Daystar Universal PowerCache P33 in SE/30

[K Trueno]

Hi, I recently installed a Universal PowerCache in the SE/30 but the "PowerCache" portion doesn't seem to be doing anything.  Toggling "PowerMath" seems to increase or decrease performance in Speedometer, but toggling "PowerCache" seems to have no effect (with or without restarts) in between.  I've tried empty Control Panel and Extensions and that sort of thing.  

Using Speedometer I'm getting around 9 for CPU when it should be around 12..

Running the DayStar PowerDemo shows the line "PowerCache external cache is disabled".  I can enable it through the menu, but it seems to do nothing, and once I put the Demo App in the background and come back it seems to toggle back to disabled automatically.

I'm running System 7.1, SE/30, stock rom.  Any thoughts?  Or could it be a broken PowerCache?

Thanks!

 

[K Trueno]

Darn outside of edit window.  Using a TwinSpark btw.  I have tried version 1.3, 2.0, and 2.2.  Control Panel is loaded, with some validation as PowerMath toggle works and affects the PowerStar demo and Speedometer.  

Command-Control clicking the PowerCentral logo shows everything as it should as far as I can tell. 

If you have one working, may I ask what behaviour is with Command-T, toggling the PowerCache and the status message is when you put the application in the background?

One more clue, this one did not come with an on-board FPU, socket is empty.  It shows 50mhz 68030 and 16mhz 68882 in the status window of PowerStar demo.  

 

[ktkm]

I believe it is the Twinspark. I had a similar problem that the one you describe but with a P34 and a TwinSpark II, with and without the onboard FPU. I had no trouble using the same PowerCache in an IICX. Is the System stable after 10 minutes? Mine wasn’t. 

 

[K Trueno]

Hmm interesting!  As far as I can tell it was running stable for the duration I was testing it, but I was swapping around Control Panels/Extensions a lot and rebooting often.  I will give it a try again tonight.  

Hopefully it is a software problem and not a hardware problem.  How does the PowerCache know to enable or disable the L1 cache on the card?  Does it look for for specific addresses and intercepts them before it goes out onto the bus?  Is that in SW or HW?

Any chance that an FPU is required for the L1 Cache to be enabled?

 

[Bolle]

The cache is usually on by default. It will work without an FPU on the PowerCache.

All the cache logic is done in hardware. Actually I am not totally sure how it is enabled/disabled. Probably the software calls a specific address which flips a bit in one of the registered GALs.

Not sure if I have seen a 50MHz P34 yet, are you sure it is running at 50MHz? If so what are the speed ratings on the CPU and SRAM/Tag chips?

 

[ktkm]

Not sure if I have seen a 50MHz P34 yet, are you sure it is running at 50MHz? If so what are the speed ratings on the CPU and SRAM/Tag chips?

Sorry, forgot to mention the P34 was 33MHz not 50MHz.

 

[K Trueno]

To be fair, I'm only comparing numbers of Speedometer that I've seen from others/records file (~9 vs ~13) vs. my own accelerator, the PowerStar Demo status output (always says "external cache disabled"), and that the toggling the PowerCache toggle in the control panel has no effect on those numbers.

Does anyone know if the toggle work instantly btw?

Probably the software calls a specific address which flips a bit in one of the registered GALs.

Interesting.. thats probably it.  I noticed this in a patch for NetBSD for mac68k:

+inline void
+enable_daystar(void)
+{
+ disable_daystar();
+ printf("Enabling DayStar External Cache...\n");
+ asm("tstb 0x52070000 | enable daystar external cache");
+ asm("tstb 0x52050000 | enable writeback cache");
+ flush_daystar();
+}




Maybe going too deep.. but could a GAL fail?  The motherboard is recapped but not the PSU or Analog board too, could bad voltages leave the cache off?

 

[ktkm]

The motherboard is recapped but not the PSU or Analog board too, could bad voltages leave the cache off?

I think so, at least with the Turbo 040. I tried two different PSU:s for the P34, the Sony stock configuration with no re-cap, and then when I got desperate, I went off and bought the Enhanced Power Supply from Artmix. Made no difference for the cache part, but the screen looks great!

 

[K Trueno]

Cool it made the screen better?  Brightness or do you mean like stability?

If I may ask, do you have benchmarks from your SE/30 setup?  Do you recall the behaviour of the PowerCache toggle and the PowerCache demo?

 

[Bolle]

GALs can theoretically go bad but I am 99% sure your card won't work at all if one of the GALs was bad.

Score in Speedometer 3 is around 9-ish for me as well. PowerDemo behaves just like it does for you.

Enabling/Disabling the external cache doesn't change the CPU benchmark score in Speedometer either. That doesn't mean a lot though as long as we don't know what exactly that benchmark is doing.

The routine probably fits into the internal caches of the 030, so the external cache isn't even used.

 

[K Trueno]

GALs can theoretically go bad but I am 99% sure your card won't work at all if one of the GALs was bad.

That would make sense.  Glad its not the hardware 

Score in Speedometer 3 is around 9-ish for me as well. PowerDemo behaves just like it does for you.

Cool thanks for confirming!   Maybe it is just how the Universal PowerCache behaves in an SE/30.

Because, from @JDW's video comment (Socketed PowerCache - "with Control Panel it goes to 11-13"),  and these posts about a DiimoCache (@50mhz):





It seems like it should be benching faster.  For example, this is my current Benchmark compared with a DiimoCache (@Knez):



Better than his original 0.87 yes, but 1.06 vs 2.38 is a big difference.  FPU does match his original which makes sense (0.14 ~= 0.16 matches since I don't have the FPU on the card).

Pretty interesting!  Funny trying to squeeze that last bit of performance out of a 16/50mhz machine :-D  

 

[K Trueno]

 

[JDW]

Version 2.2 of Power Central is here:

http://lowendmac.com/daystar/download/software/upgrades_68030/soft_powercache030.hqx

 

[K Trueno]

Thanks JDW, I have tried 1.3, 2.0, and 2.2 to the same effect.  

 

[JDW]

Your Speedometer 3 performance score of 9 is about what I get with my CPU-socketed version Daystar PowerCache 50 MHz accelerator when the Power Central control panel is NOT loaded at boot time. And by “socketed” I mean my accelerator connects to the CPU socket on the motherboard, and no Artmix adapter is required. 

Low or unstable voltage should not have any affect on whether the cash is activated or deactivated. If anything, unstable voltage  causes freezes. 

What version is your Artmix adapter?  I would say that part of the equation is suspect.  Have you contacted Manabu Sakai of ARTMIX about this problem?

 

[JDW]

I have the original TS Adapter, photos of which you can see here:

https://www.flickr.com/search/?user_id=66071596%40N00&sort=date-taken-desc&text=twinspark&view_all=1

 

[K Trueno]

Your Speedometer 3 performance score of 9 is about what I get with my CPU-socketed version Daystar PowerCache 50 MHz accelerator when the Power Central control panel is NOT loaded at boot time. And by “socketed” I mean my accelerator connects to the CPU socket on the motherboard, and no Artmix adapter is required. 

Ah yes the unobtainum socketed version (not jealous at all  )
How does it bench with the Control Panel loaded at boot time say in these scenarios:

• PowerCache On, PowerMath On
  
• PowerCache Off, PowerMath On
  
• PowerCache On, PowerMath Off
  
• PowerCache Off, PowerMath Off (I would assume 9 here)
  

I also wonder how PowerDemo behaves for you!

What version is your Artmix adapter?  I would say that part of the equation is suspect.  Have you contacted Manabu Sakai of ARTMIX about this problem?

I have the blue one (looks like the one in your pictures in shape, but blue).  I had not thought about this as being a possible issue!  

So, I happen to also have an original IIsi version of adaptor and tested it out, and I am getting the same results.  

But perhaps I am overthinking it, could that last bit just having no FPU, and the PowerDemo thing a red herring (as it happens for @Bolle too?)

 

[K Trueno]

Totally safe. ;-)  

 

[JDW]

Please post a nice photo of the FRONT and BACK of that PDS slot adapter.  I've not seen one before.  I've only scanned a photo of the Daystar version here:





You also are forgetting the need to benchmark ENHANCED PowerMath, which I mention in the following thread (which I will benchmark for you later):