BIOS modding for GPU OC fun and profit!!

This is not imperial evidence, but I thought I'd report my findings after playing with some of the BIOS settings,
1) Thermal limits seem to work
2) reducing memory speed (from 1333 to 800) has no noticeable affect on battery performance.


1) Thermal limits:
My ears report less frequent high-temperature kick-ins coming from my 1645. I'm an Aaron Sorkin fan so i have all the DVDs for The West Wing, SPorts Night and STudio 60. Every night I watch an episode or two in bed (DVD, LCD brightness 100%, AC plugged in), and most nights my 1645's fan will kick into high gear after about 30 mins of usage (remember that the exhaust port is blocked).

The past couple of nights, i found that my fan was not rigorous as it normally is. That is to say, the fan never went into the "high speed" mode, it remained in the "normal speed" throughout. Maybe it's because I set the thermal limit to 79C, or maybe it's because the temperature has been much cooler the past while. Could be either one, but to help me sleep more peaceably at night I'll say it's the former.

2) reduce memory speed
I thought that by reducing the speed of my RAM, from 1333MHzto 800MHz, I might get a bit of extra battery time from my 9-cell. I wasn't expecting mounds of extra battery juice (or would gallons be a better metaphor?), but anything like 15 minutes would have been enough for me to smile about it. Alas there was nothing like that. I still range anywhere from 2:45 to 3:30+, all depending on how much I'm using my 1645 and what I'm using it for. If there _is_ any battery savings from a slow memory clock, I would say that it is not noticeable under normal circumstances (for me).

@gpig,
i found the same thing, that my normal thermal limit has not changed: I think it still goes off around 55-60C, but i am a little optimistic (as explained above) that the high fan active trip point was actually changed from 70C (IIRC) to 79C

 

When I tested out the fan speeds, I monitored the temperatures- but I only was paying attention to the fan's lowest and highest setting. Perhaps the fan's second highest setting is actually affected by the BIOS Control (high fan trip point).

Unfortunately, for me, I'm really only interested in the lowest and very highest setting. My fan goes into the mode you call afterburner mode (same speed as the fan goes during a BIOS update- that loud) when playing a game. That's the value I want to increase by a notch.

 

Just a bump. Has anyone had any more success with the fan change? I've tried changing the cooling policy from active to passive and it makes no differences either.

We've had ambient temps of around 35 for a week so I got used to the fan but today it's only 21 degrees and the fan is off until I open up a webpage then turns off (CPU is reporting 45 degrees). I'm finding this more annoying than having it on low all the time.

 

yeah i'm still on the fence about changing the cooling fan settings. my previous post still stands - i can't entirely tell if changing the top limit from 71C to 79C.

not to sound like a downer, but the settings in the BIOS don't seem to have any effect. but maybe it was worth a try?

 

Regarding the FAN trip-point setting, the settings does work but the FAN might not work as we imagine it should work. You can check the changes you did in BIOS using AIDA64 (ACPI Tool). According to my investigation it does follow ACPI specification. The objects you might want to check are:-
" _AC0" - high fan speed
" _AC1" - low fan speed
- The value is in tenths of degrees Kelvin. 3000 means 300.0 Kelvin


(Using my HWmonitor screenshot as an example) The " TZ01" temperature is responsible in cooling policy:-


Based on my testing on my Studio, the FAN speed somewhat does change according to the FAN trip-point settings in BIOS. (I can monitor FAN speed in AIDA64) The FAN on my Studio, I believe have 4 speeds at least; 22XX-23XX, 27XX-28XX, 32XX-33XX & 4XXX RPMs. When the temperature reached high-fan trip-point, the FAN only goes up to 32XX-33XX RPM. It only reached 4XXX RPM when I stress test the CPU (I don't remember the temperature when it kicked to 4XXX RPM - too many temperatures to monitor at that time, forgot to wrote it down).

The only explanation I can think of right now why FAN doesn't work as we imagine it should work:-
1) BIOS contain buggy ACPI - this is not shocking news as it usually does
2) Windows have their own cooling policy which override the BIOS settings
3) The thermal zone sensors have it own mind - have autonomy control on the FAN

The passive cooling trip-point does work. When I set it to lower value, the processor does throttled down.

EDIT: I like to see if there is someone with linux installed on their Studio XPS, can check the FAN behavior on linux for us.

 

For what's it worth, watching HWmonitor it shows that when TZ01 hits 47 degrees all cooling turns off for me. From there onwards it's:

55 degrees and fan level 1
62 degrees and fan level 2
68 degrees and fan level 3
74 degrees and fan level 4

In BIOS I've got the fans starting at 79 degrees.

I then loaded Ubuntu from USB and ran few of the commands.

cat /proc/acpi/thermal_zone/TZ00/trip_points was 127 degrees.
cat /proc/acpi/thermal_zone/TZ01/trip_points were:
- critical (S5): 85 degrees
- passive: 95 degrees: tc1=0 tc2=1 tsp=2 devices=CPU0 to CPU7
- active [0]: 79 degrees (devices=fan1)
- active [1]: 79 degrees (devices=fan0)

When I checked the temp, it initially showed 50 and then dropped to 49 degrees and the fan was on the lowest setting (not off).

So it appears to read it from BIOS correctly and 'behaves' similarly, so all I can assume is that TZ01 has it's own mind and doesn't care about BIOS at all when it comes to engaging the fan.

 

Gosh darn it.

 

it's gotten warmer again around here, and it seems that my fan is back to normal. my optimistic results from last week really were from the cooler temperatures it seems.

look on the bright side: ...


oh, wait. right. nevermind.

in all fairness we can't expect great things from this BIOS "revision" (which i'm calling it). it was amazing that seeker_moc (and thanks to kizwan!) was able to unlock these options within the BIOS. they don't appear to really work, but it was worth a shot and now we know.

 

Nice work seeker_moc.

Can you "create" a personalized BIOS with other custom GPU/vMEM clock?
I'm testing 150/250Mhz (with 1.2v..) and 800/900Mhz, but it is interesting if clock are like this

150 250 0.9
150 250 0.9
150 250 0.9
150 250 0.9
150 250 0.9
150 250 0.9
800 900 1.2

Extreme power saving but also high performance

Thanks

 

Under-clocking your memory by simply reducing the frequency will not increase your battery life as each DIMM will still eat the same amount of power(1.5 or 1.35 volts, can't remember which.) You would need to actually reduce the voltage being sent to the RAM before you would see any power savings. Unfortunately, RAM doesn't use much power to begin with and it's likely you wouldn't see much of an increase in battery life if any.

 

I could, but it might not save as much as you think. Instead of having graduated steps of power/performance, it will be either all or nothing. Those settings won't lock you into 150/200 most of the time like you think. What would likely happen is that every time your GPU needs a small amount more power, it will kick all the way up to full blast, using more power than normal.

We could try a few things, like lowering the clocks and/or voltage for the lowest clock setting, but leave some of the intermediate steps available. Or, if you wanted full power saving possibilities, I could set it to ONLY use 150/200, and then you could use the GPU clock tool to set the clocks higher manually if you need them.

 

Sorry that I've been out for a while, I haven't really had time to test out all the BIOS options I'd like to. I'd like to test out the fan triggers a little. Has anybody tried to lower the trip points at all? We've already established that the RAM speed setting will allow you to under, but not overclock the RAM. I wonder if it's the same with the fan temps. Can we lower them, but not raise them?

Also, has anybody tried changing the DTS setting? The description of the fan trip points indicates that the DTS setting is somehow related. Maybe changing it will make the fan settings work.

 

Yeah, I tried this.. putting them to the lowest setting which from memory was something like 37 (the one before disabled), no changes. It still turns itself off when I hit 47 degrees.

If you mean DTS SMM and enabling it... It does mess up HWmonitor and it's TZ01 monitoring. It still works, but changes temps much differently and unfortunately, the fan still keeps behaving the exact same way (you can follow the core temps and assume it hit 47 degrees or 55 etc).

 

About the " DTS SMM", assuming " SMM" stand for " System Management Mode". According to the information I found, SMM have flaws/problems. So, I didn't enabled it. Already tried to enabled it but no changes on FAN behavior. I already set "Active" & "High-Fan" trip point to the second lowest ("High-Fan" higher than "Active").

 

If I've PowerPlay enabled you don't use 800/900 clock. I see that there is a "big" temperature difference between 220/300Mhz 0.9v and 150/250 1.2v, like 5°c in idle. There are others voltage options?
If I use rivatuner, I can't manage voltage, so I've always 1.2v and ATT don't works.

 

I can set voltage to whatever you want. The only problem is that there's no way to test it out with software beforehand, so there is some risk involved. Let me know if you want to try something out, if not, I'll build you the clocks that you asked for a few posts back.

 

I've no problems to test these clocks:

100 250 0.9
100 250 0.9
100 250 0.9
100 250 0.9
100 250 0.9
100 250 0.9
800 900 1.2

It's interesting that you can select other voltages, but I think that 1.3v are too much and increase throttling issue (I don't have this problem..)

 

I see your clocks are at 845-945Mhz... is that OCCT stable? The best I can do is 750/940. If you try higher volts, let us know how you go...

 

830-930Mhz are OCCT and Furmark stable without errors.
845-945Mhz are stable under a lot session of 3dmark vantage/06/03/01 and acquamark. I don't have run OCCT with these frequency

 

Hey all, thanks so much for your work - I've flashed Seeker's BIOS without problems, and it works beautifully!

Just one question: Is it possible to change the vBIOS DeviceID from 9488 to 949C (Firepro 7750)? I have soft-modded the driver already to work with the DeviceID 9488, mainly to enable the 10-bit per channel display option.

However - I tried modding the BIOS (I used the Hex locations in RBE to change both instances of 88 94 to 9C 94) and allowed RBE to recompile it, then put the modded rom into PhoenixTool. But, when I flashed the BIOS, I ended up bricking the computer - it turns on, and beeps 6 times, and the screen never displays anything. I can hear everything else going on (well, not really, but the HDD light on the front flashes, indicating that it's going through the Windows boot process just fine) - but without a functional videocard, things are problematic. Thanks again to ppl on this forum, I've restored things to the way they were before.

Is it possible to change the vBIOS DeviceID without bricking? If so, is it possible to post a modded bios on the front of this thread, with the mild overclock, for 1645, with the 4670 DeviceID changed from 9488 to 949C?

Thanks heaps!

cheers,
dave

 

Sorry it took me so long, but here you go. 1645_A12_unlocked_4670_800_900_other_100_250.zip

 

this is a bit off-topic, but it has to do with the new BIOS options available in the modded BIOS.

one of hte options in the AHCI options in the BIOS allows you to enable/disable port multipliers for the SATA devices. has anybody tried these yet? or at least, does anybody know which SATA device is the eSATA port?

if not, i'll give them a go and see which one is for the eSATA port.

i ask this because I have a ThermalTake DuetX external hard drive dock which supports 2 simultaneous HDDs over eSata or USB. unfortunately you need port multipliers for this - my desktop has no problem, but so far the 1645 has had no luck seeing 2 HDD's docked at the same time.

 

Thank you very much!!!
It works fine at the moment

 

I think it's port 5, or at least that's the only SATA channel marked as hot swap-able by default.

 

Seeker,
Thanks again for all of your help and work on this. I figured I would wait to ask for a custom unlocked A12 800/1000 vA09 for the 1645 running a 4670 until you worked through some of the questions and stuff. Hopefully, this is a good time. Thanks!

 

1000mhz on memory? wow, over 955mhz I've artifacts

 

Yeah, I had the same problem but this might help...I removed the dust filter from the bottom plate directly below the fan and use an external cooling fan. Without these 2 things especially removing the dust filter, it probably wouldn't work.

 

Sorry to ask again but may have been lost on previous page - is there any way to change the DeviceID for the Mobility HD 4670 in BIOS successfully?

Aiming to change to Firepro 7750 (both RV730 chipset) - device ID should go from 9488 (or 88 94 in the BIOS) to 949C (or 9C 94 in the BIOS). I have tried it a couple of times, correcting the checksum, but both times ended up having to blind recover the BIOS.

Any suggestions?

cheers,
dave

 

According to this post I found (from the creator of RBE, so I'd think he knows what he's talking about), DeviceID change - techPowerUp! Forums it seems like you're doing everything right. However, some laptops check for all hardware when booted, and if something's not on the 'whitelist' it won't work. It seems like that might be the case with you. I'm not sure how to modify the system BIOS to recognize you're (assumed) correctly modded vBIOS. You should ask over on the gaming forum, some people there (like nando4, niffcreature, or moralhazard) might be better able to help you.

Also, the 7750 is based on the desktop version, not the mobile one. Though they're more or less the same chip, this may be an ID issue for booting the vBIOS. The FirePro M5725 is actually based directly on the mobile 4670 (not to be confused with the V5725, which is based on the 3670), you might be better off trying to change the DeviceID (which I don't know) to that instead.

Edit: I pulled apart the HP EliteBook 8540w (which comes with the M5800) BIOS to look for the M5800 vBIOS to see if I could glean anything from it that would assist us with this. Strangely, although there were 3x Nvidia vBIOSes in there, I couldn't find an ATI one...???

Do you know of any laptop that comes with the M5725 or M5800 besides the EliteBook 8540w???

 

sorry its been a while since ive posted here. How do i flash an older bios and where do i get those bios from? i am trying the ones from dell but they will not flash since i am already on a12

 

Is there any reason you want to flash to an older BIOS? The new mod BIOSes that I posted on the first page, or any new custom ones I make, will flash regardless of what your current version is, so you no longer have to flash to an older version first.

If you still want to flash an older stock BIOS for whatever reason, the instructions can be found here: http://forum.notebookreview.com/dell-xps-studio-xps/450888-up-downgrading-your-1645-bios.html

 

Here you go: 1645_A12_unlocked_4670_A09_800_1000.zip

 

Hey seeker... seems like people have been keeping you busy since I last checked this thread.

You made a custom OC for my 1647... 750/1000. I noticed that the current OCs are 650/800.

Is there a reason for this?

 

Stock is 650/800. Current "generic" OC is 750/900. I turned it down because some people were having trouble with mem clocks over 900. That doesn't mean that your BIOS is bad. If you're not having problems with it, then enjoy. I never really intended for many people to use the "generic" OC BIOS, it was more like a starting point, where people could try it out, with clocks low enough that it should work for everybody, do testing, then request a higher-clocked custom BIOS that works for their particular laptop.

 

i need to go back and reflash my a12 because my cores aren't being recognized correctly.

 

I'm pretty sure you could change the 'base/mild OC' to 780/940..that'd be another 5% on top of the existing OC. By the looks of it, I have the worst (at least one of the worst) cards and that's the OCCT limit for me.

add: temps are a non-issue.

 

Thanks for the reply.

M5725 is based on the M86GL chipset (as opposed to M96 for the Radeon Mobility HD4670), and M5800 is based on Madison - so those with the HD5730 Mobility could emulate this one... Unfortunately, I haven't been able to find any Firepro card that have the same M96 codename as the 4670...

Perhaps this means there's no way to BIOS-mod the Device ID to a FirePro?

For the time being, I'll stick with soft-modding ;p

Thanks heaps anyway... Please post any further ideas that you might have though! Am willing to consider and try!

cheers,
dave

 

Thanks, my friend!

 

Has anyone had any more progress with controlling the fan speeds?

 

On DeviceID mods:

An interesting proposition for owners of the Mobility HD5730 version would be a mod to the M5800 as Seeker has already suggested.

The DeviceID would be 68c1 - (instead of 68c0)... If anyone is interested in trying that out in order to unlock the FirePro capabilities! I understand that the majority of the unlocking is done at a driver level anyhow, rather than at the BIOS level - but if the DeviceID could be changed in BIOS so that the driver installs properly (as downloaded, without softmod), then theoretically the HD5730 can be made to run exactly as a M5800 Firepro...

Unfortunately I'm still unable to find a suitable equivalent for the Mobility HD4670 - and as a result, bricked my laptop twice trying to mod it... (no huge loss, easy to recover, but disappointing still)... I had tried the desktop Firepro V7750, as it seems this is the card with the closest resemblance, but apparently not close enough...

cheers,
dave

 

are there any modified bios for xps 15? and i looked through the unlocked bios screenshots and didnt really recognize the options. are there any potential performance-increasing options in there asides from overclocking?

 

I read in a much earlier post that there hadn't been success mod'ing the BIOS for the 1640. Is that still the case?

I don't care about OC'ing my 1640 as much as I just want it to stop throttling all the frekin time... Am I better off just using a software solution like ThrottleStop or can i do something at the BIOS level.

Thanks, and great thread. Tons of excellent info....

Ck

 

Sorry, meant to include:

XPS Studio 16 - model 1640
3670 gpu
RGBLED
BIOS = A14
130w Power Supply

 

Not really. Most of the extra options are only useful in a few specific situations. Also, I can't do the mods required to unlock the extra options, you'll have to ask Kizwan over in his thread (linked to in the first post). Although I haven't done it yet, I could mod the vBIOS for the new XPS laptops for overclocking if anybody is interested.

 

The 1640 uses the Core2Duo CPUs, which unlike the core-i series, is capable of undervolting. Undervolting the CPU could drop temps significantly, eliminating throttling, without decreasing performance. No BIOS mods needed. Read about it over in this thread: http://forum.notebookreview.com/har...arket-upgrades/235824-undervolting-guide.html

 

Thanks for the great info/idea/link... I will try that first.

I hate to think that i have a decent laptop (w/ T9550 dual core) and can barely use it as Dell likes to basically shut itself down at the first sign of mid-range temps.

I did read that the CPU and GPU in the 1640's share the same heat sink. If thats the case, then the undervolting will certainly help.

Will post results soon as i get chance to try it out..

Thx again...

 

What firepro capabilities are we talking about here?

 

The difference is that FirePro unlocks 3D acceleration in professional apps, like CAD or other design/modeling/encoding type programs. I won't make any difference in games or home/office work.

 

Ah I see. Good call. I haven't had any issues with your OC thus far *knock on wood*. When I really push my GPU, my core temp gets to 65-70C ish with the fan on max. Would be curious to see how much more I could squeeze out of this.

 

A lot more really. 70C max is quite low.