View Full Version : ATI Radeon driving 2 displays = i controller or 2 w/pci bridge?

02-18-2002, 10:40 PM
Does the ATI Radeon series of PCI card have one controller capable of driving two displays, or are there two controllers with a PCI Bridge?

I want to drive a pair of Apple Studio Display LCDs from a single PCI card in a 20th Anniversary Mac (TAM). Like the 6500, the TAM has a buggy PCI controller, which is unfriendly to PCI bridges as found in USB+FW cards and expansion chassis.


my free 9500 is nickel-and-diming me to death!

02-19-2002, 05:21 AM
How do you tell which is which? I have a Radeon 7K coming in in a few days and can check for you.


02-21-2002, 09:48 PM
There may not be any way to tell from examination of the card. Apple System Profiler or PCI Prober might provide some insite though.

While some bridge bearing cards are made with a separate bridge chip and then two of whatever device it supports (or separate Firewire and USB chips for the combo cards) it is possible that ATI integrated a Bridge and two Radeon chips all on the same piece of silicon.

02-21-2002, 11:39 PM
The 3Dfx V5 has a bridge chip in it - pci bridge to be more accurate. It only supports one monitor though. There were issues getting the second GPU to work on vintage PCI macs - even 95/9600s. Those cards are still great for UT! http://macgurus.com/infopop/emoticons/icon_biggrin.gif

The ATI RAGE128 VR and the old RADEON support dual displays - via the S-Video. I think the RADEON 7K is the same - more or less, but it has two VGAs out and adaptor(s) for hooking it to DVI, VGA or sandard S-Video I think.

Maybe no bridge chip?

The Cost Of Freedom:
Every bit of energy.
Every ounce of courage.
Every drop of blood.

Sic Semper Tyranus!

02-22-2002, 01:48 AM
The 3Dfx V5 has a bridge chip in it - pci bridge to be more accurate. It only supports one monitor though.
There were issues getting the second GPU to work on vintage PCI macs - even 95/9600s. Those cards are
still great for UT!

Is that why I can run two monitors with the Voodoo5: because I'm using the mobo VRAM in my 8500 to run the second one?

02-25-2002, 05:33 PM
Thanks for all the responses, especially regarding the Voodoo bridge.

I'll spend a few more bucks to get a Radeon 7000 locally so that i can take it back without a hassle if it doesn't work..

02-26-2002, 10:20 AM
My VooDoo 5 card used to show up as 2 cards in Apple System Profiler. I had it in a B&W, they had 4 PCI slots and ASP would show 5 PCI slots is this because of the bridge?

Anyway the Radeon 7000 only shows as one card in ASP.

02-27-2002, 12:33 AM
Interesting... I never really noticed this before:

Slot C1
Display card
Card type: display
Card name: 3dfx,Voodoo5
Card model: Voodoo5
Card ROM #: Not available
Card revision: 1
Card vendor ID: 121A

Screen Size (pixels): 832 x 624
Grays/Colors: Millions
Resolution (dpi): 72 x 72
Additional Information: Main screen

Slot C1
PCI Slot
Card type: Not available

Card name: 3dfx,Voodoo5,Slave1
Card model: Not available
Card ROM #: Not available

Card revision: 261748411
Card vendor ID: -1

I wonder if this is due to the dual processors?

02-27-2002, 12:41 AM
yeh mine does almost the same...except i have it in slot a1 and in the second a1 info box my card model is a "-1" and the card revision is "not available"... i thought it was odd when i first got this card but didn't know enough to say it was odd but now that others have come out i think i'll say it is odd...odd huh?

02-27-2002, 08:55 PM
Definately odd. http://macgurus.com/infopop/emoticons/icon_biggrin.gif Like Dragon-x said though: it's still http://www.macgurus.com/ubb/kickass.gif in UT

03-21-2002, 03:43 PM
So I installed a Radeon 7000 in my Twentieth Anniversary Macintosh (TAM) with limited success. In 1 hour of testing things appear OK if you want VGA, SVIDEO or both simultaneously, using OS9.1 on legacy PCI Mac.

TAM configuration before Radeon 7000 install. 128M RAM, Toshiba 30G HD, Sonic CSII Ethernet, with Sonnet Crescendo L2 G3-500, Rage II motherboard video driving a built-in 800x600 LCD. ATI drivers were present from MacOS install, plus I was surprised to find that the 9.1 install CD from a 533-dual G4 Digital Audio had already installed ATI's "minimum" OpenGL v1.2.2 with 9.1. However I upgraded to OpenGL v1.2.4 copied from a Pismo 9.2.2 extensions folder.

The Radeon 7000 comes with PCI card, manual, reg card, SVIDEO cable, and adapters for
-15-pin mac-to VGA,
- DVI to VGA

I was Mighty Pleased(tm) to discover the Radeon 7000 does not have a fan, but rather a low-profile heat sink!. My TAM is nearly silent, and does not accommodate a very thick PCI card. I was not looking forward to additional noise in my environment, nor leaving the case open. At least one support chip becames too hot to touch, which was a painful discovery. I may add a heat sink to that chip to aid system longevity.

The Radeon 7000 docs state clearly and often that the user should install the ATI software before the PCI card. I have contempt for hardware that crashes without drivers, so I installed the 7000 card with two VGA LCDs, using both the VGA (Display #2) and DVI (Display #3) connectors of the 7000, since ATI kindly provided a DVI-to-VGA adapter amongst others.

I rebooted, resetting PRAM from keyboard three times before allowing full startup.

The TAM crashed on boot, at the 9.1 ATI Graphics Extension, reporting a System Error "Illegal Instruction"/Restart dialog. When I clicked OK, the machine repeated the same warning, but changed from Illegal Instruction to "Error Type 10" [sdmJTInitError - SDM Jump Table could not be initialized] and froze utterly, requiring a 3 finger restart.

Rebooting with the shift key was successful. The TAM internal video was fine, and the LCD 2 on the 7000's VGA port showed the extended desktop. Oddly, the LCD 3 on the DVI port showed a green (good sync) LED, but remained black, displayed no image. Monitors control panel was unlaunchable. However, a video utility i wrote back in the NuBus days, ftp://soundmark.net/pub/mac/WindowMaker4.bin , beeped 3 times, indicating I had 3 video devices recognized!

I disabled all ATI extensions and rebooted to 9.1 successfully. The monitors control panel confirmed that I had 3 screen devices, and the green LED was lit, but DVI still displayed no image. I installed the Quicktime 5.0.2 and ATI drivers from the 7000 CDROM and rebooted.

There was still no display on the DVI-connected LCD #3, still the green sync OK LED was lit. VGA display worked as expected, and SVIDEO output worked as expected. However, even with ATI drivers installed, DVI+VGA and DVI+SVIDEO and DVI alone did not produce image on the DVI-VGA screen. I had to go to work before I could test 3D gaming perfomance.

So far I have no joy running two monitors from the card like I hoped. Preliminary results indicate that I can run VGA and SVIDEO simultaneously, but SVIDEO is not why i spent my $116.

Coming soon:
- Testing the 7000 with a real DVI display (no adapter) in TAM
- Testing the 7000 in a 9500
- Testing the 7000 in a Beige G3-266 Rev A
- Testing the 7000 with ATI cards Rage 3D PRo and Mach64 in multiple PCI slot Macs.
- Testing the 7000 with MacOS 8.6 and 9.0
- Testing the 7000 frame rate and 3D performance in TAM


my free 9500 is nickel-and-diming me to death!

[This message has been edited by ricercar (edited 21 March 2002).]

03-21-2002, 04:20 PM

I haven't tried that yet w/ my radeon - at the moment, I have it connected up to a SGI monitor and the Rage 128 that came w/ my Yikes is connected to a 19" Hitachi.. Maybe I'll try doubling it out..


03-22-2002, 08:41 PM
Damn, damn, damn! ATI says that legacy Apple monitor sense is NOT supported on Mac Radeon DVI-VGA adapters. ATI has no intention to support my two Apple VGA Studio Display LCDs on a Radeon 7000.

Door #1 - refund for Radeon 7000.
anyone have experience with MacWarehouse and RMAs?
[later: $%! 15% restocking fee!]
Door #2 - Buy/ DVI LCD or non-apple VGA LCD
ha ha ha! not this pay period....
Door #3 - Trade Apple VGA LCD for DVI LCD or non-Apple LCD.
Door #4 use Radeon 7000
a third PCI video card in an environment with only two displays.. great.

at least the rest of my life is fantastically easy...

[This message has been edited by ricercar (edited 23 March 2002).]

03-26-2002, 03:06 AM
You may have bigger problems than a Mac Video Sense code issue.

Isn't the TAM based on the 6400 or 6500 design? It is my understanding that the 6400/6500 does not work properly with PCI-PCI Bridges. And, being a dual function card, there is a high probability that the R7000 has a PCI-PCI Bridge built in somewhere. So you could be experiencing the 6400/6500 incompatibility with PCI bridges.

I'm not certain about that. I have never personally tested a 6400/6500, and I don't *know* that the R700 has a PPB built in, but that is my hearsay based understanding.

03-26-2002, 05:30 AM
If the tone of this post seems peeved, it's because I'm annoyed at ATI and MacWarehouse. My good humor is strained by hours of unsuccessful testing, and i'm too stupid to go to bed without venting. Advance apologies, Trag; I'm not aiming at you.

Yes, the TAM shares the same motherboard with a 6500 or 6400, I forget. However, the TAM successfully drives two displays as long as neither is DVI, as decribed earlier. If there were two controllers behind a PCI bridge, the TAM would drive neither display. To wit, a generic USB+FW dual function card in the TAM failed to enable either USB or FW functionality despite successful add-in card driver installation.

Another consideration is that ATI's OEM 7500 AGP card uses the same controller of the retail 7000 PCI. ATI cannot afford to cripple AGP performance--even on a low-end controller--with an internal PCI bridge between it and the AGP bus--no way. NVIDIA's TwinView architecture doesn't use a PCI bridge; ATI's dual-monitor technology would not be able to compete with--or prevail in testing--against NVIDIA's AGP cards if ATI AGP bandwidth was limited by a PCI bridge.

Finally, my issue is virtually proven to be a monitor-sense/DDC issue. I came here this morning in the wee hours to report the Radeon 7000 didn't drive the DVI-VGA adapter properly in my 9500, my other 9500, or my Beige G3, none of which have the PCI bridge issue of the TAM/6x00.

I thought computers were supposed to make my life easier. Grumble grumble.

[This message has been edited by ricercar (edited 26 March 2002).]

03-26-2002, 02:49 PM
No offense taken, Ricercar, and, in fact, I appreciate the info that there isn't a PPB in the R7000. I've been wonderign about that but unable to confirm it one way or the other from any source.

That gives hope that the thing might work in the lower slots of the S900/J700, though it still may not.

03-26-2002, 04:49 PM
Does the UMax S900 have the G4 PCI slot issue that the 9500/9600 has?

03-28-2002, 02:02 AM
No, the S900/J700 does not have the G4/lower three slot issue. That issue appears to only occur in machines with dual Bandits.

However, the S900/J700 have a plethora of lower slot problems (any but the top two slots) related to those slots being behind a PPB.

My tests imply to me that the PowerSurge architecture (Powersurge = x500, x600, PTP, PW, S900/J700) has a PCI implementation bug which does not handle more than one level of PPB properly. So the lower slots of the S900/J700 mostly work correctly, but if you install a card with a PPB on board creating a hierarchy of two PPBs the machine will freeze during PCI initialization in most cases. There are some exceptions that aren't worth going into unless you're really interested. This also implies that the PowerSurge architecture is not actually PCI compliant as Apple claimed...unless the PCI-Bridge supplement came later than the PCI 2.0 standard from which Apple appears to have worked.

It sounds like you have (had?) some card designing experience or at least software experience. I'd be happy to read any thoughts you have on the topic.

I really should make the time to read through the PCI Developer's kit from Apple some time. I have a couple of project ideas that would require that knowledge. But I doubt I'm ever going to do them before the march of obsolesence makes them irrelevant. :-)

03-28-2002, 04:50 PM
hello trag... what is a bandit? Is it an MPC106 controller aka Grackle from Moto?

03-28-2002, 11:48 PM
The Bandit is commonly called the PCI controller in the PowerSurge machines. These machines predate the Beige G3 by one generation. I believe that the Grackle (MPC 106) integrates both a memory controller and PCI controller onto one chip. But at any rate, they're definitely not the same. The Bandit was a custom chip design for/by Apple.

In the PowerSurge architecture, the top level bus, the CPU bus, has the CPU(s), the Hammerhead Memory controller, the Bandit(s) and in the 75/76/73/85/8600 machines, a CHAOS video controller.

The Hammerhead arbitrates who gets control of the address adn data busses (memory, CPU, Bandit 1, Bandit 2 or CHAOS) at any given moment. The CPU is the CPU and of course there may be one, two or four of them.

The Bandit is called a PCI controller but it is really more properly called a Bridge. The Bandit bridges data from the CPU bus to the PCI bus. So on one side the Bandit shares a bus with the CPU and memory and operates at your CPU bus speed. On the other side, the Bandit shares a bus with your PCI cards and operates at 33 MHz.

The CHAOS is analogous to the Bandit, but it's downstream side has built-in video circuitry instead of PCI devices.

The Beige G3 has a similar architecture, except the Hammerhead and Bandit is combined into the Grackle chip. There is no Chaos, because the built-in video is just another PCI device on the PCI bus. It's just soldered down, instead of in a PCI slot.

According to Apple's documentation, their design allowed for up to four Bridges on the CPU bus. So, in theory, there could be four Bandits (at least 12 PCI slots) or three Bandits and a CHAOS. It is also true that a Bandit can support more than three PCI slots. On the Apple Network Server, Bandit 2 supports four PCI slots, and Bandit 1 supports six PCI devices, though only two of them are actual slots.

So, in theory, one could build a PowerSurge machine with four Bandits, each supporting six PCI slots (well one of them would need to only have five because of the Grand Central Chip). So that would be 23 PCI slots.

In reality, it's unclear whether the Hammerhead has arbitration lines for four Bandits (does the CHAOS in a 7500 use the same arbitration lines as the second Bandit in a 9500--I don't think so, but am unsure). It's unclear whether there are interrupt lines available on Grand Central (interrupts go through Grand Central) to support additional PCI slots. I don't know enough about Open Firmware to tell whether you'd have to tell the machine that the extra slots exist and which interrupt lines they're connected to. There are Aliases built into the ROM of the PowerSurge machines but they only cover the two Bandits and the normal six PCI slots. It's unclear whether the ROM of the PowerSurge would be able to connect drivers for PCI cards up to the memory space for PCI slots on additional Bandits.

But if all the "ifs" above could be satisfied, one could build a machine with a lot of slots. It'd have slow memory by today's standards, but what a lot of slots.

03-29-2002, 12:14 AM
With thanks. If the Grackle aka MPC106 is a bridge and memory contrller how does this relate to the jumper block at J16 in the Beige G3s (and the B&Ws). Some confusion abounds. Some appear to have suggested that the jumper 5-6 sets the "grackle" (ie Bekkome's site). If true the question becomes "Sets it to do what?". I don't get it... And then there is all sorts of stuff at the Moto site regardeing PLL, VCO and PCI. Again... I don't get it.

All I know is that my voodoo 5500 does funny stuff in my G3 when ASP looks at it.

03-29-2002, 08:49 PM
I don't know much specific stuff about the Beige and the Grackle, but from what little I read I believe those jumpers set various clock speed ratios and the voltage level supplied to the CPU.

For example, PCI runs at 33 MHz. So there may be a 33 MHz clock on the Beige's motherboard. With one jumper setting you might run the CPU/memory bus speed at 2X the PCI speed or 66 MHz. With another setting you get 2.5X or 83 MHz. Similarly for the ratio between CPU speed and bus speed. Various jumper settings might provide a 5X CPU to bus ratio, or 5 X 66 = 330 MHz. Another setting might get you 6X or some such. I'm not actually sure whether the jumper block on the Beige has a CPU ratio setting or whether that's on the ZIF. But I am sure the Grackle has a PCI:CPU Bus speed ratio setting which is on that jumper block.

03-30-2002, 12:41 AM
moved off-topic jumper settings to new thread http://www.macgurus.com/ubb/Forum14/HTML/000151.html

FWIW I'm totally unsurprised each time Apple has PCI bridge issues. They've screwed up _something_ in the PCI implementation nearly every generation since VillageTronics discovered the first-generation PCI macs were incompatible--in certain slots--with my then-employer's bi-endian video controllers.

[This message has been edited by ricercar (edited 30 March 2002).]

03-30-2002, 01:32 AM
OK. Based upon your interpreation of the xlr8yourmac stuff...

7-9 sets the PCI clock spd?

5-6 sets the PCI to system bus ratio?

1-4 sets the PLL? or is this wrong? 1-4 appears to set the multiple/ratio used to determine the cpu clock based upon the selected ratio and the system bus speed.


03-30-2002, 08:37 AM
moved off-topic jumper settings to new thread http://www.macgurus.com/ubb/Forum14/HTML/000151.html

[This message has been edited by ricercar (edited 30 March 2002).]

03-30-2002, 09:54 PM
thnx... i'm real bad at stayin focused...sorry

04-02-2002, 01:11 AM
No problem; I'm a recovering college English instructor. Can't help myself when it comes to organizing the written word...

So I'm buying Gregory's 15" Apple DVI LCD. Looks like I"l be able to put that second external display on my 20th Anniversary Mac after all.

[This message has been edited by ricercar (edited 02 April 2002).]

04-03-2002, 07:44 PM
Just a minor follow-up to this (re: two VGA displays on a Radeon 7000). I just did this on my Yikes / G4/500 box w/ the stock video + a Radeon 7000. I used the DVI-VGA adapter to drive a Hitach Superscan 812 19" + the normal output to drive an SGI monitor and all was fine under OS X 10.1.3...

Worked really well - except that I'm playing with HUGE graphics files (i.e molecular models etc) and having the VRAM halved really screwed up the ability to roll things around quickly.. so I reverted back and now have the Hitachi driven by the Radeon and the SGI under the stock card.

very slick...

Now - this SGI monitor (which I think is a rebadged Viewsonic) is just a CRT but it's seen better days and I need to replace it (it's wiggling))


04-04-2002, 02:14 AM
Chris, if you're just getting some wiggling like near one of the edges of the display, that may be a capacitor that needs replacing. That's fairly easy to do, but the trick, of course, is knowing which one to replace. If you're so inclined, you might take a look around http://www.repairfaq.com. There's a bunch of stuff there on repairing monitors and other things.

Oh, and don't electrocute yourself. :-) Insert standard warning about high voltages present inside monitors and TV sets.

04-04-2002, 07:21 AM
Thanks Trag. A colleague basically gave me this display - it's odd - I can make it "not shake" if I drive it at a higher scan rate...I think it was actually meant to be an RMA to SGI but they (SGI) didn't seem to care about getting it back, so I ended up with it.

I have another 17" Dell nee Sony sitting as an extra head on a PC w/ a Matrox G450 card so I can scam that one if need be

I'll take a peek at that site


04-10-2002, 12:37 AM

I hooked up a DVI and a VGA to my 20th Anniversary Macintosh, and they both operate as expected: Three displays, one 20th anniversary Mac.

*sigh.* Dreams do come true.

Apple System Profiler reports video memory as the expected 32M RAM, and adds "LCD and CRT displays in use." which I have never seen before.

On the devices and Volumes page ASP calls the Radeon 7000 a display card, but on the description it's ATY,DDParent and ATY, RC100Parent, and shows only one display connected. I wonder how the child is connected. The display reported is the DVI, not the VGA.

Problem report:
When I switched the VGA resolution from 1024x768@75Hz to 832x624@75Hz, the Studio Display began to go black for 0.5 seconds in 2 second intervals. I'll follow up here if I find an explanation or cause.