Mac Quadra 950 - NuBus slot position for a graphic card

[Classicman]

Hi to everyone! this is my first post here

I'm a Mac user since 1992; now have recovered a Mac Quadra 950 with 64 MB of RAM and this is fantastic!

My question: I have recovered a NuBus graphic card (Formac ProNitron 80.21 with vram expansion), into the Mac Quadra there are five NuBus slots.
I remember that there are some slots into the Quadra 950 that deliver more energy (in watts) to Nubus cards.
Please, can any of you advise me where to install that card? in which slot can it receive the more energy (and perhaps works better)?


Thanks a lot in advance.

 

[slomacuser]

In Quadra 950 the closest Nubus to the PSU (Power supply) is first to read/load in Quadra 650 is the farther one, first from left ...

 

[Classicman]

Thank you slomacuser! beyond the position of the card and the importance in terms of read/load, so in terms of energy there is no difference between the various slots?

 

[cheesestraws]

Welcome to the forums!

I think you may be overthinking this; I just sling cards in wherever and I've never had an issue . The 950 has an enormous PSU and graphics cards aren't that power-thirsty (unlike new ones).

 

[jeremywork]

Some sources say the bottom two slots provide 25 watts each, while the top three provide only the standard 15 watts. Others indicate it's just 95w total over the five slots.

Either way, as @cheesestraws mentioned there are very few nubus cards which require more than 15 watts. Offhand I can only think of the Radius Rockets, which still work fine in older machines that don't have the beefier 25 watt slots, so long as all the other slots aren't also loaded down.

Enjoy your 950

 

[Unknown_K]

My 950's are stuffed with cards and I never paid much attention to what went where.

I know AVID had a map for what cards went where for the PCI setups but that was because of the way Apple setup PCI machines with more then 3 slots.

 

[Unknown_K]

A rocket won't run on a IIsi with Nubus slot adapter for power reasons.

 

[cheesestraws]

Some sources say the bottom two slots provide 25 watts each, while the top three provide only the standard 15 watts

Yeah, I've heard this too, but never investigated, because none of my cards need more juice than a standard slot provides.

 

[Trash80toHP_Mini]

A rocket won't run on a IIsi with Nubus slot adapter for power reasons.

Interesting, have you seen documentation or put together sources adding up to that limitation? It makes some sense, but I'd thought the anemic cooling budget in that cramped, low profile form factor was the reason. For either or both reasons, Radius excluded the IIsi Machine ID in the installer, saying it wasn't supported.

Back on topic: I've seen statement's that it's best to put a video card at the Slot ID associated the first available interrupt. Don't recall when, where or reliability of the source I read though. It made some sense to me from what I knew on a technical level at that time.

 

[cheesestraws]

I've seen statement's that it's best to put a video card at the Slot ID associated the first available interrupt

Probably not worth the effort unless you are full of cards all of which are doing stuff at interrupt-time.

 

[Trash80toHP_Mini]

Makes sense. ISTR it was in the context of placing the bus mastering (other card accelerating) VidCard released with the IIfx at $9.

 

[slomacuser]

Interesting, have you seen documentation or put together sources adding up to that limitation? It makes some sense, but I'd thought the anemic cooling budget in that cramped, low profile form factor was the reason. For either or both reasons, Radius excluded the IIsi Machine ID in the installer, saying it wasn't supported.

I can confirm that IIsi PSU can't run more power consumption NuBus boards. I red in one of Macworld magazines but can't remember which one. I also tried Radius Thunder GT video card in IIsi and IIsi didn't start at all ...

 

[cheesestraws]

Yes, but the OP doesn't have a IIsi, the OP has a Quadra 950, which will cheerfully drive their nubus card in any slot, I'd imagine.

 

[Melkhior]

Makes sense. ISTR it was in the context of placing the bus mastering (other card accelerating) VidCard released with the IIfx at $9.

TL;DR: NuBus slot numbers mostly don''t matter

NuBus cards have two mechanisms to take the initiative; slave devices assert their (private) /NMRQ line while bus master device assert the shared /RQST and put their slot numbers ($9-$E) on the /ARB[3..0] lines, which may cause an arbitration cycle if more than one master request the bus at the same time.

For slave, I don't recall seeing the priority being documented in DCDMF3, it just comments on them sharing a CPU interrupt usually through VIA2.

For master, the priority is descending - highest slot number wins. Master cannot request if /RQST is already asserted (and loser keep the line asserted), so only completely simultaneous accesses require such arbitration. Starvation is not possible; the highest-priority device will not get the bus back until lower priority contender have had their turn - worst case delay with two masters is two rounds (e.g. if $E accesses the bus and $9 want the bus in the following cycle, then $9 can't assert /RQST but has to wait until $E is finished; at this point, $9 will be able to assert but if $E also asserts then $E wins and $9 has to wait but is now asserting; when $E second cycle finishes, $E is not allowed to take the bus again because $9 is asserting so $9 wins by default). More masters will increase delay has more than one master could be waiting.

So if you have a bandwidth-hogging master on the bus, you want it in the lowest-numbered slot so it doesn't create too much delay for other masters. If it's alone on the bus it really doesn't matter where it is. Dumb framebuffers are slave-only usually (and will generate interrupts for VBL only so somewhere in the 60-85 Hz range). Graphics accelerators will be bus master if they need to access the main memory and/or other NuBus device like dumb framebuffers, and are not necessarily high priority compared to e.g. high-speed I/Os (SCSI, Ethernet, ...) so should probably go in lower-numbered slots (i.e. the quoted rule makes a lot of sense).

So for performance, the position of slaves is essentially irrelevant and the position of masters is only important relative to each other.

... at least that's how I understand the specifications

 

[Trash80toHP_Mini]

Thanks for that explanation. My understanding of such things is limited, to say the least. Somewhere in there is the explanation as to why the 950, its forbears and those following all have onboard video located at $E.

@cheesestraws agreed, further Rocket in IIsi discussion needs to be moved to its own thread. I covered a lot of that ground in a topic somewhere already. If I find it I'll reboot the discussion there, thanks for keeping this OT for the OP.

 

[Classicman]

Thank you all for the precise and technical responses!!! all my doubts are now solved.