Hello,

You use a via mvp3 board which is fine for 100mhz operation. For overclocking i recommend an Ali board since i`ve read that some do 120mhz stable. Asus p5a is a very good one, Gigabyte is good for this too but don`t know which one it was. (asus you need version 1.03 or 1.04 otherwise the plus chips don`t work, so not the 1.05 or 1.06 board)

Most chips do about 600mhz stable. Going much over 600mhz WILL be a challenge but i guess that is what you want 🤣

Extreem cooling is allways better but not needed for the plus chips, they don`t heat up as much as the older k6-2 and k6-3 (newer proces), and even a good socket A cooler has much more airflow compared to the coolers used back then.

Higher fsb is possible, higher clockspeed (stable) is very hard for the chip because of the very short pipeline. The fastest one from amd was a 570mhz i believe and there was a good reason for it.

2,9v will do nothing for stability, i bet it works just the same with 2,5v and i would keep that as a max.

I hate killing my old hardware otherwise i would try it but i think on my p5a my k6-2+ will be stable enough for a benchmark at 620mhz with only 2,3v. It runs perfect with 2,2v at 600mhz (even de k6-3+ does it and i use that now).

Please let us know if you can hit 700mhz with it, i`d love to see that!

Holy shit nuggets. This interests me highly, I don't think anyone's ever managed to hit 700MHz on a K6-2/3+ chip. Heck, even 600MHz is quite high for this chip. I can only help you by providing some pretty basic ideas about overclocking that we always seem to forget.

A) The chip itself. Every chip has different overclocking potential. Sometimes this can depend on stepping, but it really all boils down to luck. That's why top overclockers usually buy chips in loads, usually 10-20 I believe, to find the one that will overclock the best. My K6-3+ for example won't even POST at 600MHz. Best I can do is 550MHz, regardless of voltage.

B) The Motherboard. It is very common to see different motherboards perform much better, even if they're using the exact same chipset. You haven't told us what mobo you're using!

C) Cache on the motherboard. Cache that resides on the motherboard (L3 cache when using K6 plus CPUs) can get really unstable. I'm sure you know this, but just in case you don't, you should disable it through the BIOS when exceeding 100MHz FSB. It's just not rated to work this fast.

D) RAM. I hope you're using 133MHz sticks!

By all means, if you ever manage to hit 700Mhz, holy ballz I want to see that!

Oh, and if you want the absolute best cooling: carefully take the heatspreader off the cpu.

Then it looks very much like the old athlons and you have direct contact to the core (my k6-3+ was allready that way when i bought it used).

F2bnp wrote:

Holy shit nuggets. This interests me highly, I don't think anyone's ever managed to hit 700MHz on a K6-2/3+ chip. Heck, even 600 […]
Show full quote

Holy shit nuggets. This interests me highly, I don't think anyone's ever managed to hit 700MHz on a K6-2/3+ chip. Heck, even 600MHz is quite high for this chip. I can only help you by providing some pretty basic ideas about overclocking that we always seem to forget.

A) The chip itself. Every chip has different overclocking potential. Sometimes this can depend on stepping, but it really all boils down to luck. That's why top overclockers usually buy chips in loads, usually 10-20 I believe, to find the one that will overclock the best. My K6-3+ for example won't even POST at 600MHz. Best I can do is 550MHz, regardless of voltage.

B) The Motherboard. It is very common to see different motherboards perform much better, even if they're using the exact same chipset. You haven't told us what mobo you're using!

C) Cache on the motherboard. Cache that resides on the motherboard (L3 cache when using K6 plus CPUs) can get really unstable. I'm sure you know this, but just in case you don't, you should disable it through the BIOS when exceeding 100MHz FSB. It's just not rated to work this fast.

D) RAM. I hope you're using 133MHz sticks!

By all means, if you ever manage to hit 700Mhz, holy ballz I want to see that!

I didn't know about the cache, I'll try that later!

I have some ram sticks that I've tested in my athlonXP machine good for PC-155 @ CL2 and using those in here. I'm using a FIC VA-503+ on VIA Mvp3. So no one knows if these chips perfer +5v or +12v? I'll probably favor the +5v heavy power supply and try that one first then. Also (I'd have to go dig it up) but there's someone over on hwbot.org that has managed to get one of these chips to 750 Mhz before but they had to use liquid nitrogen for it. So I'm not expecting to achieve it with water cooling, but I'm going to try anyway.

meljor wrote:

Hello,

You use a via mvp3 board which is fine for 100mhz operation. For overclocking i recommend an Ali board since i`ve read that some do 120mhz stable. Asus p5a is a very good one, Gigabyte is good for this too but don`t know which one it was. (asus you need version 1.03 or 1.04 otherwise the plus chips don`t work, so not the 1.05 or 1.06 board)

I already have this chip running at 112 Mhz bus in this board, I've had a lot of difficulty getting it much higher, we'll see what happens today.

meljor wrote:

Extreem cooling is allways better but not needed for the plus chips, they don`t heat up as much as the older k6-2 and k6-3 (newer proces), and even a good socket A cooler has much more airflow compared to the coolers used back then.

I've already tried an AthlonXP heatsink on this thing a few months back, at 620 mhz and 2.9v the heatsink even with a fan gets really, really hot and it doesn't even really do an effective job of cooling the actual chip (I've taken the heatsink off after an hour of gaming and realized the cpu was still very warm to the touch). So I sort of shelved this project until I could get enough spare water cooling parts from my other 2 machines to water cool it and try again, so here we are.

meljor wrote:

2,9v will do nothing for stability, i bet it works just the same with 2,5v and i would keep that as a max.

Actually as I posted above in my first post, this chip will not even turn on / POST at all with anything less than 2.9v for 620 Mhz. So yes it is required both to function and stability. It won't even come on at all with 2.5v, I've tried that. I may have to try all the way up to 3.2v with higher clocks, maybe.

K6+ voltage range is about 1,6v-2.0v so yess, at 2,9v it will be pretty hot as it is a 50%!!! increase in voltage.

If most of us can get these stable at around 600mhz@2,1 or 2,2v it shouldn`t need 2,9v for 20mhz more. My k6-2+ doesn`t get hot at all and my k6-3+ is just a little warm under a socket a heatsink (nothing fancy).

I think it is time to try a different board and/or cpu, 620mhz is not extreme but you allready need extreme cooling and voltage...

meljor wrote:

K6+ voltage range is about 1,6v-2.0v so yess, at 2,9v it will be pretty hot as it is a 50%!!! increase in voltage.

If most of us can get these stable at around 600mhz@2,1 or 2,2v it shouldn`t need 2,9v for 20mhz more. My k6-2+ doesn`t get hot at all and my k6-3+ is just a little warm under a socket a heatsink (nothing fancy).

I think it is time to try a different board and/or cpu, 620mhz is not extreme but you allready need extreme cooling and voltage...

Yeah I'm about ready to just give up entirely on this. This motherboard is just extremely stupid flakey and sometimes works sometimes doesn't, and really unstable all around at anything other than 100 Mhz FSB. Maybe it's an old defective board, I don't know. It was a nice idea and I feel I've spent too much time in it already, I'm just going to abandon it and go do something else instead. I wish I could water cool stuff like an AthlonXP chip and try that, but I'd have to come up with some sort of bracket to mount the waterblock on the CPU right. I have a spare A-XP board here, just no way to get this block on it.

Anyway.. thanks for looking, but this is the last post in this thread and I'm just going to consider this project dead for now, maybe if I can find a different Super Socket-7 board some day I might re-visit it.

If you remove the IHS, be extremely careful with the die. It chips very easily. I tried once and even with foam supports at the corners it chipped and died.

Frankly I'm not sure what to expect here. These CPUs don't consume much power so cooling isn't a challenge. The only method I would expect to bring results would be phase change or Peltier cooling to dramatically reduce the operating temp and improve the properties of the semiconductor. Back in 1998-1999 there was a company called Kryotech that built refrigerated K6 machines with a factory overclock.

You guys made me remember of this http://redhill.net.au/c/c-e.html#k6-3-450
The last entrey, regarding AMD K6-III+ 450 😀

These chips with the AGP card powered by the AGP slot, is this going to use more of the +5v rail on a power supply, or more of the +12v rail on a power supply?

According to this AGP pinout:
http://pinouts.ru/Slots/agp_pinout.shtml
I only see 1 pin on the slot which provides 12V, so it can't be doing much. There are a few 5v pins, and lots of 3.3v pins.

[edit: the onboard 3.3v rail is driven by the PSU's 5v rail. PSU 5v does basically everything on your board. See next post.]

Even among later video cards that had auxiliary power connectors, it wasn't until several generations later that they started to pull a majority of their power from the 12V rail. I know on the ATI side, most of the power came from 5v up through the Radeon 9800. The "X" series was the first to primarily use 12v. There was an xbitlabs article which measured this. I don't know when the transition occurred on the nVidia side.

I have an FIC VA-503+ stored, so I got it out and poked around with a multimeter. I was curious and might as well share what I found.

There is an onboard regulator that provides onboard 3.3v power. It draws power from the 5v supply. The 3.3v output connects to the AGP and DIMM slots. I didn't check where else, but I would presume it provides 3.3v over the whole board.
The ATX connector's 3 3.3v pins don't even have continuity to each other. This strongly suggests they lead nowhere and aren't used. This makes sense since the legacy AT connector doesn't have 3.3v, and the board needs to work with either type of PSU.

There is an empty capacitor position near the AGP slot at CT14. It is connected to the onboard 3.3v rail which also leads to the AGP slot. It may have been intended to help stabilize the 3.3v supply at the AGP slot, but was omitted in production. Adding a capacitor here couldn't hurt, but I wouldn't expect any miracles.

There is another onboard regulator for the CPU Vcore. Again 5v power is used. ATX 5v leads through an inductor, to MOSFET Q7, and the output goes through another inductor on it's way past the Vcore caps.
Q7's gate is controlled by U19, which is an SC1101CS voltage regulator. This seems to be a simpler voltage regulator than what was used on P2/later boards. Vcore gets pulsed upward by the 5v supply, but it appears nothing pulls it toward ground. P2/later boards regulate by rapidly switching Vcore in both directions, this just works one way. It's probably typical of socket-7 boards.
The switching behavior is smoothed out partly with the inductor and also by the capacitors. My Vcore caps are cheap junk made by "I.Q.". Bad caps weren't as problematic on socket-7 as they were on later boards, but I'd probably replace them if I was about to use the board. It might help smooth the Vcore better and protect against the risk of later failure (which can sometimes damage other components), but I don't know if it would actually help overclocking results any.

But to answer the PSU question, this board uses the PSU's 5V rail for essentially everything. 12v would surely lead to the ISA slots and might have some other minor function, but it's not going to anything significant. It appears 3.3v isn't even connected.
I've read that there are some PSUs which use joint regulation on the 3.3v and 5v rails. That might be bad news with a board like this, where 3.3v has no load and 5v has the heaviest.

Thank you for your information, that's good to know. I have a couple power supplies here, some slightly older ones in good shape that give more capacity in the +5v side, around +35 amps. I have another +5v one that gives 55 amps for +5v, but it's in my overclocked AthlonXP system. And then I have a modern one that has more +12v capacity and very small load for +5v (A thermaltake 430 watt), I think it's like +35 amps for 12v and only 20 amps for +5v.

So I know what to use when overclocking these older systems now. I still think I need to find that Asus one based on the ALI chipset before trying to overclock K6-III again. I saw one on ebay but I don't have any spare money at the moment, trying to fund water cooling for my two main computers at the moment.. which is probably going to be another $1k and 9 months, so this will just have to be shelved for a later date. Maybe I could try overclocking my Slot-A athlon chip if I could ever figure out how to get the casing off and get access to the chip. Those things might benefit from water, they run really hot. That's a different topic though.