386SX wrote on 2022-02-28, 12:49:

Sphere478 wrote on 2022-02-28, 12:43:

386SX wrote on 2022-02-28, 12:33:

This is impressive! Anyway were some dos low level cache benchmarks tested to check the full cache status or stability?

Earlier in the thread it was discussed how to set up memtest for a cache burn in test

Any info from the Phil dos cache testing benchmarks like CHKCPU or CACHECHK or SpeedSys? 😉

If you feel like one of them would be a good test, let us know 😀

3faltigkeit wrote on 2022-02-28, 10:24:

Hi, I wasn't at home over weekend. Which information do you need exactly? The CPUID regarding aida32 is 05D4h (extended CPU revi […]
Show full quote

Hi, I wasn't at home over weekend. Which information do you need exactly?
The CPUID regarding aida32 is 05D4h (extended CPU revision: 06D4h).

I am afraid at the moment I can't do risky experiments. Because I only have one Socket 7 Board and only one K6-2+.
Furthermore I don't have a multimeter at hand to messure.

I'm surprised to read that the 350AUZ is o5D4, but thanks. That fills a gap in my database 😀

Sphere478 wrote on 2022-02-28, 13:34:

If you feel like one of them would be a good test, let us know 😀

I'm thinking I might try modding the K6-2+ 500 cause I'd not risk the K6-2+ 550 for now.. maybe I'll try to see how easy it would be on a regular cheap K6-2 but I've never removed the metal cover so I need to find some thin tool for it.

Back in the day some guys reported success with using dental floss to go through the corners... have to use some effort though, not like normal flossing, like the dental hygienist does to make your gums bleed so they can say you should floss more often 🤣

BitWrangler wrote on 2022-02-28, 17:55:

Back in the day some guys reported success with using dental floss to go through the corners... have to use some effort though, not like normal flossing, like the dental hygienist does to make your gums bleed so they can say you should floss more often 🤣

😁 😁

Not sure it's mentioned in this thread, but https://www.tomshardware.com/news/new-mod-dou … 2-cpu-from-2000
putting the link in here just in case 😜

Tetrium wrote on 2022-03-01, 09:56:

Not sure it's mentioned in this thread, but https://www.tomshardware.com/news/new-mod-dou … 2-cpu-from-2000
putting the link in here just in case 😜

Yeah! we made the tom’s hardware news! I heard!

Thermalwrong wrote on 2022-02-25, 00:19:

Yeah, verified. Thanks for confirming it probably isn't indium solder. It was just the silicon all the way on top of the core. I took the heatgun to it thinking it was indium, then it just unceremoniously popped off - it still has a core! That's wonderful! 😁
I confirmed that the edge silicon was cleared with some dental floss.

Something like 260C made it pop off with a little bit of pressure from the craft knife.

Can you give an update on what happened with your mod for the database?

Sphere478 wrote on 2022-03-01, 10:00:

Tetrium wrote on 2022-03-01, 09:56:

Not sure it's mentioned in this thread, but https://www.tomshardware.com/news/new-mod-dou … 2-cpu-from-2000
putting the link in here just in case 😜

Yeah! we made the tom’s hardware news! I heard!

Time to save some K6-2+ now that the world knows...

Tetrium wrote on 2022-03-01, 09:56:

Not sure it's mentioned in this thread, but https://www.tomshardware.com/news/new-mod-dou … 2-cpu-from-2000
putting the link in here just in case 😜

Oh shit...decoupling capacitor, it's not. That was painful to read.

fool wrote on 2022-03-01, 17:21:

Tetrium wrote on 2022-03-01, 09:56:

Not sure it's mentioned in this thread, but https://www.tomshardware.com/news/new-mod-dou … 2-cpu-from-2000
putting the link in here just in case 😜

Oh shit...decoupling capacitor, it's not. That was painful to read.

I can ignore the capacitor/resistor confusion. We first assumed it was a capacitor as well. But the whole article appears badly researched. That was what was disappointing to me.
Especially comparing a 450MHz Pentium II (from 1998) to a 500MHz K6-III+ (from 2000) and stating the K6-III+ "changed the panorama back at the time". The Athlon changed the panorama, the Pentium II was rivaled by Intel's own Mendochino Celeron as we know.
The mod would have hardly changed anything back in 2000. (Apparently it doesn't even work on the earlier stepping!) The faster Celerons and Durons significantly beat any K6-III+ (in many applications even clock-for-clock!) and I can hardly imagine a K6-2+ being significantly cheaper than a much faster Duron or Celeron when they were still being produced.
The article kind of reads like the author at first thought this was a K6-2 (from 1998) that got unlocked and later realised his mistake and partially rewrote the article.
Oh well, worse things happen. 🤷‍♂️

I took it as meaning the K6-III original, which was fastest CPU for like 2 months. Half speed cache did not look good against on die cache.

BitWrangler wrote on 2022-03-01, 18:37:

I took it as meaning the K6-III original, which was fastest CPU for like 2 months. Half speed cache did not look good against on die cache.

That would support my half-ass-rewrite theory since the article states: "The K6-III+ changed the panorama back at the time. A 500 MHz core clock and an L2 cache size bump to 256KB outperformed the top-end Pentium II model".
The K6-III came in 333, 400 and 450MHz. (No 500MHz.) It was faster in many desktop applications but slower in most 3D games than the 500MHz Pentium III.

The 450MHz K6-III supposedly cost 476USD at launch btw. That is significantly cheaper than the Pentium III when it launched but an overclocked Mendocino Celeron was still the obvious price-performance champion.

BitWrangler wrote:

The mod would have hardly changed anything back in 2000.

Yep, difference between 128Kb vs 256Kb L2 cache on K6-2 architecture is rather small. Although probably still nice, if you're trying to squeeze all performance you can have from the bottom of the barrel CPUs in early 2000s.

Next person to try this mod try it with the resistor(0 ohm link) removed first there seems to be conflicting reports on wether or not this results in the 256k showing up, or a non working chip. (Reports seem to indicate no harm is done, just that the resistor may have to be installed again)

I think it may be a easy way for novice modders to accomplish the mod. Just plop a solder ball on it and wipe it off,

This deserves a Skalabala award 😀
Nicely done. I will try it out, my K6-2+ 570 hits 720MHz on air. If its stable with 256Kb well you know what's coming next 🤣

Skalabala wrote on 2022-03-03, 18:04:

my K6-2+ 570 hits 720MHz on air.

How much Volt do you need to reach 720MHz? 😮

Skalabala wrote on 2022-03-03, 18:04:

This deserves a Skalabala award 😀
Nicely done. I will try it out, my K6-2+ 570 hits 720MHz on air. If its stable with 256Kb well you know what's coming next 🤣

720 on air?!? Wow. Did you just get lucky? Or did you dig through a bunch of chips?

W00t Skalabala award!!!!!

RaiderOfLostVoodoo wrote on 2022-03-03, 18:10:

Skalabala wrote on 2022-03-03, 18:04:

my K6-2+ 570 hits 720MHz on air.

How much Volt do you need to reach 720MHz? 😮

Um, dude, your SN and Profile pic. Amazing.🤣

I also am wondering what the voltage is.

Got a bunch of K6-2+ 450ACZ today. Picked one of them and was able test this mod successfully.

AMD-K6-2+/450ACZ
2.0V CORE/3.3V I/O
A 0030MPMW

-Max stable clock without mod: 550MHz @ 2.0V
-Max stable clock with L2 Mod: 550MHz @ 2.1V
-does not run at 600MHz (not even at 2.4V)
-The die color is red instead of blue like my 500ACZM

@Sphere478
I read your pm and keep an eye on it. Can't reply to your pm yet, because of my newbie status (need more posts).

Fritzchens Fritz wrote on 2022-03-03, 20:35:

Got a bunch of K6-2+ 450ACZ today. Picked one of them and was able test this mod successfully. […]
Show full quote

Got a bunch of K6-2+ 450ACZ today. Picked one of them and was able test this mod successfully.

AMD-K6-2+/450ACZ
2.0V CORE/3.3V I/O
A 0030MPMW

2022-03-03_185604_393___57W_2.0V_550MHz_128KB___450ACZ_0030MPMW.png 2022-03-03_203944_045___60W_2.1V_550MHz_256KB___450ACZ_0030MPMW.png
2022-03-03_192310_355___54W_2.0V_550MHz_256KB___450ACZ_0030MPMW.png

-Max stable clock without mod: 550MHz @ 2.0V
-Max stable clock with L2 Mod: 550MHz @ 2.1V
-does not run at 600MHz (not even at 2.4V)
-The die color is red instead of blue like my 500ACZM

@Sphere478
I read your pm and keep an eye on it. Can't reply to your pm yet, because of my newbie status (need more posts).

Cool, thanks for that. This is the first to confirm my theory about the CPUID changing as well. It'll be good to see if this is the same for all CPUs this works on.

Sphere478 wrote on 2022-03-03, 19:21:

720 on air?!? Wow. Did you just get lucky? Or did you dig through a bunch of chips? […]
Show full quote

Skalabala wrote on 2022-03-03, 18:04:

This deserves a Skalabala award 😀
Nicely done. I will try it out, my K6-2+ 570 hits 720MHz on air. If its stable with 256Kb well you know what's coming next 🤣

720 on air?!? Wow. Did you just get lucky? Or did you dig through a bunch of chips?

W00t Skalabala award!!!!!

RaiderOfLostVoodoo wrote on 2022-03-03, 18:10:

Skalabala wrote on 2022-03-03, 18:04:

my K6-2+ 570 hits 720MHz on air.

How much Volt do you need to reach 720MHz? 😮

Um, dude, your SN and Profile pic. Amazing.🤣

I also am wondering what the voltage is.

Hahaha 🤣 Its completely unstable though. 2.4V
But with forced cooling it will do well :p