VOGONS


First post, by theelf

User metadata
Rank Oldbie
Rank
Oldbie

Hi, this days i read a lot of inaccurate information about how to slowdown a Pentium 2+ computers using Throttle

My computers is a Pentium 3 socket 370 1GHZ, board DFI CA61 VIA 596B. Before build this computer i have a katmai 650, a Pentium 2 450, and a mendocino 366, all with similar results, just one thing, all NON Via chipsets have 8 levels of slowdown, VIA 16 levels, is much better, but not a big problem, even the mendocino with intel board that was not very good, i can slowdown to XT levels

I test with 3dbench and topbench, i use setmul to make L1/L2 cache off, and throttle http://www.oldskool.org/pc/throttle/DOS

here all results, i compare with the ones in this post 3DBENCH CPU Benchmark Database

3DBench (1c, for low speeds 1.0)

P3 1GHZ all cache ON no throttle fullspeed 512,2
Cache L1 OFF 6,5 - If i compare with results from low end 386 is similar
Cache L2 OFF L1 ON 434,3

IMG_20240410_143452_1.jpg
Filename
IMG_20240410_143452_1.jpg
File size
155.89 KiB
Views
243 views
File license
Public domain
IMG_20240410_142646.jpg
Filename
IMG_20240410_142646.jpg
File size
156.17 KiB
Views
243 views
File license
Public domain

Cache L1/L2 ON (top speed)

Throttle 1 478,5
Throttle 2 444,4
Throttle 3 410,6
Throttle 4 376,7
Throttle 5 342,8
Throttle 6 309
Throttle 7 275
Throttle 8 241,3
Throttle 9 207,3
Throttle A 173,4
Throttle B 153,8
Throttle C 120
Throttle D 86,1 - 5x86 levels
Throttle E 48,2 - DX2 levels
Throttle F 18,4 - Slow 486 levels like DX33

Cache L1 ON L2 OFF (cache 1 on cache l2 off make computer more or less 20% slower)

Throttle 1 405,7
Throttle 2 376,7
Throttle 3 348
Throttle 4 319,1
Throttle 5 290,3
Throttle 6 261,7
Throttle 7 232,8
Throttle 8 203,8
Throttle 9 175,1
Throttle A 146,6
Throttle B 121,6
Throttle C 89,1
Throttle D 65,7
Throttle E 44
Throttle F 15,4 - similar to SX25

L1/2 Cache OFF

Throttle 1 5.9 - This is similar to my 286 16mhz
Throttle 2 5.2

With Cache off throttle from 3 and up make too slow for 3dbench, i decide to check topbench https://dosbenchmark.wordpress.com/

TOPBENCH

Cache OFF - 61 386SX33/DX20
Cache OFF Throttle 9 - 16 286 6mhz
Cache OFF Throttle D - 6 NEC V20 7mhz
Cache OFF Throttle E - 4 8088 4.77mhz

IMG_20240410_153120.jpg
Filename
IMG_20240410_153120.jpg
File size
155.46 KiB
Views
243 views
File license
Public domain

Of course this are benchmaks, but i tested with real work games and software, for example Zaxxon works OK in Cache OFF Throttle E, but works great in Throttle D, similar to my real XT turbo and too fast in Throttle C and up, like old hardware

I hope this benchmark help people to have right information about slowdown a Pentium2+ computer and how effective is. IS easy to reach XT, 286, 386 and 486 levels without much effort, just some batch files in problematic software

I tested some of results with Doom, and timedemo results are consistent with results from 3dbench, then i decide no continue more testings

Reply 1 of 4, by tauro

User metadata
Rank Member
Rank
Member

Hey @theelf, thanks for sharing your results!

I don't think you can disable L2 and L1 independently on a Pentium III, from what I've read and experimented, it's impossible. I might be wrong.

Other than throttle, you can also use CPUSPD. With it you can manage L1 and L2 independently (on processors that support it) and more options.

My main concern with this approach (throttle) is choppiness. I've had mixed results with it.

What's been your experience in regards to smoothness and throttle?

I'm currently exploring with Socket 462 Athlon XPs. The slowest I was able to get it was a 3 multiplier and 50 MHz BUS, for a 150MHz native speed. It's about as fast as a P133. For slower speeds without disabling L1, CPUSPD (throttle) comes into play. With t2 it's slightly slower than a 486 DX2.

What's your "smoothness" test? I've been using Jazz Jackrabbit (among others) and it's been a good benchmark.

For benchmarking, I recommend you to also try Speedsys (any speed), Sysinfo, DrHardware 3.1 (Pentium and slower), Checkit (slowest speeds).

Reply 2 of 4, by theelf

User metadata
Rank Oldbie
Rank
Oldbie
tauro wrote on 2024-04-15, 06:01:
Hey @theelf, thanks for sharing your results! […]
Show full quote

Hey @theelf, thanks for sharing your results!

I don't think you can disable L2 and L1 independently on a Pentium III, from what I've read and experimented, it's impossible. I might be wrong.

Other than throttle, you can also use CPUSPD. With it you can manage L1 and L2 independently (on processors that support it) and more options.

My main concern with this approach (throttle) is choppiness. I've had mixed results with it.

What's been your experience in regards to smoothness and throttle?

I'm currently exploring with Socket 462 Athlon XPs. The slowest I was able to get it was a 3 multiplier and 50 MHz BUS, for a 150MHz native speed. It's about as fast as a P133. For slower speeds without disabling L1, CPUSPD (throttle) comes into play. With t2 it's slightly slower than a 486 DX2.

What's your "smoothness" test? I've been using Jazz Jackrabbit (among others) and it's been a good benchmark.

For benchmarking, I recommend you to also try Speedsys (any speed), Sysinfo, DrHardware 3.1 (Pentium and slower), Checkit (slowest speeds).

Hi! sorry delay to answer, I test jazz in my taste is too fast to really test smothness to my eyes, but anyway, i found in my P3 1ghz scroll "smoth" at 1GHZ, with throttle at 25, 50 and 75% with cache on, without cache works fine but little slowdown, similar to my real 386SX.

A game i use to test is vinyl goddess from mars. Have a exelent scroll works fine in my pentium 3 without any slowdown, but i can slowdown until 386 SX33 and still is smoth, but some in the middle speeds are not, for example if i throttle to 486 DX2 speed scroll is not smoth, like jazz, 25 and 50 works fine in my case

I really dont use throttle to middle speed exept some cases like Alien trilogy, most of the time i just use throttle to slowdown from 386 levels to XT levels, more new games I always found a patch or works fine in 1ghz speed. But of course im hapy to test any example game

I tested more than 300 games, and i found very little problems, besides the p3 i have a DX4, a cyrix 586, a 386DX, and a Pentium MMX 233, and in my test the P3 is overall the most wide compatible, close is cyrix 5x86, but this cpu is too slow for nw stuff and emulators. Thats why i decide to keep the P3 because is solid in all areas

I really want to test a Pentium 4 or athlon but i dont have any motherboard with ISA

Reply 3 of 4, by tauro

User metadata
Rank Member
Rank
Member

Athlon XP (or even Athlon 64, S754 and S939) and Pentium 4 (or even C2D/C2Q!) are the ultimate all round retro builds.

There are some S462 boards with an ISA slot, but you're limited to early, slower processors.

It's easier to find a newer motherboard with a compatible chipset (VIA or SiS with a southbridge up to 963), and an appropriate OPL3-compatible PCI sound card, such as Yamaha YMF7x4, ESS Solo, ESS Allegro, ALS4000, FM-801, and probably others. That means you can reach really fast speeds, but you also have to find ways to slow them down.

Reply 4 of 4, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

There are some S462 boards with an ISA slot, but you're limited to early, slower processors.

Quite a lot of old KT133A boards support any Athlon XP, including Barton. The only limitation is FSB.

I must be some kind of standard: the anonymous gangbanger of the 21st century.