[Raion]

I often seem to be misquoted on problems with Fuels, so I figured I'd explain all of the problems I've figured out with the Fuel's design, explain my reasoning behind it, and encourage discussion. Should someone do fog tests with Fuels or prove alternative causes, that's great. Without further adieu:

The Fuel's PSU is a common failure item, and it seems to take the environmental monitoring chip with it. The chip is in a precarious place, that requires hot air rework and shielding from all of the surrounding plastics. 

The environmental monitoring chip covers every fan in the system, monitors the thermal probes, and shuts down the system in the event of thermal problems. It also seems to control fan speeds. Without it, the system will not boot.

A lot of people, because of this issue, turn off env monitoring and continue business as usual, but this presents a problem: Without fan controls or monitoring, the system is vulnerable to overheating. I am not 100% on this, but when i ran a Fuel without env monitoring, it didn't default to max fans. I have owned four Fuels, all of which broke down in various ways. 

Because the Fuel's airflow flows in two directions, this causes turbulence inside the case. The central fan tunnel pushes air into the case, that's why the fan shroud is shaped like it is. If it pushed it out of the case, that shroud shape makes no sense. The VPro/XIO2 fans and PSU fans vent. This creates turbulence, because air doesn't like to change directions. This can cause something called fan stall. Fan stall is when a fan stops moving air. It's similar to a compressor surge in a gas turbine, and while it doesn't damage the fan, it can cause overheating. 

So the consequence is that heat-sensitive components on the board do not get adequate ventilation, such as the VPro, and this results in component death. The three fans on the front of the VPro tunnel in this configuration are not sufficient carriers of air in this configuration, in my opinion, especially with the risk of fan stall. 

By comparison, the Tezro has all of its fans moving in a singular direction, from front to back. This tells me they learned their lesson from the Fuel, (Fuel came in 2001, the Tezro in 2003) and thus are aware of design flaws. 

I have had a bad run with Fuels, some of us have not such as Larbob (and I'm glad he hasn't, it's painful), and in frustration I sometimes say that the Fuel sucks. I stick by the fact that at current market prices (often over a grand) it's not worth it at all. Back when you could get a reasonable fuel for 200-300 in 2014, sure! But now? not at all. 

The main concern I come with this is the destruction of irreplaceable parts like the VPro cards, and the driving up of those prices, which has happened a LOT recently. $500 for a V10 on eBay now? You kidding? 

So please don't misquote me. I said what I said about Fuels with the best intentions in mind. If someone has a fog machine and wants to test (or use a cigar or incense or anything else that produces smoke), sure. A thermal camera is also helpful, but AFAIK those do not accurately record air temperatures or airflow, just surface temperatures.

[micrex22]

It should also be emphasized that the fuel's case was just re-used from the SGI Visual Workstation 230. So rather than building a case for the fuel, the fuel had to fit inside the Visual 230 to the best of its ability. Which meant the VPro WAS ALWAYS going to have sub-optimal cooling. If you look at the original thermal requirements of the VW230 you can see it's much less of a problem with the intended PC clone guts: http://hardware.localhost.nl/pictures/20..._board.jpg

Honestly, Richard Belluzzo didn't give a sh*t and just had the things pumped out. The fuel would have worked just *long enough* for the intended purpose and that's all they wanted.

Fixing the cooling should be doable but I'm not paying $2000 (!!!!!!!) for a computer that's not at all worth that much. Anyone who sells Fuels the same price as Tezros has a problem.

Oh yeah, and those surface mount electrolytics are a huge problem nobody brings up. I guess the tezro has some too but, SMT electrolytic capacitors do not hold up well as their compromised size means they have a short lifespan.

[tomvos]

I think the case of the fuel was a thinly customized off the shelf case.  In that time the SGI Fuel and SUN Blade 2500 used the same case.  Here are two images where you can see this:





I think there was even a thread on Nekochan where someone identified their generic common ancenstor case.


That said, I still like the Fuel.  It's not too loud and because it's based on the Origin 3000 architecture - at least from my point of view - it feels snappier than any other SGI system.  Having a build in DVI does not hurt either.  I've bought one of Kubatyszko's power supply adapters.  So when the power supply fails, I can use a commodity power supply.  I hope the V10 does not die anytime soon.

[Raion]

The Chimera architecture also includes the Tezro and Origin/Onyx350. The Tezro feels a lot faster, and I'd argue a Fuel is slower than a hi spec Octane.

You do NOT want to wait for that power supply to fail, because according to weblacky's theory that's what takes out the ENV monitoring chip.

[hamei]

The Chimera architecture also includes the Tezro and Origin/Onyx350. The Tezro feels a lot faster, and I'd argue a Fuel is slower than a hi spec Octane.

You do NOT want to wait for that power supply to fail, because according to weblacky's theory that's what takes out the ENV monitoring chip.

Tezro has more than 1 p   Fuel is faster for single-threaded tasks than an Octane. Especially an 800 mhz Fuel. They are okay, but have some really annoying weaknesses. If you don't depend on it for daily use, the Fuel is okay.

I believe weblacky's theory is incorrect, my environmental monitoring went out a good year before the power supply. I never did fix it, just ran without. Indigo. Indigo2, Octane, O2 (another hasslicious machine, probably worse than the Fuel) didn't have env monitoring ... I ran mine daily for years until the ps thing finally pissed me off so much that I stripped it into parts and gave them away. Fricking thing cost $3500, too.

What really kills Fuels is the stupid crap SGI did to the power supply so that users could not just throw in a cheap peecee supply. Towards the end I bet most users wanted to beat SGI in the head with a baseball bat. Starting with Rick-baby it went from shit, to worse shit, to unbelievable shit.

Thanks to kuba from all Fool users, that's a great piece of work.

[weblacky]

Hi,
Just to reiterate my theory, for those that didn’t read it initially:

My theory is the DS1780 chips go out just before the PSU goes out (of tolerance) because the data sheet for the DS1780 very clearly states the maximum allowable voltage of VCC is 5.75v for current DS1780 ICs.

As PSUs ages in the Fuel workstations most of the posts I’ve seen showing L1 voltage stats show a steady (slow) rise in voltage, especially the 5v line! It seems the voltage goes to 5.80v or more before the L1 starts complaining about voltages out of tolerance, bad reference voltage lines (again just going on old Nekochan posts), or simply refusing to boot.

My current belief is the SGI engineers set the voltage power-off system too high. The Fuel’s core mainboard obviously has no problem dealing with higher voltages but if the DS1780s do take 5v directly off the PSU then they are running just above their tolerance when the L1 warning system even kicks in.

Well after 5.75v...your DS1780 should have gone POP!

Just a theory until someone else comes up with something better.

EDIT: This theory also fits Hamei's description, first your environmental monitoring systems goes out (but L1 says nothing about voltage tolerance issues), but shortly afterwards the L1 starts to complain about the DC voltages of the PSU and performs shutdowns shortly after boot, and then eventually the L1 will prevent you from even booting due to how far off your voltages are at boot. However, since I don't have access to a Fuel right now, I can't probe the VCC pin of a DS1780 IC to show it's directly connected to the known 5v PSU pin. If that is shown true, then I believe this theory has good legs on stand on.

[Raion]

Yeah, it seems reasonable.

[hamei]

Yeah, it seems reasonable.

Maybe. But I don't believe the voltages in my Fool ever went out of tolerance. The stupid psu just refused to turn on.

I did look once in a while and everything read normal, but the dumb env would shut down the computer if you left it on for more than a few minutes. But during the short period of time that it worked, everything looked okay.

Assuming the env chip was working properly, that is. And since it shut the thing down, obviously it wasn't.

Chicken and egg, I guess.  Both components suck

[jan-jaap]

IIRC there was simply a "bad year" in the production of the DS1780 chips. There have been quite a few people who replaced DS1780 chips, I never heard anyone come back with an "OMG they burned out again and this time it took out the entire system" or similar.

The problem with the Dallas chips doesn't just affect the Fuel, the early O3000 bricks had the same problems. There just aren't many people operating an O3K from their basement.

It's always the systems with the over-complicated power systems that give you the biggest headache. Ask anyone with a Challenge or Onyx.

[lunatic]

It's always the systems with the over-complicated power systems that give you the biggest headache. Ask anyone with a Challenge or Onyx.

I guess we can safely agree that broken power supplies are the one failure that is most common across many years of SGI products.