Jump to content
OGXbox.com

1.6 128MB RAM Upgrade Tutorial


Prehistoricman
 Share

Recommended Posts

Intro

Since the beginning of time (2004) it was known that the 1.6 was missing the unpopulated footprints for the extra 4 RAM chips and this made the upgrade impossible. I was thinking recently that if all the signals for the extra RAM chips are still available on the board, we can add RAM to the 1.6 by just connecting all the necessary signals. After the tutorial segment I'll go into why this mod works.

I'll also post a video on YouTube at some point.

Be warned: this is more difficult that a 1.0 - 1.5 RAM upgrade

Tutorial

Required items

  1. Everything mentioned in the 1.0 - 1.5 RAM upgrade tutorial: https://consolemods.org/wiki/Xbox:RAM_Upgrade
  2. 4x 90mm long mod-wire. I used enamelled wire from a transformer.
  3. ~8 hours spare time :)

Step 1: flash the modified XBlast OS BIOS

XBlast OS (as a BIOS) is super helpful in doing this install correctly. It will boot with any number of extra RAM chips and includes a test for the additional RAM. XBlast OS will not show the 128MB test option on a v1.6 Xbox, so I've gone ahead and rebuilt it from source with that check removed.

xblast_os_v0.60.bin

Alternative Download Link: https://mega.nz/file/eeAwTZKQ#aaFcbACj3htA6cEtKwoiy1L7uK93o4YwUSqkGBEsUz8

If you don't trust my build, this is the line of code that needs removing:

image.png.0eaf1f1105cde38ddb04b074031a6999.png

At this point, I suggest booting it up and running the test. The test is under Settings > Tools > 128MB RAM test

Step 2: Remove the motherboard from the Xbox

Step 3: Place the new RAM chip on top of an existing chip

You'll want to solder down the corner pins first to get a good mechanical bond and so that you can adjust the alignment. I do this by pushing down pins 30 and 31 and soldering them down to the chip below. Then I check the alignment of the rest of the pins. If it's bad, then the chip can be nudged into position while reflowing the two pins. Then solder the two pins opposite.

Step 4: Push down the rest of the pins

Except pin 28! That is the chip select pin and it is unique for each RAM chip. Leave this pin floating in the air and don't solder it yet. Pin 30 is marked on the PCB.

The pins come in a kind of ‾‾\__ shape. When you push them down, the whole pin will rotate around this end -> ‾‾\__ . I like to also push down the tail ‾‾\__ <- of the pin to get more contact with the chip below.

Step 5: Start soldering all the pins

Very little extra solder is required. Use lots of flux. I like the tacky flux that usually comes in syringes. I also like the drag-soldering method. If there's any excess solder, you will get shorts between pins and this extra solder can be wicked away.

The new pins won't naturally touch the pins below even after pushing them down.

Step 6: Check for shorts and missing connections

For the first pass, I like to go in with a microscope and nudge each pin (of the new chip) with a cold soldering iron tip. If the pin moves easily, it is not soldered. Also check for shorts at the same time.

If you have the patience (and want it to work first try without frying your Xbox), check for shorts and continuity with a multimeter. I skipped this the first time and got bitten! Best to clean the pins and pads with IPA before checking with the multimeter.

Step 7: Add the chip-select wire

First, figure out which bank you are adding:

image.thumb.jpeg.07ed705f3116b0852fca537c7cdbe72e.jpeg

Then, find the chip-select solder point:

image.thumb.jpeg.f72abd0eaefacea69d338936a1c13611.jpeg

Bank 1 CS is left of the label for C4P10

Bank 2 CS is between the labels for C4R12 and C4R13

Bank 3 CS is left and above the label for C4P21

Bank 4 CS is between and above the labels C4R1 and C3R1

For banks 2 and 3, the chip-select wire can go through the hole to the left of the GPU. For bank 2, 90mm may be too short, so check the wire length and cut it as short as possible. I removed the GPU heatsink to access this hole.

At this point, it should look something like this:

image.thumb.jpeg.b2414b8ebbb990602a77a02fd3bf2145.jpeg

Step 8: Reassemble and test in XBlast

For a quick test, I don't bother with adding the fan, GPU heatsink, HDD, or DVD drive.

The test in XBlast should look like what I first posted on reddit: 

 

If the Xbox reboots twice and FRAGs, check the bios selection (if you have one available). A normal BIOS will not accept between 1 and 3 extra RAM chips.

If the Xbox reboots three of four times (and it's faster than a normal FRAG sequence) check for shorts. This happened to me.

Step 9: Repeat for the other 3 chips

The first one is the hardest.

image.thumb.jpeg.dcd1211fb53289de521f425165ff0f08.jpeg

Step 10: Reflash with your favourite BIOS

Some 1.6 BIOSs won't support 128MB. The X3 BIOS works just fine and it's the only one I tested so far.

xblast_os_v0.60.bin

  • Like 14
  • Thanks 5
Link to comment
Share on other sites

The Theory

There are some awesome scans of the Xbox motherboard online for versions 1.1 and 1.6 (they have a sickmods.net watermark). I traced all the signals of a RAM chip and found these results:

  1. A0 - A11 address pins are connected to the same pins of the chip under it
  2. BA0, BA1 bank select pins are connected to the same pins of the chip under it
  3. Same for CK, ~CK, CKE, ~RAS, ~CAS, ~WE, VREF, MCL, and DQS
  4. However, the data pins DQ00 - DQ31 are in reverse order
  5. The data mask pins DM0 - DM3 are also in reverse order
  6. Since both the data pins and data mask pins are in reverse order, we can reverse both and achieve a compatible interface
  7. The chip select pin ~CS is unique for each RAM chip

After this, I started tracing the ~CS signals back to the northbridge/GPU on the 1.1 motherboard. Then, I found the same pads of the northbridge on the 1.6 motherboard. Luckily, all 4 of the CS pins are exposed on the bottom of the board!

At that point, I had to confirm that the 1.6 northbridge wasn't actually incompatible with the extra RAM. The chip is labelled XGPU-B which is unique to the 1.6. Maybe NVIDIA removed the extra chip select logic to save on die area?

So I connected up my scope to 2 CS pins of the existing chips and 2 CS pads on the bottom. My expectation was that during the 128MB RAM test, the existing RAMs will be deselected most of the time and so their CS pins would be high. During normal operation, the extra chips should be deselected.

image.thumb.jpeg.43c0da1b625ccddb2888b934003796b6.jpeg

I don't have any scope screenshots, but here was my setup with the scope probes and flying wires. That was 3 weeks ago before I ordered the extra chips from China :D 

The experiment confirmed the expectation and the rest is history.

  • Like 3
  • Thanks 2
Link to comment
Share on other sites

3 hours ago, Prehistoricman said:

The Theory

There are some awesome scans of the Xbox motherboard online for versions 1.1 and 1.6 (they have a sickmods.net watermark). I traced all the signals of a RAM chip and found these results:

  1. A0 - A11 address pins are connected to the same pins of the chip under it
  2. BA0, BA1 bank select pins are connected to the same pins of the chip under it
  3. Same for CK, ~CK, CKE, ~RAS, ~CAS, ~WE, VREF, MCL, and DQS
  4. However, the data pins DQ00 - DQ31 are in reverse order
  5. The data mask pins DM0 - DM3 are also in reverse order
  6. Since both the data pins and data mask pins are in reverse order, we can reverse both and achieve a compatible interface
  7. The chip select pin ~CS is unique for each RAM chip

After this, I started tracing the ~CS signals back to the northbridge/GPU on the 1.1 motherboard. Then, I found the same pads of the northbridge on the 1.6 motherboard. Luckily, all 4 of the CS pins are exposed on the bottom of the board!

At that point, I had to confirm that the 1.6 northbridge wasn't actually incompatible with the extra RAM. The chip is labelled XGPU-B which is unique to the 1.6. Maybe NVIDIA removed the extra chip select logic to save on die area?

So I connected up my scope to 2 CS pins of the existing chips and 2 CS pads on the bottom. My expectation was that during the 128MB RAM test, the existing RAMs will be deselected most of the time and so their CS pins would be high. During normal operation, the extra chips should be deselected.

wiTEVRD9MYTYVT0iDBzv84AftUPzjgckBJnRtHxBX_jzNm6ftZs687KYWUSayI_SB5G_1SAfMH4cPN4fMRNANM6IT7mRe6TD6H75OdaLUom6Cd6YfuW0R6Y1M0zjaJ_FOndX2xXeMocLhYuOPtisA1lTRTxFozZJ4KGUMIheWlHTD2Nmb6o8P8LW_1GpC_2IWaF8Rt1hAPLliD2At_VK8YlXCXsIHsJwosUsW1ZaPQm5GKYj0edsRrQctI6XEnkcw_H6XG--D-E4wSey15asQlTfvD9aY0uM9D_-sJpzfIp36qES1aE6hNpCqu36sXE61aNvqQG2j569X3U63jdTIUOJS67bobMH75dFFC41vpv9pmu8MyEg21Tnjw82pWFhobTn4H2vGWBgiO99mgUvHVUoMegGVttHf5kE6LP2IviiVNKmUoXm0EPa03ipx1jUXIZEQX1cmEYdS0oU1WOxDTXBqKdLYMtT5xNArnvXplNyu1yAfwvHmYszy4dGJkNueMTNJ-kIHgU_NAizCZ0IIYsAhMp68HcXKOVTi-Z8fuId7P9gSIFqwbfmX38HX-zx9LdmxgmWd4a9glUSto0bm-UnlzQ-ddnXbxw98xh4HFaslCzSV90F8Csa8f0QPOPNYtAIRQoEboRRwv1UF7cFdJev2n3inIAmxA0o3xrlp7lVIqPJhYnZqh_7TnZm5kfN_R0Ey-phIJyUzcoOvhnF_RGe7DohIOTd-5FdlFWGrVtq3vUCXCFj37Y63gxNLw=w2405-h1353-no?authuser=0

I don't have any scope screenshots, but here was my setup with the scope probes and flying wires. That was 3 weeks ago before I ordered the extra chips from China :D 

The experiment confirmed the expectation and the rest is history.

Is there to be an image in this post or not?

  • Like 1
Link to comment
Share on other sites

3 hours ago, big F said:

Well this puts another project on my new work bench in my new workshop once my leg is outta its cast.

Hmm I did forget to add functional legs to the required items list!

1 hour ago, Stopme3931 said:

this is amazing may i suggest that maybe changing xblast rev or adding something that state on the menu that it supports 1.6 ram that way people whould know the different straight up?

I was thinking about that. I almost added a line to the main menu to say that I modified it. It would be good to get an 'official' release from the XBlast dev but it has been 4 years since the last update.

  • Like 1
Link to comment
Share on other sites

5 minutes ago, ChaoticAnarchyX said:

Does this opens up possibilities to do this on regular boards 1.0 - 1.5 as well??

You can do it, but it's pretty pointless when the normal RAM footprints are there. You can't combine this mod with a 1.0 - 1.5 Xbox to get 256MB because this mod is just using the same pins as a 1.0 - 1.5 128MB mod.

Link to comment
Share on other sites

1 hour ago, ChaoticAnarchyX said:

Does this opens up possibilities to do this on regular boards 1.0 - 1.5 as well??

As its custom silicon in there and not re packaged pc North and South bridge chips I doubt there are any paths for more than 128. 
unless some one comes out with a dev box that says otherwise I doubt M$ would have bothered.. That said the properly proper prototypes may but we will never see those as they'll be locked up in what ever storage facility M$ have for all their skunkworks prototype stuff.

Link to comment
Share on other sites

2 hours ago, Donnie-Burger said:

Haven't found a 1.6 bios that supports 128mb yet.  Gonna keep trying. 

That might skupper things….

Never got into bios modding or building on the Xbox.  On a PC it was often just a case of loading up a hex editor and changing text and values.

Just had a quick google on it and theres lots of pages telling you how to modify an X2 bios etc, butt thats just size and splash screen stuff.

I could only find stuff regarding Qemu and obscure mention to repositories on M$ sites probably long gone. Also most mentioned needing a full on Dev box.   I only have a debug, so guessing that will not help.

Theres gotta be a member her who is better placed to answer etc. 

  • Like 1
Link to comment
Share on other sites

6 minutes ago, big F said:

That might skupper things….

Never got into bios modding or building on the Xbox.  On a PC it was often just a case of loading up a hex editor and changing text and values.

Just had a quick google on it and theres lots of pages telling you how to modify an X2 bios etc, butt thats just size and splash screen stuff.

I could only find stuff regarding Qemu and obscure mention to repositories on M$ sites probably long gone. Also most mentioned needing a full on Dev box.   I only have a debug, so guessing that will not help.

Theres gotta be a member her who is better placed to answer etc. 

OP said X3 worked.  Waiting for Cerbios to drop 1.6x support.

Link to comment
Share on other sites

x3.3294.v16plus - works 128MB

x2.5035.v16plus.bin - works 128MB

evox.m8plus.v16 - can only use 64MB but still works

256k.yoshi.k2005.v16 - works 128MB but it's a debug BIOS

For some reason my X3 flashbios now has no video output, not even a black screen. After about 20 seconds the front LEDs go orange. I'll see if I can deselect the extra RAM by resoldering the chip-select points and maybe this fixes it.

  • Like 2
  • Thanks 1
Link to comment
Share on other sites

8 minutes ago, Prehistoricman said:

x3.3294.v16plus - works 128MB

x2.5035.v16plus.bin - works 128MB

evox.m8plus.v16 - can only use 64MB but still works

256k.yoshi.k2005.v16 - works 128MB but it's a debug BIOS

For some reason my X3 flashbios now has no video output, not even a black screen. After about 20 seconds the front LEDs go orange. I'll see if I can deselect the extra RAM by resoldering the chip-select points and maybe this fixes it.

Yeah same here -  evox.m8plus.v16 - can only use 64MB but still works.  Hopefully the Cerbios team has 1.6 bios plans.

I had bank 2 acting funny and xblast would show 64mb sometimes.  Retouched up bank 2 and cable select wire. So far so good.

Edited by Donnie-Burger
  • Like 1
Link to comment
Share on other sites

14 minutes ago, sweetdarkdestiny said:

I will see that I can patch the MakeMHz versions today (TITAN and regular). After that I will post a link here for download. :)

How ever, BIG THANKS for testing these Donnie. 

You are welcome.  Is there anyway to incorporate xiso support into titan, seems like make mhz and xiso is the only goodies missing, or am I asking to much? 

Any bios you want tested let me know and thank you again.

  • Like 1
Link to comment
Share on other sites

No XISO support for EvoX. I know that Rocky5 has made a BFM EvoX XISO but I don't think that I can patch that. Maybe if you ask Rocky5 for an updated version of his BFM bios he will do it. 

And the MakeMHz versions will come later. I will send you a PM with them to test and when you say they are good, I will drop them here. But, atm I need more coffee to come to life. ;) 

Edited by sweetdarkdestiny
Like usual, typos.
Link to comment
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
Unfortunately, your content contains terms that we do not allow. Please edit your content to remove the highlighted words below.
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.

 Share

Board Life Status


Board startup date: April 23, 2017 12:45:48
×
×
  • Create New...

Important Information

By using this site, you agree to our Terms of Use.