cyclone3d wrote on 2021-12-03, 23:58:I have a dual socket 5 SBC and a dual socket 7 motherboard.... I also have a couple of those interposers that came with AMD K6-2 CPUs. I will be trying both of those setups with the AMD CPUs but I have very little hope that they will work in dual CPU mode.
🤞
Recently a few others and I have been digging into what exactly all those resistors and capacitors under the lids of k6 cpus are.
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
Which led to a wonderful discovery^
Today I was examining more of fritz’s pics
And,
This one was of particular interest…
https://www.flickr.com/photos/130561288@N04/22772542500/
What ever could all those missing smds be for…
Perhaps a smp option that was never enabled?
Sphere478 wrote on 2022-03-06, 06:53:Recently a few others and I have been digging into what exactly all those resistors and capacitors under the lids of k6 cpus are […]
Recently a few others and I have been digging into what exactly all those resistors and capacitors under the lids of k6 cpus are.
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
Which led to a wonderful discovery^
Today I was examining more of fritz’s pics
And,
This one was of particular interest…
https://www.flickr.com/photos/130561288@N04/22772542500/
330DEE55-535B-465E-BD5A-C2C89A98C0C2.pngWhat ever could all those missing smds be for…
Perhaps a smp option that was never enabled?
👀
https://en.wikipedia.org/wiki/OpenPIC_and_MPIC
I keep hoping for chances for this, Ninkeo came up with this link,
It brings up a interesting point.
Several sources seem to conflict as to wether or not the k6 has outputs for smp functions.
Oxcats/catsay is very knowledgeable about this subject and brought some very valuable information to the subject.
there is also another another angle to this, it seems several players were talking about smp at the time this all went down (socket 3/4/5/7 era) apparently there were competing standards for how to implement the smp protocols, so even if the chips supported it, it may not work in my intel apic equipped mobo
Oh well, I got excited when I saw all the open smd pads.
The OP question sites k6-2 450 cpus running at 400
I now own three non matched 450 cpus. I tried them earlier today in my tyan s1564d
Post card shows - -
I’ve tried all sorts of cpus in it. Cyrix, rise, you name it. Nothing,
Other than actual pentium/pentium mmx chips the closest I’ve come to finding another chip that works is a pair of pentium mmx tillamooks, which I also tried to crack smp on and ultimately ended up confirming the opposite. The function appears to have been physicallt removed from the package, the core, or both.
Tillamook SMP and multiplier unlocking investigation thread.
At some point I’d like to see if mmx over drives work. That would be cool.
In the tillamook smp mod thread several intel smp pins were identified and tested.
Those pins were dead on the tillamook
But most appear to possibly be attached on the k6 to something.
I’m testing with a DVM on a cracked k6 chip right now, they all seem to show connection to something.
K6-2
catsay wrote: The only way to tri-state the buses on K6 is to drop the FLUSH# signal on reset.
This essentially however makes it impossible to toggle between the two processors for a functional SMP system (which one is active on the bus) since they cannot enter and exit tri-state mode at request without loosing all state. Only on a reset transition with cache flush held low.
catsay wrote: Pentium processors have complex additional logic that allows one of the processors to be isolated from the bus to a chosen extent when another performs addressing or other data/cache transcations
<mr_slug wrote: You may want to check out the Compaq Proliant 2000/4000 system. This supports FOUR socket 4 CPUs i think the 2500 was socket 5. It has its own HAL for NT. From the little i know the apic is all external on these systems. There was an update the the Acer J3 (drool) that had PCI, don't think it was special in any way though. IIRC the Compaq systems CPUs have their own RAM, so perhaps there it potential for dual K5 etc
mr_slug wrote: Infact Sphere the windows HAL list may be a way of tracking down servers that do AMD/MMX dual CPU. But finding these is hard. And on top of that they will be socket 4/5.
Looking on eBay and the listings for the 2500 CPU boards are Socket 8.
cyclone3d wrote on 2022-03-07, 02:32:Looking on eBay and the listings for the 2500 CPU boards are Socket 8.
Yep you are right. it was the 2000/4000 that was updated to take socket 5. Here is what i have on it:
http://66.113.161.23/~mR_Slug/deviceInfo/comp … e2/plnt2_4.html
It looks like that CPU card is for it, i'd have to check the p/n.
This source indicates only the 486 cpu boards are quad friendly, but there are multiple places in PC world that talk about 4 Pentiums in this system. As i remember the 4000 and the 2000 are the same board. 4000 are 4CPU, 2000 are 2CPU, up gradable to 4CPU. Anyway. The theory is, 1. this system does not use Intel SMP, it does it's own thing. 2. As it does its own thing, maybe the CPUs think they are running alone, 3. It follows that it *might* work with AMD cpus. YMMV:-)
I'm sure there would be some other issue that arises. Finding one of these things is trickey. Also EISA only.
I was speaking with a group of people on TRW’s discord server and we were rehashing this idea. It seems to me from talking to various members of this forum and from all that I’ve read that some sort of SMP or AMP with non-Intel logic is the only way to approach this. I think mR_Slug is absolutely correct, and it’s possible that this could happen, but unlikely.
If it even exists, the REV to SMP 486 upgrade is an example of such weirdness executed in a way that avoided Intel’s oft-limiting designs.
Truth be told, the chances are slim that something like this would work, but if non-a Intel SMP or AMP is possible via some obscure but extant logic ICs, like those used in some Compaq SystemPros and whatnot, then it’s a step toward a theoretical dual K6 setup. 😁
If no one implemented it then you wont find magic lost chips, the best scenario is someone reverse engineers whatever AMD designed (does AMD specification exist at all for it?) and deploys in FPGA.
