Sorry if this is a dumb question, couldn't find it mentioned skimming through... Is a pre-plus K6-2 "mobile" expected to respond to multiplier change by software?

I've got this K6-2 1.8V core that everything says is a mobile variant.

BitWrangler wrote on 2021-05-25, 18:36:

Sorry if this is a dumb question, couldn't find it mentioned skimming through... Is a pre-plus K6-2 "mobile" expected to respond to multiplier change by software?

I've got this K6-2 1.8V core that everything says is a mobile variant.

I'm also interested in this, I have a K6-2/366 that is apparently a mobile CPU as well, this would make using it a lot more interesting..

It's been quite some time since those questions were asked, but setmul makes use of PowerNow!, which was introduced in K6-2+/K6-III+ CPUs. Earlier K6-2 with lowered voltage are not supported as they provide no means to control multiplier via software. You will be able to disable L1 cache on this CPUs, but not L2 cache which is on motherboard.

So I cannot change the multiplier of my PII and PIII Slot-1 processors. But I can disable L2 and (most important) L1 cache with SetMul in these processors? My Aopen 440BX chipset based does not have the L1 enable/disable option in the BIOS.

Hey Guys,

I've been finding all kinds of uses for this tool. Again, many thanks for your work on this!

I found a way to run executables from CONFIG.SYS. This allows me to run things like SETMUL, prior to memory management loading. Please see the attached tool.

-Brian

I've been using this program to disable L1 cache on my Cx5x86 for so long now, but only today I realized it can actually manipulate the multiplier.. *facepalm*

I'll do some tests later on and see how granular I can get with speed adjustments. Sweet.

Is setmul applicable and fully functional to a Socket A Athlon 1000C and Abit KT7 (non raid) build?

Don't think Thunderbirds did it, but on those if you've got motherboard clock multi settings you can set them to lower multipliers. However if you stick any old Palomino in there and do an L5 mobile mod, it will respond ... provided the motherboard lets it ... some say it's only guaranteed on KT266A up, but I've seen it work on a KT133 board... I think it's whether the maker did something specific to block it, or didn't allow it by coincidence i.e. shall I tie this "useless" line high or low, heads or tails.

Understood thank you.
I have already installed the board on a vintage brand new ATX case and recently I bought the CT4760 with Live Drive II the CT4860 and the Digital DIN header bracket, so it's time to give it life.

I don't know which GPU should I install. I have a Voodoo 3 on a K7 build. Any Matrox or S3 better than Nvidia of the time?

PhilsComputerLab wrote on 2016-04-09, 01:49:

lvader wrote:

After an aborted attempt of building a socket 7 K6 III+ system (CPU correctly detected but the system keep rebooting instead of […]
Show full quote

After an aborted attempt of building a socket 7 K6 III+ system (CPU correctly detected but the system keep rebooting instead of booting the OS) I've decided to try VIA C3 as it looks like it gives me even more flexibility. I've found a suitable mobo with 3 ISA slots and I have a few CPUs on the way, an Ezra 800 Mhz, Ezra-T 1 Ghz and a Nehemiah 1.2 GHz. Would I be right in saying the Ezra-t would be the one to go for for the greatest flexibility at the lower end of performance? Also is the multiplier locked to it's rated speed (i.e.133 x 7.5) or can I attempt to go higher (i.e. overclock). Presuming the multiplier isn't locked, With I cache disabled is it likely to overclock more than if all caches enabled (i.e. full speed) and give me more working multiplier options?

Also I'm I right in saying the main benefit of Ezra-t vs a regular Ezra is that the FSB isn't locked?

Thanks
Jeff

Keen to hear how you go with all of this. What motherboard did you end up going for?

I had the same problem as Ivader on my Intel 430TX chipset Asus TX97-X motherboard with K6-III+ 500 running at 6x83 with a patched 0112X_J2 BIOS from The Unofficial AMD K6-2/3+ Page. I haven't seen this issue and (simple, but not obvious) solution documented anywhere so it seemed worthwhile to comment on when I saw Ivader's post from many years ago.

Like Ivader said, the problem was: "CPU correctly detected but the system keep rebooting instead of booting the OS". The issue ended up being that my computer wouldn't boot with a K6-III+ and incompatible USB keyboard. If I switched to a Pentium MMX 233 with the same USB keyboard then the problem went away. Additionally, if I kept the K6-III+ and switched keyboards then the problem went away.

The final solution for me was to switch keyboards. I was initially using a Logitech USB keyboard that was apparently incompatible and switching to a Corsair USB keyboard solved the problem with the system continually rebooting. I eventually switched to a Filco USB & PS/2 keyboard that worked even better for my situation by allowing me to use PS/2 on retro systems and USB on my newer systems that support USB at boot.

JSO wrote on 2021-07-22, 05:05:

I don't know which GPU should I install. I have a Voodoo 3 on a K7 build. Any Matrox or S3 better than Nvidia of the time?

It depends what do you want. By performance that time (end of the 2000) nVidia GeForce 2 family was the fastest in most cases (in 16-bit; and low resolution 32-bit), but with most games in 32-bit high resolution Radeon DDR was faster. If the image signal quality the most important, the Matrox G400 MAX would be the best choise.
Some Geforce 2 GTS and Radeon DDR benchmark results:
https://www.tomshardware.com/reviews/ati,212.html

bbuchholtz wrote on 2020-11-30, 18:29:

I'm looking for a way to disable CPU cache, before the memory manager loads. A lot of folks use QEMM and DPMI. May be a worthwhile feature to add.

SetMul works fine with QEMM loaded. Just avoid loading QDPMI on CONFIG.SYS unless you really need it, since it has incompatibilities and most DOS applications that requires a DPMI server uses CWSDPMI.

I have tested K6-2+ with setmul. I am able to play some 386 games. So it works (multiplier set to 2 on FSB 100MHz, disabling BIOS cache and L1/L2 cache). But one thing I don't understand: Why Lotus still crashes with divider overflow? Another machine Pentium MMX 200MHz does not have this issue.

The original Bogomips on Linux were a timing loop calibration, not really a benchmark, and Pentium and MMX would get close to clock x 0.4 whereas K6 class would get clock x 2.0... however, this was later "fixed" at some point resulting in Pentiums also getting clock x 2... Anyway, what I'm saying is that there's a timing loop thing where if done a certain way the K6 "runs" 5x faster than a Pentium, so some software has problems with this... Windows 9x needs a K6 patch, and this game apparently also has problems.

BitWrangler wrote on 2021-10-24, 13:04:

The original Bogomips on Linux were a timing loop calibration, not really a benchmark, and Pentium and MMX would get close to clock x 0.4 whereas K6 class would get clock x 2.0... however, this was later "fixed" at some point resulting in Pentiums also getting clock x 2... Anyway, what I'm saying is that there's a timing loop thing where if done a certain way the K6 "runs" 5x faster than a Pentium, so some software has problems with this... Windows 9x needs a K6 patch, and this game apparently also has problems.

Ok, I have fixed it by hex editing lotus.dat and changing FB F7 F1 to FB 90 90 (nop) according to Vogons thread Lotus fix. It works now.

Can someone clarify a few things for me? I'm trying this out on an Intel 430HX with a Pentium MMX 166Mhz. I am testing this specifically with Wing Commander 1.

I can't see any different when I use SETMUL to disable L1 cache. I do, however, see a large slowdown when I disable "System Cache" from the BIOS. The readme says that it can't disable L2 cache that's on the motherboard, but also that there should be a much bigger difference when disabling L1 cache over L2.

Why then do I only see a slowdown when disabling L2 in the BIOS, but not at all when disabling L1 with SETMUL. I tested with and without L2 disabled in the BIOS.

Kahenraz wrote on 2022-10-28, 22:52:

Can someone clarify a few things for me? I'm trying this out on an Intel 430HX with a Pentium MMX 166Mhz. I am testing this spec […]
Show full quote

Can someone clarify a few things for me? I'm trying this out on an Intel 430HX with a Pentium MMX 166Mhz. I am testing this specifically with Wing Commander 1.

I can't see any different when I use SETMUL to disable L1 cache. I do, however, see a large slowdown when I disable "System Cache" from the BIOS. The readme says that it can't disable L2 cache that's on the motherboard, but also that there should be a much bigger difference when disabling L1 cache over L2.

Why then do I only see a slowdown when disabling L2 in the BIOS, but not at all when disabling L1 with SETMUL. I tested with and without L2 disabled in the BIOS.

20221028_184715_resize_1.jpg

I’ve never tried to do that on that processor. But try central tweaking unit also

Kahenraz wrote on 2022-10-28, 22:52:

Can someone clarify a few things for me? I'm trying this out on an Intel 430HX with a Pentium MMX 166Mhz. I am testing this spec […]
Show full quote

Can someone clarify a few things for me? I'm trying this out on an Intel 430HX with a Pentium MMX 166Mhz. I am testing this specifically with Wing Commander 1.

I can't see any different when I use SETMUL to disable L1 cache. I do, however, see a large slowdown when I disable "System Cache" from the BIOS. The readme says that it can't disable L2 cache that's on the motherboard, but also that there should be a much bigger difference when disabling L1 cache over L2.

Why then do I only see a slowdown when disabling L2 in the BIOS, but not at all when disabling L1 with SETMUL. I tested with and without L2 disabled in the BIOS.

20221028_184715_resize_1.jpg

I believe System Cache in the BIOS *is* L1 cache, not L2. What SETMUL command are you using to disable L1 and how are you testing that it's not working?

I was playing a lot of old games recently from 1984-1990. I noticed some strange results with a couple and it took me ages to narrow it down to setmul and disabling cache. Tried two separate computers to rule out hardware. Both stable with no problems. A 440bx with Piii750 and i815 with celeron 600mhz, dos 6.22. Also a clean dos boot (not even himem.sys loaded and only pc speaker used for theses old games)

Batman caped crusader -1988. Using setmul and disabling L1 cache, this game freezes on dos screen when loading. It works with cache enabled but runs too fast and Joystick config is impossible.

Alien syndrome - with L1 cache disabled the game will freeze with around 20 seconds of loading and playing if a mouse driver is loaded (ctmouse or MS mouse driver). Works fine with L1 cache enabled and mouse driver loaded (but again runs too fast).

A few other old games have this same problem of freezing after a couple of minutes of playing if L1 disabled and a mouse driver is loaded. They work fine without running setmul and leaving L1 cache enabled (but run too fast).
I don't know why a mouse driver would cause a problem after L1 cache is disabled.

I have a suggestion with little to no firm reason to back it up.

Are you loading the mouse driver and then turning off L1 cache? Try turning off L1 cache and then loading the mouse driver.