QFP Cx586 and Am5x86 transplants are up next in my pipeline. Hope my 1 year old solder paste is up to the task. I kept it in the fridge.

feipoa wrote on 2025-09-24, 08:23:

QFP Cx586 and Am5x86 transplants are up next in my pipeline. Hope my 1 year old solder paste is up to the task. I kept it in the fridge.

The attachment QFP_Cyrix_5x86_and_Am5x86_transplant_1.JPG is no longer available

Ugh, I wish. Despite my efforts I have been able to locate exactly zero of any suitable interposers, seems like any supply from China has dried up. Tearing apart a working one with an AMD is just not in the budget. It makes me wonder if a motherboard with pads for a qfp 486dx would be a suitable candidate. At one point I owned an ATC-1415 that has these - It was pretty decent board, the only issue I remember it had was also having a hole in the cacheable range 56-64mb like some others with similar bioses do. I wonder if I obtained another if I could make that qfp cyrix work on it. Unfortunately, I have never seen a photo of one of these with a qfp actually in place to compare - I imagine that the boards sporting a socket 3 (all of them?) may be missing a few parts necessary for the qfp to work which may make things more complicated

bertrammatrix wrote on 2025-09-30, 17:13:

feipoa wrote on 2025-09-24, 08:23:

QFP Cx586 and Am5x86 transplants are up next in my pipeline. Hope my 1 year old solder paste is up to the task. I kept it in the fridge.

The attachment QFP_Cyrix_5x86_and_Am5x86_transplant_1.JPG is no longer available

Ugh, I wish. Despite my efforts I have been able to locate exactly zero of any suitable interposers, seems like any supply from China has dried up. Tearing apart a working one with an AMD is just not in the budget. It makes me wonder if a motherboard with pads for a qfp 486dx would be a suitable candidate. At one point I owned an ATC-1415 that has these - It was pretty decent board, the only issue I remember it had was also having a hole in the cacheable range 56-64mb like some others with similar bioses do. I wonder if I obtained another if I could make that qfp cyrix work on it. Unfortunately, I have never seen a photo of one of these with a qfp actually in place to compare - I imagine that the boards sporting a socket 3 (all of them?) may be missing a few parts necessary for the qfp to work which may make things more complicated

I have had several 486sx chips on qfp to pga adapters that came in OEM branded systems.

If your motherboard allows voltage adjustment they “should “ work

rmay635703 wrote on 2025-10-01, 03:33:

bertrammatrix wrote on 2025-09-30, 17:13:

feipoa wrote on 2025-09-24, 08:23:

QFP Cx586 and Am5x86 transplants are up next in my pipeline. Hope my 1 year old solder paste is up to the task. I kept it in the fridge.

The attachment QFP_Cyrix_5x86_and_Am5x86_transplant_1.JPG is no longer available

Ugh, I wish. Despite my efforts I have been able to locate exactly zero of any suitable interposers, seems like any supply from China has dried up. Tearing apart a working one with an AMD is just not in the budget. It makes me wonder if a motherboard with pads for a qfp 486dx would be a suitable candidate. At one point I owned an ATC-1415 that has these - It was pretty decent board, the only issue I remember it had was also having a hole in the cacheable range 56-64mb like some others with similar bioses do. I wonder if I obtained another if I could make that qfp cyrix work on it. Unfortunately, I have never seen a photo of one of these with a qfp actually in place to compare - I imagine that the boards sporting a socket 3 (all of them?) may be missing a few parts necessary for the qfp to work which may make things more complicated

I have had several 486sx chips on qfp to pga adapters that came in OEM branded systems.

If your motherboard allows voltage adjustment they “should “ work

I initially thought this as well, however unfortunately 486sx comes in a qfp with less leads then a qfp 486dx2/4/5x86 has, making the interposers unusable for those CPU

486SX has a different pin out and pin count.

I sent a follow-up about the interposers, but don't get your hopes up.

Maybe you can assemble one of RayeR's interposers?

feipoa wrote on 2025-10-01, 08:58:

486SX has a different pin out and pin count.

I sent a follow-up about the interposers, but don't get your hopes up.

Maybe you can assemble one of RayeR's interposers?

Are these the interposers?

https://rayer.g6.cz/hardware/retropc2.htm#486DX5_INTERPOSER
https://rayer.g6.cz/hardware/retropc2.htm#486 … 5_INTERPOSER_V2

Ya, that's it. There's a post about this interposer somewhere on VOGONS but I'm having trouble locating it. I think it has 6 A VRM.

The Chinese sourced interposers you see in the above photo are on the noisier side with respect to Vcc.

120 mV noise without any caps
60 mv noise with 2x10 uF tantalum and 4x100 nF ceramic

whereas with the IBM Thinkpad interposer you see in the photo, 16 mV. I'm going to try increasing the tantalums from 10 uF to 100 uF to see if it helps any.

feipoa wrote on 2025-09-23, 11:04:

Has anyone tested a QFP Cyrix 5x86-120 with markings G5A9549N?

Yes, it's S0R5. Unfortunately, it only worked for one session on the interposer. I think there might be a bad solder joint that lost contact after the interposer was pulled out of the socket. When I have more time, I'll check all the pins again and hope that it will work again, but currently, I have neither the time nor the motivation to check all the 208 contacts again 🙁

Stesch wrote on 2025-10-02, 07:15:

feipoa wrote on 2025-09-23, 11:04:

Has anyone tested a QFP Cyrix 5x86-120 with markings G5A9549N?

Yes, it's S0R5.

That's unfortunate, but it's the answer I was expecting.

Which interposer are you using?

With regard to a bad solder joint, for the most part, it is easier just to run a bead of flux paste along the pins and do a one-by-one tap with a fine tipped iron. I often will lightly press the pin inward and down with the iron tip to see how high the pin is off the substrate. If I pull away quick enough with the iron, the pin may get soldered closer to the pad. For any pin lacking solder, 0.38 mm solder can be added. There's also 0.15 mm solder rolls that I use, but I think it only comes as lead-free.

I've populated my interposers, but testing will come later; I'm engaged in an "assemble-athon" with various pending projects. Here's how I landed the caps:

The 100 uF caps are of the 2917 footprint with 75 m-Ohm ESR. Will the low ESR cause issues?

The solder pads on the interposers with the silkscreened Y-1 need 2917 and I only have new 100 uF and 330 uF. The other interposer can use the 2312 footprint.

I've soldered some interposers with 47 uF and 300 m-Ohm ESR, and others contain 10 uF and 450 m-Ohm ESR.

I have some 2917's at 330 uF and 50 m-Ohm ESR, but are only rated for 6.3 V. Is it correct that we generally need tantalums with at least 2-3x the target operating voltage? I've seen some interposers which output 3.3 V get away with 6.3 V tantalums. Target voltage is in the 3.3 V - 4.2 V range.

I also have some 220uF Niobium Oxide caps in footprint 2312, but only with a 6.3 V rating.

The IBM Thinkpad units shown with heatsink/fan run well at 2x66 MHz, but are S0R5. The plan is to replace those with S1R5 which will do 2x66 or 3x50.

feipoa wrote on 2025-10-02, 08:48:

That's unfortunate, but it's the answer I was expecting. […]
Show full quote

Stesch wrote on 2025-10-02, 07:15:

feipoa wrote on 2025-09-23, 11:04:

Has anyone tested a QFP Cyrix 5x86-120 with markings G5A9549N?

Yes, it's S0R5.

That's unfortunate, but it's the answer I was expecting.

Which interposer are you using?

With regard to a bad solder joint, for the most part, it is easier just to run a bead of flux paste along the pins and do a one-by-one tap with a fine tipped iron. I often will lightly press the pin inward and down with the iron tip to see how high the pin is off the substrate. If I pull away quick enough with the iron, the pin may get soldered closer to the pad. For any pin lacking solder, 0.38 mm solder can be added. There's also 0.15 mm solder rolls that I use, but I think it only comes as lead-free.

Thank you for the tip!

I'm using RayeRs Interposer, the same BobocoCZ used for his Cyrix QFP CPU. Right now, 0.5mm is the smallest solderwire I have, perhaps I'll order some thinner ones. Hopefully I'll have more time soon 😀

I have an IBM 5x86C (3V3100GF) currently overclocked to 120MHz. I've enabled the enhancements via Peter Moss utility and it shows enabled for the values but I'm not seeing any difference in benchmarks. Alternatively, when I use the IBM utility if I set things like FP FAST to from 'x' to '1', it'll revert to 'x' after pressing F3 to save to registry. What am I doing wrong?

On some systems or CPUs, I have also experienced that with the IBM utility. I'm not sure why it happens on some and not others. Just use the Peter Moss or Evergreen utilities.

What enhancement are you enabling and which benchmark shows now improvement? I suggest looking at my benchmark charts to identify which software showed what improvement with different Cyrix enhancements.

feipoa wrote on Yesterday, 04:19:

On some systems or CPUs, I have also experienced that with the IBM utility. I'm not sure why it happens on some and not others. Just use the Peter Moss or Evergreen utilities.

What enhancement are you enabling and which benchmark shows now improvement? I suggest looking at my benchmark charts to identify which software showed what improvement with different Cyrix enhancements.

I'm pretty much just doing the "default" from your document minus LINBRST. I know that even the IBM utility is doing *something* because enabling that does indeed crash my system. But in benchmarks, it's pretty much exactly the same as enhancements disabled. Even 3DBench is exactly the same (76.9). I can't remember what the Speedsys CPU value was before, but memory remained the same before and after. Is there a particular benchmark that is best to see if any difference is happening?

Thanks for the quick response!

Did you review the pdf document at the bottom of the original post Cyrix 5x86 Register Enhancements Revealed

This report, direct link here: download/file.php?id=9947 , goes into which benchmarks show what percent improvement for different register features. Refer to the spreadsheets on pages 13-16 of the linked report.

feipoa wrote on Yesterday, 05:27:

Did you review the pdf document at the bottom of the original post Cyrix 5x86 Register Enhancements Revealed

This report, direct link here: download/file.php?id=9947 , goes into which benchmarks show what percent improvement for different register features. Refer to the spreadsheets on pages 13-16 of the linked report.

Ahh whoops, I didn't get that far 😅

But everything seems to be "working" then, BTB on and LOOP off is pretty in line with the spreadsheet but I probably won't run it like that for normal use. Thanks again!

BTB enabled within DOS should be OK for most Cyrix/IBM 5x86 chips, depending on your clock rate. If you want to run BTB in Windows, you'll want S1R3 revision CPUs, which were normally made prior to and including Week 44 of 1995.