I can't quite read the markings on the chips but I did spot LM385 there - that's voltage reference. Chances are this device uses a non-standard VCC - like, say, 3.45V instad of 3.3V - to improve stability or power dissipation. So the regulator is actually 2 or 3 parts (reference, op-amp and power transistor). Other stuff could very well be simple glue logic.

Doesn't this approach seem like a more costly way of building an upgarde module? No wonder Gainbery went out of business way before Evergreen. I remember there being boxes and boxes of unopened Gainbery units for sale online some time ago.

90% of the cost is that CPU I'd say. The PCB is cheap and so are the parts used, the only problem - maybe - is that plastic pin adapter. It's rather unique and parts like that are always a bit more expensive.
That being said, these upgrades had issues not related to the assembly or part count - mostly BIOS/mobo compatibility. And, in general, so-so performance. I mean, by the time these high-clocked 486 were made in sufficient numbers to make them cheap and available, Intel already had Pentium on the market - and it was cleary a superior chip. So the upgrade had to be cheap to make existing 486 mobo owners get one over a new Pentium build, and you can't make a lot of money by selling product with low margin while facing tough competition.

I could never find any of these upgrades offered in shops, while everybody had the AMD DX4. The boxes you saw probably arrived "too little, too late".

Nobody has one of these gainbery QFPs? Were they that obscure?

Turned out that the Gainbery 133 MHz was an AMD 5x86 chip.

I begun testing the Gainbery Cyrix 5x86-100QP and noticed that several pins were loose. For the most part, I could heat up the pin and have the contact reflowed, except for 6 other pins. I ended up having to lay conductors for these pins to get the chip working again. Also one pin was broken and had to be soldered using a donar chip's pin.

The photo shows SUSPA# wired, but I later discovered that the chip wouldn't power on with this wired. By default, CLKMUL is not connected and left floating so the chip runs at 3x. If I connect the pin and try to run the chip at 4x, I hear a beep as if the system wants to start, but the screen stays blank. I suspect that this chip is 4x capable but cannot cope with 133 MHz. I took a gamble with these two Gainberys hoping for a Cyrix 5x86-133 but it didn't work out.

I also landed two boxed Gainbery IBM 5x86c-100HF chips. One was in the Gainbery box stating it was a 75 MHz part, but I guess they used the IBM 5x86c-100HF chips all around in these.

IBM made a 75MHz 5x86c...maybe that's what's in your 75MHz gainbery.

If you set the 4x enabled cyrix-100 part to 25MHz bus, will the system boot?

The IBM heatsink on the "75 MHz" upgrade shows 100HF.

I'm not sure why 2/4x wasn't working previously at 33 MHz, but it is now and it shows only 66 MHz. So I guess the chip isn't 4x-capable, although its datecode is in line with what 4x should be. Perhaps QFP's are different.

EDIT: I did notice something interesting about this interposer board though, if I leave the motherboard jumpered to 3.68 V, the CPU still works. According to my multimeter, the CPU is receiving 3.53 V. Some interposers don't work at all for such a small Vin to Vout delta.

Not sure how I missed this thread

I bought a gainbery maximiser 133 a few years back and was the amd version.

Also I've had gainbery 5x86-120/100/75 upgrade kits, they were all blue ibm 5x86-100 heatsinks.

All the same....

Same here. So I was pretty excited to find a Cyrix QFP on a Gainbery.

I have the same Gainbery CPU, only it's fried. I got a replacement SMD chip and a new voltage regulator. However, the legs on my chip are damaged, and I can tell, there are some pins, that appear to be deliberately bridged. Would it be possible to have a photo of the legs each side of the chip, so I could match the bridges, after resoldering? It's such a fine job, I will most likely fail, but I want to try it at least.

Did you zoom in on the photo I uploaded? There appear to be 3 bridged pins on the bottom.

I can see 2 of the 4 sides in the photo. I regret not taking a photo of mine before I started messing with the soldering iron. I'll attempt to replace the chip this weekend. As far as I know, bridges should be present on the PCB itself. It's not something that's added later. They only appear to be manually placed, but I heard there's are actual traces underneath. Will find out soon enough. I also have a replacement Evergreen CPU coming in a few weeks. I will probably do a manual job on Gainbery, but not test it till I try Evergreen first, because if I botch this, and end up frying the motherboard, it would be a stupid thing to do, before the other chip shows up. Only if the other one doesn't work (seller claims guaranteed functionality, but no visual proof, so who knows). Another regret is that I did not record original pin settings on the motherboard that came with the dead 5x86. The settings for my motherboard (Matsonic M601) are undocumented. So I have to guess, which chip matches the right configuration. Probably the same setting as DX2-66. I don't think the chip can be fried with incorrect jumpers. The worst that can happen, is I'll disrupt the VLB. The motherboard only runs 5V to the CPU, so aside from 5x86 upgrades with 5 to 3.3 regulator, there are no options above DX2. Overdrives don't work. I tested many. For me incentive to upgrade is significant, because despite having a chip selection all the way up to DX4-120, I can use none of them. I tried ordering 5V to 3.3V socket adapters awhile ago, but it turned out to be a scam. I thought I wasn't going to go back to the ruined Gainbery, but now it's either that, or the Evergreen, which should be an identical copy. I want to bring the system to Pentium level. Maybe even overclock to 160. Back in the day I had a P-100, and this 486 is the closest thing to what I remember from the 90's. My P-100 was bottom of the line, and the 486 I made to replace it, is top of the line, except the CPU, which is a half to a third too slow, compared to what I had. 5x86 is advertised as a P-75 equivalent. I have my doubts, but it's still better than a DX2-66. I'm less worried about the CPU and more about the voltage regulator. I hope replacement chips I got aren't fakes. If there's anything that could put an instant end to my motherboard, it would be a fake voltage regulator that shorts the 5V line, or sends it straight through to the 3.3V line.

Quick question - there is a missing pin on one of the sockets. Is that accidental damage, or deliberate? I don't know the 486 pinout. I assume some pins are redundant or NR.

Did you view the whole thread? There is a high-res photo of the gainbery unit showing all 4 sides very clearly. download/file.php?id=78619&mode=view

There shouldn't be any missing pins. The pins on these interposers break very easily, unlike the pins on a standard 486 ceramic package.

Yes. I looked a the photo. Bottom row is easy to see. The rest is a bit hard to distinguish. Also, I don't have any botch wires on my chip. It's more like in the top photos. Anyways, I think I can even wait till my evergreen arrives. If the adapter is similar, I can copy bridges off there.

Gainbery didn't survive my transplant, but at least it didn't damage anything else. I lost some PCB traces when desoldering the old chip. It's beyond repair, but that's alright. I've got other alternatives.

Now I'm testing Evergreen 5x86 133MHz. On first try it booted up using DX2-66 settings, but it would hang right after the BIOS. I remember reading on forums somewhere that the CPU didn't like 256K cache. I disabled cache entirely just to check, and surely, Evergreen booted up, but speed-wise it was practically the same as the DX2-66 it was replacing. Not unexpected. I then removed half of my cache chips, bringing the motherboard down to 128K and re-enabled external cache in BIOS. from there on, everything ran fine.

It's an old motherboard, so it cannot identify the processor. Thinks that it's a DX2 running at 100MHz, but speedsys reads it properly. 3Dbench went from 30.4 to 41.7. DOS games feel about a third faster, which matches my expectation. The motherboard and the graphics card are the bottleneck. Don't think I can squeeze any more performance out of that 486.

Tried overclocking by raising FSB to 40Mhz. Got 50.0 points in 3Dbench and 42.43 in Speedsys, however the CPU was only stable at below 30C at that speed. Motherboard offers no voltage adjustments, and they would've made no difference anyways, since the Evergreen regulates the input down to 3.3V

I have got hold of a Gainberry 133MHz CPU too - how should I take the heatsink off to see what the actual CPU is? Just warm it up?

Cheers,
Steve

If you want to know if it's AMD or Cyrix, just run speedsys. No need to remove the heatsink.

Interesting little bottom pins for the socket adapter. I wonder if those little holes allowed hot air jets to solder it all together in production?