Daystar PowerCache Quadra

[Fizzbinn]

I was fortunate enough to pick up a Daystar PowerCache Quadra and have just started to dig into it. I've long wondered if L2 Cache would improve performance on 68040 machines as much as it does for 68040 and PPC ones.

From the above linked archived data sheet:

FastCache Quadra 128K Static RAM Cache for Quadras and Centris 610​

​

DayStar’s FastCache Quadra increases performance 20 to 40% in almost any Quadra and the Centris 610. A low cost cache card, the award winning FastCache Quadra puts the “zip” back in your Mac—it even makes a Quadra 700 or 900 run as fast as a Quadra 950! From desktop publishing to spread- sheets or database software, it will make your machine run faster and more efficiently, saving you time and money.​

I lucked into the 33Mhz rated version:



I plan test it out in my 25Mhz Quadra 700, Centris 650* and Quadra 610. The Quadra 610 requires an adapter board, I believe I can use the one from my 486 DOS compatibility card. I also have an untested 20Mhz Centris 610 logic board that this will give me an excuse to try. I don't have a 33Mhz 68040 Mac that has a compatible PDS slot, but am considering modifying my Centris 650 to be a Quadra 650 (clock chip and resistor mod).

*Spoiler, it locks up during start up in my Centris 650 when loading the QuadraControl v2.3 control panel under System 7.1 and 7.5. Hmmm, looking at the above linked datasheet the Centris 650 is not mentioned... only the Centris 610: "almost any Quadra and the Centris 610". Seems very odd but perhaps more reason to try the mods to make it a Quadra 650.

Will post my Quadra 700 test results shortly.

 

[Fizzbinn]

Working on MacBench and Norton System Info results next...

 

[jmacz]

Lucky! Been looking for one of those! Gratz!

 

[lobust]

This is interesting.

I have one of these, but as discussed here, in a 800 the difference it makes is so marginal I wasn't even sure it was working. I don't remember if I ever tried it in my 700 as that has a PowerPC card in it, but looking at your results it seems that it would be far more useful in a 700...

Edit: Also interesting that it locks up in your Centris 650, since I can confirm no such issues on the Quadra wombat. I would have assumed the differences were way too minor to matter - it can't be CPU related as the 610 has an LC'040 too. Curious...

It's not just got some incompatible data saved in the slot pram from a previous PDS card? You tried clearing the PRAM when testing?

 

[beachycove]

I looked for one of these to no avail on eBay for years several years back, so I’d love to know what I am missing. Keep the benchmarks coming!

 

[Unknown_K]

I almost snagged one of those off LEM Swap back in the day, but someone beat me to it (was like $5 back then).

If the code you are running fits into the cache it speeds thing up quite a bit.

 

[Fizzbinn]

I looked for one of these to no avail on eBay for years several years back, so I’d love to know what I am missing. Keep the benchmarks coming!

It just came up when I manually checked an eBay saved search I have for "Daystar Digital". It was listed as "daystar digital 105013-109 33 Mhz Card 01-FC040-001", BIN $35. Super lucky timing I suppose as I'm sure lots of people have similar saved searches/alerts set up.

Other then the printed DayStar Logo on the front of the card this info is on the back in PCB solder mask text:

DAYSTAR DIGITAL​

COPYRIGHT 1992​

MADE IN THE U.S.A.​

​

01-FC040-001​

along with two stickers:

105013-109​

​

33 MHZ​

There are pictures of the 25Mhz version in this reddit thread, from an identification prospective everything the same except its stickers:

104316-1​

​

25 MHZ​

Someone trying to sell one of these is unlikely to post it as a DayStar "PowerCache Quadra" unless they do some digging.

Working on additional benchmarks!

 

[Fizzbinn]

Edit: Also interesting that it locks up in your Centris 650, since I can confirm no such issues on the Quadra wombat. I would have assumed the differences were way too minor to matter - it can't be CPU related as the 610 has an LC'040 too. Curious...

It's not just got some incompatible data saved in the slot pram from a previous PDS card? You tried clearing the PRAM when testing?

I don't think I tried clearing the PRAM, will try that tonight.

I do think it could just be a quirk of memory timings or something specific to the 25MHz Centris 650, the other Wombat's that apple shipped are 33MHz. And as mentioned above the 1993 product data sheet leaves out the Centris 650, while including the equally old Centris 610 and the newer Quadra 650/800 (Wombat based):

System Requirements​

– Apple® Macintosh Quadra 610, 650, 700, 800, 900, 950 and Centris 610​

– Quadra/Centris 610 require Adapter 610 for operation​

 

[Fizzbinn]

On my Centris 650, clearing the PRAM made no difference, neither did deleting the QuadControl System Preference file that I found in another thread as a suggestion.

Back on the Q700 doing some baselines in Norton System Info tests before putting the FastCache back in and I’m suddenly getting 10% lower video scores than I did in prior stock baselines… ugh. Apple Personal Diagnostic reports the VRAM test is failing. I do have 6 VRAM expansion SIMMs installed and when I got the system years ago I recall having video issues that resolved with cleaning the VRAM SIMM edge connectors. Guess I’m now working on that again!

 

[jeremywork]

Did you reset PRAM on the 700 in between the tests? 24-bit addressing may have been re-enabled and that will impact scores.

 

[eharmon]

I wouldn't be surprised that it's a lot more effective on a 700, as the Wombat memory systems are significantly better so there's less of a bottleneck. The 800 especially has tighter memory timings than the other Wombats to match its 60ns memory.

Sadly I can't find the data sheet for those static RAM chips, but I assume they're 15ns rated based on the label.

Neat benchmarks!

 

[Fizzbinn]

Did you reset PRAM on the 700 in between the tests? 24-bit addressing may have been re-enabled and that will impact scores.

<sheepish grin> After removing all the VRAM SIMMs and testing adding 2 of the 6 expansion SIMMs back at a time they all checked out. That's got to be the mechanical contacts, but the 10% performance drop remained ....and then I saw your post. <forehead smack>. Thanks @jeremywork!

Running MacBench 4.0 now!

 

[Fizzbinn]

Speedometer 4.02 results are interesting:

 

[Fizzbinn]

And Norton System Info 3.21:





\

 

[Fizzbinn]

For Mac Bench 4.0 I played with the QuadControl control panel "Quadra Boost" ROM in RAM setting, its seems the "MAX" setting (MQB below) provided some increased benefit with the FastCache Quadra over the default "Good" setting (GQB below) but not so much without the card installed.

 

[beachycove]

Well that sure is disappointing!

 

[David Cook]

These results are consistent with what I found earlier this year.

ROM-in-RAM for Quadra performance boost

I've long wondered, why doesn't a ROM-in-RAM utility for the Quadras exist? It was common for 030 and 040 accelerators, and the 68k AV machines had it natively (enabled with a PRAM switch, which AV Turbo ROM can flip for you). (I know you're thinking -- is it because performance changes are...

68kmla.org

In some specialized situations, such as running a compiler, the L2 cache can boost performance by 30%. But, in most circumstances, it isn't very effective.

It seems like in most cases, the larger 68040 internal L1 cache (compared with the IIci's 68030) was big enough for most classic applications. So, adding a cache card to a 68040 computer just didn't have the same impact as it did on a IIci (or as built in on the IIfx).

There's a hint about this being the case, as my performance testing on 32KB vs 64KB vs 128KB cache cards on a IIci showed little difference.

Mac IIci & cache card questions

Speaking of... https://www.ebay.com/itm/134167942890 $10 start... Will be interesting to see where it lands. 🤔

68kmla.org

This suggests that maybe even a '4KB' IIci cache card would have been effective.

There is another possibility. The Macintosh OS has a routine called 'BlockMove' that was often called by the memory manager, OS, and applications. If you move more than 12 bytes of memory, it clears the instruction cache. They introduced a fix in System Update 3.0, but it is 'off' by default in the sense that the calling application needs to set a flag to avoid clearing the cache. This is a long way of saying that the Mac OS may be clearing the instruction cache far too often for even 32KB to make a difference.

On the PowerPC, code is handled differently, so larger caches on those machines would be much more effective.

 

[slomacuser]

Nothing to add just a little price list