VOGONS


Reply 20 of 34, by appiah4

User metadata
Rank l33t++
Rank
l33t++

Pass-Through is probably for disabling Cache, which was not too uncommon as a lot of cheap boards shipped with no cache installed at all.

Retronautics: A digital gallery of my retro computers, hardware and projects.

Reply 21 of 34, by mpe

User metadata
Rank Oldbie
Rank
Oldbie

This is almost certainly setting for L1 cache.

Don't have experience with this particular adaptor. But I am not sure disabling cache is what the "pass through" means. That wouldn't make any sense and not sure if a passive adaptor even could do that. I would read it as "do not do anything" or "respect motherboard setting".

Many motherboards do have settings for L1 cache mode or multiplier. If you do the same on the interposer this can lead to undefined behaviour as you are effectively setting high/low/floating the CPU pin twice in serial. On the other hand no interposer can reliably enable WB unless the motherboard has support for it, so "WB" setting on the interposer wouldn't make any sense.

If I were designing an adaptor like this I'd certainly want to have option to either force write-through (by floating WB/WT pinout) or passthrough (do nothing). And the same for the multiplier.

derSammler wrote on 2019-09-02, 08:25:

The L1 cache is inside the CPU. The rest of the system won't care or even notice whether it's WT or WB.

This is simply not true. In order to be able to run L1 in WB you need to have solid provisions for cache coherency in place. So the rest of your system (chipset and possibly other bus masters let alone other CPUs) has to be well aware of the fact you are running WB. There has to be mechanism for invalidating cache during WB cycle. That's what these extra WB pins are there for and there was such a big deal about L1 WB support and only newer 486 boards can support it (if lucky). Otherwise the system would crash really fast.

Blog|NexGen 586|S4

Reply 22 of 34, by Brickpad

User metadata
Rank Member
Rank
Member
mwenek wrote on 2020-01-02, 00:03:
Hey all. I have one of these versions of the Am5x86-P75 (refer to photo!) Any idea what the cut wires lead to? My guess is an […]
Show full quote

Hey all. I have one of these versions of the Am5x86-P75 (refer to photo!)
Any idea what the cut wires lead to?
My guess is an old fan that was cut off as it is an Evergreen Overdrive model.
I appreciate any insight before I fire it up.
Edit: Does the WB/WT cache option on this chip itself refer to its internal cache or the MB cache?

I'd say it was for a cooling fan, but that sloppy solder work does not look like factory. Either that the end user attempted to desolder the leads, or they wanted to source power from that cap.

Reply 23 of 34, by dionb

User metadata
Rank l33t++
Rank
l33t++
mwenek wrote on 2020-01-02, 00:03:
Hey all. I have one of these versions of the Am5x86-P75 (refer to photo!) Any idea what the cut wires lead to? My guess is an […]
Show full quote

Hey all. I have one of these versions of the Am5x86-P75 (refer to photo!)
Any idea what the cut wires lead to?
My guess is an old fan that was cut off as it is an Evergreen Overdrive model.
I appreciate any insight before I fire it up.
Edit: Does the WB/WT cache option on this chip itself refer to its internal cache or the MB cache?

Yep, the wires are for a fan.
s-l300.jpg

And the WB/WT refers to the internal L1, not the motherboard L2 cache, which the CPU has no way of knowing about or influencing.

Regarding the 'Write-through' vs 'Pass-through' discussion in the original thread, pass-through means exactly that: in that mode the adapter will pass on whatever setting the pins get from the motherboard (WT or WB). Setting to WT is an override in case of issues.

Edit: hadn't seen there was a page 2 and mpe had already explained the pass-through in more detail 😮

Reply 24 of 34, by Intel486dx33

User metadata
Rank l33t
Rank
l33t

Well, I could NOT get this CPU to work in any motherboard. And I tried several.
I also tried two of these CPU’s just to make sure it was NOT the CPU at fault.

So I think these CPU’s are NON-functional.

Beware, sold on eBay.

Reply 26 of 34, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t

I really hate the Trinity Works Power Stacker chips. At most, they needed 4 jumpers to setup the CPU, but instead they went with 14 tiny DIP switches that you can only access by removing the module from the CPU socket. To make matters even worse there were multiple versions of this thing and they have slightly different jumper settings. I remember buying one second hand in the late 90s, and by that time the jumper settings had already been scrubbed from the Trinity Works webpage so I couldn't use it. I wasn't clever enough to check archive.org and ended up getting rid of it.

Did PNY buy Trinity Works?

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 27 of 34, by sorphin

User metadata
Rank Newbie
Rank
Newbie
Keatah wrote on 2020-05-25, 03:12:

Who is the seller? And exactly what CPU does this refer to? The PNY PowerStacker?

I bought one and did a quick POST, it worked, I used the default switch settings.

Based on the pics and the fact that was following this cpu on ebay, i'd say its the ones from seller 'gfsi '.

Reply 29 of 34, by Keatah

User metadata
Rank Member
Rank
Member

I bought two of these recently from gfsi and they seem to work. I gave one of them a more thorough test by playing some games and stuff. The other one I just POST-tested.

I really wish I had known about these back in the day. But I had my head buried in anything and everything "pentium" at that exact time when these were popular. So for that reason and likely others, they never showed up on my radar.

Someone said they hated these because of all the options with the switches. IDK, but I'm happy to have those to help cover all situations.

And yes it sucks (sucked) that settings and information was removed from the website. These days before I get vintage hardware I get the necessary documentation and configuration information. Whether it be original PDF's and notes from Wayback Machine or retyped stuff on a web forum like this. For without that information life can be a real pain.

Reply 30 of 34, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t

But most of the switches don't really do anything useful (like setting four switches to enable WB mode). Like I said, 4 jumpers could have covered all the bases.

In 1995/1996 I was aware of the Cyrix and AMD 5x86 chips being sold in new systems. But, what wasn't communicated very well was that you could easily make them work on a regular 486 motherboard with an interposer. I remember around that time I was in communication with the store where my 486 was purchased in late 1992. I asked them what my upgrade options were, and they told me I could only use Overdrive ODPR chips (my board had a 487sx socket, and standard non ZIF 486 socket). I tried to order both a DX2 and DX4 overdrive from them on two separate occasions, and both times they were "sold out". 3.3V chips weren't even on the radar because they were "incompatible" according to them. If you tried asking a question on usenet, you'd get 12 different answers from "experts", so back then it was better to play it safe and not invest in expensive upgrade products that might not work and probably can't be returned.

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 33 of 34, by uridium

User metadata
Rank Newbie
Rank
Newbie

Kingston and a couple of other companies also used this CPU board ..without a heatsink but with a fan (!). I have 3 all different manufacturers.

Attachments

Reply 34 of 34, by uridium

User metadata
Rank Newbie
Rank
Newbie

Some of the switches I recall when setting up one system were for wait-states and multiplier. This lets me run one at 3x50mhz instead of 4x33mhz with two VLB cards (a TekRam 690CD and an S3 805).