68040 upgrade cards in IIci

[equill]

In the last month-or-so I have acquired three 68040 accelerator cards for use in a IIci PDS (cache) slot, and other Macs through an appropriate adapter: DayStar Turbo040, MicroMac Carrera040 and Mobius Speedster. I installed the DayStar card and QuadControl CP in my second IIci (with OS 7.6.1), replacing a DayStar PowerCache 50MHz. Off went the Turbo 040, and has never looked back. That's the way all CPU upgrades should be, no?

Tell that to the fairies. The Carrera arrived with one 22µF 6V capacitor (the kind that trag likens to 'little fuel tanks') loose in the bag, one ditto connected by one leg only to the card, a third properly in place, and the fourth completely missing. The solder pads showed no sign that there had ever been a cap there. From the dirt adhering to the MicroMac-branded cooling-fan's blades I gauged that the card had been in use for some time. With as many as three ineffectual caps? I replaced those three defective/missing caps with 22µF 25V tantalums. Then the Carrera went into my third IIci, with Carrera040 CP and Startup Carrera Extension (under OS 7.6.1). TattleTech 2.84 and MicroMac's Speedometer saw the CPU as '68030 25MHz', remarkably like the native CPU, rather than the expected 68040 40MHz. No loading order changes for the CP and extension let the system 'see' the new CPU, whether the startup volume was the internal SCSI 0 or an external SCSI 6.

So, out with the Carrera and in with the Mobius. The main card of both MicroMac and Mobius upgrades is, as nearly as may be, identical physically and (presumably) electrically, but the Mobius card did have all four 22µF caps in place as tantalum electrolytics. Only the 128kB cache cards differed significantly, in apparent design and in appearance, between the two accelerators. The same CP and extension as for the Carrera have been reported to serve for the Mobius. However, the Mobius gave the same result as the MicroMac: the CPU again showed as 68030 25MHz.

Part way through the battle of wills I replaced the IIci's Asanté 10Base-T NIC with an AsantéFAST 10/100Base-T card, to speed up transfers of software from the other two IIci Macs for troubleshooting. To complete that upgrade I installed OT 1.1.2 over the existing 1.1.1. Damme, if after restart, what does TattleTech show for the CPU but 68040 40MHz. Speedometer confirmed this. I still don't know whether the attached 128kB cache card is active, but one victory per skirmish is not to be sneezed at. What possible connection might there be between a CPU and a NIC that could wreak such a result?

Next step is to pick up jaw and re-install the MicroMac to see whether such a nearly identical card shows the same behaviour.

de

 

[Unknown_K]

OT with the Daystar 040 gave me bombs, have you ever tried them together?

 

[trag]

still don't know whether the attached 128kB cache card is active, but one victory per skirmish is not to be sneezed at. de

Have you tried Cache-22? It will usually correctly detect the cache (if any) on upgrades.

Beats me what OT has to do with it. Perhaps something in the PRAM got reset when the new/different network card was installed?

Jeff

 

[equill]

... Have you tried Cache-22? It will usually correctly detect the cache (if any) on upgrades.
Beats me what OT has to do with it. Perhaps something in the PRAM got reset when the new/different network card was installed?

Jeff

Cache-22 is PPC only, as opening it in a System running on a 68K Mac will quickly tell one, in red. Strangely so, because RAMometer, GaugePRO and even Metronome work, sort of, but without mention of L2 cache. However, your suggestion prompted me to look at Cache Switch CP. The same version (7.0.1?) is installed with System 7.1 and OSs 7.6 and 8.0, and has no effect in this instance.

Something that I forgot to mention above was that, having installed OT 1.1.2 and discovered that 68040 operation of the Mobius card was thereby enabled, I also upgraded (updated?) from the the AppleShare Client 3.6.5 installed with OS 7.6.1 to AS 3.7.4. The 68040 was immediately disabled again. Pressing on like a good soldier I then installed AS 3.8.3 (the version recommended by Apple for 7.6.1 in artnum=16145), which runs happily under 7.6.1 with either a DayStar 50MHz PowerCache card or a Turbo040. The Mobius card was again enabled.

¿Que?

de

Erratum: 'Carrera' corrected to 'Mobius' in second-last line

 

[bigD]

It dismays me that System 7.x (or heaven forbid 8.x) is mentioned in the same post as a IIci.

God intended that a IIci be run under 6.x. Booting a IIci into 7 or later is like putting a sweater on your cat. Yeah, you may think it's cute, but your cat hates you for it. Trust me, your IIci hates you for making it boot into 7. System 8? It's trying its best to cut its own motherboard traces.

 

[Unknown_K]

A IIci with an 040 runs 7.x just fine. If you want to run OS 6 then a IIcx would be a better machine in my opinion.

 

[bigD]

You're right of course, but as this board's protector of all things System 6, it's my responsibility to make sure that everyone with a 6.x capable Mac run the right OS. System 7 is NEVER the right OS for a Mac II series machine.

 

[Bolle]

...its best to cut its own motherboard traces.

EMOci

system 6 on a 040 upgraded IIci is nice and fast... but not taht useful.

 

[equill]

Very skeletal, but here are some Speedometer 4 figures for IIci-2 (50MHz DayStar PowerCache); IIci-1 (33MHz DayStar Turbo040 with and without FPU), and IIci-3 (40MHz Carrera040 not 'seen' by system, ie still running a 25MHz 030; and 40MHz Mobius Speedster 040). OS 7.6.1 throughout.

• IIci-3 (68030/68882 25MHz, although MicroMac Carrera040 was installed)
  CPU = 0.43
  Graphics = 0.36
  HDD = 2.22 (Quantum Fireball 2GB)
  Maths = 1.50
   
  IIci-2 (68030/68882 50MHz DayStar)
  CPU = 0.86
  Graphics = 0.57
  HDD = 2.78 (IBM DDRS 4GB)
  Maths = 3.11
   
  IIci-1 (68LC040/33MHz DayStar)
  CPU = 1.17
  Graphics = 0.75
  HDD = 3.29 (Seagate 2GB)
  Maths = 5.25
   
  IIci-1 (68040/33MHz DayStar)
  CPU = 1.31
  Graphics = 0.77
  HDD = 3.27 (Seagate 2GB)
  Maths = 12.14
   
  IIci-3 (68040/40MHz Mobius)
  CPU = 1.62
  Graphics = 0.77
  HDD = 2.62 (Quantum Fireball 2GB)
  Maths = 24.11

Added on 3 May 2008:

• IIci-3 (68040/40MHz Carrera without cache, but Cache Switch CP ON)
  CPU = 1.54
  Graphics = 0.73
  HDD = 2.18 (Quantum Fireball 2GB)
  Maths = 23.86
   
  IIci-3 (68040/40MHz Carrera without cache, but Cache Switch CP OFF)
  CPU = 1.54
  Graphics = 0.73
  HDD = 2.18 (Quantum Fireball 2GB)
  Maths = 23.86

Ten iterations of each test. CPU, Graphics (Radius 24nn) and HDD relative to Quadra 605 = 1. Maths relative to Quadra 650 = 1.

Under System 6.0.8, the system reported: 'No coherent System found. Just random scattered primitives.' :-*

de

Erratum: DayStar Turbo040 corrected to 33MHz, which may have some bearing on the comparison with the Mobius card (40MHz).

 

[bigD]

Under System 6.0.8, the system reported: 'No coherent System found. Just random scattered primitives.' :-*

de

LOL! Touche!

Interesting benchmarks. It brings up a question though - didn't all the Quadra 650s have a full 040 in them? It's benching as though it doesn't have an FPU. Or did you transpose the 0 and 5 and that same Quadra 605 is being used for the FPU baseline?

 

[equill]

The 605, having an LC040, is used only for the non-maths-intensive tests. Effectively, the number for each IIci test is a ratio of the number of flops (or outcomes) compared to the 605. Because the 650 has a full 040, it is used in place of the 605 in maths-intensive tests, but again the IIci figures are ratios. No account is taken of CPU speeds, either relative or absolute (or so I divine) in the tests, which are just raw performance indicators.

de

 

[trag]

Curious that there's such a difference between the Daystar and the Mobius with the same CPU and speed in the Math and to a lesser extent the CPU tests. I wonder if the Mobius manages to operate with fewer wait states.

Oh, hey. Are you using one of the newest revision Turbo040s with the big ASIC or the old style with all the little GALs on board?

 

[Unknown_K]

Is there a difference in speed between the original Daystar Turbos and the later ones that didn't have a cache slot (cache built in)?

 

[equill]

OT with the Daystar 040 gave me bombs, have you ever tried them together?

Sorry. I didn't mean to ignore your query, but I did overlook it while I was preoccupied with the MicroMac and Mobius cards. All three IIcis of mine have OT as well as OS 7.6.1. IIci-2 (DayStar PowerCache) and IIci-1 (DayStar Turbo040) already had OT 1.1.2 when this exercise began, but IIci-3 had only OT 1.1.1, which was upgraded as in my first post. IIci-1 has behaved with propriety ever since the Turbo replaced the PowerCache that was already there. Unless OT is installed out-of-place (no support), I cannot think of a reason for it to misbehave, even though you have found it so. What System version were you using?

de

 

[equill]

Curious that there's such a difference between the Daystar and the Mobius with the same CPU and speed in the Math and to a lesser extent the CPU tests. I wonder if the Mobius manages to operate with fewer wait states.

Substitution of the Mobius with the MicroMac has enabled me to determine that the 128kB card of the MicroMac is the villain in the story. Under the same circumstances in which the Mobius performs faultlessly (now), although I am still not sure that the Mobius cache card is active, the MicroMac cache freezes startup immediately after the Carrera CP loads. Neither does TattleTech 'see' the cache. Removal of the MicroMac cache card restores normality. Inspection of the MicroMac cache card reveals corrosion on the pins of some ICs, but none is bridged, and the 'fuel-tank' electrolytics show no sign of bulging or leaking. However, a wash and a recap are now in the to-do list. Ideally, testing with a new cache would be the way to go, but where to find one now (apart from on the chimeric MicroMac website)?

Oh, hey. Are you using one of the newest revision Turbo040s with the big ASIC or the old style with all the little GALs on board?

There is no forest of GAL chips. The complement is:

2 x Xilinx multipin (40?) with different type numbers

1 x GAL

4 x P53SJ

1 x 88196DW80

1 x AT17C65

2 x 51C9XXK

1 x P528A

and what I take to be an 80MHz oscillator.

de

 

[trag]

There is no forest of GAL chips. The complement is:

2 x Xilinx multipin (40?) with different type numbers

1 x GAL

4 x P53SJ

1 x 88196DW80

1 x AT17C65

2 x 51C9XXK

1 x P528A

and what I take to be an 80MHz oscillator.

de

Not on the Daystar card, I don't think, as Daystar did not use Xilinx FPGAs, although I am braced to find myself wrong and learn something new here.

That sounds like the chips on the Mobius.

 

[Quadraman]

It dismays me that System 7.x (or heaven forbid 8.x) is mentioned in the same post as a IIci.
God intended that a IIci be run under 6.x. Booting a IIci into 7 or later is like putting a sweater on your cat. Yeah, you may think it's cute, but your cat hates you for it. Trust me, your IIci hates you for making it boot into 7. System 8? It's trying its best to cut its own motherboard traces.

68000 and 68020 machines should run System 6. So should any 030 machine less than 25mhz but from 25mhz onward System 7 can run acceptably.

And 030 Macs like the IIci can't run System 8. 7.6.1 is as far as they go if they are 32 bit clean, 7.5.5 if they are not. If you are tricking it into 8 using an 040 upgrade, it's technically a Quadra now.

 

[TylerEss]

I have something to mention about the difference between the Mobius and DayStar 40MHz 68040 accelerator card benchmark results... it looks to me like the 'CPU' and 'Maths' scores are being inflated on the Mobius card by software that causes 'software-mode' SANE calculations to be performed on the FPU anyway (and at 80-bit instead of 96-bit precision).
DayStar calls this feature 'PowerMath' and when you enable the trancendental functions 'Enhanced PowerMath.' Check to make sure that feature is turned on on your DayStar results.

Well, scratch that idea!

In my SE/30 with 48MHz Turbo040, with Enhanced PowerMath I get the following results: 39.8 CPU, 9.33 FPU (Norton System Info 3.5.3). With Enhanced PowerMath turned off, I get the same results.

You can really see how important memory bandwidth is to some of the tests that they're using, though. With the 128k cache turned off, my scores dropped to 30.1 CPU, 8.1 FPU. Obviously a lot of their tests fit wholly into 128k L2 cache but not into the (32k? I don't remember exactly) L1 cache.

Re: IIci with OS 8... You're totally right about the IIci loving System 6... That said, mine seems to like running OS 8.1 from time to time; it's like going window-shopping and it's glad to get back to System 6 when it can. Since it's used for testing, it's napping 99% of the time anyway.

Also, 68030 Macs run MacOS 8 just fine; the biggest problem is making sure to have enough RAM. http://home.earthlink.net/~gamba2/os8_68030.html

 

[equill]

Not on the Daystar card, I don't think, as Daystar did not use Xilinx FPGAs, although I am braced to find myself wrong and learn something new here.
That sounds like the chips on the Mobius.

Yes. My apology. The wits were out to lunch when I looked at the MicroMac instead of the DayStar Turbo040. The DayStar (a 33MHz, by the way) has a proprietary DayStar larger chip (and replaceable ROM, I presume) but only two probable GAL chips and a variety of others that I didn't make a census of in a quick dive inside a moment ago.

As I mentioned above, the MicroMac and Mobius cards proper (both 40MHz) are almost identical physically, and probably electrically, but their cache cards are widely different in layout and content. I hesitate, without a lot more investigation, to try the Mobius cache with the MicroMac card.

de