VOGONS


Reply 61 of 268, by feipoa

User metadata
Rank l33t++
Rank
l33t++

If the BIOS can switch the clock from 200 MHz to 180 MHz, then I think it is adjusting the FSB from 33.3 MHz to 30 MHz, with the CPU multiplier remaining fixed at 6x.

Plan your life wisely, you'll be dead before you know it.

Reply 62 of 268, by ibm5155

User metadata
Rank Member
Rank
Member

hmm it could be...
when I finally build my infrared cable, I'll transfer setmul into this netbook and check if the other option for slowing down the cpu by half is changing the cpu multiplier or not (or I could make some asm program for detecting the multiplier and the bus frequency, since there's the debug program inside the netbook [but in this case, I really don't know how to get that info...])

Reply 63 of 268, by feipoa

User metadata
Rank l33t++
Rank
l33t++

The DOS program CHKCPU should be able to determine the correct FSB and multiplier on GXm chips. Be sure to download the latest version though, as I found some issues with Cyrix 5x86 and GXm chips which the author was able to correct.

Plan your life wisely, you'll be dead before you know it.

Reply 64 of 268, by gerwin

User metadata
Rank l33t
Rank
l33t
ibm5155 wrote:

I'll transfer setmul into this netbook and check if the other option for slowing down the cpu by half is changing the cpu multiplier or not

Note that the SetMul versions I uploaded will only allow switches to be used on matching CPUs. If you want to experiment you would need a modified version.

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

Reply 65 of 268, by ibm5155

User metadata
Rank Member
Rank
Member

I used speedsys and this is what it shows:
100% cpu clock : 200MHz
75% cpu clock: 142MHz
50% cpu clock: 90MHz

on chkcpu these are the only values showed:
100% cpu clock : 200MHz
75% cpu clock: 143.9MHz
50% cpu clock: 87.2MHz

Actually the FSB is 66 and the clock 3 acording to http://www.cpu-collection.de/?l0=co&l1=Cyrix&l2=MediaGX
EDIT: I got the dos version and now it shows the fsb is 33.2 and the clock mul 6
OK Here's what I got with checkcpu:
clock speed-------fsb------multiplier
100%---------------33-------6 (198MHz)
75%----------------24-------6 (143.8MHz)
50%----------------74.4-----6 (446,7MHz) - But I belive since it says 50% the fsb is near to 16,5 (16 * 6 = 96, and that is equal to the checkcpu32 cpu clock)

EDIT2:
Ok setting the FSB to 16 and disabling the L1 cache with setmul (now even the dos prompt is slow, even in fullscreen)
I got a cpu performance on speedsys of: 6.43 😳
and the max cpu performance as: 78.48

Last edited by ibm5155 on 2016-05-30, 01:37. Edited 4 times in total.

Reply 66 of 268, by clueless1

User metadata
Rank l33t
Rank
l33t

I've got a very special (to me) CPU. I don't know if all POD 200MMXs are like this, but all of the TR switches work on mine. I think I've worked out all the combinations:

setmul.png
Filename
setmul.png
File size
27.86 KiB
Views
2664 views
File license
Fair use/fair dealing exception

Because my motherboard does not have any L2, this makes the TR switches exceptionally useful.

On my Pentium 120, CCD and DCD to not have any effect, which takes all 486 performance levels out of the equation.

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 67 of 268, by Tertz

User metadata
Rank Oldbie
Rank
Oldbie

Would be good to have a TSR which allows same by hot-keys. Some games may to have problems only on start or concrete places, while switching off the chache makes the machine too slow sometimes.

DOSBox CPU Benchmark
Yamaha YMF7x4 Guide

Reply 68 of 268, by clueless1

User metadata
Rank l33t
Rank
l33t
Tertz wrote:

Would be good to have a TSR which allows same by hot-keys. Some games may to have problems only on start or concrete places, while switching off the chache makes the machine too slow sometimes.

Batch files have worked pretty well for me. I name them go.bat and they live in each game's folder. For example:

setmul dcd
game.exe
setmul pfe

or

setmul l1d bpd vpd
game.com
setmul l1e pfe

So the effect only lasts during that gaming session.

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 69 of 268, by gerwin

User metadata
Rank l33t
Rank
l33t
Tertz wrote:

Would be good to have a TSR which allows same by hot-keys. Some games may to have problems only on start or concrete places, while switching off the chache makes the machine too slow sometimes.

I looked into this some months ago, for a simple tool to turn off L1 cache. The problem is that 99% of the games out there take over the keyboard and timer interrupts. So the TSR cannot be called by timer or keyboard without using advanced tricks. PCXdump is the only program I know that has such a TSR stealth mode (not open source).

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

Reply 70 of 268, by Tertz

User metadata
Rank Oldbie
Rank
Oldbie
gerwin wrote:

The problem is that 99% of the games out there take over the keyboard and timer interrupts. So the TSR cannot be called by timer or keyboard without using advanced tricks. PCXdump is the only program I know that has such a TSR stealth mode (not open source).

Picture grabbers, TSR game cheating software like GameWizard, etc. - control the keyboard with games run, some in protected mode games too. Maybe their sources are available.

DOSBox CPU Benchmark
Yamaha YMF7x4 Guide

Reply 71 of 268, by j^aws

User metadata
Rank Oldbie
Rank
Oldbie
gerwin wrote:
-Pentium P54C test register "TR12" options. Parameters: BPD - Disable Branch Prediction VPD - Disable V Pipeline L1DX - Disable […]
Show full quote

-Pentium P54C test register "TR12" options. Parameters:
BPD - Disable Branch Prediction
VPD - Disable V Pipeline
L1DX - Disable L1 cache exclusively
CCD - Disable L1 code cache
DCD - Disable L1 data cache
PFE - Pentium Features Enable; Resets the above TR12 options to default.
The status of register TR12 cannnot be read by design.

FYI, I've come across a P55C - Pentium MMX 233MHz, that responds to all TR12 test registers. It makes this CPUs speed range smoother, especially when using CCD and DCD. And is similar to using ICD with VIA C3 Ezra CPUs.

Reply 72 of 268, by clueless1

User metadata
Rank l33t
Rank
l33t
j^aws wrote:
gerwin wrote:
-Pentium P54C test register "TR12" options. Parameters: BPD - Disable Branch Prediction VPD - Disable V Pipeline L1DX - Disable […]
Show full quote

-Pentium P54C test register "TR12" options. Parameters:
BPD - Disable Branch Prediction
VPD - Disable V Pipeline
L1DX - Disable L1 cache exclusively
CCD - Disable L1 code cache
DCD - Disable L1 data cache
PFE - Pentium Features Enable; Resets the above TR12 options to default.
The status of register TR12 cannnot be read by design.

FYI, I've come across a P55C - Pentium MMX 233MHz, that responds to all TR12 test registers. It makes this CPUs speed range smoother, especially when using CCD and DCD. And is similar to using ICD with VIA C3 Ezra CPUs.

Awesome! I thought I was the only one with a CPU that supported the test registers. 😀 Check about 5 posts up and you'll see a chart I made documenting my speed zones. If you get an opportunity, it would be cool to see something similar for your 233MMX.

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 73 of 268, by j^aws

User metadata
Rank Oldbie
Rank
Oldbie
clueless1 wrote:
j^aws wrote:
gerwin wrote:
-Pentium P54C test register "TR12" options. Parameters: BPD - Disable Branch Prediction VPD - Disable V Pipeline L1DX - Disable […]
Show full quote

-Pentium P54C test register "TR12" options. Parameters:
BPD - Disable Branch Prediction
VPD - Disable V Pipeline
L1DX - Disable L1 cache exclusively
CCD - Disable L1 code cache
DCD - Disable L1 data cache
PFE - Pentium Features Enable; Resets the above TR12 options to default.
The status of register TR12 cannnot be read by design.

FYI, I've come across a P55C - Pentium MMX 233MHz, that responds to all TR12 test registers. It makes this CPUs speed range smoother, especially when using CCD and DCD. And is similar to using ICD with VIA C3 Ezra CPUs.

Awesome! I thought I was the only one with a CPU that supported the test registers. 😀 Check about 5 posts up and you'll see a chart I made documenting my speed zones. If you get an opportunity, it would be cool to see something similar for your 233MMX.

I just had a look at your table above - did you vary the FSB, or are the results for a fixed FSB? If you can vary the FSB, then you should get smoother scaling. You have a steep drop from Speedsys 123 to 21.

I also have a POD 200MMX on a Socket 5 board, and it can POST the Pentium at around 16MHz using 1x multi and 16MHz FSB. Moreover, the board can also use Turbo to slowdown, as well as these test registers.

I need to do more tests with this Socket 7 233MMX, it has a very smooth range, where all the registers respond significantly - from least to most: VPD, BPD, CCD, DCD and L1D. I've also tried L1DX, but it seems to behave the same as L1D...

Reply 74 of 268, by clueless1

User metadata
Rank l33t
Rank
l33t
j^aws wrote:

I just had a look at your table above - did you vary the FSB, or are the results for a fixed FSB? If you can vary the FSB, then you should get smoother scaling. You have a steep drop from Speedsys 123 to 21.

Fixed FSB and multiplier. On the system used in that chart, I have no way to change the FSB or multiplier except by jumpers. And the whole point of SETMUL (to me) is to be able to change speeds through software, without having to open the case. So that chart is the speed profile running at 200Mhz

j^aws wrote:

I also have a POD 200MMX on a Socket 5 board, and it can POST the Pentium at around 16MHz using 1x multi and 16MHz FSB. Moreover, the board can also use Turbo to slowdown, as well as these test registers.

I thought the POD was multiplier locked (and I've never seen 1x as an option--1.5x is the lowest multiplier I've seen on socket 5), and how do you have a 16Mhz FSB? Typically the only options are 50, 60, and 66.

j^aws wrote:

I need to do more tests with this Socket 7 233MMX, it has a very smooth range, where all the registers respond significantly - from least to most: VPD, BPD, CCD, DCD and L1D. I've also tried L1DX, but it seems to behave the same as L1D...

Please share when you do those tests. 😀 And in my case, DCD is faster than CCD.

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 75 of 268, by gdjacobs

User metadata
Rank l33t++
Rank
l33t++

If you don't have a turbo switch, putting a set of toggle switches on a drive bay plate to control FSB would be a useful modification.

All hail the Great Capacitor Brand Finder

Reply 76 of 268, by j^aws

User metadata
Rank Oldbie
Rank
Oldbie
clueless1 wrote:
j^aws wrote:

I just had a look at your table above - did you vary the FSB, or are the results for a fixed FSB? If you can vary the FSB, then you should get smoother scaling. You have a steep drop from Speedsys 123 to 21.

Fixed FSB and multiplier. On the system used in that chart, I have no way to change the FSB or multiplier except by jumpers. And the whole point of SETMUL (to me) is to be able to change speeds through software, without having to open the case. So that chart is the speed profile running at 200Mhz

As mentioned by gdjacobs, you can setup simple external switches (many ways to do this) to access internal switches like FSB, Turbo, Multi etc. in conjuction with using Setmul to gain more flexibility, and not having to open cases.

clueless1 wrote:
j^aws wrote:

I also have a POD 200MMX on a Socket 5 board, and it can POST the Pentium at around 16MHz using 1x multi and 16MHz FSB. Moreover, the board can also use Turbo to slowdown, as well as these test registers.

I thought the POD was multiplier locked (and I've never seen 1x as an option--1.5x is the lowest multiplier I've seen on socket 5), and how do you have a 16Mhz FSB? Typically the only options are 50, 60, and 66.

The POD 200MMX behaves like the Socket 3 POD83, i.e. when the cooling fan is removed, it defaults to 1x multi. You can setup a simple switch to enable/ disable the fan, and thereby switch from 1x to 3x multi. The Socket 5 board also has hidden FSB settings as low as 16MHz, and a working Turbo header as well.

clueless1 wrote:
j^aws wrote:

I need to do more tests with this Socket 7 233MMX, it has a very smooth range, where all the registers respond significantly - from least to most: VPD, BPD, CCD, DCD and L1D. I've also tried L1DX, but it seems to behave the same as L1D...

Please share when you do those tests. 😀 And in my case, DCD is faster than CCD.

Yeah, will do - either here or a new thread when I get around to it.

Reply 77 of 268, by Schyz

User metadata
Rank Newbie
Rank
Newbie
clueless1 wrote:
j^aws wrote:
gerwin wrote:
-Pentium P54C test register "TR12" options. Parameters: BPD - Disable Branch Prediction VPD - Disable V Pipeline L1DX - Disable […]
Show full quote

-Pentium P54C test register "TR12" options. Parameters:
BPD - Disable Branch Prediction
VPD - Disable V Pipeline
L1DX - Disable L1 cache exclusively
CCD - Disable L1 code cache
DCD - Disable L1 data cache
PFE - Pentium Features Enable; Resets the above TR12 options to default.
The status of register TR12 cannnot be read by design.

FYI, I've come across a P55C - Pentium MMX 233MHz, that responds to all TR12 test registers. It makes this CPUs speed range smoother, especially when using CCD and DCD. And is similar to using ICD with VIA C3 Ezra CPUs.

Awesome! I thought I was the only one with a CPU that supported the test registers. 😀 Check about 5 posts up and you'll see a chart I made documenting my speed zones. If you get an opportunity, it would be cool to see something similar for your 233MMX.

Hi, following Clueless1 report on his Pentium Overdrive 200 MMX, I decided to try on my Toshiba Libretto 110CT, which has a Pentium 233 MMX.

I'm happy to report that all TR12 options have some effect and I obtain values very similar (slightly faster) to those reported by Clueless1.

[Speedsys]
Nothing: 176.34
VPD: 167.57
BPD: 151.19
VPD+BPD: 146.64
CCD: 38.46
DCD: 31.03
L1DX: 13.86
L1D: 13.85
*ALL DISABLED*: 12.04

This small laptop is now perfect for retro-gaming.

Reply 78 of 268, by clueless1

User metadata
Rank l33t
Rank
l33t
Schyz wrote:

This small laptop is now perfect for retro-gaming.

Good deal. Cheers, man!

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 79 of 268, by clueless1

User metadata
Rank l33t
Rank
l33t

I moved my POD200MMX into a SS7 motherboard which let me overclock it to 83x3=250Mhz. Here is a new speed profile. One thing I noticed in this retest is that Ultima 7: The Black Gate, which is well-known to be immune to disabling L1 cache, is NOT immune to the test registers BPD (branch prediction), CCD (L1 code cache), DCD (L1 data cache), and VPD (V Pipeline). So if your CPU supports the test registers, you can disable combinations of these to get Ultima 7 to a playable speed! I wrongly assumed previously that CCD+DCD is equivalent to L1D, which proved out wrong when I ran the game with "setmul L1D" and watched it run at full speed. On my system, U7 played perfectly with the setmul switches CCD+DCD (which I give a "Performs Like" rating of 486DX-25).

I also threw in results for my real 486DX2-66 for reference. Lastly, I ranked the results by the sum of the real game benchmarks of Wolfenstein 3D, Doom, and Quake, then rated each speed zone as a ratio of the real DX2-66 game sums.

Notably, many of the synthetic benchmark results vary wildly with different combinations of these test registers.

The slowest result on the chart plays Wing Commander I too fast with few enemies on screen, and with a screenful of enemies it *almost* plays at perfect speed, still a bit too fast. So that seems to jive with the "Performs Like" rating of 486SX-20.

P250MMX_SETMUL.jpg
Filename
P250MMX_SETMUL.jpg
File size
135.67 KiB
Views
2264 views
File license
Fair use/fair dealing exception

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