cyclone3d wrote on 2022-02-21, 06:06:

BitWrangler wrote on 2022-02-21, 06:02:

There's the contract fulfilment thing too, X hundred thousand processors by date Y for Z amount of money, or the lawyers built in penalties. What are you gonna do, say "Hold back that thousand or two, they're too good, risk the penalty clause." ??

AMD did the same sort of thing with the Slot-A Athlons. But that was even with the retails ones.

There was more demand for the lower cost ones and the yield was so good that you would almost always get a 150-200+ Mhz higher die than the CPU was sold as.

I hope I'm wrong and I'm mostly going by the rumor that k6 yields were not good. Where the general consensus has been something like only 1 in 10 k62s would overclock more than 50mhz. As in a 450 would usually get 500mhz but not 550. Which means that a K6450 is not simply a gimped 550, but actually a binned cpu due to bad yields.

K6III+ is also a much more rare cpu than a K6II+ and higher clocked versions of K6III+ are also more rare than lower clocked ones. Which leads me to believe there really was a yield issue.

looking at my 570 here..

what is the thinnest razor blade you can order?

these maybe?

you probably want just a single sided like the ones you use on windows.

A sturdy piece of string may be a good alternative to a razor blade when cutting the thin layer of silicone. Should be harder to damage anything with it too.

Doornkaat wrote on 2022-02-21, 06:40:

A sturdy piece of string may be a good alternative to a razor blade when cutting the thin layer of silicone. Should be harder to damage anything with it too.

a piece of thread actually may not be a bad idea

Sphere478 wrote on 2022-02-21, 06:44:

Doornkaat wrote on 2022-02-21, 06:40:

A sturdy piece of string may be a good alternative to a razor blade when cutting the thin layer of silicone. Should be harder to damage anything with it too.

a piece of thread actually may not be a bad idea

Yeah; thread, not string.
Guess who's not a native speaker!😉😅

Doornkaat wrote on 2022-02-21, 06:46:

Sphere478 wrote on 2022-02-21, 06:44:

Doornkaat wrote on 2022-02-21, 06:40:

A sturdy piece of string may be a good alternative to a razor blade when cutting the thin layer of silicone. Should be harder to damage anything with it too.

a piece of thread actually may not be a bad idea

Yeah; thread, not string.
Guess who's not a native speaker!😉😅

Ehhh.. they are ceramic, not like the Pentium III and Celeron S370 CPUs that have the traces close to/on the top layer.

I would like to see somebody try to use string to cut that glue. I'll stick with my razor blade method.

I'm a native "English" speaker and string/thread are interchangeable though thread is usually used more in relation to sewing.

cyclone3d wrote on 2022-02-21, 07:09:

Ehhh.. they are ceramic, not like the Pentium III and Celeron S370 CPUs that have the traces close to/on the top layer. […]
Show full quote

Doornkaat wrote on 2022-02-21, 06:46:

Sphere478 wrote on 2022-02-21, 06:44:

a piece of thread actually may not be a bad idea

Yeah; thread, not string.
Guess who's not a native speaker!😉😅

Ehhh.. they are ceramic, not like the Pentium III and Celeron S370 CPUs that have the traces close to/on the top layer.

I would like to see somebody try to use string to cut that glue. I'll stick with my razor blade method.

I'm a native "English" speaker and string/thread are interchangeable though thread is usually used more in relation to sewing.

Thanks for the explaination 'string vs. thread'!👍
I cut the glue on a K6-2 using some rigid sewing thread before. If it's the same glue on K6-2+ CPUs it should work just the same.
I'm not concerned about the substrate that's obviously ceramic but rather about damaging the capacitors by accident. They are less rigid than the substrate. Maybe I'm just overly cautious.🤷‍♂️

Until I have seen actual proof I am very suspicious of a (capacitor) actually being able to strap cache config bit.
You can't use capacitors as pullup/downs for GPIO/Register bit mask strapping.

Unless that is not a capacitor, but a resistor disguised as one.

I'll be trying this myself on my K6-2+ 500ACZM.

Well, that's interesting... I have no real use for all 3. Got those for $68 including shipping so maybe one could be sacrificed for science, let's see.
I remember delidding one K6-2 just for fun over 20 years ago. Used a thin knife or razor blade and it worked after all. Glue texture might have shifted by time, possibly not to the direction of more elastic.

0xCats wrote on 2022-02-21, 08:06:

Until I have seen actual proof I am very suspicious of a (capacitor) actually being able to strap cache config bit. You can't us […]
Show full quote

Until I have seen actual proof I am very suspicious of a (capacitor) actually being able to strap cache config bit.
You can't use capacitors as pullup/downs for GPIO/Register bit mask strapping.

Unless that is not a capacitor, but a resistor disguised as one.

I'll be trying this myself on my K6-2+ 500ACZM.

Well, cats I’m happy to say that it’s confirmed.
(May not be a capacitor, I haven’t measured it)

I used one of the shaver blades like I talked about earlier,

Got all four corners.

Tried heat, nothing, it’s rubber like high temp rubber maybe kinda like the rubber they use on windshields. Best way to scrape it off is with a piece of rough fabric or finger nail.

I’ll try and see if some RTV isn’t basically the same. (That may be exactly what it is…)

They used the same stuff on the die and the corners on mine.

Perhaps I can order some thermal adhesive and some rtv.. will update.

Anyway after cutting all four corners it simply came down to putting a normal razor under the lip on an entire side and leveraging that side up until it went “pop” and there it was!

I cleaned up the solder a bit between the installed pic and the pic on the wood.

I gotta say, that resistor/capacitor was not easy to move guys. The processor sucked up a lot of heat and made it hard to get it all to flow. But I got it after about 20 min.

I kinda feel like I may have opened pandora’s box here leading to the possible distruction of many chips.

Guys, transplanting that cap/resistor is not easy, so be sure you are up to it. I made use of a hot air station, a soldering iron and flux paste and a lot of patience and many tries to get it.

As a retro enthusiast I must emplore you guys, please take the time to reinstall the IHS and don’t chip the die. There are only so many of these left. The more you destroy the more expensive they will become.

Enjoy, and I am very glad this hunch paid off!! A real victory for socket 7 enthusiasts today!

Don’t forget to scrape off the rtv on the die it’s a thin film. Use your thumb and a flat fabric

I've just gone and measured the part and it is indeed a resistor (0 Ohm link)

0xCats wrote on 2022-02-21, 08:34:

I've just gone and measured the part and it is indeed a resistor (0 Ohm link)

I suspected. Makes sense.

So what do we think that one at the top is for, and the other empty spaces?

The 570 2+ is IMHO the perfect chip for this. It’s a high clock chip, and there seems to be a near endless supply of them for sale right now on ebay. It’s a shame to molest collectable chips like a 2+500, 550, etc.

So mine also unlocks to K6-3
But I now need to run it at 2.4V and just 66Mhz FSB 2x(6x) Multi to get it to run stable.

0xCats wrote on 2022-02-21, 08:52:

So mine also unlocks to K6-3
But I now need to run it at 2.4V and just 66Mhz FSB 2x(6x) Multi to get it to run stable.

Bummer.

Maybe I should do some tests before reinstalling my IHS I guess…

Okay, here we go.

570 for the win! (At 600mhz/256k)

Just started a super pi 1m will run 32 m if that passes

So far, no issues.

spoke too soon super pi error, bumping by 0.1v to 2.2v it's worth noting that the heatsink I am using isn't the best it's getting kinda hot. but that in of it's self is kinda a good test. I don't wanna put one of my better heatsinks on it because the die is fragile till I get the ihs back on this one is light pressure

update 2

tryin again with more airflow over heatsink and 2.2v so far so good

update 3 ran pi for like 5 min no issues, rebooted and lowered voltage back to 2.1v (lowest board can go)

update 4 all seems fine, just a cooler getting too hot. this chip (570 2+) works fine at 600mhz and 256k

1m spi

Memtest 86+ can be set up to test the cache by doing this btw.

Press C
Select 2
Select Upper address range
enter 262144 bytes
enter
0 to continue

Figured out thanks to cats help

Warlord wrote:

K6III+ is also a much more rare cpu than a K6II+ and higher clocked versions of K6III+ are also more rare than lower clocked ones. Which leads me to believe there really was a yield issue.

Yields are a thing, but it's no an issue for us, because we can add more voltage and it will most likely will work stable. Most K6 with cache can work at 550 Mhz, but voltage may vary.

Disabled cache might not be down to bad yields, not strictly. By that time AMD had a good idea how well the chips would turn out and perhaps designed a fast cache a bit on the optimistic side. Simply put, they knew not every chip is going to turn out great but some will, and those that will be sold as III+. The rest will be downgraded to 2+ to meet the specs. That's the price for lower latency (which also comes with lower clock ceiling but those are mobile chips after all, perf/W is what they were made for).

We have some results now and already we can see the mod might require better control over temperature, or higher voltage, or both.

BTW if I was desiging that chip I would make both halfs of the cache configurable. I mean unless it is partly configured in the silicon, there should be two jumpers on the CPU, to be set depending on which half of the 256k cache is performing better. So some 2+ chips might require different jumper to be moved, but the final position would be like what III+ parts have.

That's fantastic news !