WarpSE: 25 MHz 68HC000-based accelerator for Mac SE
- Thread starter Zane Kaminski
- Start date
I tested it today but found HeapTool 1.4 doesn't resolve the low memory issue that triggers the error message in SCSI Director Pro 4.0.
The author of HeapTool made it for System 6 and says it "probably" wouldn't have any meaning under System 7 due to the way 7 handles the heap in a much better way. But he left the door open for the possibility it might help. Even so, after getting the memory error in SCSI Director Pro, I quite the app and opened the HeapTool control panel and clicked the PURGE button. All that does is create a little more white space at the far right of the bar for System Software. The total RAM allocation doesn't change. So in the case of your screenshot above, the 1,602K would not change, and the Largest Unused Block number would not change after clicking the PURGE button. It only clears out some RAM that is already allocated for System Software, which SCSI Director Pro cannot touch. That's why after clicking PURGE and then relaunching SCSI Director Pro 4.0, the same memory error occurs.
I am guessing it's a coding quirk of SCSI Director Pro 4.0. Even though I have given it 1024K in the Get Info Box, if it sees the Largest Unused Block (of RAM) to be less than about 1500K, then it complains it is out of memory (after the initial tests complete, just before the graphing tests begin). So long as I can keep the Largest Unused Block higher than about 1500K, then SCSI Director Pro 4.0 is content to perform all the tests, including the graphing tests.
Anyone can simulate this problem easily, without fiddling with INITs by trial and error. Just increase the Disk Cache in the Memory control panel to be 384K instead of the default 128K. That usually will drop your Largest Unused Block of RAM to below 1500K, thereby triggering the out of memory issue in SCSI Director Pro 4.0 (just before the graphing tests begin).
And for whatever reason, no, adjusting the Minimum and Preferred settings in the Get Info box for SCSI Director Pro 4.0 doesn't solve the problem. I think it's a bug in the app, ignoring those settings. The only way to prevent the error is like I said. You have to increase the Largest Unused Block of RAM to above 1500K or so.
I don't know a huge amount about the SuperMac cards - they look pretty cool. I'd be interested to see some benchmarks if you get a chance (<cough>Norton System Info</cough>).
Regarding this thread and other projects - when you're working on your own or in a small team on something like this it sometimes takes time. Life gets in the way, or a barrier to progress is frustrating in a way that you just... leave it for a while.
Things only stay fun if you can put them down and walk away for a while and it doesn't matter, if there is a pressure to deliver, it can take the joy out of things.
Its been cool to see the development documented here! If it results in a new, fast, awesome accelerator, all the more awesome
I should be getting the SuperMac SpeedCard in a few days. I will try and get benchmarks posted soon.
I got the SuperMac SpeedCard installed and benchmarked using SpeedOMeter 3.23. I am having issues taking pictures of my screen so I just summarize the results.
CPU: 233% increase (2.33 Index)
Graphics: 75% increase (1.75 index)
SCSI: No real difference (I was getting 1.5 and now around 1.6 index)
Math: 445% increase (4.45 index)
KWhetstone: 85K
No FPU is detected using any software I have tried. That is to be expected as I recall a 68000 can only treat a 68881/2 as a peripheral device. All math performance increase is simply by intercepting SANE calls and routing through the 68881. I am using the v1.7 of the SpeedCard Control Panel that I downloaded from the Macintosh Repository. Overall a nice performance boost. I think would get better I/O performance with a Brainstorm accelerator, but since I am using an external BlueSCSI v2, that is not really an issue.
CPU: 233% increase (2.33 Index)
Graphics: 75% increase (1.75 index)
SCSI: No real difference (I was getting 1.5 and now around 1.6 index)
Math: 445% increase (4.45 index)
KWhetstone: 85K
No FPU is detected using any software I have tried. That is to be expected as I recall a 68000 can only treat a 68881/2 as a peripheral device. All math performance increase is simply by intercepting SANE calls and routing through the 68881. I am using the v1.7 of the SpeedCard Control Panel that I downloaded from the Macintosh Repository. Overall a nice performance boost. I think would get better I/O performance with a Brainstorm accelerator, but since I am using an external BlueSCSI v2, that is not really an issue.
Thanks for logging this!
It'll be exciting to compare it to the WarpSE once it's out.
I will try and get full results this weekend and posted with pictures using SpeedOMeter 3.23.
I realised after my post that System Info doesn't actually work in System 6, doh!
Anyway, pretty serious speed increases. The graphics must be amazing.
Do you have a PGA 68882 in anything? They should be a direct swap for the 68881, but give you a good bump to the FPU performance. It would be an interesting experiment if you have one in an SE/30, or a IIx or similar. Given you only need a 16MHz one they're also pretty common on ebay.
I do have a PGA 68882 in couple of things. I may give it a try someday, but based on past experience, the increase would not be that great. I have tried both a 68881 and 68882 in the FPU card on Macintosh LC. The speed difference was 5-10% at most, depending on the benchmark. When I post some more detailed scores this weekend, i will test System 6.0.8, 7.01 and 7.1 for comparison.
I posted a thread about the SuperMac SpeedCard along with the results here: