Jump to content
LaptopVideo2Go Forums
aapenaa

Help Undervolting Old Geforce 6150 IGP via flash

Recommended Posts

aapenaa

Hi Everyone!

 

As the title explains I'm starting this post to find help in flashing an old geforce 6150 (nforce 430) IGP, this card is very common in old hp laptops (Pavilion Tx1000, Tx2000 and some Dv's) and it is well known that such models commonly fail by overheating.

 

I own one Hp Tx2000, and I managed to bring it back to life by the old heatgun trick, after that, I've been very careful with the temps and managed to lower them significantly, but I'm still afraid that the GPU will fail eeventually, I've been looking for undervolting software compatible with the card but had no luck so I though I could then flash it to UNDERvolt or UNDERclock the factory settings to lower the temps but haven't found any solid information of the procedure on this card.

 

I've already looked at nvflash compatibility list an it says it is compatible with Geforce 6 series, also I've managed to exctract the GPU bios using GPU-Z and I can open it using NiBitor, So I think I can modify the clock and voltage values.

 

the questions is, will it be possible to flash the card using the modified bios, or by any means will I end up (again) with a useless tx2000? any one has done it before, not necessarily on the go 6150, but any other onboard GPU?, did it work?

 

I would like to hear any suggestions or advices, and people intested in help are very welcome.

 

Thanks In advance.

Share this post


Link to post
Share on other sites
milone

I have a tx1220us and am curious if this will work. I modified the VBIOS by reducing 3D to 1.15 (-0.05) and Thrtl to 1.05 (-0.05), and adjusted temperatures (critical = 95, Throttling = 80, Fan boost = 70) because they were all set very high (120-130) and I've learned that 80+ for a period of time can be enough to kill the GPU and require reflow.

 

That said, I am too afraid to try flashing this. I do have another motherboard that I apparently killed playing Angry Birds (I had successfully reflowed twice in the oven, but reflow didn't help after an Angry Birds crash.) I bought a replacement board on ebay for about $60, which is what I'm using now.

 

I used RMClock to undervolt the CPU, and was able to get CPU idle temps around 37-40 degrees C. My CPU voltages were 0.8000V (4x) to 0.1750 (11x), and I was able to lower them to 0.6875 (4x) to 1.0250 (11x). I have a Turion 64 X2 TL-64, which has a 35W TDP. I just ordered a Turion 64 X2 TK-57 for $6 on ebay, which is a slight performance decrease but is only 31W TDP.


I also modified my heatsink. I originally tried creating separate heatsinks for the CPU and GPU, since one heating up causes the other to heat up. I had fit a 24mm 5V fan above the GPU heatsink I created, and drilled holes to let it vent out of the bottom of the case. The fan was loud, and didn't cool well enough and that's when Angry Birds killed the GPU while testing it out.


Since then, I went back to a stock heatsink. Similar to the "copper shim" trick, I bought a sheet of copper and created a shim that goes all the way down to the fan housing and has holes drilled to help get ventilation from the fan. I also use thermal conductive silicone pads (TP-260, 6.0 W/mK) on the GPU (between GPU and copper sheet) and between copper sheet and the CPU heatsink. This seems to work better than regular thermal paste.


Right now with a cooling pad I'm getting idle CPU temps of 38-39, and idle GPU temps of 62 degrees C. Without a cooling pad those temps go up almost 10 degrees C. The CPU under load (running ORTHOS) will max around 93 and stay there thanks to the undervolting (max operating temp for it is 95.) Prior to undervolting I couldn't run ORTHOS for longer than 30 seconds without temps going 95+. I have not stress tested the GPU because I know the consequences. If I can get this thing stable, I only plan on using it for web browsing, etc. I just put Windows 8 Pro on it, so hoping to get it cooler so I can run Windows 8 apps without killing it. I have AIDA64 overlay temps in the corner of the screen, but this only works in desktop mode, not in metro... so I can't keep an eye on them when using apps.


I'd love to find the parts to make a water-cooling kit for the GPU, even if I had to add something to the bottom to house components. It would have to be a 5V pump, and would hurt battery life... but at least it would stay alive. I'm not sure how I would mount it to stay firmly against the GPU, but I'll worry about that when I find a small enough 5V pump.


Anyway, please post results if you try flashing. I've never flashed VBIOS but I'm guessing you just boot to DOS and run Nvflash? It's too bad most people who had these laptops/tablets gave up on them, and most a******* on ebay charge $100 for their motherboards.


UPDATE: I found motherboard schematics for the TX1000, I believe. If only I knew what I was looking at. I was thinking that if I could find where the VBIOS is, I would have a backup in case I brick the current motherboard by flashing it. This is all assuming I understand what I am talking about. I know people who have flashed the VBIOS on discrete video cards were able to recover them by taking VBIOS chips from other ones. From what I understand, it is probably going to be a chip with 8 connectors, 4 on each side... unless it is done differently on laptops. If anyone has any idea, please feel free to chime in. My other option is buying another motherboard from ebay if I ruin this one, but I'd like to not spend any more money on this thing.

 

UPDATE 2: Out of curiosity, I tried backing up the VBIOS using nvflash and no matter what command I try to run, nvflash returns "Error: Could not enable the PROM pins"... so it may not be possible to flash the VBIOS. Even running "nvflash --protectoff" (or "nvflash -i0 -r") returns the same error. Strangely, I am able to back it up using GPU-Z, and nvflash does show the GeForce Go 6150 when I use "nvflash --list" (or "nvflash -a"). I found the closest thing I could to motherboard schematics for the tx1000 and uploaded them here. I also read this post and tried locating the VBIOS EEPROM, with no luck. I was hoping that if I could find it I could always take it off of the bad motherboard in case of a bad flash.

Edited by milone

Share this post


Link to post
Share on other sites
aapenaa

Hi Milone, I'm glad you posted on the topic, I thought it was dead, it seems that no one cares. well basically we've both tried the same thing, I've too extracted the VBIOS using GPU-Z, modified it with NiBitor and tried to flash with the same error 'Could not enable PROM pins', I have also been lookiing for cheap motherboards to test the flashing procedure with no luck so far, what I can advice for sure to lower your temperatures is to use the least powerful of turion's supported by our mobo which will be TL-50, right now I'm Using it undervolted to 0.6875 on 4x and 0.8750 on 8x and the temperatures stay at 30~40 °C idled and gpu on the mid 50°C idled, under heavy load (Running ORTHOS and Furmark) the whole system gets to about 85°C (GPU and CPU), This has to do with a mod I did to the heatsink, basically if you cut the white wire on the heatsink fan, the fan will stay running at full speed always, The noise is unbearable but at least it cools down the system a bit.

 

I am totally sure that this isn't the fix we need, what we need is a way to undervolt the GPU by hardware, I've read somewhere that in tx1000 the VBIOS was Somehow integrated in the system main BIOS, and probably that's why we are unable to flash the gpu via nvflash.

 

I have a little knowledge on electronics so I'll have a look to the MOBO schematics you attached to see if there's any hardware mod we can do to undervolt the GPU, I can't promise anything but I'll try.

 

I just wish there were more people interested on this thing, this is such a good laptop, if we can only solve this issue.

 

UPDATE:

 

Ok so I just checked both the MOBO schematics and the MOBO itself, in the mobo schematics the only mention of a ROM chip is of that of the MAIN system BIOS, then I opened up the laptop and checked for VBIOS-PROM-like chips near the nvidia gpu, but found only a  voltage regulator and a multiplexer, so my suspect is that this system has no individual PROM chip for the VBIOS ,that would explain the nvflash error message 'Could not open PROM Pins' of course it can't if there are none!, what I'm thinking now is that we can somehow edit the system main BIOS to lower the GPU VCC, I KNOW that editing the system BIOS could be done because right now I'm running on a modified BIOS wich skips the wifi card whitelist imposed by hp, what I don't know is, if the system MAIN BIOS could be edited to undervolt the GPU.

 

basically this is an OPEN CALL for anyone who knows how to edit bios files to help us find the VCC, Core clock and Mem clock values inside HP's Bios Files and edit them.

 

I really hope someone can help.

Edited by aapenaa

Share this post


Link to post
Share on other sites
milone

That's an interesting idea about cutting the white wire. I have a box of small black slider switches I bought a long time ago when I was trying to mount a fan inside, so I may try to run the white wire through a switch rather than cutting it so I can have the option of running the fan at full speed or not. Then I'll just cut a small hole in the case somewhere to mount it.

 

I also still have a couple 5V 24mm fans I may try to mount anywhere inside I can to keep air circulating. They're loud though (sort of a whistling sound) so they need to be on switches and only turned on when really needed.

 

I felt the "copper shim" wasn't enough so you can see photos of what I did here (although I made some changes after taking the photos):

 

https://plus.google.com/photos/117693056407035997676/albums/5844946768849845713?authkey=CP-7kJbbmppb

 

Above the GPU, I ended up with these layers: GPU -> thermal pad -> .51mm copper sheet -> thermal pad -> CPU heatsink (actually the small square that is attached to the CPU heatsink, I left that there.) I used 6W/mK thermal pads (a 4"x4" sheet is about $7 on ebay.) I put holes in the copper sheet over the hole in the motherboard because I assumed it was there for airflow purposes. I'm hoping that having the copper run all the way down to the fan helps cool it on that end and draw heat away from the GPU. 

 

Also, the motherboard I have now is one I bought from ebay for about $60 and I think they may have modified the system BIOS because it says I have version F.21 even though F.20 is the last released from HP. I think maybe they tried to alter the fan profile even more, although I don't notice much of a difference from the F.20 motherboard. I downloaded Phoenix Bios Editor and looked around at F.20 but realized I have no idea what I'm doing. I posted those motherboard pictures on reddit and asked for help identifying the VBIOS EEPROM, but didn't get much help and the one person who did take a look at it said it was probably internal to the chipset - meaning no flash I guess. I wish I could find an Nvidia engineer and ask him a few questions before I strangled him.

 

I have ran ORTHOS but I'm too afraid of running Furmark without liquid nitrogen on the GPU.

Share this post


Link to post
Share on other sites
aapenaa

WOW! that mod was impressive, I've saw your post on reddit while googling how to mod the GPU VCC on system Bios, I think that we should try modding the System Bios, but we need someone who really knows how to do it.

I would suggest you to leave the contact between the nvidia chip and the heat sink as clean as you can, I mean, chip-thermal compound-copper. I´ve heard of people who have soldered a copper shim to the actual heatsink to fill the gap between the heatsink and the chip and improve heat conductivity.

 

I also read somewhere that this guys: http://www.biosman.com/custom.htm would do BIOS edit for cash. 

Share this post


Link to post
Share on other sites
milone

The custom BIOS mod looks interesting, but I'm not sure how well it will work. For me, the CPU temps are no longer the problem. After undervolting, I get idle temps in the 30s (with cooling pad.) BIOS tells the fan to kick on as the CPU heats up, not the GPU. I'm not sure that BIOS even has access to the GPU temps, because I could not access GPU sensors until I installed the correct drivers. So I think being able to access GPU temps is dependent on the driver, and therefore the OS. I'm not an expert so maybe I just don't understand correctly, but my goal would be to have the fan speed increase if the CPU or GPU got hot, not just the CPU. As an example, AIDA64 would not detect the GPU thermal sensor while using the drivers Windows installed automatically... but once I installed the correct drivers the thermal sensor is found. So I'm not sure how BIOS would be able to access that. I guess it's possible to try changing the fan profile for the CPU even more, but I'd prefer something that was able to react to the GPU temps since that is what I'm more worried about.

 

As I mentioned, I have BIOS F.21 while HP's latest was F.20. There was also a sticker on the BIOS when I received it that said F.21. I think the seller is a business specializing in this area, as they had several of these laptops for sale and I think they had modified the BIOS themselves (or paid someone else to.) So I'd like to send the guys at biosman.com this BIOS I have and see what is changed from F.20, as well as seeing what they thought they could do. But I'm not sure how to save it. It seems like Phoenix is the one BIOS that is difficult, if not impossible, to backup (go figure.) I've even read that using tools that backup AMI and Award BIOS can actually corrupt a Phoenix BIOS. I tried to do a "backup only" in WinPhlash 64 and it required me to specify a new BIOS. I pointed to version F.20 (figuring it didn't matter because I was only backing up anyway) and then after warnings about flashing BIOS (which I wasn't doing), it said that the flash failed! Not to mention, no backup created! Luckily it didn't affect the BIOS in its attempt to do whatever it was doing.

 

It would be great if there was an older version of Nvidia drivers that allowed underclocking the GPU. I'm now using Windows 8, and it looks like drivers for the 6150 and the nForce chipset were created specifically for Windows 8 (though it took me a while to find them) so I'm not sure how much luck I'd have trying different versions. I've heard of others having to install older drivers to be able to modify their GPU clock speeds. I installed Nvidia System Tools and under Performance, it has a slider that lets me adjust the GPU clock speed from 425, but when I click Apply it says something about the changes could not be applied and reverts back to 425. The same thing happens with MSI afterburner. Rivatuner does nothing. I've been a software developer for 10+ years but I've slowly started trying to learn more about hardware and drivers just to one day get this laptop working correctly.

 

I plan on taking it apart again to change the CPU and mount a switch for the fan's white wire you suggested, and I'll also take your advice on getting the copper plate closer to the GPU by using thermal paste instead of the thermal pad I have now, as long as it doesn't create a gap. I could stick a small copper shim in there if it was, but I think the less number of layers, the better. The thermal padding I have seems to work really good, and I know I looked at the specs before I ordered it because I planned on using it for this laptop. The temp ranges were appropriate for GPU use. It's a soft gray material that you can cut and mold any way you need to. Compound is probably a better word than pad, but it was labeled as a pad when I bought it and came sandwiched between two sheets of plastic because it is very soft and you can easily tear it into pieces.

 

I'd also like to switch to an SSD eventually to lower the heat inside the case even more, but that's not in the budget at the moment... especially for something that feels like it could die at any moment. Maybe once I use it more I'll get more confident. I'm still waiting on a new cooling pad so I haven't used it too much, other than playing around with different underclocking and undervolting tools. 

 

The other thing I've looked into is mounting a 24mm fan inside (if I can fit it without going back to a hacked apart heatsink) and slowing it down with a resistor so it's not as loud. Maybe putting it on a switch as well so it can be ran at two speeds. I have a feeling that by the end of this, I'm going to have switches all over this thing.

Edited by milone

Share this post


Link to post
Share on other sites
aapenaa

milone, your current BIOS could be extracted by using the winphlash tool that's commonly used to flash phoenix bios inside windows, it comes with an option of 'Backup BIOS only' so you can backup your current BIOS, here's a screenshot http://itc.ua/img/ko/2002/41/2Winphlash.jpg if you want to extract your current BIOS for testing I would suggest doing that but being very careful.

 

I think you have very good ideas to lower the whole system temperatures (by hardware modding and ssd(haven't thought about that)), but if we can manage to lower the clocks/voltages in the GPU chip by providing only a 'BIOS update' that would very helpful for a wider range of people with this same laptop and issue, I think that this is what hp must've done, provide a BIOS update to lower the power drawn by the GPU and modify the fan profile to be less tolerant to high temperatures, but they failed at it, so, it seems to be our job now.

 

I would be glad to do it if I knew how, but right now when I open Phoenix BIOS editor with my bios file, the only thing I see is a bunch of hex digits which i can't fully understand, all of this is supposing that I was right on my hypothesis that the GPU VCC/Core Speed/Mem Speed values are hiding somewhere inside the system main BIOS, but I'm not an expert I can't assure that they are.

 

so the conclusion is, right now we need a BIOS modding expert, to tell us if that values are hiding in the system MAIN BIOS and if they can be modified, we must start knocking doors I guess, I suggest starting here: http://www.bios-mods.com/ .

 

UPDATE: 

 

Ive started a thread here: http://www.bios-mods.com/forum/Thread-HP-Pavilion-Tx1000-Tx2000-Phoenix-Bios

 

UPDATE 2:

 

I've opened the MAIN BIOS file with an HEX editor and the VBIOS file extracted using GPU-Z and guess what, found a match, this makes me think even more that the VBIOS is indeed inside system BIOS i just need to find exactly where and how to modify it.

 

Original System BIOS

 HP_BIOS.jpg

 

Nvidia GPU-Z extracted BIOS

image.jpg

 

The strings mv51 and 'C51MV VGA BIOS' can also be read..... SO EXCITING!!! i think we are very close to it.

Edited by aapenaa

Share this post


Link to post
Share on other sites
milone

I tried WinPhlash shortly after posting the last reply and went back and added the results - it didn't work. It seems like the "Backup only" option was added as an afterthought and the program wasn't updated correctly to account for it. Even though I selected the Backup Only option, it still required a new BIOS file and it appeared that it attempted to flash it.

 

From what I understand (and I may be wrong), the BIOS updates from HP were to create a more aggressive CPU fan profile (based on CPU temp only.) I think this was their easiest way of trying to prevent GPU overheating - since the CPU and GPU share the heatsink. But yeah the VBIOS has to be somewhere, and the system BIOS is probably the best guess.

 

I installed a switch and ran the white PWMFAN wire to it, so now I can switch from the default fan speeds (determined by CPU temp) to full speed when things get warm. I found room to mount it to the left of the power switch on the bottom of the case.  I've added some photos to the album here, but it's a small black switch so it blends in very well. While I had it open, I installed the TK-57 CPU I got from ebay for $6. It didn't work. I guess I should have spent more time checking compatibility, but for $6 I'm not too upset. I wasn't entirely sure I wanted to sacrifice any more CPU performance anyway since my CPU temps aren't bad. Still waiting for a new cooling pad to be delivered, which I have black double-sided adhesive I'll be using to attach it to the tablet since I'll always be using it. That way it won't be sliding around when I'm using it as a tablet.

 

I possibly found a better fan here (specs here) I think I am going to order. It's a blower, so it will pull in air from the ventilation holes I added right under the GPU and blow it out the side, across the GPU/copper. It looks like it might take a while to ship though. The fan is only 16mm x 16mm x 4.5mm so it may fit even without cutting the original heatsink. The only thing I'm not sure about is if the .423 cubit feet per minute it blows is even worth it.

 

Also, if you're running Windows on yours you may want to take a look at PowerStrip. I just installed it so haven't played around with it much, but it seems to allow me to adjust the GPU clock. When I moved the GPU clock from 425 down to 424, 420, and then 400, other programs (AIDA64, GPU-Z) reported the GPU to be running at the new clock speeds. I'd watch it idle at 100MHz, then run FurMark and it would jump to the new 3D speed and stay there until I stopped FurMark, in which it would return to 100MHz after a few seconds. I'd obviously prefer undervolting to underclocking, and I hate to take performance away from this already weak GPU, but I'll take anything that works. One thing I noticed is that running GPU-Z while PowerStrip was running caused repeated Nvidia driver crashes. This is in Windows 8 Pro so your results may be different. PowerStrip has options for running in the background, running on startup, applying settings on startup, etc. just like RMClock. I don't have it set to run on startup yet, as I want to experiment more with different clock speeds... but it seems to be the only thing I've found that actually changes the GPU clock. Hooray!

 

Edit: After playing with PowerStrip a little more, it seems that adjusting the 3D clock speed causes it to stick sometimes (or at least the programs reporting the speed are showing it that way.) For example, I changed the clock speed to 417MHz and ran FurMark. I stopped FurMark, but after several minutes AIDA64 is still showing the clock speed at 417MHz, where it normally should have returned to 100MHz by now (nothing else is running.) Based on the temps, it looks like the GPU is actually getting stuck in 3D/Performance mode and not returning to the low clock 2D mode. I am fairly certain that PowerStrip is in fact changing the clock speed (and not just reporting that it changed it.) I've verified clock speeds in Nvidia System Tools, GPU-Z (although it caused drivers to crash constantly), and MSI Afterburner. I'll probably edit this again, but wanted to add to what I wrote earlier. If it does stick at the 3D clock speed, restarting the computer resolves it (because I don't have PowerStrip set to run at startup.) PowerStrip also causes GPU-Z and Nvidia System tools to show the GPU memory at 1MHz (although the default shown is 0MHz so I'm guessing this is just a bug since it is such an old GPU.) Also, I am getting about 1FPS on FurMark. Does that sound right?

 

Edit 2: I just saw your last edit. Good job on finding the VBIOS. Now the tricky part is editing it, and not setting it so low that the GPU doesn't work. Not to mention that each GPU is unique and may act differently even when at the same voltage. After years though, it finally feels like some progress is being made on figuring this thing out.

 

Edit 3: Ok, another edit. It appears that just having FurMark open (not running the burn-in test) is causing the GPU to jump to 3D speeds. With FurMark closed, the GPU clock returns to 100MHz even with PowerStrip opened and the 3D clock speed changed. I know I should have tested all of this before I posted, but it's more fun this way. As of right now, PowerStrip appears to be working correctly. Whether this lower clock speed translates to improved temps is a question I'll answer in another edit. Right now I have the GPU 3D clock at 401MHz and I'm getting CPU temps of 41 and GPU temp of 67 at idle, without the fan on full speed and without a cooling pad. Under load, CPU temps flatten out around 84 (AIDA64 stability test) and I'm not sure about GPU because I shut FurMark down once it hits 75. If I turn the fan on full speed, the GPU flattens out at 74 while running FurMark. I'm comfortable with these temps, and a cooling pad will only improve them. I might just start using this thing for purposes other than trying to get it to run cooler!

Edited by milone

Share this post


Link to post
Share on other sites
zipper

The GPU temp seems to depend more on voltage than clock - my 260M  runs 65 C at 0.85 V/500 MHz and 70 C at 0.95V/250 MHz. And the highest values hover about 94 C when pushed - Furmark will go even higher.

Share this post


Link to post
Share on other sites
milone

Yeah, I realize voltage is going to have a bigger impact. But up until recently we could not underclock or undervolt. You could say that at least underclocking is a backup plan now in case we cannot get the VBIOS modified to lower voltages. It might not make a big improvement, but this project has turned into finding as many small improvements as possible.

 

On a side note, last night was the first time in a couple years I've used the tx1000 normally, without spending the whole time trying to cool it down. I switched it to tablet mode and played around with Windows 8. I did turn the fan to full speed a couple times, but not as much as I expected. Without a cooling pad, the GPU stayed around 65 on idle. (Still waiting for cooling pad delivery.) Progress!

 

Also if anyone knows of a stable way to extract/backup Phoenix BIOS please let me know. WinPhlash 64 did not work and after trying it, I do not feel comfortable using it for anything, let alone anything to do with BIOS. I emailed the business I bought the motherboard off of ebay from asking to be sent the F.21 BIOS that I'm guessing they modified, so we can maybe compare that to F.20 and see what the differences are... if they send me it.

 

Edit: aapenaa, check out RW Everything and see if you can make sense of it. I just downloaded it a couple minutes ago, but looks like it could be helpful? (Though I admittedly have little idea what I'm doing.) I was also told to check out Phoenix Tool to backup the BIOS, but I do not see any backup options (unless a backup is automatically created before flashing, which I am not interested in doing.) I'm not sure if RW Everything is capable of creating a backup BIOS... will continue playing around with it.

 

Edit 2: If we can extract the VBIOS using GPU-Z, and if the VBIOS is complete inside the system BIOS (I'm not sure of this), we could use NiBiToR to change the voltages and create a new VBIOS... then replace the VBIOS inside the system BIOS with this new one (with a hex editor) and then flash the new BIOS? Does this seem possible? The only thing I'd be scared of is not being able to test the voltages before flashing. Even lowering them the smallest amount may cause the GPU to not work.

Edited by milone

Share this post


Link to post
Share on other sites
milone

The VBIOS is inside the system BIOS, as you pointed out. I've extracted it and opened it in NiBiTor. So now I have four different versions of VBIOS (I've uploaded them all):

  1. The VBIOS extracted from the F.20 system BIOS
  2. My VBIOS extracted using GPU-Z (from BIOS F.21 which came on the motherboard I purchased on ebay)
  3. VBIOS where I decreased the voltages in (smallest increment possible) based on VBIOS #1 in this list
  4. VBIOS where I decreased the voltages in (smallest increment possible) based on VBIOS #2 in this list

My F.21 VBIOS did have some differences between it and the VBIOS extracted from the F.20 system BIOS. I'm not sure what they are. When I opened them in NiBiTor I don't see any differences with voltages between the two. 

 

So I could take either of these modified VBIOS and basically paste them back into the F.20 BIOS over the old VBIOS. Then saving the system BIOS would give me a new system BIOS with the new modified GPU voltages. Since I don't have the system BIOS F.21 to work with at the moment, I did this with the F.20 BIOS and the #3 VBIOS from above.

 

Here are the changes I made while modifying the VBIOS in NiBiTor:

  • Change 3D voltage from 1.2V to 1.15V
  • Change Throttle voltage from 1.1V to 1.05V
  • Change Core Threshold - Critical from 130°C to 110°C
  • Change Core Threshold - Throttling from 140°C to 120°C
  • Change Core Threshold - Fan Boost from 90°C to 75°C

I'm not sure if the last three things I changed will have any effect, but I figured it was worth trying.

 

 

 

If you're interested, you can download the original BIOS I used here (from HP's website, BIOS F.20)

 

and the final modified BIOS here. (<-- so flashing this BIOS should apply the modified voltages)

 

Keep in mind that we have different models, and without knowing your exact model number I cannot guarantee that we use the same exact BIOS. The model I have is the tx1220us and that is what these files are based on. If you do use a different system BIOS, you could follow the same steps to modify it.

 

Obviously, flash at your own risk. I have a backup BIOS on my old motherboard but I don't think I have the tools to remove it and solder it to my motherboard should something go wrong... so I might have to build up some courage before I try flashing. Again, I can't guarantee the GPU will even work at these voltages, but the 2D voltage was NOT modified so hopefully that would allow the GPU to startup at least and allow you to perform another flash if it crashes during 3D mode. I'm not sure if this is how it works, but I would think booting to DOS would not require the GPU at full speed (unless it runs in full speed by default and only down-volts inside Windows?) If these voltages work, we could try lowering them even more. 

Edited by milone

Share this post


Link to post
Share on other sites
aapenaa

milone, i think we've got kind of the same line of thought, here's my update from the last days which I haven't had enough time to share but, here I go.

 

right after posting my finding in the hex editor I went ahead and modified with NiBitor a previously extracted VBIOS  I've got off of GPU-Z, and saved it what I basically did was lower the clocks and lower the voltages, and then compare both the 'original' VBIOS extracted with GPU-Z and the modified one on the hex editor, I found differences in single HEX digits which I show here:difference.jpg

 

Obviously I got very excited and went ahead and opened my system Main BIOS (Hp pavilion Tx2000 BIOS ver F.8) in the hex editor, I then located the original string and changed the modified digits that came off out of the NiBitor Editor and decided to flash, what I edited in Nibitor was:

 

Lower the GPU frecuency to 350 in 3D, 300 in throttle and 2D was left default, and changed to 1,1V in 3D, 1,0 in throttle and 1,0 2d.

 

The flash did went well but at the boot up happened what I fear, my computer started and emitted a long beep, followed by two short beeps, that's a GPU failure according to phoenix beep-code, and the screen remained black (not even the HP logo showed up), surprisingly the computer showed HDD activity so I assumed It booted to windows, and confirmed it because I could see it online in my home network after a few seconds (boooting I presume). After that I prepared an 'autoflash' usb thumb drive with the unmodified BIOS, and, blinded as I was started the computer to boot up from usb, after a few seconds the flashing procedure ended (I knew because I intentionally set the computer to beep while flashing), and the Hp logo came up (phew!!!) I thought I've killed my board !!!! but got lucky in the end.

 

After these experience I went back to windows and modified The VBIOS again but this time changed fewer things (only frequencies), and I decided to flash again (As I got an emergency-recovery-thumb-drive) in case something went wrong, aaaaaannndd... it did!, by changing only the default frequencies the GPU did not came up at all... recovered again.

 

this was very frustrating but yielded some insights:

 

First: I still have to try if letting the default clocks and lowering the voltages will help, but haven't have enough time to do it but surely I will.

 

Second: Somehow the digits I've altered in the HEX editor have nothing to do with the clocks or voltages on the GPU and are maybe encoded instructions to the BIOS chip but I don't know it's a possibility a VERY UNLIKELY one but this has to be considered.

 

Third: And this is the most likely, The GPU voltages values in this pc are controlled physically by an IC, according to the Schematics you've attached previosly it is the RT8204 (datasheet here: http://datasheet-found.com/Datasheet.jsp?pn=RT8204), in the schematics it's like somehow the output of the RT8204 is feeding many inputs of the C51M chipset, I'm not an electronics expert but I've read somewhere that someone changed the GPU voltage by changing a resistor somewhere in the mainboard so maybe this can be the case.

 

As of right now the results on BIOS-editing Mod, are not promising much, mainly because I can't understand anything of how it works but as I work on it, it throws some light (It needs more testing), but i think we really need help of someone more experienced, the same applies for the mainboard mod, but right now for me this is more a challenge than a need, I want to fix this issue, the Hp Engineers couldn't, and see how hard it really was, we can call it a 'Research project' right now, and as one, it will take a little time.

 

I'm sorry for my bad English but I'm not a native speaker and trying to explain ALL of what I've done is a little difficult n a foreign language. Hope you all uderstood it well, I'll keep on working on this, Hope you do so.

Share this post


Link to post
Share on other sites
milone

Can you tell me how you setup the "autoflash" USB drive? I know I would need to set USB Hard Drive to first in the boot order (or at least before hard disk drive), but am wondering what arguments you used for WinPhlash (or whatever flashing utility you used). Did you run it from autoexec.bat? I already have a bootable DOS USB drive that I booted to when I tried using nvflash. I just want to make sure I have a backup plan ready to go if I try it. I'm actually getting pretty good temps right now, but would consider flashing just to see if it will work.

 

I was able to downclock my GPU to 400MHz (on 3D mode) using PowerStrip and it ran ok. FurMark didn't cause any crashes. So I think there may be some underclocking room at least for 3D mode. I didn't try downclocking 2D mode below 100MHz. Obviously undervolting would be preferred to underclocking, so if we can undervolt, underclock may not be necessary.

 

And as far as undervolting by hardware... yeah that's something beyond my knowledge. I had read a post by someone more knowledgeable than I am who had done this to their GPU (not a 6150) by jumping specific pins. I cannot find the post at the moment but I'll look for it again tomorrow.

 

I got this cooling pad and attached it using black double-sided adhesive. Running FurMark, my GPU temp never goes above 75 even at regular clock speeds and without having to turn the fan on full speed. So I am satisfied with it as it stands now, but am curious if undervolting is possible and I think I would ultimately try it if I'm sure I can do a blind flash to recover in case things go wrong.

Share this post


Link to post
Share on other sites
aapenaa

wow, that's a nice cooling pad, I think that n the end if the undervolting thing doesn't work our last resort would be hardware mods, also you're getting pretty good temperatures right now, I'm glad to hear that.

 

about the autoflash thumb drive, you have to make a bootable usb thumb drive with the least drivers you can so I would recommend using MINI-DOS, I've made it using Hp USB Disk Storage Format Tool, the latest version of PHLASH16 you can find (I've attached the one I've used) and your BIOS.WPH file, and make an autoexec.bat with these PHLASH16.exe /X /V /C NAME_OF_YOUR_BIOS_FILE_HERE.WPH when phlash16 starts to flash, you will hear a lot of beeping from your comp, so you can know if it is working.

 

this is usable in the case where a previous flash with bad gpu clocks don't allow the GPU to start and you're blind. you will know if this is the case because your computer will POST with a long beep followed by two short ones, what I would suggest to do in order to recover your pc from that flash is to remove your cdrom drive, start the pc press f9 right after the computer has beeped and then press the down arrow and then enter, this will boot from usb and start the automatic flash, obviously you must've edited the autoexec.bat in other computer to flash with the original unmodified BIOS.

 

If by some reason your flash procedure failed in other way (power failure, etc) and your PC can't even POST, you will need a Phoenix Crisis disk, and a guide on The phoenix CRISIS recovery procedure, or another BIOS chip, depending on the situation.

 

I haven't tried to flash with the modified voltages only but I will try soon, so I'll keep you updated on that.

Share this post


Link to post
Share on other sites
aapenaa

I forgot to attach the phlash16 ... HAHA!, here it is:

 

PHLASH16.rar

 

UPDATE: It seems like removing the black paint from the heatsink helps reducing temperature

 

links: 

 

http://forum.tabletpcreview.com/hewlett-packard/8875-modding-tx1000-part-1-changing-thermal-compound.html

 

http://baminc.co.uk/blog/2011/05/hptx1000/

 

http://www.crazyboards.org/forums/index.php/topic/34064-oh-why-do-they-paint-the-heatsink/

Edited by aapenaa

Share this post


Link to post
Share on other sites
q10
Hi!
Sorry for my bad English.
The most effective - is to remove the cover that is over the keyboard and close the heatsink. I have the temperature dropped to 9 degrees (70 to 61) at idle.
I made a hole above the heatsink and closed the net. In this embodiment, the temperature dropped by 6 degrees.
PS
You can rotate the screen in Win8?
If yes, please give a link to the drivers for win8, or send by mel-sha@bk.ru

Thanks!

Share this post


Link to post
Share on other sites
cdoublejj

Hi Everyone!

 

As the title explains I'm starting this post to find help in flashing an old geforce 6150 (nforce 430) IGP, this card is very common in old hp laptops (Pavilion Tx1000, Tx2000 and some Dv's) and it is well known that such models commonly fail by overheating.

 

I own one Hp Tx2000, and I managed to bring it back to life by the old heatgun trick, after that, I've been very careful with the temps and managed to lower them significantly, but I'm still afraid that the GPU will fail eeventually, I've been looking for undervolting software compatible with the card but had no luck so I though I could then flash it to UNDERvolt or UNDERclock the factory settings to lower the temps but haven't found any solid information of the procedure on this card.

 

I've already looked at nvflash compatibility list an it says it is compatible with Geforce 6 series, also I've managed to exctract the GPU bios using GPU-Z and I can open it using NiBitor, So I think I can modify the clock and voltage values.

 

the questions is, will it be possible to flash the card using the modified bios, or by any means will I end up (again) with a useless tx2000? any one has done it before, not necessarily on the go 6150, but any other onboard GPU?, did it work?

 

I would like to hear any suggestions or advices, and people intested in help are very welcome.

 

Thanks In advance.

 

 

I do this quite a bit at work for customers, grant it, i do not use a heat gun. even if you under volt it it will still probably fail unless you fix the cooling, you need to replace the thermal pad with a smooth lapped copper shim with arctic silver 5. that's what i did with my very own tx2000 the temps went down too. it's really the only way to keep a re flowed laptop nvidia card from failing again.

 

how ever you do have me thinking about overclocking mine.

Edited by cdoublejj

Share this post


Link to post
Share on other sites
arawsg

hi guys got a tx2000 with tl-68 as processor I too have resorted to the heat-gun temporary solution and made some mods on my heatsink similar to milone just not using copper but aluminum instead.

 

temps on stock  (no under-clocking) no mods just diy reflow. using windows 7 on balanced power management setting max processor state at 100% 800mhz - 2400ghz.

cpu  on idle around 50c

gpu  on idle round 65c

room temp 25 - 29 c

on load cpu goes about 68 - 70+ c

gpu  75 - 85+ c  can't recall

 

after a certain amount of time I had to do the reflow again.

after modding I have managed the tempertures to go even cooler  using

 

 Linux puppy:

ive under clocked mine using a scaling tool built in using Linux puppy 5.7.1 clocked at 800mhz.

 idle temps 39 - 40 c for cpu  on load never goes above 50c on 800mhz. on 1600ghz just above 55c

no monitor for gpu. if I set it any higher the tx2000 shuts down (sometimes needs to be reflowed again)

 

using windows 7 I've found that using power options I could set the gpu to maximum savings lowering  its clock to 99 MHz when idle

and around 333mhz on medium load then 425mhz on heavy load. but will quickly go back to 99mhz.

 

 for the cpu I used processor power management setting, putting maximum processor state to a certain % would not allow the cpu to reach its maximum frequency  (2.4ghz)

setting it at 65 or lower %, will  make the cpu stay at 800mhz even in heavy tasks maintaining a lower temp and effectively lowering gpu temp too. increase the % to attain 1800ghz effectively using the built in k8 amd scaling function within the cpu.

 

temps with win7 power management settings set to low power 65%-85% max processor state

idle 39-40c cpu

idle 50-55c gpu

 

on load 45-55c cpu temp on 800mhz - 1800 ghz setting (around 65%-85 % on max processor state)

on load cpu temp +10-15c = gpu temp

 

I am also looking for other solutions that can solve this problem  and not to worry about temperatures, totally killing the gpu.

I hope this helps

 

I am about to do the  wire cutting solution on the fan or make a switch as suggested.

Edited by arawsg

Share this post


Link to post
Share on other sites
TopHatProductions115
On 2/20/2013 at 6:07 PM, milone said:

The VBIOS is inside the system BIOS, as you pointed out. I've extracted it and opened it in NiBiTor. So now I have four different versions of VBIOS (I've uploaded them all):

  1. The VBIOS extracted from the F.20 system BIOS
  2. My VBIOS extracted using GPU-Z (from BIOS F.21 which came on the motherboard I purchased on ebay)
  3. VBIOS where I decreased the voltages in (smallest increment possible) based on VBIOS #1 in this list
  4. VBIOS where I decreased the voltages in (smallest increment possible) based on VBIOS #2 in this list

My F.21 VBIOS did have some differences between it and the VBIOS extracted from the F.20 system BIOS. I'm not sure what they are. When I opened them in NiBiTor I don't see any differences with voltages between the two. 

 

So I could take either of these modified VBIOS and basically paste them back into the F.20 BIOS over the old VBIOS. Then saving the system BIOS would give me a new system BIOS with the new modified GPU voltages. Since I don't have the system BIOS F.21 to work with at the moment, I did this with the F.20 BIOS and the #3 VBIOS from above.

 

Here are the changes I made while modifying the VBIOS in NiBiTor:

  • Change 3D voltage from 1.2V to 1.15V
  • Change Throttle voltage from 1.1V to 1.05V
  • Change Core Threshold - Critical from 130°C to 110°C
  • Change Core Threshold - Throttling from 140°C to 120°C
  • Change Core Threshold - Fan Boost from 90°C to 75°C

I'm not sure if the last three things I changed will have any effect, but I figured it was worth trying.

 

 

 

If you're interested, you can download the original BIOS I used here (from HP's website, BIOS F.20)

 

and the final modified BIOS here. (<-- so flashing this BIOS should apply the modified voltages)

 

Keep in mind that we have different models, and without knowing your exact model number I cannot guarantee that we use the same exact BIOS. The model I have is the tx1220us and that is what these files are based on. If you do use a different system BIOS, you could follow the same steps to modify it.

 

Obviously, flash at your own risk. I have a backup BIOS on my old motherboard but I don't think I have the tools to remove it and solder it to my motherboard should something go wrong... so I might have to build up some courage before I try flashing. Again, I can't guarantee the GPU will even work at these voltages, but the 2D voltage was NOT modified so hopefully that would allow the GPU to startup at least and allow you to perform another flash if it crashes during 3D mode. I'm not sure if this is how it works, but I would think booting to DOS would not require the GPU at full speed (unless it runs in full speed by default and only down-volts inside Windows?) If these voltages work, we could try lowering them even more. 

I happen to have a similar issue with a tx1305us running WIndows 10 64-bit. In order for any of the fixes listed here to work, I had to change some system settings (bcdedit) to allow for regular loading of unsigned drivers (undervolting). Still haven't gotten GPU-Z to work flawlessly. It screws up my display output when it first opens, forcing me to turn the display off and back on again. I'll try to post more details when I can.

Share this post


Link to post
Share on other sites
TopHatProductions115

Okay, I was doing some more research online when I came across some interesting sources. I found these two links recently:
 - https://www.techpowerup.com/vgabios/62519/hp-6150go-128-061207
 - https://www.techpowerup.com/vgabios/31614/31614

It appears as though many HP laptops with this GPU had similar voltages set in VGA BIOS. I haven't been able to find any official specs on the supported voltages, though. It appears as though one would possibly have to start meddling with PCB resistor signals/outputs to change the voltages. 

If it is hardware-based controls (an IC, like the 'RT8204' that aapenaa mentioned from the spec), that would make things difficult on the GPU undervolting end. I think this would require us to get the laptop'(s) hardware diagrams/PCB layout. However, I am out-of-practice when it comes to this sort of thing, so that rules out a few more options in my case.

I can only see one last option outside of simply maiming this netbook - liquid cooling.
 - https://www.reddit.com/r/buildapc/comments/1k97dq/build_complete_success_insane_custom_laptop/

 - https://i.imgur.com/jCz1k4C.jpg
 - https://i.imgur.com/ApTgXqI.jpg
 - https://i.imgur.com/QApvNqO.jpg
 - https://i.imgur.com/MPwaTnK.jpg

It may seem crazy, but what other options (besides putting more clock-speed leashes on the CPU/GPU) do I have?
 - https://www.google.com/search?q=laptop+water+block&newwindow=1

Guess it's time to go shopping (again) - another project begins...

Share this post


Link to post
Share on other sites
TopHatProductions115

I've started the effort of sourcing the required parts. Part(s) selection is as follows:

Water Pump(s)
 - * https://www.amazon.com/Cytec-Cooling-Brushless-Computer-computer/dp/B018E6ZWIQ/
 - * https://www.amazon.com/dp/B0713T9PRP

Reservoir(s)
 - * http://www.ebay.com/itm/Bykski-150mm-Reservoir-Tank-Water-Cooling-G1-4-Thread-50mm-x-150mm/172650911852 

Water Pump/Reservoir
 - http://www.ebay.com/itm/500L-H-10W-DC-12V-Water-Cooling-DC-Pump-Tank-250ML-F-pc-CPU-CO2-Water-Cooled/111241824048
 - http://www.ebay.com/itm/Hot-140mm-Cylinder-Water-tank-SC600-Pump-Computer-Water-Cooling-Radiator-600L-H/361465906220
 - http://www.ebay.com/itm/New-60mm-Cylinder-Water-Tank-SC600-Pump-Computer-Water-Cooling-Radiator-Set/122037372139 

Heat-Exchange Water Radiator
 - http://www.ebay.com/itm/CPU-Aluminum-18-Pipes-Water-Cooling-System-Heat-Exchanger-Radiator-120mm-Black/322384645567
 - http://www.ebay.com/itm/Computer-CPU-Aluminum-18-Pipes-Water-Cooling-Heat-Exchanger-Radiator-120mm-Black/311776535205 

Water Block(s)
 - http://www.ebay.com/itm/Water-cooling-copper-block-FOR-graphics-GPU-endothermic-head-CPU-Computer-/162423052110
 - http://www.ebay.com/itm/Copper-Water-Cooling-Block-For-Graphics-GPU-Endothermic-Head-CPU-PC-/131924520298
 - * http://www.ebay.com/itm/Oxidation-blue-Computer-CPU-semiconductor-Cooling-Block-T1-/252603791521 

Liquid Transfer Tubing
 - * http://www.ebay.com/itm/3-Ft-of-Food-Grade-Home-Brew-Beer-Wine-Making-Water-Line-Clear-Tubing-1-Metre-/181912754725 
 - http://www.ebay.com/itm/2M-Water-Cooling-PVC-Computer-Water-Cooling-Tube-8mm-ID-x-12mm-OD-Turbing-Pipe-/142384629230

* Not likely to be in the finalized parts list and/or physical build

Might actually post this project to YouTube if it goes as planned...

EDIT :: Still purchasing parts...

Edited by TopHatProductions115
updated status of fix attempt...

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...

×
×
  • Create New...