Fizzbinn
Well-known member
*** This content is from a thread lost in the 6/30/2021 outage. I'm reposting attempting to include all info (if not images) posted by others. ***
** 3/28/2021 **
@Fizzbinn
I picked up a 6500/275 motherboard off eBay for reasonable price, I’ve been playing around with upgrades for my 6360 and was interested in finding a 512KB L2 cache. It might be out there somewhere but the identifying part numbers printed on the Apple L2 cache modules have been hard to find. In case others are interested:
IDT7MPV6284 - 512KB
IDT7MPV6283 - 256KB
Fronts:
Backs:
I believe these work in any Alchemy (6360/5400/6400), Gazelle (5500/6500/TAM) or Tanzania (4400/Clones) motherboard.
I’m going to run some benchmarks in my compatible motherboards and post results here.
@jessenator
Interesting! My 256k (top) and 512k (bottom) modules are a bit different (both from StarMax machines, though the 256k looks identical to the one in my 4400). The 512k being taller, falling somewhere between the 256k height and a Sonnet Crescendo L2
<missing image>
Also, not to steal your thunder, but here are my Power Macintosh 4400/200 results next to the 8500 and 9500. I'm curious what you get on the 6xx0/5x00 systems, since I don't own them. My results (MacBench 4) are all on vTools on the 'public' volume in the macbench folder, should you feel inclined to compare
<missing image>
Though the raw compute numbers aren't terribly dramatic, my own subjective "feel-o-meter" is that the 4400, et al. are right dogs without an L2 cache. Lots of visual stuttering, general lag, or simply just "not smooth" would be my hot take on it.
@Fizzbinn
Cool, thanks! I’ll use MacBench 4 as well.
It might take me awhile but I want to test with these systems I have:
With the following L2 Cache modules:
** 3/29/2021 **
@trag
Guessing, but I bet the second cache board Fizzbin pictured used 3.3V SRAM for the cache and added a bunch of voltage level shifters. Could be wrong, but that's my guess for all those extra little chips.
Differences in the appearances of the caches is mainly because of different packaging for the SRAM chips and tag RAM (tag RAM is basically just SRAM with some comparators built into the chip).
@Fizzbinn
Just to be clear, the two pictures I posted each show both the Apple L2 Cache modules, with the 512k card above the 256k card. I just flipped the modules over for the second picture.
The 512k module has (4) "IDT 71V433" chips vs. only (2) on the 256k module, I take this to be the actual cache memory (doubled for 512k).
The 512k module has (2) "IDT 71V218" chips vs. only (1) on the 256k module, function? (doubled for 512k).
Both the 512k and 256k modules have (1) "IDT 74FCT" chip, function?
Both the 512k and 256k modules have (8) "IDT 74FST" chips, function?
The Motorola Starmax versions @jessenator posted show the 256k front and back and then the 512k front and back. It looks like the 512k module has double the number of "IDT 71256" chips, with what looks to be the same number of the two other chips (one chip has a sticker blocking, but the "IDT 71216" is the same).
Seems like Apple and Motorola apparently accomplishing the same thing with different parts.
@rsolberg
Here's my 256KiB IDT branded IDT7MP6071 cache module that I have installed in my 6360. When I bought it from eBay some time ago, I believe it was listed as for a 7600. It has eight IDT 71256 15ns 5v 32k*8 SRAM ICs, one IDT 71216S12PF 16k*15 tag SRAM IC with integrated comparators that appears to be specifically catered to PowerPC and other RISC platforms and rated up to 66MHz. The large IC is a mystery to me. It looks like the later IDT7MP6071A module (like in Jessenator's photo above) is lower profile and uses a significantly smaller IC.
<missing image>
<missing image>
@rsolberg
IDT71V433 is a 32k*32 bit 3.3v SRAM, making them 128KiB each. I believe it's 11ns based on your photos.
These are 3.3v cache tag SRAM in an 8k*16 bit configuration, making 128k bits/ 16KiB each. Yours appear to be 12ns
The IDT74FCT163373 is a 3.3v 16-bit transparent latch, which I suspect is being used to buffer the tag SRAM, or could be used for part of the level shifting circuitry
The IDT74FST3xxx ICs are level shifting bus switches. As Trag suspected, the SRAM in your modules is 3.3v, but the cache logic on our logic boards is 5v. Basically most of the extra ICs are to manage this difference. The use of 3.3v SRAM may have been a necessity for economics or availability since your modules have 1997 date codes compared to 1996 for mine, or it could be to create compatibility with systems that use 3.3v logic for the cache. With the level shift bus switches, they should be compatible with both 3.3 and 5v logic on the bus.
@jeremywork
I have a couple 512k modules; slight variation to the tag number on @Fizzbinn's 512: IDT7MPV6289
Otherwise I haven't spotted any differences between this and IDT7MPV6284.
<missing image>
@beachycove
Those 4400 benchmarks are very interesting. The 603ev was a great processor — precursor to the G3, was it not?
@Fizzbinn
...after looking at these responses one thing I can't believe I didn't pick up earlier on is the company name "IDT", "Integrated Device Technology". Their stylized large integral symbol for the letter I seems obvious now.
512KB:
IDT7MPV6076
IDT7MPV6284
IDT7MPV6289
256KB:
IDT7MPV6071
IDT7MPV6283
@jessenator
IIRC, it was the first commercial CPU to hit 300 MHz (in the 6500), but I would wager if you stacked the instructions per clock/cycle, the 604e and 604ev (it had a different letter designation than "v" but I can't recall offhand) would come out on top—I want to see numbers though. These later "Mach V" 604-based CPUs that broke the 200 MHz barrier were pretty fast.
I'd like to see a kind of development timeline for all of these later-90s CPUs—when one was started, when/if they were released. IIRC there was a 620, which was supposed to be a 604 successor, but the improvements there led the 620 to not see any Apple spotlight.
It's interesting that the Wikipedia article for the 7xx series, under the 740 and 750 says the 740, specifically, is pin-compatible with the 603... I'd like to see a source on that, but interesting if true.
** 4/27/2021 **
@Fizzbinn
...and the MacBench 4 results are in!
Really no big surprises here, more L2 cache = better and the same L2 cache slot G3 upgrade runs pretty much identically on these platforms (even though the 6500 series uses a faster 50MHz bus - the large 1MB G3 L2 cache likely makes that moot).
One thing you'll notice is there are no results for the Umax SumperMac c500/180 with 512KB L2 cache, I didn't skip that one, I just couldn't get it to boot successfully with my IDT7MPV6284 512KB L2 cache module installed (which worked fine on the other platforms), it would always lock up right before the extensions loaded.
I tried to keep the parameters of the test machines consistent outside the L2 cache slot contents.
@Trash80toHP_Mini
Very nice workup there, great job!
** 5/2/2021 **
@Fizzbinn
Additional MacBench test results
Floating Point - Pretty much inline with the Processor test:
Graphics - Somewhat surprising that the older models, lacking ATI chipsets, faired better with the G3 upgrade, of course they can't support higher resolutions/bit depths with only 1MB of video RAM. The test config was pretty low end, 640x480 and 16-bit thousands of colors:
Disk - Looks like the Power Macintosh 4400's IDE bus lags the others. The test config used the same 160GB IDE hard drive on all systems/tests:
CD-ROM - Again the Power Macintosh 4400 lags the others, this time more significantly. The test config used the same 12x SCSI CD-ROM on all systems except the 4400 which has a 12x IDE CD-ROM:
Other Notes:
The Umax SuperMac c500's original 256KB L2 Cache module was the IDT7MPV6071. While I could not get it to boot with my 512KB L2 Cache module IDT7MPV6284, I wonder if it might have worked with the older (IDT7MPV6076) or newer (IDT7MPV6289) version.
Also my Power Macintosh 4400's stock 256KB L2 Cache module was the IDT7MPV6283.
** 3/28/2021 **
@Fizzbinn
I picked up a 6500/275 motherboard off eBay for reasonable price, I’ve been playing around with upgrades for my 6360 and was interested in finding a 512KB L2 cache. It might be out there somewhere but the identifying part numbers printed on the Apple L2 cache modules have been hard to find. In case others are interested:
IDT7MPV6284 - 512KB
IDT7MPV6283 - 256KB
Fronts:
Backs:
I believe these work in any Alchemy (6360/5400/6400), Gazelle (5500/6500/TAM) or Tanzania (4400/Clones) motherboard.
I’m going to run some benchmarks in my compatible motherboards and post results here.
@jessenator
Interesting! My 256k (top) and 512k (bottom) modules are a bit different (both from StarMax machines, though the 256k looks identical to the one in my 4400). The 512k being taller, falling somewhere between the 256k height and a Sonnet Crescendo L2
<missing image>
Also, not to steal your thunder, but here are my Power Macintosh 4400/200 results next to the 8500 and 9500. I'm curious what you get on the 6xx0/5x00 systems, since I don't own them. My results (MacBench 4) are all on vTools on the 'public' volume in the macbench folder, should you feel inclined to compare
<missing image>
Though the raw compute numbers aren't terribly dramatic, my own subjective "feel-o-meter" is that the 4400, et al. are right dogs without an L2 cache. Lots of visual stuttering, general lag, or simply just "not smooth" would be my hot take on it.
@Fizzbinn
Cool, thanks! I’ll use MacBench 4 as well.
It might take me awhile but I want to test with these systems I have:
- Performa 6360 - 6360/160 motherboard
- Performa 6360 - 6500/225 motherboard
- Performa 6360 - 6500/275 motherboard
- Power Macintosh 4400/200
- Umax SuperMac c500/180
With the following L2 Cache modules:
- None
- 256KB
- 512KB
- Sonnet G3 400Mhz/1MB
** 3/29/2021 **
@trag
Guessing, but I bet the second cache board Fizzbin pictured used 3.3V SRAM for the cache and added a bunch of voltage level shifters. Could be wrong, but that's my guess for all those extra little chips.
Differences in the appearances of the caches is mainly because of different packaging for the SRAM chips and tag RAM (tag RAM is basically just SRAM with some comparators built into the chip).
@Fizzbinn
Just to be clear, the two pictures I posted each show both the Apple L2 Cache modules, with the 512k card above the 256k card. I just flipped the modules over for the second picture.
The 512k module has (4) "IDT 71V433" chips vs. only (2) on the 256k module, I take this to be the actual cache memory (doubled for 512k).
The 512k module has (2) "IDT 71V218" chips vs. only (1) on the 256k module, function? (doubled for 512k).
Both the 512k and 256k modules have (1) "IDT 74FCT" chip, function?
Both the 512k and 256k modules have (8) "IDT 74FST" chips, function?
The Motorola Starmax versions @jessenator posted show the 256k front and back and then the 512k front and back. It looks like the 512k module has double the number of "IDT 71256" chips, with what looks to be the same number of the two other chips (one chip has a sticker blocking, but the "IDT 71216" is the same).
Seems like Apple and Motorola apparently accomplishing the same thing with different parts.
@rsolberg
Here's my 256KiB IDT branded IDT7MP6071 cache module that I have installed in my 6360. When I bought it from eBay some time ago, I believe it was listed as for a 7600. It has eight IDT 71256 15ns 5v 32k*8 SRAM ICs, one IDT 71216S12PF 16k*15 tag SRAM IC with integrated comparators that appears to be specifically catered to PowerPC and other RISC platforms and rated up to 66MHz. The large IC is a mystery to me. It looks like the later IDT7MP6071A module (like in Jessenator's photo above) is lower profile and uses a significantly smaller IC.
<missing image>
<missing image>
@rsolberg
21 hours ago, Fizzbinn said:
The 512k module has (4) "IDT71V433" chips vs. only (2) on the 256k module, I take this to be the actual cache memory (doubled for 512k).
IDT71V433 is a 32k*32 bit 3.3v SRAM, making them 128KiB each. I believe it's 11ns based on your photos.
21 hours ago, Fizzbinn said:
The 512k module has (2) "IDT71V218" chips vs. only (1) on the 256k module, function? (doubled for 512k).
These are 3.3v cache tag SRAM in an 8k*16 bit configuration, making 128k bits/ 16KiB each. Yours appear to be 12ns
21 hours ago, Fizzbinn said:
Both the 512k and 256k modules have (1) "IDT74FCT" chip, function?
Both the 512k and 256k modules have (8) "IDT74FST" chips, function?
The IDT74FCT163373 is a 3.3v 16-bit transparent latch, which I suspect is being used to buffer the tag SRAM, or could be used for part of the level shifting circuitry
The IDT74FST3xxx ICs are level shifting bus switches. As Trag suspected, the SRAM in your modules is 3.3v, but the cache logic on our logic boards is 5v. Basically most of the extra ICs are to manage this difference. The use of 3.3v SRAM may have been a necessity for economics or availability since your modules have 1997 date codes compared to 1996 for mine, or it could be to create compatibility with systems that use 3.3v logic for the cache. With the level shift bus switches, they should be compatible with both 3.3 and 5v logic on the bus.
@jeremywork
I have a couple 512k modules; slight variation to the tag number on @Fizzbinn's 512: IDT7MPV6289
Otherwise I haven't spotted any differences between this and IDT7MPV6284.
<missing image>
@beachycove
Those 4400 benchmarks are very interesting. The 603ev was a great processor — precursor to the G3, was it not?
@Fizzbinn
...after looking at these responses one thing I can't believe I didn't pick up earlier on is the company name "IDT", "Integrated Device Technology". Their stylized large integral symbol for the letter I seems obvious now.
512KB:
IDT7MPV6076
IDT7MPV6284
IDT7MPV6289
256KB:
IDT7MPV6071
IDT7MPV6283
@jessenator
20 hours ago, beachycove said:
Those 4400 benchmarks are very interesting. The 603ev was a great processor — precursor to the G3, was it not?
IIRC, it was the first commercial CPU to hit 300 MHz (in the 6500), but I would wager if you stacked the instructions per clock/cycle, the 604e and 604ev (it had a different letter designation than "v" but I can't recall offhand) would come out on top—I want to see numbers though. These later "Mach V" 604-based CPUs that broke the 200 MHz barrier were pretty fast.
I'd like to see a kind of development timeline for all of these later-90s CPUs—when one was started, when/if they were released. IIRC there was a 620, which was supposed to be a 604 successor, but the improvements there led the 620 to not see any Apple spotlight.
It's interesting that the Wikipedia article for the 7xx series, under the 740 and 750 says the 740, specifically, is pin-compatible with the 603... I'd like to see a source on that, but interesting if true.
** 4/27/2021 **
@Fizzbinn
...and the MacBench 4 results are in!
Really no big surprises here, more L2 cache = better and the same L2 cache slot G3 upgrade runs pretty much identically on these platforms (even though the 6500 series uses a faster 50MHz bus - the large 1MB G3 L2 cache likely makes that moot).
One thing you'll notice is there are no results for the Umax SumperMac c500/180 with 512KB L2 cache, I didn't skip that one, I just couldn't get it to boot successfully with my IDT7MPV6284 512KB L2 cache module installed (which worked fine on the other platforms), it would always lock up right before the extensions loaded.
I tried to keep the parameters of the test machines consistent outside the L2 cache slot contents.
- Same 160GB IDE hard drive running the same install of Mac OS 8.6 (originally done on the Power Macintosh 4400)
- AppleTalk off, Virtual memory off, Disk Cache 2560KB, Monitor set to 640x480 and 16-bit thousands of colors
- Same two 64MB 5v EDO DIMMs except the for the 4400 where I only have 80MB total at present of 3.3v kind it uses
- For consistent MacBench CD-ROM access even moved the same 12x SCSI CD-ROM between all the machines except the 4400 which has a 12x IDE CD-ROM
@Trash80toHP_Mini
Very nice workup there, great job!
** 5/2/2021 **
@Fizzbinn
Additional MacBench test results
Floating Point - Pretty much inline with the Processor test:
Graphics - Somewhat surprising that the older models, lacking ATI chipsets, faired better with the G3 upgrade, of course they can't support higher resolutions/bit depths with only 1MB of video RAM. The test config was pretty low end, 640x480 and 16-bit thousands of colors:
Disk - Looks like the Power Macintosh 4400's IDE bus lags the others. The test config used the same 160GB IDE hard drive on all systems/tests:
CD-ROM - Again the Power Macintosh 4400 lags the others, this time more significantly. The test config used the same 12x SCSI CD-ROM on all systems except the 4400 which has a 12x IDE CD-ROM:
Other Notes:
The Umax SuperMac c500's original 256KB L2 Cache module was the IDT7MPV6071. While I could not get it to boot with my 512KB L2 Cache module IDT7MPV6284, I wonder if it might have worked with the older (IDT7MPV6076) or newer (IDT7MPV6289) version.
Also my Power Macintosh 4400's stock 256KB L2 Cache module was the IDT7MPV6283.
Last edited: