VOGONS


Reply 120 of 268, by MERCURY127

User metadata
Rank Member
Rank
Member

My P3 800 machine give 7.56 Speedsys score. With setmul or with bios settings - one score, 7.56. Its low? Possible its related with slightly buggy memory, which all 768 MB work at 100 MHz w/o errors, not 133... Also, WC run only with emm386... else i get halt.

Reply 121 of 268, by clueless1

User metadata
Rank l33t
Rank
l33t

@MERCURY127 - P2 and P3 cache performance are more highly optimized, so disabling them has a bigger effect. These cpus are much less flexible for cache manipulating than P54C, P55C, K6-2, K6-3+, VIA C3, etc.

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks

Reply 123 of 268, by gerwin

User metadata
Rank l33t
Rank
l33t
dr.zeissler wrote:

Is there a way to use powernow in dos on older XP-Mobile CPU's?

SetMul does not do PowerNow for AMD Athlons and FalcoSoft's tools seem to support Athlon64 only. I don't own any Athlon XP Mobiles myself, just regular ones.

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

Reply 125 of 268, by Falcosoft

User metadata
Rank Oldbie
Rank
Oldbie
gerwin wrote:
dr.zeissler wrote:

Is there a way to use powernow in dos on older XP-Mobile CPU's?

SetMul does not do PowerNow for AMD Athlons and FalcoSoft's tools seem to support Athlon64 only. I don't own any Athlon XP Mobiles myself, just regular ones.

Let me inform you that in v1.1 of A64LOWP I have added partial support for Athlon 4/Athlon XP Mobile. On Athlon 4/XP Mobile only the multiplier/frequency can be modified, the voltage can not.
Also contrary to Athlon64, run-time chipset hacks may be required on desktop systems. More info can be found in the A64Lowp.txt file.
Since I do not own such CPU myself I worked blind so without the help of to S. Miltenberger (testing and useful information) it could not be done.
http://falcosoft.hu/dos_softwares.html#a64lowp
Note:
the multiplier table is very different from Athlon64, and you have to make the conversion manually:

1st Parameter -> Athlon 4/XP Mobile CPU Multiplier
13x -> 4x
6x -> 5x
7x -> 6x
8x -> 7x
9x -> 8x
10x -> 9x
11x -> 10x
4x -> 11x
5x -> 12x
14x -> 13x
15x -> 14x
16x -> 15x
17x -> 16x
18x -> 17x

You can use TSCDOS for testing if multiplier setting worked or not. The more testers the better.
http://falcosoft.hu/dos_softwares.html#tscdos

Website, Facebook, Youtube
Falcosoft Soundfont Midi Player + Munt VSTi + BassMidi VSTi
VST Midi Driver Midi Mapper

Reply 126 of 268, by gerwin

User metadata
Rank l33t
Rank
l33t
Falcosoft wrote:

Let me inform you that in v1.1 of A64LOWP I have added partial support for Athlon 4/Athlon XP Mobile.

Nice work, thanks for the update!
Wonder if I actually have any mobile Athlon XPs...

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

Reply 127 of 268, by amijim

User metadata
Rank Member
Rank
Member

Hello,i am wondering if anyone has tried setmul to enable l2 cache on a tillamook pentium mmx mobile cpu .There is a lot of mystery around how this cpu disables l2 cache on some mobos.Tillamook is based on a p55c architecture.

Iwill ZmaxDP
Arima4way
Tyan s2885
Iwill MPX2
Gigabyte GA-7DPXDW+
Compaq SP700
Compaq ml350

Reply 128 of 268, by appiah4

User metadata
Rank l33t++
Rank
l33t++

A few questions from a new user of this utility (previously I relied on motherboard BIOS for this).

1. Is it me or does SETMUL L2D not work with P54C CPUs? It seems to have no effect on performance that I can discern and CACHECHK can still find 256K of cache when I try.
2. My P133 after using SETMUL L1D scores around what Speedsys says a DX2-66 should score (~25) and when I set it to 50MHz FSB I get around what a DX2-50 would score (~20). However DOOM runs painfully slow in full screen, probably the way it would run on a 386DX (<10fps) and certainly much slower than I remember it running on my DX33. For reference, I ran Phil's Doom benchmark for faster machines (fullscreen and high details), and got 2134 gametics and 1159 realtics with Cache On vs 2134 gametics and 6143 realtics with L1D. Is this because the L1D option disables L1 and L2 together? Should I have used L1DX instead, and is this supported on a P133?
3. As an extension of the above, seeing as the Speedsys results aren't really representative of my system's performance for whatever reason, which benchmark is a good gauge of where my system is after playing with SETMUL options?
4. Will using my K5 PR166 give me more control over the L1/L2 caches with SETMUL? There are a lot of references to SETMUL and K6 but not many people seem to have tried it with K5?

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

Reply 129 of 268, by clueless1

User metadata
Rank l33t
Rank
l33t
appiah4 wrote:

A few questions from a new user of this utility (previously I relied on motherboard BIOS for this).

1. The setmul documentation says "SetMul allows disabling the L1 cache on any x86 processor from the 486 onwards.
SetMul allows disabling the L2 cache on the K6 Mobile and VIA C3. Note that
the C3 Samuel 1 has no L2 cache"
So L2D only works with K6 Mobile and VIA C3, not P54C.

2. Use a variety of gaming and non-gaming benchmarks to get a better overall feel:

time_machines.png
Filename
time_machines.png
File size
107.64 KiB
Views
2281 views
File license
Fair use/fair dealing exception

This is my P55C (Overdrive) and DX2-66 cpus combined onto one chart.

3. See above.

4. I have no idea about the K5, sorry. 😀

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks

Reply 130 of 268, by appiah4

User metadata
Rank l33t++
Rank
l33t++

I see. Even though Speedsys says the computer benches at around the DX2-66 mark the Doom scores are more in line with something inbetween a 486DX33 and DX25 at 66MHz FSB. At 50MHz FSB it's pretty much in line with a DX25. I suppose disabling the L2 cache in bios would bring it down to around 386DX33 levels? Well, looks like I can play Wing Commander and Ultima 7 on this PC if need be.

I am very keen to try the K5 PR166 as well at this point, I have a gut feeling its performance will land squarely on the 486DX33 and 386DX33 mark.

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

Reply 131 of 268, by gerwin

User metadata
Rank l33t
Rank
l33t

As a general rule, There is no way for SetMul to toggle the motherboard cache on/off. Only for on-die L2 cache did I find the necessary CPU register options. It figures.

appiah4 wrote:

I am very keen to try the K5 PR166 as well at this point, I have a gut feeling its performance will land squarely on the 486DX33 and 386DX33 mark.

The K5 is a little boring regarding speed options IMHO. You may want to consider the The Cyrix 6x86L PR200, as it can run at 1x multiplier (through Jumper settings or maybe BIOS options).

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

Reply 132 of 268, by appiah4

User metadata
Rank l33t++
Rank
l33t++

It's also a hotter chip, seems like TDP is 15W or thereabots compared to P133's 12W, so running it without a fan is also not an option.

The turbo button on the case right now is wired up to toggle between 50MHz/66MHz FSB but I could use it to toggle the L2 cache 0K/256K, not sure how much slower a 66MHz FSB with L2 disabled would be compared to 50MHZ FSB with L2 enabled though. (Of course, I can always disable L2 from BIOS but it's not half as convenient).

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

Reply 133 of 268, by gdjacobs

User metadata
Rank l33t++
Rank
l33t++

I tended to correlate between SpeedSys, NSSI, and others to get a sense of where performance was.

L2 manipulation is very important once L1 is disabled (see my chart), but you ultimately want both FSB and L2 manipulation.
Let's benchmark our systems with caches disabled

All hail the Great Capacitor Brand Finder

Reply 134 of 268, by appiah4

User metadata
Rank l33t++
Rank
l33t++

Yeah, since I can do the L2 Cache in BIOS (I can actually disable either L1 or L2 in BIOS, but SETMUL is pretty handy to deal with L1 on the fly for things like Ultima 7, Underworld etc.) so I will probably use Turbo for FSB, BIOS for L2(/L1) and SETMUL for L1.

Out of curiosity, does SETMUL persist when launching PITA games like Ultima 7?

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

Reply 135 of 268, by gerwin

User metadata
Rank l33t
Rank
l33t
appiah4 wrote:

Out of curiosity, does SETMUL persist when launching PITA games like Ultima 7?

SetMul does not persist (edit: I mean it does not stay resident, but the CPU registers will remain as set, until the next system reset/reboot).
Ultima 7 is indeed a particularly troublesome game as it re-enables disabled L1 cache, and I was thinking of how to fix that one, when vvbee came up with his solution:
U7DCP - Ultima VII DOS Cache Patch
Haven't tested it myself yet.

Last edited by gerwin on 2018-05-28, 18:57. Edited 1 time in total.

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

Reply 136 of 268, by clueless1

User metadata
Rank l33t
Rank
l33t

Test Registers (if supported) persist with Ultima 7. So on my 200Mhz POD, if I use the DCD test register, it drops speeds to slow 486 levels, which runs U7 great -- AND it persists with U7!

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks

Reply 139 of 268, by appiah4

User metadata
Rank l33t++
Rank
l33t++
vvbee wrote:

Don't let that post-v1.0 software infect your machine.

I don't get this.. 🙁 it's a joke? right?

Ok, so I have quite a bit of cache disabling testing planned for my Pentium 133 PC, here is the Google Docs file I have set up for this test. Hopefully I can find ways to achieve my speed targets (Basically 486DX4-100, 486DX2-66, 486DX33 and 386SX25).

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