Hey Guys, Have a Quadra 700 with 68MB RAM, 2MB VRAM, and also have an available PowerPC 601 PDS Card. I have a Performa 466 with the 68030 and 32MB RAM to play games that require a 68030 or below and I want to use the Quadra for the more demanding 68K games. Will adding a NuBus video card accelerate graphics in these games at all? Any idea what the difference would be over onboard video with the 2 MB VRAM upgrade? I'm mostly using a M1212 Apple Color Trinitron 13" monitor at 640x480 but I do have higher resolution and screen size VGA monitors that I could use, just prefer to use the matching Apple monitor.
-
Updated 2023-07-12: Hello, Guest! Welcome back, and be sure to check out this follow-up post about our outage a week or so ago.
Gaming performance increase with NuBus Video card on Quadra 700?
- Thread starter 9646gt
- Start date
It really depends what games you are trying to play. Will a graphics card help? Yes, to an extent. It probably wont be night and day difference though. I like playing descent on my q950, which requires the PPC card. To me, there is a noticeable difference between the onboard video and my thunder/24 card. However, on these older machines, processor speed also plays a big role in how well the game plays. I'm looking at doing a crystal oscillator swap on mine since my overclock nubus addon card died.
Thanks for the reply. Mostly FPS, sim games like SimCity and Civilization stuff like that. I too would like to overclock mine down the road but am looking for the clip style solution so I don't have to modify the board.
Byrd
Well-known member
There is no such thing as a faster Nubus card for games; in the day such higher end Nubus cards were designed to make working with high colour depths and resolutions with acceleration bearable. Onboard video is as good as you'll get in terms of raw performance, for the Quadra 700 a PPC card will make Marathon I run well but Marathon II/Infinity Doom, Duke 3D are still pretty terrible on low Mhz PPC CPUs. SimCity, Civilization - should run fine.
When it and the 900 were new, the 700 outperformed every then-extant NuBus video card, so "probably not".
EDIT/Add: Some of the reason for this is because Quadras, and in general all 040 Macs, have their graphics connected directly to the CPU via PDS, and not further out in the topology, like on NuBus.
Your best bet here might be a very late-stage NuBus video card like the Apple/Radius 24AC video cards. Although, as mentioned, that was mostly for things like "I need more than 1152x870" or "I need to accelerate Gaussian Blurs in Photoshop slightly faster (or whatever plugins might use a card's DSP, if it had it.)
In general, in this era VRAM doesn't impact performance per se because 3d and texture memory wasn't a thing. So, more VRAM in the early '90s is mostly about being able to drive bigger displays and use more colors.
I'm not aware of any NuBus cards that had any claim to improving game performance. The Macintosh 630 is said to have contained some stuff that was optimizeable-for, but I don't know if there's a list of what supported it and if it's ever been tested.
EDIT/Add: Some of the reason for this is because Quadras, and in general all 040 Macs, have their graphics connected directly to the CPU via PDS, and not further out in the topology, like on NuBus.
Your best bet here might be a very late-stage NuBus video card like the Apple/Radius 24AC video cards. Although, as mentioned, that was mostly for things like "I need more than 1152x870" or "I need to accelerate Gaussian Blurs in Photoshop slightly faster (or whatever plugins might use a card's DSP, if it had it.)
In general, in this era VRAM doesn't impact performance per se because 3d and texture memory wasn't a thing. So, more VRAM in the early '90s is mostly about being able to drive bigger displays and use more colors.
I'm not aware of any NuBus cards that had any claim to improving game performance. The Macintosh 630 is said to have contained some stuff that was optimizeable-for, but I don't know if there's a list of what supported it and if it's ever been tested.
Last edited by a moderator:
jeremywork
Well-known member
There's an important distinction to make between Color QuickDraw performance and raw bandwidth.
The way Nubus video cards accelerate the desktop experience is by handling QuickDraw instructions in place of the CPU, thereby allowing the CPU to immediately parse the next instruction while the video processor on the Nubus card goes about completing the draw task. In cases where the instructions are relatively straightforward to understand but take a longer time to actually carry-out, the performance is considerably greater. Scrolling through a full-screen finder window in 24-bit color is a good example of this: The CPU only needs to tell the Nubus card about the lowest bit of picture, the one that appeared from the bottom when you scrolled down- otherwise, the Nubus card just needs to take the rest of the window (which is already live in the card's VRAM) and move it the correct number of pixels higher. Since the Nubus card's controller is optimized for video tasks and has a very high bandwidth connection to the card's onboard VRAM, this can be completed much more quickly than a 68040 makes the same change to the Quadra's internal VRAM. As a bonus, the 68040 was able to go back to other computational tasks immediately after delegating the command to the Nubus card, further enhancing the speed of operation.
Where things get interesting is when the screen is being drawn with mostly unique information. Since there's no elegant way to utilize QuickDraw commands to draw a complex image, the 68040 sends a stream of small instructions, essentially manually going through each pixel and relaying its proper value. The Nubus controller can still translate this data to its VRAM with little issue, but the bandwidth shared between all Nubus devices is only 10 or 20MHz, compared to the 25, 33, or 40MHz dedicated connection between the 68040 and its onboard VRAM. As a result, the built-in video is noticeably faster at handling motion video type content.
As a quick test, you can open the 'Jigsaw Puzzle' app, generate a new puzzle with large pieces, and set your monitor(s) to 'Millions of colors.' Piece the puzzle together without letting it snap into the background, and you'll be able to drag the world-map around the frame in real-time.
The redraw movement is visibly smoother on my Quadra 950's built-in-video than the second screen on my Radius LeMans GT Nubus card, which is significantly faster than the 950 in QuickDraw tasks.
In general, most games of the period used 640x480x8bpp to minimize the drawing overhead of full-frames anyhow, so the lack of Nubus bandwidth shouldn't make anything unplayable, but don't expect video to be universally faster even if the card is said to be faster than the built-in video.
The way Nubus video cards accelerate the desktop experience is by handling QuickDraw instructions in place of the CPU, thereby allowing the CPU to immediately parse the next instruction while the video processor on the Nubus card goes about completing the draw task. In cases where the instructions are relatively straightforward to understand but take a longer time to actually carry-out, the performance is considerably greater. Scrolling through a full-screen finder window in 24-bit color is a good example of this: The CPU only needs to tell the Nubus card about the lowest bit of picture, the one that appeared from the bottom when you scrolled down- otherwise, the Nubus card just needs to take the rest of the window (which is already live in the card's VRAM) and move it the correct number of pixels higher. Since the Nubus card's controller is optimized for video tasks and has a very high bandwidth connection to the card's onboard VRAM, this can be completed much more quickly than a 68040 makes the same change to the Quadra's internal VRAM. As a bonus, the 68040 was able to go back to other computational tasks immediately after delegating the command to the Nubus card, further enhancing the speed of operation.
Where things get interesting is when the screen is being drawn with mostly unique information. Since there's no elegant way to utilize QuickDraw commands to draw a complex image, the 68040 sends a stream of small instructions, essentially manually going through each pixel and relaying its proper value. The Nubus controller can still translate this data to its VRAM with little issue, but the bandwidth shared between all Nubus devices is only 10 or 20MHz, compared to the 25, 33, or 40MHz dedicated connection between the 68040 and its onboard VRAM. As a result, the built-in video is noticeably faster at handling motion video type content.
As a quick test, you can open the 'Jigsaw Puzzle' app, generate a new puzzle with large pieces, and set your monitor(s) to 'Millions of colors.' Piece the puzzle together without letting it snap into the background, and you'll be able to drag the world-map around the frame in real-time.
The redraw movement is visibly smoother on my Quadra 950's built-in-video than the second screen on my Radius LeMans GT Nubus card, which is significantly faster than the 950 in QuickDraw tasks.
In general, most games of the period used 640x480x8bpp to minimize the drawing overhead of full-frames anyhow, so the lack of Nubus bandwidth shouldn't make anything unplayable, but don't expect video to be universally faster even if the card is said to be faster than the built-in video.
Last edited by a moderator:
Unknown_K
Well-known member
The better Nubus cards were sold for specific tasks and gaming was not one of them.
http://www.ralentz.com/old/mac/hardware/dale-adams/video-quad700-900.html
Has some interesting information of the Q700/Q9x0 video.
For gaming of that era moving bytes around in RAM is all that was needed (high bandwidth) and built in video was the fastest for that.
http://www.ralentz.com/old/mac/hardware/dale-adams/video-quad700-900.html
Has some interesting information of the Q700/Q9x0 video.
For gaming of that era moving bytes around in RAM is all that was needed (high bandwidth) and built in video was the fastest for that.
ArmorAlley
Well-known member
Don't expect too much from NuBus cards, even on a Q700. I bought a Thunder/24 card so that I could drive 1200x1600 monitor (a nice IBM LCD screen) from my IIfx. I wanted to play Civilization on the biggest monitor it could support. The screen redraw was soooo slooow that I had to reduce it down to 1024x768. I was asking too much of my then 25-year-old computer.
i wouldn't buy the latest and greatest NuBus card. If gaming is what you want, get a card that supports 256 colours at 640x480 or maybe 1024x768 at 256 colours. I think the onboard video of the Q700 supports that, doesn't it?
What might give you a performance boost in speed is a RAM-disk or a SCSI2SD. You have 68MB RAM and you only need about 16MB at most. Install the RamDisk+ control panel, copy across the system folder and game in question to the RAM-disk and boot from it. The contents of the RAM-disk are volatile, so only do a soft-restart. Since all of the software you need is in RAM, the system won't be slowed down by reading or writing to the hard disk.
jeremywork
Well-known member
Yes- the video connector on the card is driven by the RAMDAC reading from the card's VRAM every screen refresh, and likewise the internal video is drawn from the onboard RAMDAC, which pulls data from the expandable VRAM on the logic board. The RAMDAC reads from the VRAM extremely fast- 640x480x8-bit@67Hz is already 20MB/s, and a nicer Nubus' 1152x870x24-bit@75Hz is 215MB/s! (this is why high end RAMDACs and large VRAM frame buffers were expensive.)
Even if you just wanted the Nubus graphics processor write to the onboard VRAM during instructions, it would require interrupting the CPU (the VRAM is tied directly to the pseudo-PDS, not in a DMA channel.) This penalty would almost certainly negate any CPU time saved by the Nubus controller, not to mention the Nubus interface the data would have to move through is much slower than the CPU's VRAM interface. This is why Radius Rockets perform so much better with video cards that support Nubus block transfer, rather than onboard Quadra video. Aside from last generation machines like the 840AV and x100 PowerMacs, Nubus can exchange data significantly faster between two cards via block transfer than any one card can talk to the host CPU.
The one feature that may work with no monitor connected would be the DSPs on Thunder cards. In the context of Photoshop filters, DSPs were implemented more traditionally as logic coprocessors, saving the CPU from having to render the complex output, but still relaying back to the CPU to handle that output. Radius and Supermac both made standalone cards with only the DSPs, too.
Last edited by a moderator:
Trash80toHP_Mini
NIGHT STALKER
QuickDraw Acceleration only does specifics things well: for example it makes scrolling across a large image or illustration in high magnification far faster when using the scroll bars, but if you hit the space bar to use the hand tool for movement as almost everyone did/does, you've giving up the advantage of QuickDraw acceleration entirely.
For gaming, nothing beats built-in video. At around the time of the Q700, NuBus was reaching its limits for Graphic Card performance for anything but graphic design. In the PC world, 16bit ISA had long become a millstone around the neck of graphics performance. This led to the scattershot Local Bus Video phenomenon that culminated in the VESA Local Bus standard which ruled the PC world until the advent of the PCI architecture's break from the new millstone of the 486.
The Mac had been there and done that with built-in Video starting with the IIci. It was terrible in terms of available resolutions, but fast as all get out pushing gaming pixel quantities at gaming color depths. Built-in Video is the equivalent of the later Local Bus Video on the PC, running at processor clock on main, unified I/O bus in the 20-40MHz era.
Don't get me wrong, NuBus cards driving 24bit on multiple 21" CRTs or 20" CRT and Portrait Display with which I started was my world, but it was without any doubt the wrong world in which to be gaming.
_________________________________________
When Apple went half-assed PPC in the NuBus architecture X100 series, passable built-in video was provided on a ludicrous connector. However three frame buffer cards running at CPU clock in the PDS were provided. At that point fairly high resolution gaming would have been practical but for the hard PC-centric tack of game development.
For gaming, nothing beats built-in video. At around the time of the Q700, NuBus was reaching its limits for Graphic Card performance for anything but graphic design. In the PC world, 16bit ISA had long become a millstone around the neck of graphics performance. This led to the scattershot Local Bus Video phenomenon that culminated in the VESA Local Bus standard which ruled the PC world until the advent of the PCI architecture's break from the new millstone of the 486.
The Mac had been there and done that with built-in Video starting with the IIci. It was terrible in terms of available resolutions, but fast as all get out pushing gaming pixel quantities at gaming color depths. Built-in Video is the equivalent of the later Local Bus Video on the PC, running at processor clock on main, unified I/O bus in the 20-40MHz era.
Don't get me wrong, NuBus cards driving 24bit on multiple 21" CRTs or 20" CRT and Portrait Display with which I started was my world, but it was without any doubt the wrong world in which to be gaming.
_________________________________________
When Apple went half-assed PPC in the NuBus architecture X100 series, passable built-in video was provided on a ludicrous connector. However three frame buffer cards running at CPU clock in the PDS were provided. At that point fairly high resolution gaming would have been practical but for the hard PC-centric tack of game development.
Unknown_K
Well-known member
Don't bring the PC into this.
Nubus had a burst speed of 40MBs and 10-20MBs sustained. Nubus 90 (Quadra 840av and PPC era) did 70MBs burst and 30Mbs sustained.
VLB started at 100 MBs for CPUs with a 25 MHz bus, increased to 133 MBs at 33 MHz and 160 MBs at 40 MHz, and reached 200 MBs at 50 MHz. Granted getting cards that worked at 50Mhz was not easy let alone using more then one at a time.
So while the built in video an a Quadra would have a faster frame buffer then an ISA based 486, a VLB based 486 would blow the doors off a Quadra especially if the 486 had cache which they all pretty much did.
Trash80toHP_Mini
NIGHT STALKER
Exactly, though the details of relative speeds between platforms and even the time frame are irrelevant and confusing given the point of my post. I tied to illustrate the limitations of NuBus vs. onboard video by using a general, parallel example of expansion bus limitations for video. Sorry if that was confusing.
A better example of this might be the SE/30 and IIsi with their PDS based video expansion cards vs. NuBus expansion. Most 030 PDS PDS VidCards "lack" QuickDraw Acceleration which is seen here and on LEM as a major drawback vs. the many NuBus solutions that do at higher resolutions and greater color depths. Running at processor speeds of 16MHz and 20MHz, the low resolutions and color depths supported, PDS VidCards don't really have a pressing need for "acceleration" as is also the case for onboard video.
Much of the confusion on this issue is based on LEM's misconceptions about QuickDraw acceleration and suitable applications for it.
Wow, great information from the replies everyone. I'm a hardcore Mac guy these days, but I was straight PC in the dos and early Windows days (was all we could afford really). So I learned a ton from what you guys have said 
Thank you for all the help. You guys are a wealth of knowledge.
Thank you for all the help. You guys are a wealth of knowledge. Similar threads
