6970m and original M17x (a.k.a. R1)

[imsolidstate]

Hello everyone, I'm new to the site. I ended up here looking for 6970m VBIOS and can't download due to the 5 post restriction.

I have a Dell/Alienware M17x, the first one with Socket P and Nvidia chipset. I thought it was an R2 until I tried flashing an R2 BIOS and bricked it. I had to pull the flash memory from the motherboard and re-program the BIOS/EC code to get it working again. Anyway, I had bought a 6970m to go into it until I figured out I had an R1.

The laptop doesn't even post once the 6970m is installed. When you plug in the power adapter the alienhead lights for about half a second and then goes off. If you press the power button the lights and fans come on but the screen stays black. No video on the VGA connector. I'm going to check power to the GPU at the LDO to make sure there isn't some power issue. It has a power good output that indicates if the voltage is stable and within limits. The alienhead flashing off so quickly makes me think there might be something with the power supply to the card.

I bought the card off eBay. It's a Dell card, supposedly flashed with an undervolted vBIOS. That's all I know at this point.

The flash memory that contains the vBIOS is a PM25LD010 SPI flash in an SO-8 package. When I get a few vBIOSes to try, I will pull the chip from the board, dump it's contents, and then verify whether or not the chip is in-circuit programmable. That will simplify trying different BIOSes.

In my research figuring out how to re-flash the BIOS/EC, I became pretty familiar with the EC (IT8512E). I have yet to get the datasheet for the chipset and try to start figuring out where the connection is to the video card and how it is recognized by the system. I don't think the EC is involved here, but it controls the fan PWM. I suspect the Nvidia chipset is the issue.

I want to get this working on my R1 because the GTX260m that's in it sucks. I guess I will post here as I find out more information, until I get to my five posts and can try some different vBIOSes and try to figure out what is preventing the system from working. I also have yet to get a datasheet for the 6970m to have an idea what the BIOS code represents. I registered with AMD but have yet to hear back from them.

I've done a pretty extensive web search but can't find anything that isn't about the R2. If anyone has some information I would be glad to read it. If I figure this out I will post an article about how to do it on my website. Thanks to everyone in advance.

[svl7]

Hi, and welcome to T|I!

The R1 works with the 5870m... but I never heard about anyone trying a 6970m in it. I assume you're already on the latest BIOS, right?

Is your card an OEM version or ES?

In-circuit programming of the 6970m works, did this several times. If you somehow manage to get a datasheet of the 6900m series, let me know, would be really interested in seeing this. Regarding the M17x R1 schematics... got something which might interest you . Though I think you can't download it at the moment...

Alienware M17x R1 schematics.pdf

[Michael]

Subscribing. Very interested to see further results.

Good luck!

P.S @imsolidstate you have full permissions now. Welcome to TI :-)

----

Asylumized!

[imsolidstate]

@svl7 and Michael: thank you for the warm welcome! Those schematics will be incredibly useful, chasing traces with all of the buried vias on motherboards is impossible. Thanks a lot.

I had a chance to mess with it this afternoon, and I have some clarifications. First, the alienhead flashing briefly was because I forgot I had the battery in. With the battery out it dims on and off like normal.

Second, I measured the voltages at the LDO regulator for the GPU core. With the machine powered on there is no VIN, so the processor isn't powered up at all. However there is voltage at VCNTL which means the card is asking for power. There is full voltage available (19VDC) at the large power contacts on the edge of the MXM connector. Not sure about the other power sources yet (5V, 3V3).

There are other pins in the MXM 3.0 connector that are reserved for OEMs to interface with the motherboard. Thanks to svl7 I can check some of those signals and see if maybe power is being denied to the card because it doesn't have some signal pin that the OEM (Dell) defined to lock out this card. That would be convenient but I am skeptical it will be that easy.

To do: Measure what voltages are present at the card connector and verify against the schematic. Then make an educated guess about where power might be missing.

To answer svl7's questions the MB BIOS is A07, which is the latest I found at Dell's website. I don't know if the card is an ES, but it seems unlikely since it has a Dell sticker with the P/N 06W46K. There is another sticker with what I assume is an AMD P/N: 102C2961300 000001.

Thanks for the help.

[svl7]

Always glad to provide some help for a cool project

I don't know what's the issue with the card... it would be really good if you could test the card in a system which is known to work with the 6970m, if the card is dead all the trouble isn't worth it.

Also reseating the card can really help, I experienced it several times that a card didn't POST after I put it in a system, but reseating it solved it.

[imsolidstate]

I'd love to be able to test it and make sure I'm not wasting my time, but I don't really have the capability. I don't know anyone with a platform I could test it on.

I messed with it again today. Here is the results:

Verified high voltage supply to the card (19V on adapter, 12V on battery)

Verified 3.3V_RUN.

Verified 5V_RUN.

Measured ACAV_IN_MXM at 1.5V with the adapter, .3V without.

Verified MXM_PRESENT# signal to MCP is low.

Verified MXM1_PWR_EN from MCP. (3.3V)

PWR_GOOD / MXM1_PWROK from the card is incorrect. It's low. However I know it is getting power.

Observations:

Fans spin for about a second at power up with PCIE 2.0 enabled, then shut down. Fans run continuously on high if PCIE 2.0 is disabled.

I figured out the MB will run only on battery with the keyboard disconnected. The AC adapter input is disabled as long as the keyboard is disconnected.

Todo: Try a few different VBIOSes and see if I can get different behavior. I don't really know how the VBIOS affects the GPU since I can't get hold of a datasheet. I might try a completely wrong BIOS but one that is supposed to work with the system (HD5870) and see what happens.

Many thanks again for the datasheet, I wouldn't be getting nearly as far without it.

[svl7]

You could test it in any recent Clevo (and reseller) system, AW, or even some Acer and Samsung notebooks... but true, might not be easy to find someone who's willing to let you do this.

I can't remember all the specs, but if you got PWR_EN asserted but you don't get PWR_GOOD then it's possible that some of the internal power regulators aren't within the required tolerance.

What I can guarantee you is that your card won't work with a 5870m vbios.

Also you have to assume that either your system isn't compatible with this card due to some BIOS restrictions or that your card is defective. It's highly unlikely that this is only caused by the vbios. In order to prevent the system from booting you need either to use the completely wrong vbios or totally mess it up while modifying it.

If you dump your vbios you can upload it here, then I can take a look at it.

[imsolidstate]

PWR_EN without PWR_GOOD and input power correct does have me worried. I am hoping that PWR_GOOD is a signal output from the GPU, and not from the on-board regulators.

Yeah, I know it wouldn't work right with a 5870m vBIOS but I am curious what it would do, even if it is nothing. You can't really disassemble vBIOS code without a datasheet so it's just trial and error I guess.

I'm not understanding where the BIOS would be placing a restriction on the cards used in the system. The BIOS runs on the EC, and the EC has an SMBUS connection to the MXM slot and an LPC interface to the MCP. Are you suggesting that the BIOS/EC code communicates with the card through SMBUS or through the MCP/PCIE to determine if it's authorized?

I dumped the vBIOS but I get an invalid file error when I try to attach it here. I will try to upload it to my website. (Nevermind, it worked as a zip file.)

6970_as_received.zip

Comparing the dumped vBIOS and the Clevo vBIOS from the R2 tutorial here reveals quite a few differences. I've attached a screenshot of the first few hundred bytes that show the version info. There are more differences farther in the file. What is interesting is that the part number for the dumped vBIOS sort of matches the number silk screened on my board: 109-C29647-00.

I will try programming the card's flash with the Clevo vBIOS when I finish modifying my programmer code for the 1Mb flash. It's currently set up for 16Mb and I have only modified the read code so far. Maybe tomorrow.

[imsolidstate]

Got the code working for programming the Pm25LD010, the flash on my 6970m. Successfully programmed with the Clevo BIOS I got here. Will test it later.

I posted the programmer source code files to my website if anyone is interested in ISP flash programming a 6970m.

6970m vBIOS ISP / Pm25LD010 » imsolidstate

And the programmer:

SPI programmer for SST25VF016B » imsolidstate

[naldor]

Interesting stuff, sometimes makes me regret I decided to go for econometrics , might pick a minor electronics engineer, since I still need to do a minor

[mikecacho]

Very cool, as i own an R1 as well as R4, but would love to breath new life into the R1 again with the 7970M. Ill keep an eye on this thread to see if you can get it to run on the R1 (is CF in the planned future once you get the single card working).

[imsolidstate]

Clevo VBIOS has the same effect.

I was testing a system BIOS that I modified, and it looks like I messed up an offset because the machine isn't posting. The problem is that even after flashing a correct BIOS image to the flash, the machine still isn't posting. I know the EC is reading the flash, because if I remove the flash the computer is unresponsive. When I replace the flash it starts to boot, spins the fans, and turns on the lights.

The power button works and the LED alien head light dimming still works. The keyboard has no response, and no beeps. There is no video. Power is good.

I have a suspicion that the MCP loaded some code from the modified BIOS image that is causing it to be unresponsive. It seems like the EC is going fine until it tries to bring up the video. I am guessing that the BIOS code is hanging waiting for the MCP to respond. The power button is unresponsive when trying to shut down the machine. Clearing the CMOS seems to have no effect.

Does anyone know how to recover or reset the MCP, or how to get into the manufacturing mode? When I changed out the flash chip the first time I corrupted the BIOS it automatically entered manufacturing mode and reprogrammed some stuff. I can't figure out how to get back into it, but it might be farther in the boot code or something, after the video initialization. I guess I'll have to look for it in the BIOS.

I don't think it matters but the only thing I did different this time was in-system programming the system BIOS flash.

[svl7]

Does it shut down again after powering on or do the lights stay on? Usually the AW systems indicate POST issues with flashing Caps/Scroll/Num.

If you just get a black screen it might be a vbios / bios issue, it's possible that you simply don't get any picture.

I assume you dumped the whole BIOS flash memory before starting to modify it, right? So as long as you completely flash your dump back you should be at the same state as before you started modifying. I'm not familiar with the R1 and can't remember all the details, does it have a separate flash memory for the EC?

[imsolidstate]

The lights stay on. The power button is unresponsive, you have to pull power to shut it down. That's why I suspect the EC is sitting waiting for something, with interrupts disabled. The Caps/Scroll/Num lights are always off.

If there was just a no picture issue, I should still get a beep from F2 to go into the BIOS settings, or a beep if I held down a key for too long. FN+B/FN+F should also do something. I tried getting into the crisis recovery mode but nothing happens. Also the power button would shut the computer down. And I don't have a card in at the moment, I'm trying to run it from the integrated graphics controller; however my GTX260m doesn't work either, even though it powers up and asserts a power good signal.

I flashed the original BIOS back into the flash chip that's dedicated to the EC. That's why I don't understand what's going on. I programmed and verified the image, it's A07 from Dell's website.

I'm worried that I possibly messed something up by in-system programming the EC's flash. I don't think that it would cause any issues, but after flashing back the original A07 which was working previously it won't post. Every power rail I've checked is good though.

I don't know how the integrated NB/GPU loads and stores code, so I also have to assume that it loaded an invalid segment of BIOS code from my modification that is rendering it unresponsive to the EC. The EC loads from the flash every reset, I don't know if the MCP does that as well, or just when it's commanded to reload. When I got into the manufacturer mode the very first time I changed the flash chip and loaded A07 it took about a minute to program various things. I'm wondering if only a small portion of the BIOS code is actually used during startup and the rest is in flash memory on all of the distributed chips. The EC may not load the entire code at boot, just enough to post. It happens awful fast for 16Mb even at 80MHz. So I'm afraid the MCP needs to reload it's flash but I can't tell it to since it's flash is potentially corrupted and it doesn't know what to do. I have to assume that the MCP has some kind of on-board configuration memory. I thought the CMOS clear signal for the MCP would do it, but nothing changed.

[imsolidstate]

Also figured out that on battery only the alien head is a different color, and the EC responds to a held-down power button by shutting down.

[svl7]

Hmm, I see. But let me ask again, did you take a backup of the whole BIOS chip or did you simply flash a downloaded BIOS afterwards?

[imsolidstate]

I did not back up the flash chip. I corrupted the flash a few weeks ago because I thought it was an R2 and I flashed an R2 BIOS without verifying the motherboard. When I tore down the computer to get at the flash I ran the motherboard part number and figured out it was an R1. Since the flash was already corrupt I just replaced the flash chip with a new part and programmed it with the A07 BIOS I downloaded from Dell. It has had that BIOS for the last few weeks. I modified that BIOS and flashed it the other day, and started whatever is going on. Once I couldn't get it booting I reflashed the chip with the A07 BIOS that I programmed it with a few weeks ago.

[svl7]

Thing is, the BIOS image from the Dell site isn't necessarily complete. E.g. it doesn't contain the service tag, which is stored in the same flash memory if I'm not mistaken. There might be other code as well which is in your case now missing, I'm not sure.

[imsolidstate]

Well, it worked when I did it a few weeks ago. I flashed the A07 image and it worked fine. Only difference is it somehow got into a manufacturing mode and spent a minute programming stuff.

I tried flashing a different BIOS image and it does the same thing. If I disconnect the MXM fan I get a brief flash of scroll/caps/num on startup.

I doubt I'll be able to recover this motherboard at this point. I either messed up something by ISP, or the MCP needs to be reprogrammed manually, presumably through the "ITP" connector that looks like a factory thing.

I wish I hadn't changed two things at once. One of the things that could potentially be a problem is the EC's connection to the flash. I didn't think about it earlier, but programming the flash means powering up the same rail that powers the EC. That means the EC's GPIO pins that drive the flash were active. My buffer IC that programs the flash evidently overpowered the EC signals because it programmed and read properly. That means there is a chance that the EC's I/O pins were overdriven. If I can get my logic analyzer working I want to put it on the flash's SPI connections and see what the EC is doing with the flash.

I'm not sure that there is anything actually wrong with the SPI interface though because with the flash removed, the computer is totally dead. If I put the flash back in, it powers on the lights, flashes the alien head, even charges the battery. That means that the flash is likely getting read. I have no way of knowing if the EC is actually getting the data latched though.

Guess I might be looking at eBay for a new MB.

[svl7]

I see... that's a nasty situation.

Btw, if you disconnect a fan and try to boot it won't POST, thus the flashing LEDs.

[imsolidstate]

Ok I confirmed my suspicion that ISP is the problem. I did some more digging on the MCP and manually asserted CMOS clear, since pulling the RTC battery won't work, there's a regulator feeding the RTC_CELL rail to the MCP when the battery is absent. I also verified that the MCP is strapped to boot from the LPC interface, which means it boots from the EC's flash. So it doesn't look like it stores code.

I put an oscilloscope on the EC flash interface signals. They looked terrible, and the EC was stuck requesting the same block continuously. Seriously SCK and SI looked like a mountain range. SCK was kind of a 1VDC triangle wave with a 1VDC offset.

The EC eventually gave up and isn't trying to read the flash at all now. I'm going to order another ITE8512E to replace it. Hopefully there isn't any initial code in the EC that tells it to boot from the flash. If I remember right it looks to the flash on reset. And hopefully I can actually replace it, it's a very fine pitch package. I wish I still had my hot air rework station.

In summary: DO NOT IN-SYSTEM PROGRAM THE BIOS FLASH ON AN R1 MOTHERBOARD. Bad things happen.

[svl7]

You'll probably need a hot-air station, yes... it's definitely easier Let us know how it goes, good luck!

[imsolidstate]

I was successful at replacing the EC. Unfortunately there appears to be some initial configuration for the EC, because the activity on the flash SPI bus is not SPI. The EC can't communicate with the flash. I already tried contacting ITE but they don't give their tools to anyone but manufacturers. I wouldn't have the programming file anyway so I guess I'm stuck getting a new motherboard.

[imsolidstate]

So before I ordered a motherboard from eBay I called Dell to see if I could buy it from them. They quoted me almost $700.00 for a replacement motherboard. Words fail me. I paid that for the entire computer.

So who works on Alienware systems? How do people normally get their Alienwares fixed? And do they pay that kind of ridiculous price? I'm starting to think I should start an Alienware repair shop.

[mikecacho]

Solidstate, Dell always sells their items individually for full market price regardless of age. When i wanted to get a QX9300 (about 8-10 months ago) for my R1 dell quoted 1K... i laughed and searched ebay for a week or so for one, ended up winning the bid for a new OEM one for $265.00!