Undervolting - The difference is SHOCKING - over 2000 points in 3Dmark2005

I used 3Dmark 2005 because for me it's a better barometer for the capabilites of the M1530 w/8600M GT. The program used for the undervolt is of course Rightmark CPU Clock Utility (RMClock). For reference, the undervolt settings are as follows:

T5800(2.0ghz)
6.0x 1.050v
7.0x 1.062v
8.0x 1.075v
9.0x 1.087v
10.0x 1.100v

This test is with default GPU clocks of 475/702

The results of 3Dmark 2005:

No undervolt - 6842
With undervolt - 8848

A difference of 2006 points

That pretty much says it all. Take that extra current provided the power supply from the CPU and give it to the GPU and presto. A gain of more than 2000 3Dmark 2005 points. And it's not just the points. The graphics are as smooth as glass during the run. You will most definitely see this in your games.

 

That does not sound right. A difference of 200 points maybe, but 2000 would be an major issue. How many benchmarks did you do at either point.

 

Andy, wouldn't that be only valid if the GPU was overclocked, thus 'overvolted'?...since the GPU would already limited.?

 

The amount of current supplied to the device does not affect its performance in any way, other than if not enough current is available the device will not function.

Something else must be going on, perhaps your notebook is/was overheating during the first run.

 

Andy, not sure what your comment was about, but the reality is that I looked at the power consumption of the system under full load (both CPU and GPU). It totals 88 watts. That is too close to the rated output of the adapter (90 watts). That is not enough headroom for the PSU.

It doesn't take much deduction to figure out that if you see such a massive increase in performance from undervolting (taking small amounts of un-needed voltage from the CPU), then you give more available power to the GPU. Power = Volts x Current. If you have more power available to the GPU, then the power supply is giving more current to the GPU.

There is nothing off base about what I stated.

 

Completely false. For example, if you take a desktop system that has an 8600 GT that when overclocked fails, then after the introduction of a power supply with more power it succeeds, then the extra current provided by the PSU helped the GPU maintain stability because it had extra power to provide.

In this very typical situation, the OCed GPU was failing because the 12 volt rail of the PSU sagged in voltage due to the fact that it cannot provide the needed current for the GPU. A PSU capable of the needed current will not sag it's supply voltage and thus solves the problem.

These rules apply to ALL computers and electronic devices.

Edit: if you were to connect the Dell 130 watt PSU to the M1530, you would not need undervolting at all. You would have more than enough power to run the system even while overclocked.

 

Ran the benchmark 3 times. I did not change anything other than the undervolt active/innactive. Driver 180.84. Bios A09.

 

One more thing. You 3 guys were awefully quick to jump on this thread with your objections (I had just posted this).

Go run the tests yourself and see what your results are. Follow the undervolting guide, run 3Dmark 2005, and record your results. Then let's talk about it.

 

I have and saw near to no difference. I have run more benchmarks than you can dream I spent a month tweaking my voltages for undervolting as well. 3DMark has as much as 250 point moe on the tests I ran. Taking the averages and factoring a reduced moe, nada...

Edit: Did you check your resolution? That seems to be a likely cause if your test battery supports your initial score.

 

I ran the tests at 1024x768 so that there would be no bottlenecks other than available power.

By 'battery' do you mean as in 'a battery of tests' or did you literally mean the notebook battery? If you meant the notebook battery, then I would say no. There is no way that the battery could supply as much current as the power adapter.

56 watt/hour battery
battery = 11.1 volts at 5.04 amps = 55.944 watts
PSU = 19.5 volts at 4.62 amps = 90.9 watts

Edit: Full load M1530 = 88 watts. You can clearly see that if you game on your battery that it simply doesn't have the power to run full performance. You will get much worse frame rates on battery.

 

I would like to add that I am running on a properly cooled system. Under full load my temps do not exceed 80c. I have done the thermal compound mod on my M1530. If you cannot reach 8700 points in 3Dmark 2005 while undervolted, you may have heat issues. Temperatures over 85c +/- 2c will cause downclocking on the GPU and lower your score.

 

After looking again at my data, I realized something else. When the system is not undervolted, the GPU reached a max of 74c. When the system was undervolted, the GPU reached a max of 80c. That 6c rise in temperature is easy to explain.

While the system was not undervolted, the GPU had less current available, it consumed less current, it's power consumption was lower, and thus it's max temperature was lower.

On the other hand, the undervolted system provided more available current to the GPU, it consumed more power, thus the GPU produced more heat.

Whatever a given chip consumes as power, it gives off as heat. If it consumes more power, it gives off more heat.

What this translates to is that when more power is available to the 8600M GT, it uses it. The result is a ridiculous gain in performance.

 

I don't know whether this is true or not but assume that you are right, then this should cook up the video card and shorten its lifespan, therefore I don't think it is recommended to do something like this =)

 

Testing without the battery yields lower temperatures...and I have meant battery as 'set of tests'.

 

just a short question, how did you measure the power consumption of yor laptop? did you use the normal power adapter?

 

But that's largely irrelevant. With identical hardware across tests, the only factor which affects performance is clockspeeds. And if your clockspeeds stay constant across all runs, it doesn't matter how much power you make available to your system.

It's worth noting that components don't perform worse if they don't have enough power. They simply fail to function. Therefore, your system is likely not starved for power and I'd be very surprised if the system was smart enough to detect insufficient power and automatically compensate.

Undervolting your CPU will not cause the GPU to consume more power. Computer components have a rated TDP - and will not magically exceed these design parameters unless you go ahead and overvolt/overclock your components. It's possible that you have some software installed which can automatically overclock your components if they have enough headroom to do so.

A likely explanation is that one of your components may have been throttling itself due to excessive temperature or voltages, thus reducing performance. Reducing voltage may have reduced throttling, and thereforce increased performance.

 

Sc4Freak, I hate to call you out but it just looks like you want to offer nothing but assumptions with no technical data and no laws of physics/electricity to back up anything you have said. I on the other hand have not only offered my true and unbiased test results, but I have also provided the laws in which all electronic devices operate.

The example that I gave with the desktop card isn't fictional. And certainly isn't irrelevant. I actually had an 8600 GT with a power supply that sagged (lowered in voltage) on the 12 volt rail because it did not have the recommended current capabilty. Nvidia suggests a 12 volt rail with 25 amps for that card (and the 8800 GT), or two twelve volt rails with 13 amps or better each. I OC'ed the card, it went to pull additional current to get the work done, and since the PSU didn't have the extra curent to give the frame rate hit the dirt. After replacing that PSU with a good one with 28 amps on the 12 volt rail the OC'ed card sailed through the same tests with perfect frame rates. End of story.

As stated from the begining my tests were done with standard clocks. Furthermore your assumption is just plain wrong. Notebooks are engineered so that the PSU will be very close to it's limits while under full load. The M1530 with CPU at 100% consumes 68 watts. The 8600M GT consumes 20 watts at full load. If you hit a 90 watt PSU with a full load of 88 watts or more, you have reached the design limit of the PSU. Every PC manufacturer knows this. That is why guys that build gaming rigs get huge PSU's. It is far better to run a 600 watt PSU at 70% load than to run a 450 watt PSU at 100% load. So with my test, the excellent results come from the fact that with undervolting, you free up a little bit of current from the PSU, the GPU uses it, and muscles through 3Dmark.

I'm not stating any of my opinions in this thread. These are the facts -

I conducted the tests that I mentioned in the begining of this thread. I posted the results.

I didn't want this thread to turn into an electronics lesson but you guys took it there. Let's calm down with the objections. People reading this thread might appreciate my findings and this thread has gone horribly in the wrong direction.

Instead of arguing, just do the test yourself. Once again, your results will not be like mine if your card is overheating. Not one af you people that are cluttering this thread with unfounded objections have even bothered to post a 3Dmark score of over 8700 with no overclocking. Until you can do that ...

Post your results. Let's see those scores.

 

1. Yes, everything I have said is true. I really did do these tests and the results are posted. Those 3Dmark scores are real.

2. I wouldn't worry about the extra 6c of heat generated by the GPU. I never passed 80c even while beating the GPU to death. There are people running their machines in the 90's. Now that will do some harm. They need to fix their cooling systems.

3. The extra heat - This just reinforces everything that I have said. The GPU, while the system was undervolted, produced 6c of extra heat. This is elementary. P=IxE : power = current times voltage. The GPU consumed the same voltage (max design voltage does not change) but the current draw of the GPU did change. Power used by the GPU did increase and therfore the GPU produced extra heat. It can't be any more simple to understand wher the added heat came from. The GPU used the additional current provided by undervolting the CPU which was previously using that extra current.

Edit: I would also like to add that I did these tests with 2 different M1530's. Both had the thermal compound mod. same results on both machines. I do not own a 'voodoo box'.

 

I can image two reasons:

1. Speedstep
2. Thermal Management of either CPU, GPU or both

Let's talk about Speedstep first. Running RMClock changes the behavior of the Speedstep transition triggers as can be seen an configured in Profiles / Performance on demand:



Running RMClock will result in different patterns of FID/VID swichtes under certain load levels compared to Vista's default behavior. To rule out that possibiliy you have to disable all FID/FID indices exept one (the highest one - in your case 10x). And in Vista change the power plan and adjust min processor throttle level = max processor throttle level = 100%. Now that Vista and RMClock don't use Speedstep repeat the benchmark and see whether it still differs.

The performance impact of Speedstep varies.

Now let's talk about Thermal Management. There are many different ways the system has to reduce power consumption if temps of certain devices climb too high. The control logic which of them is enabled when, changes from BIOS version to BIOS version.

There isn't a single temperature, when thermal management kicks in. As I said: This varies from BIOS version to BIOS version. Also thermal management takes the time window into account how long a certain temp threshold has been exceeded.

Anyway: In order to rule out thermal management, you need to be able to read it. I give you two examples:

TM1: Thermal Management 1 is a CPU based one. It injects idle cycles in the execution pipes. If it kicks it, it looks like this:



The important bars are the two on top: The red and purple one. If they differ TM1 is active. The red one tells the core clock and the purple one is the clock minus the injected idle cycles. The system performance is determined by the lower purple one.

TM2: FID/VID is the successor of TM1. It's more efficient in reducing energy consumption while still maintaining a high performance level. It looks like this:



The yellow FID switch on a regular basis and the temp is quickly following this behavior. It maintains a temp of 95°C in this example.

Don't judge by the temps to tell whether thermal management is active. There are more variables that trigger thermal managment than most care. Only tell by looking at the graphs.

Similar technics are used by the nVidia GPU. You have to use tools such as Rivatuner to tell. I noticed the nVidia GPU will fall back to a lower frequency if it get's too hot and often stays at this level till next reboot.

 

7oby, first I would like to thank you for your contribution to this thread. You presented a valid point and backed it with some data and graphics.

I am actually not running Vista. All test were ran under XP and I probably should have said that at the begining of my thread.

Yes, I did mention downclocking and I found that the trigger seems to be anything above 85c. My cooling system keeps the GPU under 80c no matter what you throw at it so I'm safe from downclocking.

Please run the same tests that I did, record your scores and temperatures and post them here. I think we are getting too deep into the technical aspects involved here when we really just need to look at the results.

Vista may have different behavior as you point out so the comparisons may only be valid under XP. I will run the same tests under Vista (I've got dual boot) and post my results. You may be on to something.

For all who would like to contribute to this thread:

1. Run 3Dmark 2005.
2. Record score and temperature of GPU (maybe even add CPU temp also).
3. Undervolt by 100 milliamps or better (at highest multiplier).
4. Re-run 3Dmark 2005.
5. Record score and temps.
6. Post your data.

Let the real world tests speak for themselves. And above all, have fun

 

That doesn't change much except in XP you have to select a power scheme without adaptive modulation of FID/VID:


http://www.microsoft.com/whdc/archive/winpowmgmt.mspx

Always on or Home/Office Desktop on AC would be a good choice if you want to perform some testing without speedstep.

You don't want that: I'm on X3100 on a M1330. My 3DMark would be somewhere around 800.

In general undervolting doesn't alter performance. Only the sideeffects that I described already.

 

Didn't know that. I'm sorry that you won't be able to participate in this experiment. But I do appreciate your comments.

I hope that everyone that wants to participate just follows the undervolting guide provided right here on these forums. That way everything is standard for each person. However, in light of what you have said 7oby, I'm going to mess around with some things in RMClock. I wonder just how much I can tweak things.

 

Once again, for anyone who wants to throw in their numbers :

1. Run 3Dmark 2005.
2. Record score and temperature of GPU (maybe even add CPU temp also).
3. Undervolt by 100 milliamps or better (at highest multiplier).
4. Re-run 3Dmark 2005.
5. Record score and temps.
6. Post your data.

I am waiting to see the first poster that claims that they can get 8700 points or higher with undervolting. Everyone knows that out of the box you will get around 6800 with an M1530. Let's push these babies!

BTW: please just use default of 1024x768 for all testing. That way all of the people who only have the 3Dmark 2005 demo can participate with valid numbers (you can't change resolution on the demo).

 

This doesnt make sense my t8300's multipliers are at .9500 with rmclock and im not getting any gpu performance difference...looking at your results slowdown117 2000 points of a difference is ALOTand i mean a couple hundred would be believable but 2000 is just not possible unless you have change to a faster gpu (which is impossible if you didnt know already).

 

I know, it just doesn't seem possible. But I assure you that it happened.

I am at work right now, but when I get home I will post the 3Dmark score screenshots that I took for normal and undervolted.

I have absolutely no reason to lie.

BTW: My 3Dmark 2005 score while overclocked is over 9800 (at 1024x768).
Keep in mind that I have moded my cooling system. I think my good temps are a huge factor as to why my card can do this. I did the thermal compound mod (remove the crap, stock Dell thermal pads, clean up, and apply good thermal compound. I reduced my max temps from 94c to 76c on one M1530, and from 94c to 80c on the other.

I did these tests on both machines with identical specs. Same results.

 

oh okay makes sense but would you tell me how you modded your cooling system since i would like to do the same with my system

 

Sure thing. Me and some other members were sorting this all out on this thread: http://forum.notebookreview.com/showthread.php?t=337934

and this one: http://forum.notebookreview.com/showthread.php?t=337729

Visit those links. There is good info in those threads

Recap of my instructions:
1. You need experience with this sort of stuff. One mistake can brick your machine.

2. Remove bottom cover.

3. Remove heatsink (7 screws).

4. VERY carefully clean the remains of the thermal pads with cotton swab and Isopropyl alcohol.

5. Be VERY careful around the tiny capacitors and resistors surrounding the GPU. Knock one of those off and the GPU will artifact, or it will be toast.

6. DO NOT disturb the blue pad on the North Bridge chip. You must reuse it. I found out why it's needed but I'll put that in another thread.

7. Apply some decent quality thermal compound. Use very small drops. Do not drown those chips with grease and do not let any grease contact the capacitors and resistors around the GPU. If you do get some grease on them, carefully clean it up with alcohol.

8. You must ensure that your heat plates are sitting flush on top of the GPU and CPU by pushing down on the heatsink over both chips, lifting up the heatsink assembly, and examining the greas patterns that you will see on top of the chips and bottom of the heat plates. You want good coverage and an even looking pattern.

9. Be brave, but cautious. You should be fine with this mod.

For reference, I used Dynex Silver Paste that I bought from Best Buy. If you can't get some really good stuff, this will work just fine.

 

BTW: Hey fluff, what are your present max temps of the GPU under full load, and what do you measure them with? I would use Rightmark Hardware Monitor.

I would like to determine if your GPU is downclocking due to high temps.

 

My default 3dmark05 result is 8600 without undervolting, are you sure this is related to power and not cooling?

max cpu 77, max gpu 76 after running 3dmark05.

 

mc1, that is very interesting and I'm glad that you did the tests. I will go home and triple check (I'll run the tests 3 times more just to be sure).

When I am not undervolted, I get a max temp of 74c on the GPU. One thing that I thought of is the possibilty that I have a weak power adapter that only allows good performance while undervolted.

I either have to somehow measure the DC voltage coming from the adapter under full load, or buy another power adapter and see what happens. If there is a voltage drop, that indicates that the power adapter can't handle the load. If the 19.1 vdc drops to let's say less than 18.0 vdc under load, that would indicate a whimpy power adapter.
Your score is right on par (only 240 points less) than my score while undervolted.

Thank you for your contribution to this thread. Good info
Your post made me question the integrity of the power adapter. Maybe that is responsible for such a huge margin between scores. Hopefully more peeps will post there scores. I did use the same power adapter with both M1530's that I ran tests on. I am becoming very suspicious now of the PSU.

Edit: mc1, could you also run the test while under a moddest overclock (say 600/800)? That would be most helpful. I am curious to know if you will break 10,000. My best so far @ 600/800 is 9847.

 

I have run a test when undervolted, other forum posts suggested that 1.15 was lowest stable setting for my t7500.

T7500, Windows Vista32, Nvidia 179.28

Dell recommended power settings:
3dmark05: 8556 cpu=77 gpu=76

RMclock 11x @ 1.15 (down from 1.212)
3dmark05: 8549 cpu=73 gpu=73

I have only done one test and I am pretty confident the results would stand over multiple tests i.e. the gpu performance is equal.

One thing I have noticed is the cpu idle's more when running with rmclock power settings, under the dell settings the cpu clock speed graph spikes a lot lot more.

How much would that overcock increase my gpu and system temps? I am trying to nurse this mobo/heatsink for a while since my last one reached 98c after 9 months of moderate/heavy usage before i got it replaced.

 

Alright, thanks. I've got to get to the bottom of this. My results are very strange when compared to yours. i would like to see some more people jump in here with some numbers.

 

I used to keep track of this kind of thing, I don't have the numbers around here, but I remember little to no change

 

I'm hearing several people now say the same thing. Around 200 - 300 points at the most is all that undervolting did for them. I think I smell a rat. I need another power adapter to connect and rerun my tests.

Well if my testing was flawed, at least you guys helped me find an answer. When I have the results with a new adapter I'll post. It still puzzles me though as to why I would get an awesome score of 8848 when I undervolt.

I'll run these tests again tonight and if something changes I'll post that info as well.

 

Ran the tests again. Same exact results with the scores. No undervolt = 6820, undervolt = 8840. There was one change however - the max temp for the GPU is now only 74c . What I think happened is that my thermal compound settled in and the temps dropped.

So at this time my results for the scores have not changed. I need more people to chip in on this thread with their numbers. I would like to get to the bottom of this as to why my score is so terrible without the undervolt. At the start of this thread I had attributed the performance gain to the undervolt. But it is looking like it's more like I have a deficiency without the undervolt.

As the title says - in my case the difference is shocking. I think everyone can agree on that. Does anyone know of a piece of software that will monitor the DC voltage at the power adapter input? If I can measure the voltage, and it sags while not undervolted, that would explain everything. I would hate to resort to hacking my adapter to read the voltage with a meter.

 

Are you still using rmclock for testing at standard voltages? if not could your advanced power settings in vista be limiting your cpu?

 

Good thought, however I am running XP Pro. To my knowledge there does not exist in XP a setting that would alter the max speed of the CPU, but I see your point though. I have tried standard voltages two ways - RMclock still active but undervolt profile is disabled, and shutting down RMclock to restore defaults to the OS.

That was a good idea though.

 

I'm spinning my wheels right now trying to think where I would get a place to put the probes on with my multimeter and measure the PSU voltage. I want to monitor that voltage while I load up the GPU and undervolting is disabled. I wish there was a program that would measure that voltage at the input to the mainboard.

 

Alright, I read this page http://newyorkermen.multiply.com/journal/item/85 and they were able to gain nearly 1000 points through undervolting. I'm not sure what GPU they were using but their top score was 3900 at default speed.

After reading that I decided to try 3Dmark 2006. Here are the results. Note that the GPU gains 551 points and the CPU gains 673 points as a result of the undervolt.

 

I just ran 3Dmark again while overclocked to 625/800 and my score was 10071. That would be the first time that this M1530 has scored that high.

Something else is wrong here. Those of you that stated there was no difference between standard and undervolted, are you absolutely sure that you disabled undervolting when you ran your tests?

 

Okay this thread is ridiculous. First what you state above and in your original post is all incorrect. You could be experiencing overheating when the CPU is not undervolted thus causing the GPU and or CPU to be throttling down to keep temps manageable that's why you have a lower score.

And second what you state above is wrong. The battery technically can supply as much power as the PSU if not more, just not as long. The battery is a 56 watthour battery meaning it can supply 56 watts for the duration of an hour or 112 watts for 30 minutes etc. 112 watt draw from the battery is a 2C draw from the battery and is not recommended but the battery can do it.

If you go by what you said, how do you explain RC airplanes and helicopters and such that use li-ion or li-poly batteries that have say a 7 watt hour battery but the heli draws 28 watts. With your explanation the helicopter would not be able to lift off the ground drawing only the 7 watts the battery can supply. But no that is wrong, the heli draws 28 watts from the battery for a duration of 15mins or less.

You should really research this stuff before saying everyone else is wrong.

And even though the full load of the M1530 is 88watts, it is still under the rating of the 90w PSU so it should still power it just fine unless you are also charging your battery which takes about 25-30W this leaving you with 60W which could affect the performance of your benchmark as both the CPU and GPU would have to throttle down to use what power it is given.

And to add to this, the battery can supply more than its watthour rating but only under a shorter amount of time than an hour (hence watt hour). But the PSU cannot supply more than its rated Watt output, that's why it is rated watt and not watt hour.

 

im using hwmonitor and Gpu-z to monitor my gpu temp. My gpu idle's at around 59c.

btw..i have replaced the thermal paste on my gpu and cpu. I recently took apart my whole M1530 to replace the lcd,cpu, and palm rest.

 

@slowdown117

If 3dmark05 can be run in window mode (= not fullscreen), do that. If not, then grab some TFT and use extended desktop (= external display is primary and laptop panel is extended). Move RMClock monitor mode and rivatuner to extended desktop or on free space in window mode.

Rerun 3dmark05 and observe whether any thermal throtteling is kicking in as I described here:
http://forum.notebookreview.com/showpost.php?p=4460983&postcount=20

The numbers of windows mode / extended will be different. But you should still observe the discrepancy between undervolted and non undervoltet. The 200 point difference others discovered most likely are due to a completely different steepstep handling. Your's have a different reason.

 

jhonny13oi - I'm affraid that you are the one that doesn't understand power and how batteries offer their power. But I haven't got time to teach you about batteries in this thread, nor do I have the time to teach you about 'headroom' when you design a power supply. If you don't believe me that's fine - just do this simple test:

1. Be sure to disable NV power miser.
2. Run 3Dmark on AC power.
3. Run 3Dmark on your battery, and wach your frame rate plumit.

That should clear things up for you regarding batteries and current. I may not be dead on about the current capabilty of the M1530 battery, but I'm close. Just do that test and it will all make sense to you. Don't argue and clutter up this thread. Just do that test, watch your frame rates hit the dirt when on battery, and then go study electronics and electron flow theory like I did when I went to school for it.

Everything else you said is irrelevant and is not getting any closer to the answer here. Everyone else on this thread has been pretty cool, but then there's jhonny...

Awsome fluff! I hope my guide was helpful.

Alright 7oby, I'll try that. But let's recap what we are hunting after -
I don't have any overheating whatsoever - 76c max. The GPU won't overheat and downclock until after 85c. I did the thermal compound mod that I have written about on these forums. Others have read my thread and afterward performed the mod themsleves. All have been successful at reducing temps. If anything, I am good at finding and correcting cooling problems. Heat is not an issue for my machine.

My performance is not suffering, it's awsome! 10,000 points in 3Dmark2005 @ 625x800! I could not achieve that number in the other M1530. There was no downclocking at any point. This performance is of course only while undervolted.

All that we really want to know is why the score drops to 6800 when I am not undervolted. After I read that link that I posted above, It seems that it's normal to see such a huge gain from undervolting.

 

On the link you posted he attributes the increase in 3dmark05 score directly to overclocking the GPU... and a frame rate decrease on battery could be attributed to BIOS/firmware control.

 

I won't argue that. But then you have to think of 'why?'.

I believe that the answer would be that the manufacturer knows that the machine would:

A) explode the battery from trying to deliver 90 watts inside of a 60 watt battery package.

B) burn up a trace on the mainboard.

C) both

If we have a 56whr battery, then we can draw 56 watts for 1 hour. We would be drawing 5.04 amps from the battery. The power consumption of the M1530 while loading the CPU, as measured by a "Kill A Watt" meter was 67or 68 watts. If the CPU is at 100% load, and the GPU is at 20 watt load (per Nvidia's specs, 8600M GT draws 20 watts at full load), then we are now drawing 88watts.

You wont typically load the CPU and GPU 100%, but you will see the CPU reach 60%, the GPU 90%, and with the chipset working hard, all that data on the busses, the hard drive churning away, the LCD very busy changing states, and the fan on full blast while you blast away in Crysis, you will be reaching the 90 watt limit of the adapter.

To draw 88 watts from an 11.1 volt 56whr (4400mAH battery we) would need 7.93 amps. That's just not gonna happen when the battery can only supply 4.4 amps. That's it. The battery is limited in it's current capacity. So the max power that the M1530 battery can produce is 48.84 amps.

Hopefully we have cleared up this misunderstanding. The battery can only give 4.4 amps max so when you run 3Dmark on the M1530 while on battery, your frame rates plumit because the 7.93 amps that the machine would need to run the GPU full bore is simply not available.

Now can we please stop talking about this and get back on track with finding the answer of the real question. I didn't want this to turn into power 101.

Edit: those of you with the super capacity battery (84 Whr) will get somewhere around 73 watt output from your battery, so you will have a power available increase. Your frame rate would be better.

 

I'm afraid to say: But it doesn't tell anything.

Running a laptop on battery (DC) is very different from running it on the PSU (AC). Different power plans run the CPU differently depending on whether the computer is powered by DC or AC. See "Home/Office Desktop" power plan here:
http://forum.notebookreview.com/showpost.php?p=4461124&postcount=22
Similar things apply to the GPU.

Your test doesn't tell anything about the capabilities of the AC and DC mode. If you know how, you can get the same performance from DC mode than AC mode.

Let me rephrase: You might not have an overheating issue, but a thermal management kicking in issue. Have a look at the RMClock picture of TM1:
http://forum.notebookreview.com/showpost.php?p=4460983&postcount=20
What's the temp? 74,6°C. What's it doing? It's throtteling !!!! Which laptop? It's an M1330 with the most recent BIOS. The thermal management of the M1530 is not that different.

If you would have read my posting more careful - it's all in there:

 

7oby, please reread my post above. The battery is limited and can only provide 48.84 watts. That is the bottleneck. It has nothing to do with power plans. Without the power the system needs to run full performance, it just can't do it.

Once again can we get back on track here.

Edit:

A score of 10,000, throtteling? I don't think so. The 8600M GT will not throttle until after 85c. At that point it severely downclocks in spurts. Whe're not even close to that temperature.

Edit: they need a smiley that is banging his head against the wall.

 

9 amps are not a problem for a Lithium Ion battery such as the 56whr Dell battery. A Lithium Ion battery of that capacity and voltage can supply up to 75 amps (!) and it would last 1/15th of an hour. Read more about that in this section:
http://en.wikipedia.org/wiki/Lithium_ion#High_internal_resistance

However draining that much current from a Lithium ion battery quickly reduces its capacity and it's aging. According to "GP Batteries" which have conducted many tests, you shouldn't draw more than 1/5th in amps of the total capacity in Ah of a LiIon battery. Meaning you shouldn't draw more than 11 amps (56Wh = 3 cells of 3,6V in serial connection of 56 Ah capacity). 11 amps * 11,1V > 100Watt.

I would like to see some backup of the assertions regarding batteries you make here. I can easily backup my data.