SGI: Hardware

And Success!!

I got these two warnings:
and


The first one comes and goes:

Not sure what to make of it, there is no 2.5V from the PSU so its either a fault in a DC/DC regulator on the motherboard or its derived from something on the ATX PSU that isn't within WTX specs.

The second problem then, the fan. Well, it wasn't spinning. I measured power between FANC and ground and it was very low, perhaps it wasn't meant to be powered from the motherboard. So I took power to the fan from a 4 pin molex. Now it spins. And after figuring out the fan wasn't giving any RPM readings, I swapped it out for the original fuel psu fan. And voila:



I'll post details tomorrow. Now I'm gonna take a picture of this mess and go to bed.

Nicely done Pontus!
This could very well be an option for fried PSU's, but it would be awesome if one could work out the fan issue altogether and get a (much) quieter Fuel PSU.

Oh, about the VT320... I'll make sure to include a bunch of VT510/VT520's when you come collect the ES45's.

Hi Pontus, nice hack!

Would be nice to know your L1 firmware version, with my blue' Fuel I could not upgrade the L1 because it failed to 'detect' the PSU after that.

Oh, and AFAIK, 2.5V is used for the (DDR1) RAM, and is derived from some other voltage (usually the 12V line I think) by a VRM on your mainboard.

recondas wrote: Outstanding!!

Very nicely done Pontus - Fuel owners worldwide you you a debt of gratitude.

Thank you! Your encouragement is much appreciated.

ramq wrote: Nicely done Pontus!
This could very well be an option for fried PSU's, but it would be awesome if one could work out the fan issue altogether and get a (much) quieter Fuel PSU.

Thank you. Regarding the fan, It's probably just a matter of taking the PSU apart and run an extra the FANM cable out from the fan to the mother board connector.

ramq wrote: Oh, about the VT320... I'll make sure to include a bunch of VT510/VT520's when you come collect the ES45's.

vt terminals is always good to stock Hmm, ES45 you say, Coincidentally I just bought a bigger car.

jan-jaap wrote: Hi Pontus, nice hack!

Would be nice to know your L1 firmware version, with my blue' Fuel I could not upgrade the L1 because it failed to 'detect' the PSU after that.

Oh, and AFAIK, 2.5V is used for the (DDR1) RAM, and is derived from some other voltage (usually the 12V line I think) by a VRM on your mainboard.

Thanks, much obliged.

Is this the version string you want?


I won't have time to do anything more with this for a while, but I'll try to get that picture and the details up a.s.a.p

Good work! Once you get a diagram worked out it should be possible to make an adapter cable.

Pontus wrote: Is this the version string you want?

Code: Select all

SGI SN1 L1 Controller
Firmware Image B: Rev. 1.12.6, Built 04/22/2002 08:13:40

Yeah, that's a pretty old revision, last is 1.48.1 I think.

Sometime soon I'll have another look at the L1 of my Fuel . It's even older (1.9.15) . If I feel brave I'll try to boot the newer L1 again.

So, what exactly have I done?
I bought one of these:

and one of these:

and one of these

The wires are easily removed from the molex using two paper staples that has been straightened out. So I just moved the necessary wires around according to the scheme below. I kept the female side as it were and moved the wires in male molex. The left hand side in the table is thus the female, ATX side and the right is the male, WTX side.

Using ATX pinout and WTX pinout (look at the alternate one) I made this mapping. I also followed the wires back into the WTX PSU since the purpose of the wires were printed on the circuit board end. And thereby I could verify the pinout.


Three positions on the male connector has get special treatment:

Pin 3:
This is +3.3V sense. If I understand sense lines correctly you have to run the sense line as close to the load as possible for proper function. The sense line on the ATX side shares pin 13 with a +3.3V wire, it is usually a brown slightly thinner cable, but you have to open the PSU to be sure. Cut it and extend it to PIN 3 on the WTX side (it is not shared with any other wire)

Pin 12:
This is FANC, it should control fan speed. Read more below

Pin 24
This is FANM, it monitors fan speed. Connect to tachyometer pin on fan:


Next you must swap GND and +12V on the P3 connector . Just to be sure, remove wires from pin 4 and 8. Like so:



Concerns:

1. The original Fuel PSU has sense lines on +3, +5V and +12V. ATX only on +3V. If accurate +5V and +12V is needed this might be a problem. But according to L1 they are well within margin when the system i idle.
2. The WTX guidlines state that a 460W PSU should deliver 45A on +3.3V. Not many ATX PSU goes over 30A. Similar for +5V
3. The FANC on pin 12 doesn't appear to work as expected. I expected that the wire would go directly to the fan header inside the PSU, but it looks like its going into one of the IC's, but I'm not sure. The FANC line stays put at +3.3V even when the system if powered down. This might be to low to start the fan from the PSU. Which might indicate that it is just a signal wire to a fan controller in the PSU.. but I don't know.. perhaps a good reason to get my oscilloscope out.
4. Will This system handle load? We'll see I suppose

So, thats it for now Here's how fun I've had:


Edit: Totally forgot the EPS12V / P3 connector pin swap table

Good work!

Meanwhile, I've had a another look at the Fuel that didn't like it's PSU.

Here's some L1 output:

So far so good. However:

Major identity crisis!

Finally:

However, I'm hesitant to blame this on the PSU and/or the wiring. It looks like the L1 didn't migrate it's config data when upgraded from a very old to the latest & greatest. I should have known

recondas wrote: The strongest argument for an SGI-only power supply is jan-jaap's experience with the PS in his prototype Fuel becoming unrecognized after a firmware upgrade - which strongly suggests there's additional outside of the ATX-spec monitoring going on. But it might also be possible that j-j's pre-production prototype Fuel didn't include a PS with ATX2.1 fan and power monitoring capabilities.

I think that, between Pontus' work, and the obvious issues between the L1 software and my IP34 board, this is open for discussion.

I may try incremental L1 software updates on this Fuel, but I'd better be extra careful because if I loose flash bank B I brick the damn thing.

One more thing:

On the O350, I have a 2MB flash chip, and a 2MB L1 image:

But on the Fuel , I have:

The 'brick' command returns a 2MB flash chip, but a 1MB image is loaded. I'll take this with a grain of salt since it doesn't even know what it's talking to, but I have the feeling that the wrong L1 image is loaded. Which would explain a few things Unfortunately, I don't have another Fuel handy, nor did anyone bother to mention 'l1cmd version' output in the 'hinv' forum ...

Edit: Google found it. The correct L1 image is the "Fuel/PE 1MB image" , not "Base 1MB image".

Pontus wrote: And after figuring out the fan wasn't giving any RPM readings, I swapped it out for the original fuel psu fan.

It might not be related your fan issues, but when I added low noise fans to the O350/V10 I recently set up, I got fan failure messages. The "quiet" fans I used included a temperature sensor, and wouldn't even spin up until the O350 generated enough ambient heat - so they always generated a fan fail message at start up. I used a similar fix by adding a fan without a thermostat. Perhaps the original fan from your WTX power supply has a similar "low noise" feature.

jan-jaap wrote: nor did anyone bother to mention 'l1cmd version' output in the 'hinv' forum ...

Guilty as charged However there are several that included the output of "flash status" run against the L1 - which gives the same info for both L1 firmware images. smj's 700MHz is at 1.44.4 , and both of Fuels I have are at 1.26.5:
viewtopic.php?f=14&t=16725501
viewtopic.php?f=14&t=16725486

Well "flash status" is part of the recommended procedure in the Wiki page hinv forum guidelines , so hopefully it will figure in more posts going forward. It might help if a link to that page were in the banner for the hinv forum, instead of the note about hinv and gfxinfo output...

recondas wrote:

jan-jaap wrote: nor did anyone bother to mention 'l1cmd version' output in the 'hinv' forum ...

Guilty as charged However there are several that included the output of "flash status" run against the L1 - which gives the same info for both L1 firmware images. smj's 700MHz is at 1.44.4 , and both of Fuels I have are at 1.26.5

Absolutely true, but the actual version number was of course not what I was after looking at L1 "version" output

For a given L1 version, there's more than one image:

The problem with my Fuel is that I flashed it with version 1.48.1 while it was running something < 1.22.2 which is not good .
As a result of this (?), my 1.48.1 image is the "Base 1MB image" rather than the appropriate "Fuel/PE/O300 1MB image", and this information is *not* in the L1 "status" output.

Successfully upgraded the L1 on my Fuel to the last rev (1.48.1) using tjsgifan's recipe (thanks!). PSU still recognized

I'll stop hijacking your thread now

Hello,

Can you more info about 'tjsgifan's recipe' ?

Regards,

Mark

jan-jaap wrote: Successfully upgraded the L1 on my Fuel to the last rev (1.48.1) using tjsgifan's recipe (thanks!). PSU still recognized

I'll stop hijacking your thread now

No problem. Great that you got the blue freak going again

Mark_G wrote: Can you more info about 'tjsgifan's recipe' ?

I linked to it twice already in this thread, but here it is: viewtopic.php?f=3&t=16721376&p=7303920

Basically, if you want to upgrade the L1 from an old version, you have to install IRIX 6.5.21 and flash to version 1.22.2 first. I did 1.9.x -> 1.22.2 -> 1.32.6 -> 1.48.1

Up to (and including) version 1.22.2, there were two L1 binaries, one of 1MB and one of 2MB. In the newer versions there's a 'Base 1MB' and a 'Fuel/PE/O300 1MB' image. It looks like the flashsc program will flash the wrong (Base) image into the Fuel if you skip he intermediate steps.

Pontus wrote: Checking the pinout on the 24 pin P1 connector it doesn't match the color coding of neither ATX or WTX specs, but the _alternate_ pinout here http://pinouts.ru/Power/wtxmain_pinout.shtml matches except for one pin. That page also states that there is a 6-pin header used with the alternate pinout for CPU power, which the Fuel uses (well, it's a 8 pin connector but 2 are not connected).

As a follow up to Pontus' excellent description of the differences between the pinout of the 24-pin connector used in alternative WTX and Fuel HiPro HP-W460GC33 power supplies, I took a few minutes and created a table that helps to illustrate the differences and similarities of the (alternative) WTX, the Fuel Sparkle FSP460-60PFN, a generic Sparkle ATX2.2 FSP460-460PFN and ATX2.2 power supply pinouts side-by-side.

The WTX pinout was taken from the pinouts.ru page linked by Pontus, the Fuel PS pinout was derived from photographs posted by 87Porsche of the 24-pin connector on his Fuel's Sparkle FSP460-60PFN 460W power supply , and the list of wire colors/voltages printed on the Fuel/Sparkle power supply label .

I obtained a non-SGI/ATX2.2 version of the Sparkle power supply that uses the same FSP460-60PFN part number as the SGI version , but with an ATX 2.2 pinout. The non-SGI Sparkle FSP460-60PFN uses the same Nidec D08T-12PU cooling fan as the SGI version, which may eliminate the fan issues Pontus experienced.

The pinout of the non-SGI/ATX2.2 power supply is included in the table for comparison, as is the standard ATX 2.2 pinout (which also provides additional clues as to the electrical values of wire colors not listed on either Sparkle label). Hopefully a side-by-side of the four color-coded pinouts will serve as sort of a Fuel Power Supply Rosetta Stone.

Obvious differences between the WTX and Fuel Power Supply pinouts:

• Pin 3 in the WTX pinout is an orange wire listed as "3.3VDC Sense", the values for the brown wire connected to pin 3 of the SGI/Sparkle aren't listed on the label, but the ATX/EPS version of the Sparkle power supply (with the same FSP460-60PFN part number) piggybacks the brown wire into the motherboard connector with one of the orange +3.3VDC wires (ATX pinouts from other sources list the brown wire as 3.3V Aux)
  
• Pin7 is listed in the WTX pinout as "+5VDC (Aux)", the purple wire connected to pin 7 of the SGI/Sparkle is listed as "+5Vsb" (sb = standby).
  
• Pins 11 (PC), 12 (FAN C), 23 (PSON) and 24 (FAN M) aren't assigned specific wire colors or values in the WTX pinout, the colors and values used in the SGI/Sparkle connector most likely provide environmental monitoring of the power supply to the Fuel L1.

Note that the WTX and SGI pinouts for the eight-pin connector used to supply CPU power (labeled "P2" in the Sparkle power supply wiring bundle) are different . The available photos of the SGI version of the Sparkle PS don't provide a clear view of that connector, so it hasn't been included yet. I'll update the post as soon as I have an opportunity to open one of the Fuels.

Please double check the values given in the table *and* the unlisted CPU power connector *before* attempting any to use an alternative power supply in your Fuel!!

The chart visible in this post is a jpeg, if the text proves hard to read in that format, the pdf version attached should be easier to read.

Most excellent. I have it on my todo-list to investigate the FAN-issues, but there are sooo many other fun things to do

Had an opportunity to extract the Sparkle FSP460-60PFN power supply from one of the Fuels and open it side-by-side with the PC-ATX version (that carries the same Sparkle part number).

First a comparison of the the labeling (the SGI version is on top). Not seen in this view, but the non-SGI/PC-ATX version is labeled "For Intel", so to minimize confusion I'll refer to them as the Intel or SGI verions.

The layout of the interior of the two power supplies is almost identical. The most notable difference between the two is a small circuit board, which is present in the SGI version but absent in the Intel version.

It would appear the extra circuit board is there to provide the SGI version environmental/fan monitoring and control, the fan cable in the SGI version is routed to the extra circuit board. Both the Intel and SGI versions have a Nidec BetaSL model D08T-12U fan - the SGI version has three wires, the one used in the Intel version two.

In the Intel version the fan is connected to the main power supply PCB: The fan header location used in the Intel version is marked on the main PCB in the SGI version, but no connection header was soldered in during assembly:

My guess is that the Fuel's L1 environmental monitoring won't work as intended without the additional circuit board (though it may be possible to wire the fan using the same method as pontus illustrated earlier in this thread).

Would be cool if some of this made its way back to the wiki.
Not me, I know little about Fuels.

R.