Do any G4 ZIF upgrades run @ 100Mhz bus?

[Byrd]

Hi,

I've a Blue and White G3 I'm fixing up with the aim of installing a G4 ZIF CPU, I've two models to try

- Sonnet Encore G4 ZIF 400/1M (I think it's auto sensing bus/multiplier) - currently in a beige G3

- Interware Booster G4Z 400/1M (with mystery SW1 and DIP switches I can't find documentation on)

From what I've read it seems pretty much all G4 ZIF upgrades (apart from the later 800+ Mhz G4 upgrades) slow down the bus speed from 100 to 66Mhz in B&W G3s due to "incompatiblity issues"; has anyone gotten theirs to work at 100Mhz bus by manual motherboard adjustments? I'd be a happy chap to get one of the above upgrades to 4.5 x 100Mhz on a B&W.

Thanks

JB

 

[Bolle]

I think you got it the wrong way around.
The Sonnet 700MHz+ upgrades slow down the bus (I think due to the weird CPLD setup on the bus pins of those upgrades). Ever noticed those are not keyed and can be plugged in in two directions depending if they’re installed in a B/W or Beige? There’s additional logic on there to move around all the signals on the pin array accordingly.

The slower upgrades work just fine at 100MHz bus speed. Running one with a 533Mhz rated 7410 chip at 600MHz in my B/W.

 

[macuserman]

I think you got it the wrong way around.
The Sonnet 700MHz+ upgrades slow down the bus (I think due to the weird CPLD setup on the bus pins of those upgrades). Ever noticed those are not keyed and can be plugged in in two directions depending if they’re installed in a B/W or Beige? There’s additional logic on there to move around all the signals on the pin array accordingly.

The slower upgrades work just fine at 100MHz bus speed. Running one with a 533Mhz rated 7410 chip at 600MHz in my B/W.

Bus speed is set by the jumpers correct?

 

[Gorgonops]

Ditto Bolle’s comment. I also vaguely remember there was some super-fast upgrade that came with the weird bus downclocking limitation, but more normal G4 ZIFs run at 100mhz. The G4 upgrade in my B&W runs at 100x4, same as the original G3. (I think it’s actually just Yikes! ZIF module, it’s physically identical to the G3 ZIF.)

 

[Franklinstein]

Yeah basically if a ZIF upgrade has an L3 cache it will downclock a B&W/Yikes! G4's bus. Otherwise the 7400 or 7410-based upgrades should work fine at 100MHz.

 

[macuserman]

Hmm I need to check my powerlogix ZIF and see if it is downclocking or not.

 

[Byrd]

Thanks all for the advice - had some good success.

Installed the PowerLogix G4 Firmware update as it appeared to be reversible if I had problems. Patch went through fine and CPU Director 1.5 was next installed, vanilla OS 9.1.

Pulled the 350Mhz CPU and set the jumper block to 4 X 100Mhz; got no chime with the Sonnet or Interware G4 400Mhz CPUs. Cleaned CPU socket with contact cleaner, put OEM jumper block in place (3.5 X 100Mhz) and the Sonnet G4 chimed @ 400Mhz, cache enabled with CPU Director. Tried to up the multiplier to 4.5 but typical pedestrian Sonnet upgrade kept at 400Mhz, but 100Mhz bus as hoped.

Put in the mystery Interware G4 - no chime - tried some SW1 DIP switch settings, nothing then to up up down down and it sprang to life at 500Mhz, 100Mhz bus backside cache 200Mhz - crashed at 250Mhz.

I’m pretty happy with that, system is working well. Goes to show you don’t need to pull that jumper block as the CPU can also control multiplier settings.

 

[Byrd]

More testing, some really odd software glitching with the G4 400 @ 500 (I was admittedly pushing it, early G4s don't overclock well), some more random fiddling with the DIP switch and it's rock solid at 450Mhz, 225Mhz cache. Seems a fast little Mac.

 

[macuserman]

More testing, some really odd software glitching with the G4 400 @ 500 (I was admittedly pushing it, early G4s don't overclock well), some more random fiddling with the DIP switch and it's rock solid at 450Mhz, 225Mhz cache. Seems a fast little Mac.

Nice that’s one of my favorite machines. What else do you plan to do with it?

 

[Byrd]

Nice that’s one of my favorite machines. What else do you plan to do with it?

Thanks - I'd like it to be an OS 9 only gaming machine. It's cleaned up well so far. Just found my old Radeon 7000, and Apple 100Mhz Pentium PC compatibility card. Possibly ... the G4 might be able to emulate a 100Mhz Pentium, but probably not that well so actual hardware might be fun.

 

[macuserman]

Thanks - I'd like it to be an OS 9 only gaming machine. It's cleaned up well so far. Just found my old Radeon 7000, and Apple 100Mhz Pentium PC compatibility card. Possibly ... the G4 might be able to emulate a 100Mhz Pentium, but probably not that well so actual hardware might be fun.

Nice that’s super fun that’s exactly what I plan for mine but I need to find some games that will actually make this thing break a sweat.

 

[Byrd]

Nice that’s super fun that’s exactly what I plan for mine but I need to find some games that will actually make this thing break a sweat.

It's the sort of Mac that needs a few cheap upgrades to make it sing. The B&W stock Rage 128 is Apple bargain basement designed for just the most basic 3D acceleration; I'm interested to see how the Radeon 7000 fares in it. In my search I found two Radeon 7000 Mac cards with same part nos! One is VGA only with 4ns RAM and the other DVI/VGA/Svideo with 5ns RAM they look the same apart from this. I'll use the VGA only one in the B&W. Any games involving the Unreal or Quake II engine should be decent, Quake III and above I'm keen to see.

Also have a 8MB 3DFX Voodoo 1 card (Mac version not PC); also will try and see if any better, I think games will look better using Glide but not as fast, perhaps.

 

[macuserman]

It's the sort of Mac that needs a few cheap upgrades to make it sing. The B&W stock Rage 128 is Apple bargain basement designed for just the most basic 3D acceleration; I'm interested to see how the Radeon 7000 fares in it. In my search I found two Radeon 7000 Mac cards with same part nos! One is VGA only with 4ns RAM and the other DVI/VGA/Svideo with 5ns RAM they look the same apart from this. I'll use the VGA only one in the B&W. Any games involving the Unreal or Quake II engine should be decent, Quake III and above I'm keen to see.

Also have a 8MB 3DFX Voodoo 1 card (Mac version not PC); also will try and see if any better, I think games will look better using Glide but not as fast, perhaps.

Nice, I have a:
Radeon 9200
Radeon 7000 DVI/VGA
Radeon 7000 VGA
Formac ProForman3
Voodoo 5500 Mac Edition

I’m currently running the 9200 but I’d like to experiment and see how the different cards stack up against each other with different games. I’ve pretty much maxed the machine out in all the ways possible. Although I’m still on the hunt for a gthree serial port for it just for kicks not because I need it.

 

[macuserman]

Nice, I have a:
Radeon 9200
Radeon 7000 DVI/VGA
Radeon 7000 VGA
Formac ProForman3
Voodoo 5500 Mac Edition

I’m currently running the 9200 but I’d like to experiment and see how the different cards stack up against each other with different games. I’ve pretty much maxed the machine out in all the ways possible. Although I’m still on the hunt for a gthree serial port for it just for kicks not because I need it.

Sorry if that was to much info it’s just that deciding on the which card is going to be the very best for my setup is sort of my last missing piece. Didn’t mean to hijack your thread.

 

[Byrd]

No that's cool @macuserman - I love a good GPU discussion I enjoy matching GPUs with the right systems, be it PC or Mac, and putting them through their paces. It would be interesting to see the Radeon 9200 benched against the Radeon 7000.

I've also 2 x Voodoo 2 cards (there was this seller in Russia IIRC 7-8 years ago who sold these dirt cheap on eBay), but from what I can gather SLI is terrible in Mac OS - so they have gone into a proper Pentium III PC.

Accelerate Your Mac! - 3Dfx Voodoo2 Drivers SLI Game Performance page!

Accelerate Your Mac's 3Dfx Voodoo2 SLI Game Performance page!

www.xlr8yourmac.com

 

[macuserman]

No that's cool @macuserman - I love a good GPU discussion I enjoy matching GPUs with the right systems, be it PC or Mac, and putting them through their paces. It would be interesting to see the Radeon 9200 benched against the Radeon 7000.

I've also 2 x Voodoo 2 cards (there was this seller in Russia IIRC 7-8 years ago who sold these dirt cheap on eBay), but from what I can gather SLI is terrible in Mac OS - so they have gone into a proper Pentium III PC.

Accelerate Your Mac! - 3Dfx Voodoo2 Drivers SLI Game Performance page!

Accelerate Your Mac's 3Dfx Voodoo2 SLI Game Performance page!

www.xlr8yourmac.com

Hmmm that's an interesting read for sure, makes me wonder based on his comments if indeed a faster processor would have allowed the SLI to really work well. I have a 1.1Ghz G3 processor in mine and he only had a 500 I think. I don't have any of those older 3dfx cards that actually work though I have one but it doesn't display anything and I need to look into fixing it. Sorta think that my 5500 would still be faster though even without SLI but would be fun to compare for sure.

 

[Gorgonops]

It's the sort of Mac that needs a few cheap upgrades to make it sing. The B&W stock Rage 128 is Apple bargain basement designed for just the most basic 3D acceleration; I'm interested to see how the Radeon 7000 fares in it. In my search I found two Radeon 7000 Mac cards with same part nos! One is VGA only with 4ns RAM and the other DVI/VGA/Svideo with 5ns RAM they look the same apart from this. I'll use the VGA only one in the B&W. Any games involving the Unreal or Quake II engine should be decent, Quake III and above I'm keen to see.

So I can speak on the subject of a Radeon 7000 verses the stock Rage 128: long and short of it is you'll probably be disappointed. The two cards are actually very much in the same ballpark when it comes to raw OpenGL triangle-ing. The 7000 *is* faster, but we're talking like, I dunno, 40-50% faster. (This figure being roughly what I remember from Quake I timedemos; note this was using an open-source OS X Quake engine, not OS 9, maybe your mileage will vary.) It's way slower than a "real" Radeon. The RV100 chipset has only a single pixel pipeline and ditches both hardware transform and lighting and HyperZ. The main advance it offers over the Rage 128 is it supports arbitrary texture dimensions instead of the power-of-two limitation of the older hardware; this is why a Radeon 7000/Mobility is the least powerful graphics card supported by Quartz Extreme.

Anyway, this was on my 400mhz G4-up/sidegraded B&W. The card was actually the stripped OEM one with just a VGA port that Apple used in Xserves, but I don't think that's going to make any difference, the variations between the cards were to my knowledge just in which video ports they soldered on. (There's empty pads on the board for the DVI and Svideo bit.)

 

[Coloruser]

it´s no wonder that the Radeon 7000 PCI doesn´t feel much snappier than the Rage 128. AFAIK, the Radeon 7000 was the lowest end of the Radeons - a 64bit interfaced card versus the 128bit of the Rage 128/Rage128 pro. It clocked higher and had DDR instead of SDR VRam to overcome the limited Chip bandwidth for a slightly higher 3D fillrate. The 7200 and the 7500 were also available as PCI versions. With their DDR 128bit memory interface and 3 times the fillrate, they run circles around the 7000.

I loved to use the 7500s in my testbeds.....

 

[EtherRad]

Hello,
Just wanted to provide an update on the ZIF G4 1GHz. I received a response from one of the original lattice engineers:

"Wow definitely some old devices, I remember specifying this family back in 1998 Thoughts:

1. Have attached the datasheet of the relevant device family. That and more info is available here https://www.latticesemi.com/en/Products/FPGAandCPLD/ispMACH4000VZ
2. Page 76 of the datasheet has the decoder for the part number. “-35” refers to the delay through the part over the commercial operating range. (3.5 nS) and “-5I” is the delay through the part over the industrial operating range. (5 nS)
3. Don’t have any specific knowledge on the accelerators that you refer to. However at 100MHz the time available between sending something on the bus and it being received is 10ns. Adding 3.5nS of delay would required slowing the bus to 75Mhz to accommodate. Three possible reasons come to mind as to why the original designers chose 66MHz not 75MHz:
   1. Actual path through the part based on the logic design is longer than 3.5nS.
   2. Addition of timing margin/safety factor.
   3. Only discrete frequency steps available for the bus speed.
4. For the part used “-35” is the fastest part so no easy update.

Hope this helps."