores
3. If the OS is requesting the highest P-state, boost however many cores to the turbo speed that the difference in current estimated consumption and the TDP allows
4. If AMD's marketing slides are to be believed, the APU can exceed the TDP as long as it stays within the temperature limits mentioned in step 2

You seem to be under the impression that Turbo Core has something to do with the temperature, judging by your Death Valley comment, but it doesn't. It is based entirely on processor load and estimated power consumption. In this discussion, heat dissipation and TDP are not the same thing. When you said that AMD must have been ultra-conservative in their calculations, I thought you meant that they were estimating the power being drawn on the high end. That might make sense, though the wattage and voltage numbers from Danube and Llano still say that turbo should be better than it is. Your idea of them being conservative with the estimated heat dissipation is wrong because they aren't measuring or estimating the temperature to decide when to turbo and because the APU has a temperature sensor, or it would not know when it exceeds temperature limits and when it has to throttle down.

 

Ordered my dv6zqe 06/30. HP website stills says it should ship out 07/13 and I have not received a delay email. Crossing my fingers it stays that way.

 

I cancelled my order. The ship date is somewhat insanely long and I need it right away so will get mine on best buy. The deal is great enough and specs. Are decent. + it ended up cheaper even with a8 than a a6 on HPA.

 

Yes they do. Temperature and power draw are not independent. The higher the temperature, the higher the leakage current (technical explanation here). If they tested the chips in Death Valley or otherwise pushed the temperature close to 100C and assigned the power draw based on that, it would result in Turbo almost not being used at all. I don't understand why they'd do that, but it's a possible explanation.

 

I am not saying that temperature and power draw are independent, just that they aren't the same thing. Obviously if more power is being drawn then the temperature is going to be higher in the same machine. But one laptop could be drawing less than another and still be hotter due to the specific cooling system, ambient temperature, etc.

And obviously higher temperatures are going to reduce the efficiency of chips. But, given the numbers in the link you gave, AMD would have had to have been almost impossibly conservative for turbo to not be working. In the link they give the example that reducing the temperature of the CPU from 85C to 30C saves 7W. Even if they had pushed the chips to their thermal limits, 100C, to decide when to turbo it shouldn't be working as poorly as it is. That 7W is with an 8-core CPU running at 2GHz, Llano has 4 cores running at speeds from 1.4-1.9GHz. Even if they tested it at 100C and it consumed 7W more than it does at more realistic operating temperatures, and it probably would be less than that due to half the cores and lower clock speeds and because that 7W is from 30C to 85C and we are talking about around 70C to 100C, there should be enough room for turbo based on the consumption numbers from reviews. It could have consumed 10W more and it still should have turboed if AMD is to be believed when they say that it can exceed the TDP as long as it stays within temperature limits.

 

Sorry if this was asked before, but I couldn't find it.

The New HP DV6Z Quad Edition, which exact model are they? Because I wanna buy one, and I customized it with "1GB GDDR5 Radeon(TM) HD Dual Graphics [HDMI, VGA]". What does that means?

I saw AnandTech AMD Llano Notebook Review and they had the A8-3500M With 6620G and the graphic card 6630M and XFire...is that what HP is selling?

 

Yes that's what the dv6z is. The APU is cpu plus graphics card. The additional card allows for hybrid crossfire. When a discrete card is added it changes the graphics model number.

 

Ok, the A8-3500M has the HD 6620G...but what's the graphic card it brings when HP says "1GB GDDR5 Radeon(TM) HD Dual Graphics [HDMI, VGA]"?

EDIT: Since HP told me: "Actually, we do not have the accurate answer for that question."

 

The 1gb GDDR5 option is an AMD 6750m, according to the service manual. This can crossfire with the integrated graphics.

 

Excellent, so it would Crossfire with the integrated 6620g? right?

 

Yes. And AMD calls the 6620m + 6750m crossfire the 6755g2, just fyi.

 

Thanks man...I knew about the 6755g2...I'm looking at this at Anandtech and that's why I was wondering what Xfire was I gonna get...

 

Notebookcheck's full review is online:
Google Translator

 

They didn't quite get the same FPS as Anandtech...

 

It shouldn't - I think the Anandtech reviews are A8's and this is an A6-3410MX, which is cheaper, slower, and has 80% the gpu (or thereabouts).

Those A6's should have been triple cores with maybe a faster GPU and CPU speed rather crippled quad cores/crippled graphics.

 

Hi.Can you help me I want to buy laptop for college and find two interest it's Dell xps 15 with i7 2630 Qm and dv6-6135 with a8 3500m which is better for work and multitasking . Thanks.

 

The dv6 also has the 6750m with GRRD5, while Anandtech's unit was equipped with a 6630m with DDR3, which is at least 30% slower. I don't know why the Anandtech numbers are better but they shouldn't be.

 

It might be my bad reading or the automated translation, but I don't think they tested the 6750M - but I admit I'm not sure whether they're testing the 6520g or the 6750M when the crossfire fails. I think they are just testing the crossfire results (all the charts say 6755G2, which is known to be buggy for DX9 games) and putting in 6520g if/when it fails.

 

I can't tell what they are saying very well, either, but I assume that is what they are doing. But in games that they didn't mention a crash for, like BC2, the numbers are lower than you would expect. In Anandtech's review, BC2 was one of the games that benefited from Crossfire but that is not the case in Notebookcheck's review. The 6750m by itself gets 77, 62 and 39 on low, medium and high, respectively, while the 6755g2 gets only 64, 48 and 35. There might be some micro-stuttering but the FPS should be higher. It might have something to do with them using the default drivers, which is 11.3. An update to 11.6 might help.