gerwin wrote:

Today I finally set-up a system with an Athlon XP-M processor. This processor supports "Powernow!", and I wonder how similar the implementation is compared to the K6 mobiles.

SetMul v1.23 now works with the K7 Athlon XP-M, and theoretically also with the Athlon 4 Mobile. It can do 3.0x to 16.0x.
What is still pending:
- Check the maximum supported multiplier for each processor.
- Do a hardware check to verify SetMul properly detects Athlons without PowerNow.
- Maybe some diagnostic info on the voltage setting, since one cannot change multiplier without setting the voltage as well.
- Look into multiplier support above 16.0x.
- Look into K8 support?

K8 support would be sweet 😀

gerwin wrote:

gerwin wrote:

Today I finally set-up a system with an Athlon XP-M processor. This processor supports "Powernow!", and I wonder how similar the implementation is compared to the K6 mobiles.

SetMul v1.23 now works with the K7 Athlon XP-M, and theoretically also with the Athlon 4 Mobile. It can do 3.0x to 16.0x.

I have a socket462 mobo with an ISA slot. Now i might have a use for it 😀

appiah4 wrote:

K8 support would be sweet 😀

A64LOWP works with K8. It can set both supported multipliers and frequencies freely. It can also show the startup and maximum possible FID/VID.
http://falcosoft.hu/dos_softwares.html#a64lowp

@Gerwin:
Nice, as I said earlier I have no Athlon XP-M/Athlon 4 Mobile to test with, so could you test also my K7 hacks in A64LOWP by running ATHLON4.BAT? (it's just a real time multiplier translator for A64LOWP).
Some other questions:
1. Have you solved the chipset settings dependency in case of Athlon XP-M and desktop boards?
https://web.archive.org/web/20170822100906/ht … html/Tweaks.htm
2. What's the problem with above 16x multipliers? According to Athlon XP-M/Athlon 4 Mobile multiplier tables e.g. 17x should work the same way as 16x.

If you want to add K8 support to Setmul I can help and also the source code of K8 FID/VID setting logic is available in A64LOWP's package.

Falcosoft wrote:

appiah4 wrote:

K8 support would be sweet 😀

A64LOWP works with K8. It can set both supported multipliers and frequencies freely. It can also show the startup and maximum possible FID/VID.
http://falcosoft.hu/dos_softwares.html#a64lowp

Oooohhh, Niiiice! 😀

Falcosoft wrote:

Nice, as I said earlier I have no Athlon XP-M/Athlon 4 Mobile to test with, so could you test also my K7 hacks in A64LOWP by running ATHLON4.BAT?

I just tested it. The bat file somehow does not want to do the commands using these %1 switches. So instead I made the bat file simpler and ran all the options there: They work, and the processor speed that I read out matches with what was indicated in the bat file.

Falcosoft wrote:

Have you solved the chipset settings dependency in case of Athlon XP-M and desktop boards?

No, I just tried it on my A7V600 motherboard and Powernow switching worked right away. But thanks for the warning and the link.

Falcosoft wrote:

What's the problem with above 16x multipliers?

Nothing really, just that my SetMul source table handling for C3/K6/K7 maxed out 16.0x. But I just increased that to 24.0x this evening. To test it I modified my mobile Athlon processor with a small cut in the L6 bridges: Now multipliers up to 23.0x are accepted for testing.

Falcosoft wrote:

If you want to add K8 support to Setmul I can help and also the source code of K8 FID/VID setting logic is available in A64LOWP's package.

I appreciate that, at least for testing the K8, when I get to that part.

I just acquired a VIA C3 667MHz CPU, not the most exciting thing ever but it's my first C3 and I'm looking forward to trying SetMul with it. 😎

Probably a Samuel 2, so not as flexible as an Ezra or Ezra 2. it'll give you a nice first taste.

Ideally you'll want a BX board having FSB clocks programmable by Rayer's SMB tool -- try to keep your eyes open for one.

gdjacobs wrote:

Probably a Samuel 2, so not as flexible as an Ezra or Ezra 2. it'll give you a nice first taste.

I don't think there is an Ezra2, and the only difference between Ezra and Sam2 is Tualatin support. Ezra also goes up to higher clock speeds.

There are two versions of Ezra with the second offering more multipliers (16 vs 12). Going by Gerwin's info, he might have a Sam2 that also supports the 12 multiplier steps of the Ezra not-T. The Ezra uses a bit less juice, though.

Either way, I think he'll find it very yummy.

gdjacobs wrote:

Probably a Samuel 2, so not as flexible as an Ezra or Ezra 2. it'll give you a nice first taste.

Ideally you'll want a BX board having FSB clocks programmable by Rayer's SMB tool -- try to keep your eyes open for one.

It's not a 667A but a 667 so I guess it's actually a Samuel1 I think?..

Spin DOS up on it and see what SETMUL can get it to do!

If anyone is interested, there are a bunch of cheap Athlon XP-M cpus for sale on a certain auction site, apparently ripped out of Acer laptops (no affiliation). They are actually quite expensive otherwise.

SetMul v1.23 is up!

• AMD K7 mobile and K8 multiplier adjustment and readback.
• Better checks in case of unexpected parameters.
• Changed CWSDPMI version to 025 of 2006 as suggested by Mercury127.
• Expanded and cleaned up SetMul.txt.

As Falcosoft wrote; It seems some motherboard support is required for K7 multiplier adjustment:
https://web.archive.org/web/20170822100906/ht … html/Tweaks.htm
Forgot to add that to SetMul.txt. Well, lets just see how it works for people.

K8 support could only be added with the help of Falcosoft, since he did the testing of the experimental versions and gave feedback to make it proper.

I'm getting strange results from PIII-600 Katmai, I can disable L2 cache in BIOS (no L1 option), Topbench drops from 453 to 69, and that's ok.
But with SetMul disabling L2 via L2D I see no effect in speed. And after disabling L1 with L2 "enabled" I'm getting same 69 result.

So, L1E L2E = 453, L1E L2D = 453, L1D L2E = 69, L1D L2D = 69
Looks like L1 and L2 swapped in SetMul and it can't toggle L1, or am I wrong?

ujav wrote:

I'm getting strange results from PIII-600 Katmai, I can disable L2 cache in BIOS (no L1 option), Topbench drops from 453 to 69, […]
Show full quote

I'm getting strange results from PIII-600 Katmai, I can disable L2 cache in BIOS (no L1 option), Topbench drops from 453 to 69, and that's ok.
But with SetMul disabling L2 via L2D I see no effect in speed. And after disabling L1 with L2 "enabled" I'm getting same 69 result.

So, L1E L2E = 453, L1E L2D = 453, L1D L2E = 69, L1D L2D = 69
Looks like L1 and L2 swapped in SetMul and it can't toggle L1, or am I wrong?

Your BIOS is wrong. The L2 disable option in your BIOS is simply mislabeled. It's definitely a L1 cache disable option. There is no x86 microarchitecture where disabling L2 cache would result in such a huge performance drop. Topbench score of 69 is somewhere on 386 DX level so a PIII should produce this score only with disabled L1 cache.

ujav wrote:

I'm getting strange results from PIII-600 Katmai...

What Falcosoft said.
Also I recommend downloading the SpeedSys 4.78 program. It gives cache measurements in the bottom-right graph, and that helps to understand which caches are actually working.

lvader wrote:

(24-7-2017) Been playing with a 5x86 build, the turbo button doesn't seem to make much difference so I tried it on the multiplier jumper. Unlike other Pentiums, here it changes the multiplier on the fly so the turbo button switches between 66 and 100 MHz and with setmul I can switch to 33 MHz as well.

So this Cyrix Cx5x86 matter is long overdue for some attention.
I had already located CPU pin R17 'clkmul' in the pin headers of the motherboard, now I added a mechanical switch to connect and disconnect it with the ground plane. Now of course this works to toggle 2x and 3x before starting the system, but like lvader wrote it also toggles the multiplier on the fly when the system is running!
I got some confusing behaviour when booting up at 2.0x (66MHz), after that it can only toggle between 1.0x and 2.0x, either with SetMul or the switch. Also that gave a system lockup once.
After booting up at 3.0x (100MHz) things worked much better: I could do all kind of sequences with SetMul and the mechanical switch: the system just gave the intended speeds, being 33, 66 or 100MHz, without reboot.

For this Cyrix I will have to add an alternate speed detection to SetMul. It is annoying to not see the speed directly. SpeedSys does it better.

lvader wrote:

Even more odd, when I invert the 3 pin turbo connector it switches between 1x and 2x.

I don't get this part.

gerwin wrote:

lvader wrote:

Even more odd, when I invert the 3 pin turbo connector it switches between 1x and 2x.

I don't get this part.

Sounds like the turbo button is plugged into some kind of multiplier selection header?

What I don't get is that there is only one clkmul pin on the CPU and thus only one clkmul pin on the motherboard. So with that one pin you can do a few things: Connect it to voltage or connect it to ground (and sometimes leaving it floating does something different). I don't know of any additional CPU clkmul pins on any 486 CPU, and that is not for lack of trying.

gerwin wrote:

What I don't get is that there is only one clkmul pin on the CPU and thus only one clkmul pin on the motherboard. So with that one pin you can do a few things: Connect it to voltage or connect it to ground (and sometimes leaving it floating does something different). I don't know of any additional CPU clkmul pins on any 486 CPU, and that is not for lack of trying.

Yeah, AFAIK all you can do is enable or disable the internal multiplier with these things..