Mussels Laptop review (worklog)

Edit: moved the review (in its near completed state) to this first post so people can find it easier.

1. Intro


The laptop we're going to be marinating and eating today is a delicious model from HP. its an exclusive variant of one of HP's regular DV6 models, for aussie retailers.

Its full model number is "HP Pavilion G6-1206AX" with some very vague specs available here:
http://www.harveynorman.com.au/hp-pavilion-g6-1206ax-laptop.html




These specs are pretty vague, and in the case of the video card*S*, rather misleading. more on that later.


*generic laptop pics from web go here - copies are in graphics artist united thread if needed*



2.Real Specs/First Impressions


First off, the real COMPLETE specs.

AMD A6 3400M quad core CPU. Its either 1.4GHz quad or 2.3Ghz *single* core, depending on turbo mode.

AMD Radeon HD 6520G 512MB integrated APU. This is not mentioned in the specs at all. *CPU-Z thinks its 1GB while GPU-Z says 512. i'll have to check.*
AMD Radeon 6470 HD 1GB Dedicated GPU.
These two cards run "AMD Dual Graphics", a crossfire variant. I'll dedicate a section to this later.

1x4GB DDR3 1333Mhz at timings of 9-9-9-24 (Mine was upgraded to 8GB when i bought it, because i haggle like a pro)
640GB Toshiba 5,400RPM SATA II hard drive.
A DVD drive that no one cares about at all.
Windows 7 home premium x64

One HDMI port, with full HD audio capabilities the same as Radeon 6K desktop cards.
One VGA port.
one 3.5mm headphone jack
one 3.5mm mic jack
Stereo Altec Lansing speakers.
3x USB 2.0 ports (one on the right near power cable, two on left between HDMI and audio jacks)
SDHC/MMC Card Reader. Drivers were for a PCI-E model, so it may well be a high speed reader.
Gigabit Ethernet (PCI-E based)
802.11N wifi - 150Mb - i can sustain about 4MB/s through this, FWIW.
15.6" 1366x768 screen. *screen is blurry at 1360x768, which is the opposite to my (supposedly) 1366x768 samsung HDTV. see note later on graphics page.
One webcam and mic - specs unknown. will add in an image later for visual quality, the mic seems good from testing but isnt worth making a recording.
Bluetooth. Hard to tell, but its either bluetooth 3.0 or 4.0 based on the drivers i installed. This worked well for my BT keyboard as a test, but has a delay of about 15 seconds
before BT devices sync waking from S3 sleep mode.


*self taken photos go here*


The laptop itself is actually very sleek looking with a very nice dark grey color scheme and a small, non obtrusive HP logo in the upper right corner of the screen with the laptop 'closed'. As a minimalist with looks, this suits my tastes perfectly.

The underside is flat and not raised at all, except for tiny rubber feet. this makes it sleeker and easier to fit into laptop bags, but also hampers breathing room on surfaces that arent perfectly solid - such as on a bed, or tablecloth.

The Screen does not seem glossy to me at all, so either its got a light gloss, or none at all. lighting does seem to reflect more clearly on it compared to my desktop monitor, but thats quite normal for laptops due to the angled nature of their screens (angled up, rather than perfectly vertical)

The keyboard is flat, with very angular keys. i had no trouble jumping to it and touch typing immediately, only the extra thin up and down arrow keys caused me some confusion when in the BIOS.

The keyboard also has a bunch of function keys for features that nobody will ever use, as well as basic media playback controls for pause/play, prev/next and volume control.
By default the function keys are the defaults and the F1-F12 keys are activated in combination with the Fn key on the keyboard - more about this in the BIOS section.

Battery life: Windows 7's estimates seemed fairly accurate on this hard to measure property of the laptop:
2D desktop with minimal use gets just under 4 hours, watching videos (in my case, 720p anime with subtitles and no hardware acceleration at all) gets about 2.5-3 hours, and

gaming is in the 1.5-2.5 hour range, varying wildly depending on the game.
Accurate power consumption results are listed on the 'power consumption' page.

Fan Noise: by default the fan always idles at a low speed, which is very hard to hear. outside of benchmarking i've never heard it spin up to higher speeds. you can disable this

so that it powers off completely, but i dont reccomend it - see the BIOS section of this review for more info.

In built speaker quality:
While they lack bass as any sound person would expect, you can max the volume out at 100% without distortion. i listened to some songs via youtube and it was always clear and
distortion free. the speakers are at the very front of the laptop which means your arms could muffle them if you are wearing heavy clothing while using your laptop outside in a blizzard. Just sayin.


Touchpad:
The touchpad on this unit is actually the best i've ever used. it does support configurable features such as multi touch, and setting zones for scrolling up/down and
horizontally via its software, but i didnt really care about using that.
What really stands out is the 'textured' design - its low friction compared to the laptop shell itself, so the second your finger slides off the touchpad you come to a screeching friction induced halt, so even using it in the dark you get some fantastic tactile feedback that
you've ran off the edge of the 'work' area.

There is also a small indentation you can double tap to disable the touchpad entirely, should you wish to do so fast and easily for a
big typing session. I know lots of people who screw up late night chat sessions because they hit the touchpad with their palm or wrist without knowing it and typed messages out
of order, or overwritten text before sending the messages.


Included Hardware: Just the laptop, battery and charging brick. nothing else included.

Included software: all neccesary drivers and a HP customised/mutilated version of windows 7 home premium x64. i formatted and reinstalled a fresh, clean OS first thing - and

found out that the drivers are nigh on impossible to find. If you get one of these laptops, make a recovery disk ASAP and back up any drivers you can early.


Graphics cards and setup:

It should be noted that the Catalyst control center (or AMD VISION center, or Graphics properties or... damnit AMD, make up your DAMN MIND ALREADY) there is an option to save

power by having the laptop automatically change the brightness of the screen to save power. This feature works perfectly for me, and i suggest leaving it enabled.
if you load a dark page, it will up the brightness - and the reverse, on a white image it will lower brightness to save power.

This laptop runs AMD's DUal Graphics technology, which boils down to:

One "low power" GPU for idle usage, 2D tasks and low 3D tasks like hardware accelerated web browsing or flash playback.
One "high performance' GPU for DX9 gaming, and to crossfireX with the low power card for DX10/11 gaming.

When you start a new 3D application, the drivers have a popup asking which GPU you'd like to use - low or high power. on this laptop its utterly meaningless, since both are the

APU. This means that outside of DX10/11 gaming, you are *always* using that GPU. if this baffles you, remember that this system was initially designed for intel based laptops,

with the intel IGP as the primary low power card.

One oddity is that this laptop has a 512MB APU and 1GB secondary, so i have no idea how that works in crossfire, if i'm restricted to 512MB, or if it uses system ram to up to 1GB.

Another problem i ran into was that by default it cloned a duplicate image between my laptops screen and my HDTV since they both have 1366x768 resolution panels - or so they
claim.
In reality my HDTV works at 1360x768, which caused some blurring on the laptop screen. The fix was easy enough, which was to set both to 1360x768 and then enable
'centered' timings for the scaling options in the AMD CCC/AMD VISION/god i hate typing that over and over again/etc. Of course its greyed out by default and unselectable, but if
you set the screen resolution to 1024x768 it enables the options and remembers what you select after you set it back to 1360x768 again. Yes its stupid, but dont fight AMD over it. last time i tried i got banned from their forums because the feature 'works as intended'




Benchmarks!

CPU Benchmarks first:

These are synthetic tests that dont hold all that much real world performance meaning, but they're very, very good for stock vs OC'd performance tests. if they dont scale as expected you can find performance bottlenecks such as system ram, or thermal throttling.

Superpi Single threaded test, to see how turbo mode compares.

Wprime to see how it fares in quad threaded apps (its just an expandable superpi that can use as many threads as you have)


Stock clocks:

Superpi: (2.3GHz turbo)
(paste image link)

Wprime (1.4GHz quad)
(paste image link)

Unigine Heaven was ran at default settings, i merely changed the DX level except for one DX11 test where i disabled tesselation.
All tests are ran on the inbuilt screen at native res of 1366x768. anything else is pointless since thats what you'd game at on this laptop.

These are the tests with stock CPU clocks
Min / Max / Average FPS

DX9:
Heaven:

DX10 single + multi GPU
Heaven:
Heaven Dual GPU:

DX11 single + multi GPU
Heaven:
Heaven Dual GPU:

these are with CPU at 2.5GHz (quad core, i should retest at new settings since it may have temp throttled)
Min / Max / Average FPS / power consumption (no battery, tested at wall - internal LCD used)

DX9:
Heaven: 8.5 min, 23.8 max, 13.2 avg, 331 score (39-43W)

DX10 single + multi GPU
Heaven: 7.9 min, 22.4 max, 12.3 avg, 309 score (36-39W) interesting how DX10 had lower power consumption but equal performance.
Heaven Dual GPU: 10.8 min, 36.4 max, 17.4 avg, 439 score (46-51W)

DX11 single + multi GPU (this adds tesselation, so performance drops accordingly). Since AMD drivers allow me to lock tesselation levels down low, in real DX11 titles i can force better performance in future titles if needed.
Heaven: 4.5 min, 19 max, 8.3 avg, 209 score (34-38W) - again, still more efficient than DX9. possible power savings advice for gamers there on battery.
Heaven Dual GPU: 5.7 min, 28.5 max, 11.2 avg, 282 score (42-54W)
Heaven dual GPU no tess: 11.1 min, 37 max, 17.5 avg, 441 score (47-54w) - To compare with plain DX9/10 scores, for 'future gaming' (dx9 vs dx11 at same settings.) the average speaks for itself. DX11 ftw.


And now thanks to a Futuremark representative who contacted me and gave me some serial keys:

3Dmark11:
Stock clock test: (i screwed up and ran 1.6GHz instead of 1.4. have to retest these)
E1486 (1024x600 'entry' test)
P977 (performance 1280x720)
X300 (extreme, 1920x1080i - my HDTV cant do 1080p)

CPU @ 2GHz/2.5GHz:
E1658
P978
X321 (0.4FPS on the combined test hurts my eyes XD)

PCmark 7 professional edition:

I ran this with every test enabled, at default settings. It should be noted that the DX10 graphics tests did NOT run in fullscreen, so only the onboard APU was used. this means the gaming tests and overall system score is lower than it should be. I will contact futuremark on this, for their thoughts and opinions.

Score at stock clocks: 1636
www.3dmark.com/pcm7/259048
*Link to test result here*

OC result: 1785 - a 9% increase from a 42% CPU increase. this test is not CPU bound, clearly.
www.3dmark.com/pcm7/259085



Battery Life:
Windows can give me some vague battery life estimations based on doing absolutely nothing, whereas Futuremarks Powermark program (anyone else think of a superhero called Mark with all these mark titles?) can give some more realistic load tests.

Unfortunately this has the same problem as PCmark 7, wherein it tests in a window. Therefore the second GPU in this setup will not activate, and power consumption and performance will be lower. So DX10/11 gaming will chew the battery faster than this shows.


Using the default "balanced" test: (a mix of all tests combined for a more 'average' result)
Stock clocks: 2 hours, 34 minutes
Tweaked clocks: 2 Hours, 52 minutes (11% increase in battery life) (mistake, wifi left on, should retest)
locked to 800MHz:

Productivity test: (Based around office/2D tasks like web browsing and office tasks)
Stock clocks: 3 hours, 30 minutes. Oddly, it was estimating about 5 hours of battery for the first half of the test i actually watched.
Tweaked clocks:
Locked to 800MHz:

Entertainment test: (Game tests and 2D video tests)
Stock clocks: 1 Hour 51 Minutes
Tweaked clocks: 2 Hours 23 minutes - an extra half an hour here, quite an improvement. 28% gain is massive.
locked to 800MHz: 2 hours 28 minutes (and i left wifi on again! damnit!) (yes, this ones not realistic)

Overclocking:
The CPU itself is easy as piss to overclock, since it has an unlocked multiplier.
The K10stat program i used can be setup in just seconds to set any multiplier and voltages you want, how you want it (each core at different speeds, turbo on/off, etc). What also helps is that this does NOT affect your idle clocks and voltages - with my max from 1.4GHz to 2.4, it still idles at the exact same voltages and power consumption.

I managed 2.8GHz out of the CPU with ease, but it overheated and thermal throttled to 800Mhz or 1300Mhz once it reached close to 95C when stress testing.
I could most likely use this speed for a single or dual core Turbo mode.

2.6Ghz would throttle occasionally, so i chose 2.4GHz which ran a lot cooler initially, but ended up going with 2.0GHz for my permanent clocks. Why? because laptops dont run in ideal situations, so adding in GPU heat and no cooling pad, and those temps would go too high and end up throttling in summer, or used on a tablecloth etc.

The onboard GPU on the other hand, cant be OC'd yet which rather sucks
OCing the secondary card is possible via MSI afterburner, but since it only kicks in for DX10 and DX11 titles, its not worth bothering most of the time and its really difficult to stress test.

These are a list of the settings (stock and tweaked) i found stable on the laptop.

B0: 2300MHz, 1.2875V
P0: 1400MHz, 1.0250V
P1: 1300MHz, 1.0000V
P2: 1200MHz, 0.9875V
P3: 1100MHz, 0.9625V
P4: 1000MHz, 0.9500V
P5: 900MHz, 0.9375V
P6: 800MHz, 0.9375V

These are the results i found from using OCCT to stress the CPU, and lowering the voltages as i went. i then gamed on the settings to make sure they were stable with the extra power/heat as well.

At a lot of the results, i checked my wall power meter to see how much i was saving. results were pretty erratic, so i included ranges (measured by me as averages, not by the meter) and a few stock vs undervolted power consumption tests along the way.

2533: 1.2625
2500MHz 1.250v (at 92/93 load!) -62W - thermal throttled. - My chosen turbo mode, since it doesnt throttle except when all four cores are loaded.
2400MHz 1.2375v - thermal throttled at times in testing
2300MHz skipped (thermal throttled)
2200MHz 1.1375 - no throttle, but over 90C. i imagine GPU load would cause throttling.
2100MHz
2000MHz 1.075v - 46w 80-82c 12m stable (my optimum 'load' setting, since it has approx 15-20C before it throttles, accounting for hot weather even in extreme load conditions)
1900MHz 1.0625 30m
1800MHz 1.05
1700MHz 1.0250
1600MHz 0.9875
1500Mhz:
1400MHz, 0.9250 31-32W
1300MHz, 0.9000
1200MHz, 0.8750 31-32W (stock 1.075v flat 36W) - this is where power consumption starts rising and voltage needs more than one notch per 100Mhz.
1100MHz, 0.8375 27-30W
1000MHz, 0.8125v (25-29W) (stock 1.0125v was 31-32W)
900MHz, 0.800V 26-27W (wattage more stable than 800Mhz - interesting)
800MHz, 0.7875V 25-28W load

The final settings i chose for my profile can be summarised best with a screenshot:


The reason some of the voltages are higher than the results i tested as stable is because i wanted to add some headroom in case of the VRM's weakening in the future (or performing worse with higher ambients) and i also have the unverified belief that smooth steps between voltages/MHz will be easier for the VRM's and clock generators to handle for long term life of the laptop.


Conclusion:

For a light gaming machine, it does well. just about any game you want will run on medium even on the APU alone, and DX11 only seems to help, not hinder performance in what i've
seen. This means the machine will degrade slower than most, because as more and more titles run DX11, you get the second GPU to help out.

For a HTPC replacement (which is what i use it for) its damn near perfect. you can watch a full HD movie on one battery charge via its own screen or on a HDTV through HDMI.

750GB hard drive is plenty of storage for whatever you want to use this laptop for, and with decent wifi N and gigabit ethernet, fast file transfers are possible as well.
The really low idle power consumption means low heat and low fan noise, so this is extremely quiet when watching movies at night.




rightio, so after lots of TPU advice, i ended up getting myself a low/midrange gaming laptop.


i intend to do a write up/user review on the beasty, but i'd like help from TPU in advance as to what should be in the review.


So far i've got:


1. Intro and specs
specs, looks, opinions on mouse trackpad, keyboard, input/outputs.

2. first impressions - keyboard layout, weight, screen brightness and clarity, touchpad impressions

3. usage scenarios - how it fares for SD and HD media playback, 3D gaming and 2D desktop use (and power consumption/battery life for these scenarios)


if people have specific SD/HD files (movie trailers etc) for easier comparison, please link me to some. no copyright material, thanks.

as for the games, please link me to some benchmarks/games i should use with easy benchmarks, i require DX9 and 11 titles. i will include AMD's 'dual graphics' (crossfire variant) in the review

ATM for 3D testing i've just got:

3Dmarks (free versions)
Company of Heroes (DX9 and 10 comparison)

4. overclocking (incl how 'dual graphics' messes with GPU OCing)

5. conclusion, derp.




any suggestions for things i've missed, and those videos/game titles to test?

not going to FRAPS anything, just want quick simple benches. will look into those ones.

excellent ideas asrock. was already thinking of some of them, but wasnt sure if they were worth doing. since you had the same ideas (ease of finding drivers, etc) then i think its worth mentioning.

oh drivers is impossible to find for this one, its a custom made model for specific aussie stores, so its not on any websites. short of using device manager and hunting for the drivers for several hours (there are no generics for many of them, such as atheros wifi) its a total nightmare.


took me around 6 hours to find all of them (still havent found the ethernet driver) and you all know i'm not a noob at that >.>

got the first say... 4 pages written up, all that will need to be edited in is gaming/benching results with and without OC later.

Damn i'm putting a lot more work into this than i originally thought. no wonder most laptop reviews feel rushed, these things take a LOT of time.


*note to self: photos*






edit 2: i'm constantly updating this post as the meat of the review, since its easier to do so swapping between desktop and laptop, and safer on TPU's servers than my hard drives. refresh this post to see updates/improvements.

nah im not going to buy that, and i cant really pirate it for a TPU review.

edit: i was given a key by futuremark, so it WILL be edited in once i get finalised OC/undervolt settings so i can do a 'stock vs tweaked' comparison.


currently messing with undervolting (working great) to see how it fares - taking 4-6W off the idle power consumption (100% load at 800MHz) should boost the idle battery life a massive amount.


default voltage for 800Mhz (idle) is 0.9375v and mines stable (so far) at 0.800
edit: BSOD at 0.750, so i'll go for an in-between voltage once i'm sure its stable. AMD really sets these voltages higher than needed, which means i might be able to drag 5ish hours of out of the battery now :3


edit: idle with monitor off (External screen, HDTV) is now down to 13W, with monitor on (default brightness) 18W.

well i've just been offered a key for futuremarks powermark program, so results will be included for that.


i'll have to run a stock/undervolted/overclocked test and tally up the complete results.

http://sites.google.com/site/k10stat/


simple program, dont forget to right click the tray icon and enable clock control there.


you can use things like this in shortcuts to load profiles automatically once they're thoroughly tested:

K10STAT.exe -loadprofile:1 -nowindow

started editing in futuremark programs and their results, 3dmark11 is done for now at stock, with OC results to follow. PC mark 7 is next, hopefully powermark will be done tonight or tomorrow.

overclock settings have been finalised as 2GHz for full quad core clocks, with 2.5GHz for turbo (i have no idea if thats single or dual core in turbo mode, whatever 'default' is).


now that thats done, i'm re-running the 3dmark 11 tests for stock vs OC'd comparison, will do stock vs OC'd on PCmark 7, and then the same with the battery life tests.


the big thing with the battery life tests is how i want to do them - i've got stock clocks, my 'modded' clocks (idle is undervolted but max clock is higher), or locking it low (say to 800MHz). i could use some input as to what settings people think would be practical here since i have a limited amount of tests i can run. ideas? thoughts?


edit: heres an FM result for those interested
(2GHz quad/2.5 turbo for this result)
3dmark.com/3dm11/2613685

turns out i screwed up somewhere along the line, i thought it was a 1.6GHz quad with 2.1 turbo, when its really 1.4/2.3

this means my 'stock' tests were actually set wrong since i'd manually adjusted the voltages for them.

time wasted, lesson learned, rebenching in progress (starting with CPU tests, then futuremark programs)

its kinda funny, but due to the good batter results i'm getting in the 2D tests (5 hours is the estimate) and the 2-3 hour recharge time... these tests are taking a looooooong time to complete.

and thank god, HP have updated their website so drivers for this laptop are available at last, at least on the .au site.

i'll think about adding RE5 into the mix, as a real gaming test.

back from holidays, so updates will resume on this soon.

started editing in power consumption and undervolting results. now that i've got stable undervolt/OC settings, i can finish the powermark testing tomorrow.