VOGONS


Reply 141 of 202, by appiah4

User metadata
Rank l33t++
Rank
l33t++
gdjacobs wrote:

Excellent, although it's possible some of these will have no effect.

I thought all P54C cores responded to TR12 flags?

Retronautics: A digital gallery of my retro computers, hardware and projects.

Reply 143 of 202, by gdjacobs

User metadata
Rank l33t++
Rank
l33t++
appiah4 wrote:
gdjacobs wrote:

Excellent, although it's possible some of these will have no effect.

I thought all P54C cores responded to TR12 flags?

As far as I've found, they will respond to the flags, but some will have no performance impact where they do with the PMMX cpus. At least that's the case with the P133 I tested.

All hail the Great Capacitor Brand Finder

Reply 144 of 202, by appiah4

User metadata
Rank l33t++
Rank
l33t++
gdjacobs wrote:
appiah4 wrote:
gdjacobs wrote:

Excellent, although it's possible some of these will have no effect.

I thought all P54C cores responded to TR12 flags?

As far as I've found, they will respond to the flags, but some will have no performance impact where they do with the PMMX cpus. At least that's the case with the P133 I tested.

That's very interesting, care to let me know which ones did not work?

Retronautics: A digital gallery of my retro computers, hardware and projects.

Reply 146 of 202, by gdjacobs

User metadata
Rank l33t++
Rank
l33t++
appiah4 wrote:

That's very interesting, care to let me know which ones did not work?

Just a generic flat top P54C. I don't think I even looked at what the stock multiplier was as I was mainly interested in approaching 286 performance if possible.

Anyway, I was somewhat underwhelmed. Lots of flags didn't have an appreciable effect. It just didn't seem as versatile as a PMMX.

p54c-setmul-performance.png
Filename
p54c-setmul-performance.png
File size
55.85 KiB
Views
2139 views
File license
Fair use/fair dealing exception

All hail the Great Capacitor Brand Finder

Reply 147 of 202, by appiah4

User metadata
Rank l33t++
Rank
l33t++
gdjacobs wrote:
Just a generic flat top P54C. I don't think I even looked at what the stock multiplier was as I was mainly interested in approac […]
Show full quote
appiah4 wrote:

That's very interesting, care to let me know which ones did not work?

Just a generic flat top P54C. I don't think I even looked at what the stock multiplier was as I was mainly interested in approaching 286 performance if possible.

Anyway, I was somewhat underwhelmed. Lots of flags didn't have an appreciable effect. It just didn't seem as versatile as a PMMX.

p54c-setmul-performance.png

Interesting. The legend may be wrong? 83MHz L2D is faster than everything else when PFE is used, is the data mislabeled?

Retronautics: A digital gallery of my retro computers, hardware and projects.

Reply 148 of 202, by gdjacobs

User metadata
Rank l33t++
Rank
l33t++

Yup, I think 83mhz and 100mhz are reversed for the L2D series. It's a graph I threw together from one year old data, so there's probably other clerical glitches in there. Even more strange, look at the performance between the L2D and L2E series with caches enabled. L2 cache actually slows down the chip here.

My point was the much more horizontal trend of the P54 wrt TR12 parameters in comparison to my PMMX. Notice on the PMMX how the CCD and DCD flags have a differing performance impact with L2 enabled and serve to fill gaps in the performance range. That's not nearly the case with the P54.

All hail the Great Capacitor Brand Finder

Reply 149 of 202, by the_ultra_code

User metadata
Rank Member
Rank
Member

Quick question: Is setmul able to disable L1 and L2 caches on Intel-chipset Socket 7 motherboards with Pentium MMX CPUs installed? I'm not worried about messing with multipliers and such, just the caches. My Intel LT430XT motherboard sadly does not have that functionality built into the BIOS.

Builds

Other:
* USB2 PCI Card in Win98 SE
* Futuremark Result Browsers

Reply 150 of 202, by appiah4

User metadata
Rank l33t++
Rank
l33t++
the_ultra_code wrote:

Quick question: Is setmul able to disable L1 and L2 caches on Intel-chipset Socket 7 motherboards with Pentium MMX CPUs installed? I'm not worried about messing with multipliers and such, just the caches. My Intel LT430XT motherboard sadly does not have that functionality built into the BIOS.

The CPU has no control over L2 on Socket 7 motherboards as far as I know, so setmul can not disable that.

Retronautics: A digital gallery of my retro computers, hardware and projects.

Reply 152 of 202, by ruthan

User metadata
Rank Oldbie
Rank
Oldbie

Hi.
Im using setmul on with my X58 (ICH10) + Xeon 5660 machine it look like there is some issue with soundcards.. Without sound card, or with Yamaha 744 soundcards setmul is working fine in EMS, NOEMS.. or realmode, protected mode (last two are not sound drivers.. because yamaha has EMS only driver), but with Aureal Vortex 1 multiple drivers version tested.. setmul giving me freeze in EMS and NOEMS mode.. it realmode its not freezing, but its not working well too.. even 286 games are too slow..Note that slowdos also not working good with Aureal drivers and its ok with Yamaha.
I always used only setmul l1D, i dont really need other options.. i make my machine 300x slower, somewhere at fast 386/slow 486 level.

I have only 1 other clue, SlowDos default dealy interval is 12, but after is Aureal driver loaded, its change to 35, maybe its messing with system timer..

Has someone similar problems, could it be fixed? Its strange.. i didnt expect that would be some sound drivers conflict..

Im old goal oriented goatman, i care about facts and freedom, not about egos+prejudices. Hoarding=sickness. If you want respect, gain it by your behavior. I hate stupid SW limits, SW=virtual world, everything should be possible if you have enough raw HW.

Reply 153 of 202, by gerwin

User metadata
Rank l33t
Rank
l33t

Setmul was not really tested in systems above a Pentium III.

PCI soundcard ISA SBPro emulation mode and cache disabling is a troublesome combination. These are not normal drivers but TSR hacks that intercept Legacy ISA communication and translate it to PCI communication, often there is also a Software OPL FM emulator or even a Midi synth in software. I found the Ensoniq/Creative labs SB16 emulation driver to stutter when L1 cache is disabled, this was with an Athlon 2GHz system. I figure that such an emulation driver requires a certain amount of processor speed that is not available when L1 is disabled.

When using SetMul only for L1 cache disabling, there are many other programs that can do that, allthough not in protected mode. you can try Cacheoff.zip found here.

--> ISA Soundcard Overview // Doom MBF 2.04 // SetMul

Reply 154 of 202, by ruthan

User metadata
Rank Oldbie
Rank
Oldbie
gerwin wrote:

When using SetMul only for L1 cache disabling, there are many other programs that can do that, allthough not in protected mode. you can try Cacheoff.zip found

Thanks, meanwhile i tested others L1D disablers and they are behaving exactly the same as Setmul, same problem.

Im old goal oriented goatman, i care about facts and freedom, not about egos+prejudices. Hoarding=sickness. If you want respect, gain it by your behavior. I hate stupid SW limits, SW=virtual world, everything should be possible if you have enough raw HW.

Reply 155 of 202, by realoldguy23

User metadata
Rank Newbie
Rank
Newbie

First a big Thank You for this cool utility!
I have an issue on my Pentium Pro build:

setmul_ppro_gpf_s.jpg
Filename
setmul_ppro_gpf_s.jpg
File size
244.74 KiB
Views
940 views
File comment
setmul121 L2D on PentiumPro in real mode (no EMM386 etc.)
File license
Fair use/fair dealing exception

It is the same if I have loaded EMM386 or if I don't load any driver (F5 at DOS startup). What could I do?

Disabling and re-enabling L1 cache works. Is it correct that it always disables (enables) the L2 as well?

Reply 156 of 202, by gerwin

User metadata
Rank l33t
Rank
l33t

Thanks for the information.
I have to look up the intel Pentium Pro and Pentium II/III datasheet again regarding the L2 cache functions. But since I don't have access to a Pentium Pro system, this will most likely end up with me disabling the L2 cache function on Pentium Pro CPU's, since in practice it does not work.

--> ISA Soundcard Overview // Doom MBF 2.04 // SetMul

Reply 157 of 202, by MERCURY127

User metadata
Rank Member
Rank
Member

i tested SetMul on I7 3770 - it WORK, i write about this here year ago.
but DONT USE cwsdpmi other than r5 00 and 02 year.
only these bulds really work w/o problems on 386, p3 and i7.
all other versions cwsdpmi is buggy.
i attach proper (tested by me) versions.

Attachments

  • Filename
    DOSX2.zip
    File size
    25.11 KiB
    Downloads
    34 downloads
    File license
    Fair use/fair dealing exception

Reply 159 of 202, by gerwin

User metadata
Rank l33t
Rank
l33t
MERCURY127 wrote:

but DONT USE cwsdpmi other than r5 00 and 02 year. only these bulds really work w/o problems on 386, p3 and i7.

That is interesting. Thanks for sharing.
It reminds me of the other cwsdpmi caveat I came across: It does not like Aureal Vortex DOS emulation driver, but DOS32A works fine with it.

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.

--> ISA Soundcard Overview // Doom MBF 2.04 // SetMul