NuBus video cards Benchmarks - SuperMac, Radius, E-Machines, RasterOps

[slomacuser]

Here are some results of my tests, that I did with Norton System Info 3.5.2 (1997). I have performed tests with Macintosh IIci with 64k Daystar FastCache card and System version 7.0.1. The cards in test were SuperMac Thunder II GX, Radius Thunder GT, PrecisionColor 24x, Radius Pivot, E-Machines Futura SX, RasterOps 24STV.











The second test were made with Quadra 700 and 950 computers. What I found is that the performance also depends of System software installed. The 7.1.2 or 7.5.5 perform better results than system 7.6.1. Interesting results, NuBus cards outperform Built-in video when using Milions of colors (32-bit). Power PC card in Q700 slows down the video so SpeedDoubler is essential software to have with it.

 

[trag]

Nice.

Just for clarity, is that the Precision Color or the Precision Color Pro?

Also, Futura SX or Futura II SX?

Lowendmac is frustrating. Or my memory is bad. The Thunder GT was marketed by Radius as the "Thunder 24GT". IIRC.

The "Precision Color" card in LEM's description is the Precision Color Pro. 7" card. Non Pro was longer, I believe.

The "Futura SX" card in LEMs description really is the Futura SX and not the Futura IISX. Futura II had no rotary switch and was a 7" card.

It'd be nice if they were consistent about including or ignoring the updated cards. Sigh.

 

[slomacuser]

Nice.

Just for clarity, is that the Precision Color or the Precision Color Pro?

Also, Futura SX or Futura II SX?

Lowendmac is frustrating. Or my memory is bad. The Thunder GT was marketed by Radius as the "Thunder 24GT". IIRC.

The "Precision Color" card in LEM's description is the Precision Color Pro. 7" card. Non Pro was longer, I believe.

The "Futura SX" card in LEMs description really is the Futura SX and not the Futura IISX. Futura II had no rotary switch and was a 7" card.

It'd be nice if they were consistent about including or ignoring the updated cards. Sigh.

Well the card has no identification of PRO marking in hardware or software. It is identifed as PrecisionColor but as it is a 7" card I guess it is Pro.

The Futura is SX not the II. It has rotary switch and it is 12" card.

Yes the Thunder GT is identified as Thunder 24 GT.

 

[derFunkenstein]

If I may add to your data point, I set up my IIci as close to yours as I can - mine has the Apple cache card rather than the Daystar, but it's running 7.0.1. I have 8x1 MB SIMMs installed, so 4MB in both banks.

My video card is an Interware GrandVimage 24-16. At 640x480 8-bit color I get a 37.1, so just a touch slower than the Radius GT but a very worthwhile upgrade from the RAM-based video chip that uses system memory. My built-in video scores dead on with your 26.0.

It's amazing how badly these RBV Macs are hampered by using system memory. Overall system was measured ~35% slower, not just video.

 

[jeremywork]

It's amazing how badly these RBV Macs are hampered by using system memory. Overall system was measured ~35% slower, not just video.

The popular mitigation is to populate the IIci's bank A with 4 x 256k SIMMs, then set the disk cache (or use RAM-Muncher) to allocate the first 768k (or more if you plan to use video modes which use less than 640 x 480 x 8-bit = 307.2k) so that only the disk cache needs to share bandwidth with the video subsystem; all other system and application memory is serviced by bank B at full speed. The IIsi has 4 x 256k soldered to the motherboard, so the same approach works there.

 

[derFunkenstein]

Hmm, never thought of that but it makes sense that if you make the memory configuration mirror the IIsi that the IIsi RAM muncher app would do what it needs to. I'm going to go try that.

 

[Unknown_K]

The expensive cards shine at high resolution and color depth (basically needs more VRAM then configured on the motherboard), Quadra video bandwidth just has too much of an advantage not having to run over the Nubus interface.

 

[derFunkenstein]

The popular mitigation is to populate the IIci's bank A with 4 x 256k SIMMs, then set the disk cache (or use RAM-Muncher) to allocate the first 768k (or more if you plan to use video modes which use less than 640 x 480 x 8-bit = 307.2k) so that only the disk cache needs to share bandwidth with the video subsystem; all other system and application memory is serviced by bank B at full speed. The IIsi has 4 x 256k soldered to the motherboard, so the same approach works there.

This did end up helping quite a bit. CPU and FP performance came up to match where it was with my Interware installed. The GrandVimage is still a faster solution than RAM-based video, though.