Yes, that is extremely helpful. The first challenge in tracking down those old parts is to find the actual part number and manufacturer. Did KEL still have a datasheet posted for the connector? That could be helpful too, with dimensions on the pad/hole placement. If the part is unobtainium, it might be practical to substitute some other connector that will fit in the same footprint.
-
Updated 2023-07-12: Hello, Guest! Welcome back, and be sure to check out this follow-up post about our outage a week or so ago.
Mac IIci & cache card questions
- Thread starter Bolkonskij
- Start date
Yes, that is extremely helpful. The first challenge in tracking down those old parts is to find the actual part number and manufacturer. Did KEL still have a datasheet posted for the connector? That could be helpful too, with dimensions on the pad/hole placement. If the part is unobtainium, it might be practical to substitute some other connector that will fit in the same footprint.
Consider though, that a IIci could be run with a total of 4MB of RAM. If you configure the IIci with 4MB of RAM and it has a 4MB cache, then everything the IIci does, actually hits the cache. Of course, diminishing returns means that might only be 10% faster than having a 32KB cache.
Perhaps the IIci is close to minimum wait states already with a more modest cache.But that's what makes it an interesting experiment.
jessenator
Well-known member
These people:
https://www.connectorpeople.com/Connector/KEL/8
Have the several hundred of the KEL 8817-140-170 available. Which doesn't really help us.
I have contacted them to ask about the 8807 part.
Here is the new datasheet from KEL. It encompasses both the 8807 and 8817. All their parts have been updated to be ROHS compliant which changed the last couple of letters a bit
The 8807 is still a current part from KEL. It might be worth contacting them.
https://www.kel.jp/english/product/product_detail/id=380&p_search=1
View attachment 8817-140-170.pdf
https://www.connectorpeople.com/Connector/KEL/8
Have the several hundred of the KEL 8817-140-170 available. Which doesn't really help us.
I have contacted them to ask about the 8807 part.
Here is the new datasheet from KEL. It encompasses both the 8807 and 8817. All their parts have been updated to be ROHS compliant which changed the last couple of letters a bit
The 8807 is still a current part from KEL. It might be worth contacting them.
https://www.kel.jp/english/product/product_detail/id=380&p_search=1
View attachment 8817-140-170.pdf
Last edited by a moderator:
jessenator
Well-known member
womp womp... guess I didn't look hard enough
yes, that's a much better datasheet, than that fax-resolution copy. Interesting note on the Amiga CPU connector, and great to know KEL still makes the part. Would it be more advantageous for someone stateside to get them then ship them across the pond? I don't imagine the parcel being particularly huge or heavy for even 50-100 of the connectors.I'm really surprised DigiKey doesn't have these... I mean they have those 030 PDS/NuBus style connectors in stock...
Last edited by a moderator:
Unknown_K
Well-known member
joethezombie
Well-known member
DiiMO was also in on the cache goodness. Quadra Cache 128k from my collection.
jessenator
Well-known member
Mr SN
Well-known member
Bolle
Well-known member
Bolle
Well-known member
There is a thread on those connectors on a1k right now where people gang up to reach the 201 pieces minimum order from the European distributor of KEL.
Price per part is 12.69€. Lead time is 10-12 weeks for fresh brand new connectors from Japan.
They have a distributor in the US as well and I'd guess that similar conditions apply for ordering a batch from them directly.
jessenator
Well-known member
Yeah, what Jessenator said. Ouch. Of course, until someone checks, we can hope that a smaller minimum and/or price applies. In 2011 the minimum was around 20, but that may have been a time when they had stock on hand. Now the difference could be that they're running off a new manufacturing batch every time.
It's too bad The Connector People didn't have them on hand. I emailed them to check about the 8807 because they had so many of the 8817 on hand, and received a kind response very quickly that they have none of the 8807.
It's too bad The Connector People didn't have them on hand. I emailed them to check about the 8807 because they had so many of the 8817 on hand, and received a kind response very quickly that they have none of the 8807.
Last edited by a moderator:
David Cook
Well-known member
For those with envy for the IIci 128KB cache and Quadra FastCache, I ran some real-world-ish performance tests.
TLDR; Very little performance improvement from either 128KB memory caches.
System 7.5.5. Disk cache 128KB (except on IIci where BankA=larger disk cache to fill video memory and force programs to run in BankB). All computers with 64MB RAM or more. Tests run on external MacSD. 640x480 8-bit color. Appletalk off.
I complied a commercial C program using Metrowerks CodeWarrior 5. It consisted of 81 files with a total of 164,000 lines of code. So, this is a processor/SCSI/memory test, not a video test.
All results compared to stock (no cache) Macintosh IIci.
First real boost is the 32KB cache card.
Next real boost is a IIci 68040 card. This is massive.
Next boost is an actual 68040 computer (Quadra 630/605/650) running at 33 MHz.
If the Quadra Cache software is not loaded, the card does not activate and makes no improvement.
With the software loaded, we only get a 7% improvement (428%/398%) from the 128KB Quadra cache card. Norton System Info test results were almost nothing.
With the PowerMath and ROM features of the software enabled, we still only get an 11% improvement from the 128KB Quadra cache card.
Moving two resistors on the Quadra 630 provides better results due to overclock to 40 MHz.
Replacing the crystal on the Quadra 650 provides the best results due to the overclock to 40 MHz. In fact, it scored better than the Quadra 840AV in Norton System Info 3.53.
The Quadra cache card did not work with the overclock. So, no results are shown for that.
In summary, expect almost no boost from a 128KB IIci cache card compared to 32KB. Expect maybe a 10% boost in certain cases for the 128KB Quadra cache card, whereas a crystal swap will give you 17%.
Even if you ignore overclocking, the Macintosh mid-range line had a 4x speed improvement in 4 years.
TLDR; Very little performance improvement from either 128KB memory caches.
System 7.5.5. Disk cache 128KB (except on IIci where BankA=larger disk cache to fill video memory and force programs to run in BankB). All computers with 64MB RAM or more. Tests run on external MacSD. 640x480 8-bit color. Appletalk off.
I complied a commercial C program using Metrowerks CodeWarrior 5. It consisted of 81 files with a total of 164,000 lines of code. So, this is a processor/SCSI/memory test, not a video test.
All results compared to stock (no cache) Macintosh IIci.
First real boost is the 32KB cache card.
Next real boost is a IIci 68040 card. This is massive.
Next boost is an actual 68040 computer (Quadra 630/605/650) running at 33 MHz.
If the Quadra Cache software is not loaded, the card does not activate and makes no improvement.
With the software loaded, we only get a 7% improvement (428%/398%) from the 128KB Quadra cache card. Norton System Info test results were almost nothing.
With the PowerMath and ROM features of the software enabled, we still only get an 11% improvement from the 128KB Quadra cache card.
Moving two resistors on the Quadra 630 provides better results due to overclock to 40 MHz.
Replacing the crystal on the Quadra 650 provides the best results due to the overclock to 40 MHz. In fact, it scored better than the Quadra 840AV in Norton System Info 3.53.
The Quadra cache card did not work with the overclock. So, no results are shown for that.
In summary, expect almost no boost from a 128KB IIci cache card compared to 32KB. Expect maybe a 10% boost in certain cases for the 128KB Quadra cache card, whereas a crystal swap will give you 17%.
Even if you ignore overclocking, the Macintosh mid-range line had a 4x speed improvement in 4 years.
Fizzbinn
Well-known member
Very cool test set! Thanks for sharing!
jessenator
Well-known member
Truth. The IPS/C lift pretty impressive. (According to wikipedia anyway) a 25MHz 040 had a MIPS rating of ~27.5 while a 50MHz 030 only scored 18 MIPS.
Powerbase
Well-known member
Ive heard before that clock-for-clock a 040 was usually twice as fast as a 030.
I think until either the late 90s or early 2000s memory was still pretty expensive. Apple must have felt the 040 was fast enough to not bother.
mg.man
Well-known member
Perhaps it's early and my coffee hasn't kicked in... but I see...
Isn't that a 23% increase over a non-cache machine? That's a pretty significant and I'd argue useful boost. I know they're getting more and more hard to come by, but I picked up my original cache card about a year ago for ~$20 inc. postage, so they are still out there. I've also seen them lurking in batt-bombed IIcis that have gone cheap. Seems worth a hunt if you don't have one or don't want to fork out the serious $$$ for an accelerator.
Isn't that a 23% increase over a non-cache machine? That's a pretty significant and I'd argue useful boost. I know they're getting more and more hard to come by, but I picked up my original cache card about a year ago for ~$20 inc. postage, so they are still out there. I've also seen them lurking in batt-bombed IIcis that have gone cheap. Seems worth a hunt if you don't have one or don't want to fork out the serious $$$ for an accelerator.
