Exactly - I though the same. Now the thing is how to prove this issue to ASUS if we'd like to argue for getting our money back or a replacement (but for what? Definately not a G51Jx...,G73 - rather impossible)...
Maybe there's a way to modify the BIOS or to get the CPU to work at proper speeds. Definately waiting couple of weeks/month for the new BIOS (which may never come out) is not an option for me.
-
-
by finding a balance that depends on screen brightness. then you return it with "brightness up -> crap performance, brightness down -> good performance". in any support book they have that should be flagged as a hardware problem. after returning it 3 times you can argue with them for a replacement or check your local lemon laws.
editing the BIOS is currently impossible and that situation probably won't change before ASUS releases a fix (if ever). -
Ok, I just finished some more testing and took screen shots. I ran everything at 720p and max everything...16gx AA 16x AS and so on. Pay attention to when i have full bots or when they are dead. Also when I am looking at water or other things that require physics(AKA CPU usage) HL2 source engine works like this. It only does physics in front of you on the screen. If it is behind you it does not calculate those. Source was made like that so computers at that time would not be bogged down with unnecessary calculations. So when i look at ground textures i get 50-70% gpu and good fps but when i look at water...especially a dead body floating in water GPU drops to 5-30% utilization, which results in 10-40fps.
Also it appears for me in CS:S the throttling happens in the GPU not the CPU like many other games. I wont dare speculate on why that happens because i don't have a clue.
Here are the screenshots. I left only the link to the album so it wouldn't make the page load slowly. I also left comments on each screenshots so feel free to add some more.
http://img22.imageshack.us/gal.php?g=css1.png
EDIT:The gay thing is...it cnat be hard for ASUS to fix the bios or leak us the tool. So why waste more money and time by having everyone send it in to tech support 3+ times to not even get fixed and then having to give us a replacement? That will cost them thousands of dollars....so why not spend 100 bucks to pay a guy to fix the bios....or better leak the damn too and spend nothing. -
theres no guarantee we can change it even if they do. we don't have any specs or info, only cowboys who make a change/flash/pray it doesn't turn into a brick (fun when $1500 isn't on the line).
looking at those results, it seems the CPU is at it's max. i'll do the test as well and see what i come up with. -
DCMAKER: If you would like ThrottleStop to report the core temperature instead of DTS then just add this to the ThrottleStop.INI file.
TJMax=100
or use whatever value is appropriate for your CPU. If you're not sure the RealTemp Settings window can show you the default TJMax value.
http://www.fileden.com/files/2008/3/3/1794507/RealTempBeta.zip -
How can the cpu be at its max....My friends old computer with a P4 2.4gphz and a radeon 9800 pro can run the game with 32 bots and still do background stuff. Now the graphics isn't as high but how would that affect CPU?
EDIT: its a 4 year old game....how can it max the cpu?
EDIT: wiki says it affects the jx as well. Does that mean the jx may be less effected since the GPU uses less then half the watts?
EDIT: Also you talked about flashing the bios. How does that ruin a computer if the bios is not exactly right? I know absolutely nothing about that. Can you give a brief class or show me a link to a brief general overview of that? It sounds interesting. -
the GPUs use apx. 58W for the underclocked 260M and 50W for the 360M. so you have an extra ~8W for the CPU before it throttles.
the BIOS is like a firmware for the entire system (or a big collection of firmwares for every device/sensor/component). like any other firmware, if there's a mismatch or problem, it will most likely brick the component (in this case the whole system).
nothing to show with BIOS editting. mostly guess-and-check work with a hex editor on modules extracted from Pheonix or AMI BIOSs. -
What i meant to say when you brick it...is it a brick forever or until you change the value?
Also...well i forgot. I'll edit if i remember. -
if it's bricked it's done. gg. as useful as a brick. no changing the value back, no way to flash it to the original.
-
wow thats ty. Well did you get any other testing done with the CS:S?
-
hopefully tomorrow (feels like i said that every day for the past week...). still have to do the vantage run for g51jx comparison, proper logging for dolphin-emu and pcsx2, make my way via transit to university to watch lost with a few friends (missed it today), regular chores back home, and then if it's earlier than 3AM i'll fire up some CS:S.
on the bright side i don't have to wake up at 6 -
Are you sure you're not just looking at a scenario where all the threads that CSS uses are saturated e.g. CPU is maxed out and is limiting FPS, so GPU utilisation% drops as GPU cycles aren't required? I've observed GPU downclocking when usage is consistently low [e.g. during the CPU tests in 3dmark06]. If you could install Rivatuner and use the Hardware Statistics Server to set up a graph showing framerates over time, that would be good (should show low fps at start of round, and fps steadily increase as bots run away & get dead?)
It seems from those screenies, higher fps is associated with being nowhere near bots/dead bots, low fps with being near them/live bots. (I assume you've just sat at a spawn point for all pics?) Lower fps corresponds with lower GPU utilisation% indicating a CPU bottleneck.
I know it doesn't much look like the CPU is bottlenecked (util% on cores 0246 between 40 and 80%) but looking at the average multipliers (being >12), there is some turboboost-ing going on making me think CSS is not running more than 2 threads, which according to the behaviour I've observed, would be swapped around all 8 cores.
---
I for one don't want a tool or fix, that when used to eliminate this throttling, would cause the motherboard to fail. I wonder if the issue is cheap components (caps etc) being forced to operate outside their tolerance if power draw is any higher, hence an intentional hard-coded limit.
Remember, despite its price, this notebook was built to an aggressive budget (considering the power of the hardware in it)
edit: missed bout the 8 posts above, hibernated with firefox open & didn't refresh :doh: -
Iunderstand what your saying but it doesn't make sense how a P4 2.4 or 3.0 GHz can run the game with 30 bots and not an issue. Also there was obvious Clock dropage too. Doesn't it mean massive throttling if the frequencies of the GPU drop.
-
it still ran stable. if it was a component issue, i would expect the system to crash, or at least the CPU. even if i was pulling only 10-12W from USB, that would give the CPU a mere 3W to work with. possible, but very unlikely.
edit: @unclewebb; is it possible that asus' scheme is switching the throttling state as many times a second as throttlestop reports the multiplier? -
then how is the cpu maxed out? The i7 has to be able to run miles of laps around a single core P4. that is like 6+ years old.
-
looks like it might be throttling, but we don't know. which is why i'll play around with cs:s and bots in a bit.
-
Well....I would say your right on 90W ^^
http://www.battdepot.com/model/asus/m60j.aspx
all they sell is a 72W car and a 90W multi one. Also i did a google shopping search and i only see 90 watts and even some 75w(might be different model or car one...was just glancing).
Off to asus site to check i'll edit
http://www.asus.com/product.aspx?P_ID=pAWDOlwZMb508qWy
Wa lah! 90w adapter....well it edits my swear lol
So i confirm the theory that throttling came from the M60 because of the 90 Watt adapter...&#%&$%@ -
A 90w adapter? That's not even close. The Studio XPS 16's from Dell are high end multimedia laptops that are ending up needing 130w psu's. Asus really just needs to send out some new adapter and a new bios just like Dell is what this means. I hope they actually do it, this machine has so much potential.
Thanks for doing all the research guys, <3 -
Also, why is it not possible for the cpu to pull power from the gpu? Someone said CS:S bottleneck was the cpu was maxed out and that the gpu just didn't need to work as hard since the cpu was slowing the game down. That just seems completely illogical to me because the cpu should have no problems handling CS:S with bots. It seems much more logical that the cpu was pulling power from the gpu. Couldn't there be more parameters set for how the bios throttles the system?
EDIT: In the wiki it says i think 50-58 watts for consumption depending on clocks. These 2 sites I found that have the watts say its 65 and 75. I also looked at the NVIDIA site and couldn't find the TDP so these are the 2 that i found. Either way the watts are higher then what you tested. My point is if there are right numbers are you even getting max performance out of your graphics card? I only get 70-73% utilization in cs:s and at best 75% in game for batman and at the peak 82ish% for the benchmark in Batman Arkum Asylum. Don't have a clue how to spell it ^^. So the point is why am i not getting max utilization in the gpu when gaming. In both games including CS:S when there are few or no bots it doesn't max out the gpu. Why is that? Is it maybe becuase we are not getting enough power to the GPU for it to push itself? I remember with my old PC it was a 3.0GHZ E edition and a 9500 pro ^^ and it was always maxed utilization whether HL or HL2 or C&C. So why is this different?
http://www.notebookcheck.net/NVIDIA-GeForce-GTX-260M.14559.0.html (75watts)
http://en.wikipedia.org/wiki/GeForce_200_Series (65 watts)
EDIT@ Bennyg I was not standing still. I was running around and playing the whole game. I only paused to take screenys where something was noticeably wrong or interesting. aka a water or a body floating in water requires a lot of cpu calculations to do the physics. -
ALLurGroceries Vegan Vermin Super Moderator
Edit: BIOS thread started @ http://forum.notebookreview.com/showthread.php?t=460583
-
I'll wait until someone else tries it first since i know this can brick a machine now lol.
P.S. this is awesome found it in steam...granted i could d/l probably 90% of those games but 50 bucks isn't anything for all of this lol. Now to the fact am i ever really going to play all of these....i wish Also i already own the first red faction and it was awesome lol. Never beat it because the bomb timer was a pain lol if anyone remembers the ending plz PM me
50 bucks!!!
Buy THQ Complete Pack
Includes: S.T.A.L.K.E.R.: Shadow of Chernobyl, Full Spectrum Warrior, Full Spectrum Warrior: Ten Hammers, Titan Quest, Titan Quest - Immortal Throne, Company of Heroes, Warhammer® 40,000: Dawn of War® - Gold Edition, Warhammer® 40,000: Dawn of War® - Dark Crusade, Company of Heroes: Opposing Fronts, Juiced 2: Hot Import Nights, Warhammer® 40,000: Dawn of War® - Soulstorm, Frontlines™: Fuel of War™, Saints Row 2, Warhammer® 40,000™: Dawn of War® II, Red Faction Guerrilla, Red Faction, Company of Heroes: Tales of Valor, Red Faction II
lol is that 18 games/expansions ^^ -
Re ths CSS bots: try turning your vid details down to minimum and seeing if that changes FPS. If the CPU is maxing out, there won't be much diff between high-vfx and low-vfx, and the GPU clock will stay fairly low.
Yeah the CPU *should* have no problem handling all the bots... but then again the CPU *shouldn't* be throttling as it is either... so it all comes down to: we don't know with enough certainty about how the CPU handles itself (turboboost, shuffling threads around cores, etc) let alone how the BIOS/motherboard distributes power between the CPU & GPU.
What happens when CPU is "maxed out", or to be more specific, is the limiting factor in how many frames per second can be drawn, is that the GPU is sitting idle waiting for the CPU to finish each frame (think the smart kid can't go on to the next question in the math test until the not-so-smart kid has finished the same question). The GPU-monitoring stuff in the driver notices "GPU running lots of idle cycles" so decides "Clock speeds don't need to be so high so we can reduce them to reduce power consumption and heat output".
As a result, at this lower clock, GPU utilisation % will be higher as the same amount of work still needs to be done in order to draw the frames, but as the "maximum output" of the GPU is now lower due to it having lower clock speeds, this is why the utilisation figure appears higher.
Re TDPs:
75W is the GTX280M (585/1463/950 clocks with 128 shaders)
65W is the GTX260M (550/1375/950 with 112 shaders)
58W is the GTX260M at the speeds in the G51J - the Asus underclocked GTX260M (500/1250/800)
So if you overclock to stock GTX260M speeds, power draw will rise to 65W, if you overclock higher power draw will be higher (like my maximum overclock, 590/1500/1050 will be much closer to, perhaps more than 75W) -
yea but thalanix said on the wiki stock asus is 50w and overclocked is 58....so thats still to low.
But still why does the gpu never go above 70-75% for games(CS:S/Batman)? -
Something was funky when I tried overclocking the 260 with Rivatuner; what program have you guys been using? The 260 has different max clocks when on a/c and battery, and even when I'm on a/c Rivatuner only shows the battery clocks (which can't be put anywhere near the a/c one's).
-
50W for the gts360m/g51jx. i estimated our underclocked gtx260m based on total usage figures in linux.
rivatuner doesn't show the 3d-perf clocks. have to use something like ntune / nvidia system tools (6.05, google) -
yea i was just about to edit. I read the wiki wrong -_- my bad
-
Okay, thanks!
-
wait a minute...do you think asus underclocked because of the outdated heatsink and because of the throttling....in an effort to hide it? Also someone mentioned that they found on their G51j that asus went skimpy on the thermal lube and they redid it and their temps dropped a lot. Does that sound legit? Also if that is so can we get a how to guide so i can do that and put some arctic silver 5 on my machine.
-
I might put some AS5 on the G51J. I wouldn't be surprised if that helps it a ton, after what I watched happen with my M1530. The Asus G51J is pretty easy to navigate internally, just open up that big back panel and take a look at everything.
*Look at this!* -
I used NVidia System tools 6.05 to change clocks, but prepare to be very very frustrated. It'll only make you think it works, some of the time. IMO it's of alpha-release quality. Very poor. The only reliable way to survive a reboot is to make sure the setting is in the osboot profile (can find in some arcane location just search for the profile names).
I can confirm on battery the highest clocks aren't reached. It only goes up to 383/766/300 (shows up as 414/828/300 in rivatuner) which is the 2nd highest powermizer profile.
Sirhcz0r I think this may have been what was happening when you thought the BIOS changed the max temp down to 75C, which is around where I noticed GPU temp hovers at around 383/766/300 during a game.
DCMAKER I've seen GPU 98-99% of the time sitting on 98-99% use during most games. (This is the GPU util% reading in GPUz, can either look at the graph as it updates in the background, or look at the log). Some games will be different (e.g. ones which are CPU bottlenecked), even then some settings will behave differently even within the same game as they use different parts of the hardware differently. -
I was on a/c with those temps, but that may very well have been the case. I wish it didn't have different clocks for a/c and battery, it's not like it gets any battery life either way lol.
-
Can someone test Batman arkum aslulym and CS:S to see cpu utilization because I only get 73% max for CS:S and maybe 80% for Batman. If you want to use my steam for Batman i can give you acccess just PM me. Also like I said before I can only reach 85% max in the batman benchmark and thats only in one part rest is high 70s or maybe low 80s
-
make a batch script with
Code:"C:\Program Files (x86)\nTune\nTuneCmd.exe" boot "C:\Users\Mike\Scripts"\max.oc.nsu"
both nvidia driver service and performance service have to be running for the auto profiles to work.
if you start on battery, the 383- clock max sticks even after you plug it back in. -
If you start on battery, unplug, and replug will it be throttled or not?
Thanks for the information thalanix. -
Thermal paste help for those who need it:
Once the back panel is off, remove the six circled screws. After they are removed, slide a flathead screwdriver underneath the cpu heat sink and gpu heat sink to separate it from the old thermal paste. Once they are loose, lift the cpu portion of the heat sink, and then the gpu portion. slide them both out to the left simultaneously. make sure the fan cable is disconnected first. clean all surfaces, apply paste, reverse steps.
If you have any questions ask away, I did the thermal paste swap today and all went well. We still need a new bios to see what it really did in terms of cooling though. -
thats cool. Have you done it yourself? If you have have you noticed a difference? My runs really hot because my room is 70-78 degrees depending on the temperature outside. Also what can you use to clean off the gpu and cpu? Also doing this does this void ASUS warranty?
EDIT: If DTS is 38 that means its 38 from throttling? So how is that for idle? Also speedfan reports 61-62 degrees celcious or is speedfan's numbers off? GPU is also 63 according to speed fan. -
Yes I did do the change, but it hasn't made much of a difference. We desperately need a new bios with more fan activity. I used the Arcticlean kit to clean everything off. I did have to remove a "void if removed" sticker, but was told by Asus over the phone that it would be okay to do the thermal paste swap.
-
ok i will also call them to get a verification and record the convo over phone ^^ thx for the info
-
Ok, I won't repost everything here - but I think my post in the G51Jx thread about CPU throttling was interesting and might add to the discussion here/people subscribed here would want to know.
. I think we have the same problems as Dell, and knowing what happened in their case can really help us. Now we can check out what users reported on the threads there and what tests they did and then reproduce them here. If we see the exact same symptoms (and I think we do) then we've got just as good a case as those users there had and they now have new power adapters that are more powerful than ours. Some are annoyed that the new BIOS is set for 130W rather than being an open limit. The guys who bought 150W and 210W adapters are a bit annoyed by this, since their overclocking is being cut short - but everyone agrees it's much better than before.
Last edited by a moderator: May 8, 2015 -
iv recently been getting some pausing every few seconds when typing (happened just now), watching youtube videos, or playing java games online. I pull up task manager and there are spikes in the cpu usage history during those pauses. This is related to the throttling right?
-
Has anyone tryed to talk with someone at http://www.engadget.com/ or other of the like ?
We have things well documented and it ins't only G50j with the problem , G60j G51Jx etc... .
We could gain massive media support and alert those users out there that don't know of the problem.
My second hard drive has just died on me, I have to send my laptop on RMA and i will list throttling has a problem. -
at CCz_Cataphract. We know its throttling and that its the bios. That has been known for awhile now and we confirmed that by looking at the M60 on asus site. The M60 has a 90w adapter and our tests show that the power limit is set to 90-100w.
EDIT:Also allurgrocies has found the tool to edit the bios and him and thalanix are seeing if they can edit the bios and fix the issue on their own.
at scothu. You need to get throttle stop and record a log and see if those slow downs are because the cpu is throttling to 933mhz or 7x multipliers. Whenever your system does that on a loud thats because of throttling. I doubt it is thottling because from what you said you shouldn't be using more then 90w.
At breezer. We have tried to talk to other mainstream media places and have been ignored. You can try emailing them about it and tring to get them to help but its unlikely. I know i emailed laptop mag and never recieved a response. -
The same issue seems to be occuring with most i7-720QM processors. I had a dell studio 1557 that did this too (paired with a slow, but exceptionally hot ati 5470)... and reading around, looking for a suitable i7 laptop turns out the same results.
It seems the i7 isn't suited for laptop use - the thermal impact of laptop chassis cooling will lower clock speed almost every time! -
I know what you mean about the 90W adapter, AllurG and me have both confirmed with Power Meters that in fact the 126-135W are being used which is beyond the amount that a 90W adapter can give. I tested my G51Jx system with multiple 90W adapters and the performance was much worse with it. With the 120W adapter and the GPU underclocked, up to 113W I am able to have no throttling at all. It is only when 114W is broken that the throttling system engages.
That is why I suspect that it is the BIOS with a limit on the Intel Turbo Boost code that is either to 90W or 120W, but to fully utilize our systems we need more powerful AC adapters, such as the 150W in addition to the BIOS fix. As was reported in Dells case. In that case they only had 90W adapters to start and are now getting 130W adapters - our systems need at least that and probably more to avoid the throttling. With the 90W adapter it was much easier to see the throttling than with the 120W ones that we have because they are closer to the edge of the throttle. That's why doing a definitive test with the 150W adapter to see what which point Wattage it hits the 7x Multiplier would be key. If it is more than 114W then we know that we need the new AC adapter AND the BIOS fix.
Peter -
Any multiplier reductions before this built in throttling point are being done by the bios and has nothing to do with protecting the CPU. Intel knows what temperature their CPUs are capable of running reliably at and their protection mechanism is second to none.
The maximum temperature that the CPU reaches when this happens is known as the maximum junction temperature or TJMax for short. If you know what TJMax is then you can use this formula to figure out the approximate core temperature:
Temperature = TJMax - DTS
If ThrottleStop is showing 38 for the DTS value and your TJMax is 100 then the core temperature equals
Temperature = TJMax - DTS = 100 - 38 = 62C
You can use RealTemp to find out what your TJMax is by looking in the Settings window.
Intel says that there is some error in how they set TJMax but have never given any details about how big that error is. For that reason, converting DTS to TJMax is not an exact science and the larger the DTS value is, the more possible error there will be in this calculation. The only thing for certain is that when DTS hits zero, it's game over and throttling will begin. Thermal shut down usually doesn't occur until about 25C beyond this point which is equivalent to a core temperature of approximately 125C.
If you would rather see ThrottleStop display core temperature instead of DTS data then just add this to the ThrottleStop.ini configuration file:
TJMax=100
Use whatever value RealTemp reports this as. It's usually 99 or 100 but each CPU can be unique. The newer Core i5 mobile CPUs tend to be 105 which is the same as most 45nm Core 2 mobile CPUs.
What Dell does with their adapters is they limit them to less than the rated wattage. On the XPS 1645, I believe the original adapter that was rated at 90 watts was limited by the original bios to only draw 75 watts DC before throttling would begin. The 130 watt adapter improves things greatly but it still throttles when it reaches a limit long before its 130 watt rating. I'm not sure how Asus does things but keep in mind that the rating on the outside of the adapter does not always equal the power that a laptop is allowed to draw. Also, a Kill-a-Watt meter or similar meter reads AC power consumption at the wall and this number will be higher than what you end up with as DC output from your adapter. -
Peter -
if it throttles at 114 and the theory is 90, an 80% efficient brick is nothing unusual. it is built on the cheap after all.
after a little googling and tomshardware, 80% is pretty good. -
thalaix is right i was about to say that but been away form the computer. The efficiency is usually around 80ish(if not lower). I would say that it is probably more likely to be 75% then 80. So the values of 114w like thalanix said is for a 90w adapter. It hasn't been until recently that psu have reached 85-90% efficiency. You can now buy the 80 plus or 85 plus and so on psu for desktops. So a 150w psu will pull closer to 190w or so. Also whoever said that these i7 were not meant to be in a laptop because they run how is a ridiculous statement. The reason these things run hot is because the morons put little thermal compound and probably a cheap one too. On top of that, they covered the intake....who knows why but they did. Thats why this weekend i am ordering that new back panel and drilling holes into this and putting AS5 on it. my room temp is from 72-78 Fahrenheit because the marine corps likes to run heat when its out outside. That's why mine ideals hot as and gets above 70c on loads. It also doesn't help that the fan doesn't kick on til really late. If you mod/fix these issues this thing wont run too bad. Now it could still use a better heatsink but its not the worst that i have seen.
-
(Reference: www.energystar.gov/ia/partners/prod...International_Efficiency_Marking_Protocol.pdf )
Therefore, assuming both PSUs go by their certification marking: The 90W adapter must take a maximum of 103.5W to deliver the marked 90W load. The 120W adapter must take a maximum of 135.6W to give it's 120W load (and 101W to deliver 90W).
I decided to be methodical and try many different settings and modes on both adapters - and so I took a couple of video recordings of the system and power meter w/each adapter to make sure that I had all the data and matched the Turbo Boost status and peak wattages at which things changed exactly.
The test was doing the FurMark Test (Stability, 4xAA, Xtreme Burn, Windowed 1440x900), using the Intel® Turbo Boost Technology Monitor + my Wall Power Meter with other values such as screen brightness and Extreme Turbo CPU mode being adjusted throughout the test. I waited for the values to stabilize in each test and cycle twice to and from Turbo Boost. I did the test once through counting results from when I started FurMark, but I realized when reviewing the video that the level that the GPU was at (whether I had just finished a run or left it idle for awhile for example) would affect the results greatly. So I did not count initial spikes while the GPU was below 80*C (and hence at a low fan power mode), but waited until any Turbo Boost activity was stable for at least 5 seconds minimum.
120W PSU rated at 87%+ Efficiency - 90W PSU rated at 84%+ Efficiency
120W Regular CPU Speed + Full brightness + Stock GPU Clocks: Avg. 113W, Peak 117W (caused by GPU Fan), Never any Turbo Boost
90W Regular CPU Speed + Full brightness + Stock GPU Clocks: Never any Turbo Boost, @ 108W Battery Mode to AC Mode pop ups keep cycling until Brightness is turned down
120W Regular CPU Speed + Lowest brightness + Stock GPU Clocks: Avg 110W, Peak 114W, Never any Turbo Boost
90W Regular CPU Speed + Lowest brightness + Stock GPU Clocks: Avg 99W, Never any Turbo Boost (CPU Core 0 was steady at 57*C and GPU steady at 87*C)
120W Regular CPU Speed + Full brightness + 400/1500/1000: 2.53GHz @ 118W max momentary peak down to 101W when Turbo Boost de-activated itself (Confirmed 115W stable for 5+ secs, CPU Core 0 temp 64*C AND GPU temp 85*C @ deactivation)
90W Regular CPU Speed + Full brightness + 400/1500/1000: 2.53GHz @ 109W max momentary peak down to 93W when Turbo Boost deactivated itself (Confirmed 108W stable for 5+ secs - CPU Core 0 temp 63*C and GPU temp 82*C @ deactivation)
120W Regular CPU Speed + Lowest brightness + 400/1500/1000: 2.53GHz @ 114W peak down to 99W when Turbo Boost de-activated itself (Confirmed 113W stable for 5+ secs - CPU Core 0 temp 65*C and GPU temp 85*C @ deactivation)
90W Regular CPU Speed + Lowest brightness + 400/1500/1000: 2.53GHz @ 104W peak down to 91W when Turbo Boost deactivated itself (Confirmed 104W stable for 5+ secs - CPU Core 0 temp 66*C and GPU temp 85*C @ deactivation)
120W Extreme Turbo CPU Speed + Full brightness + 400/1500/1000: Avg 105W, Peak 119W, Never any Turbo Boost (CPU Core 0 was steady at 58*C and GPU steady at 83*C) - (Bonus: Dim Screen 2.68 GHz @ 117W down to
103W when Turbo Boost de-activated itself - GPU 83*C, CPU Core 0 65*C at deactivation)
90W Extreme Turbo CPU Speed + Full brightness + 400/1500/1000: 95W Steady, Never any Turbo Boost (CPU Core 0 was steady at 58*C and GPU steady at 84*C)
120W Extreme Turbo CPU Speed + Lowest brightness + 400/1500/1000: 2.68GHz @ 115W momentary peak down to 97W when Turbo Boost deactivated itself (Confirmed 114W stable for 5+ secs - CPU Core 0 temp 65*C and GPU temp 86*C @ deactivation)
90W Extreme Turbo CPU Speed + Lowest brightness + 400/1500/1000: 2.68GHz @ 106W down to 91W when Turbo Boost deactivated itself (Confirmed 106W stable for 15+ secs - CPU Core 0 temp 67*C and GPU temp 83*C @ deactivation)
120W Extreme Turbo CPU Speed + Full brightness + Stock GPU Clocks: Avg 116W, Peak 120W, Never any Turbo Boost
90W Extreme Turbo CPU Speed + Full brightness + Stock GPU Clocks: Never any Turbo Boost, @ 109W Battery Mode to AC Mode pop ups keep cycling until Brightness is turned down
120W Extreme Turbo CPU Speed + Lowest brightness + Stock GPU Clocks: Avg 111W, Peak 114W, Never any Turbo Boost
90W Extreme Turbo CPU Speed + Lowest brightness + Stock GPU Clocks: Peak 108W, Never any Turbo Boost
Key Results:
90W Regular CPU Speed + Full brightness + 400/1500/1000: 2.53GHz @ 109W max momentary peak down to 93W when Turbo Boost deactivated itself (Confirmed 108W stable for 5+ secs - CPU Core 0 temp 63*C and GPU temp 82*C @ deactivation) -> For Sustained 108W Given 84% Minimum Efficiency, Minimum Actual System Wattage w/Turbo Boost Active: 90.72W
120W Regular CPU Speed + Full brightness + 400/1500/1000: 2.53GHz @ 118W max momentary peak down to 101W when Turbo Boost de-activated itself (Confirmed 115W stable for 5+ secs, CPU Core 0 temp 64*C AND GPU temp 85*C @ deactivation) -> For Sustained 115W Given 87% Minimum Efficiency, Minimum Actual System Wattage w/Turbo Boost Active: 100.05W
I spent many hours to write this post first by examining the PSUs and then researching the meanings, and ran and examined the tests very carefully and re-ran most of them 3 times to confirm the results. I'm not saying the BIOS isn't part of this - but I think that we also need to run a full series of tests on a 150W adapter. The combination Dell BIOS+Adapter upgrade may very well be what we need. I hope we can use these figures to compare with what we get on a 150W adapter doing these same things to confirm the Turbo Boost Status. Especially if it is possible to have Turbo Boost be at least relatively stable and active in FurMark with the GPU at Stock Clocks (as it was with the 400/1500/1000 underclock).
As was shown by my and AllurGroceries previous tests, you can push the 120W adapter to the expected 136W wall power draw but these tests did not do that because I did not overclock the GPU or use other USB devices - though if I had I suspect we could have reached the theoretical wall power draw limit (and possibly slightly beyond it) of the 120W adapter as we did the 90W adapter.
From the tests, CPU Turbo Boost when it disables itself, reduces total system power usage by ~15W, after which point it will reactivate itself again if slightly more than that power becomes available (fan speed reduction, screen dimming, etc...). It seems that if the system has enough power and is allowed to use it, then it should be able to support full GPU usage and full CPU Turbo Boost at the same time, or at least some level of CPU Turbo Boost and an overclocked GPU. Therefore, once and as long as we are within ~15W of the maximum power draw of the PSU, then the CPU throttles itself since there is not enough power to sustain the full mode. If it has determined that we need as much single core CPU power as we can get, as in the FurMark test, then it will keep trying to go to the full mode continually, rather than settling for a lower Turbo Boost that it can sustain with the power available.
Thanks,
Peter
P.S. Concerning the system not showing a difference if the PSU has 15+W available for Turbo Boost, in that case it would almost certainly be the BIOS limiting it. The fact that it is Intel Turbo Boost code that may be causing the restriction makes me think that Intel may have sent the same BIOS code to all the OEMs from Dell to Acer to ASUS and in that code it has a section where the OEM adds the wattage of the PSU to be included in the system. In Dell's case they had to change this value to 130W when they sent out the 130W adapters (as is shown that in those systems, even 150W adapters work at 130W levels after the BIOS fix). In ASUS' case I think it might be that the value is 120W or something similar. -
sounds like this is a lost case >_<
[Fixed/Workaround] Asus G51J(x) CPU throttling investigation
Discussion in 'ASUS Gaming Notebook Forum' started by thalanix, Jan 20, 2010.