Let's benchmark our systems with caches disabled

Discussion about old PC hardware.

Re: Let's benchmark our systems with caches disabled

Postby Kamerat » 2016-4-16 @ 13:59

Anyone who knows a piece of software that disables L1 cache on a Pentium 4, I only get L2 disabled? I want to test a Northwood @1.2GHz. :)
User avatar
Kamerat
Oldbie
 
Posts: 843
Joined: 2014-3-14 @ 19:09
Location: Oslo, Norway

Re: Let's benchmark our systems with caches disabled

Postby BSA Starfire » 2016-4-16 @ 15:55

Ignore the score I just added with a Pentium M, neither BIOS or Setmul will disable caches so it's irrelevant.
Sorry to waste a time.
edit: never mind entry deleted.
386SX 33MHz,4MB Ram,ALi M1217 motherboard,OAK OTI077 1MB. Seagate ST1144A.
386DX 25MHz,IIT 387,8MB RAM,Opti,32K cache,WDC90C30 512K.Conner CF170A.
IDT Winchip C6 200MHz,DFi P5BTX/L,Number Nine S3 ViRGE 2MB,SoundBlasterCT2260,1.2Gb Fujitsu.
User avatar
BSA Starfire
Oldbie
 
Posts: 796
Joined: 2014-3-22 @ 05:20
Location: UK

Re: Let's benchmark our systems with caches disabled

Postby clueless1 » 2016-4-17 @ 14:55

Added a result for my POD 200 with the setmul switch 'BPD' (branch prediction disable). This slows the cpu down another 16% in Speedsys over just disabling L1. Doom framerate drops by 12.6%.
The more I learn, the more I realize how much I don't know.
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks
User avatar
clueless1
l33t
 
Posts: 3651
Joined: 2015-12-22 @ 17:43
Location: Midwest US

Re: Let's benchmark our systems with caches disabled

Postby clueless1 » 2016-4-21 @ 23:37

Updated POD results with new video card--TNT2 M64 PCI replaces Voodoo3 2000 PCI. Only the results with cache enabled changed. This seems to show that when you disable cache on a CPU, the video card has much less impact on performance.

Also added results with more setmul switches--
using VPD (disable V pipeline), speeds drop from P200 to P166.
using BPD (disable branch prediction), speeds drop to P133.
using DCD (disable L1 data cache), speeds drop to 486SX-25.
using CCD (disable L1 code cache), speeds drop to 486SX-20 (theoretical, based on Doom scores).
using L1D (disable entire L1 cache), speeds drop to Am386DX-29 (theoretical, going by Doom scores).
using L1D+BPD, speeds drop to Am386DX-25 (theoretical, based on Doom scores).

These switches only work with Pentiums, and apparently not all of them. Just try to see if it works.
From setmul docs:
-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 combinations I list above are the only ones I could get to give me discrete results. Other combinations did not yield new results.
The more I learn, the more I realize how much I don't know.
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks
User avatar
clueless1
l33t
 
Posts: 3651
Joined: 2015-12-22 @ 17:43
Location: Midwest US

Re: Let's benchmark our systems with caches disabled

Postby BSA Starfire » 2016-4-27 @ 17:15

Results added for a AMD Duron 650 MHz on Dfi AK74 motherboard. 512 MB RAM & Nvidia TNT 16 MB STB Velocity 4400 graphics card.

Sorry I have been a bit slow on contributions, been really busy this last 2 weeks.
Best,
Chris
386SX 33MHz,4MB Ram,ALi M1217 motherboard,OAK OTI077 1MB. Seagate ST1144A.
386DX 25MHz,IIT 387,8MB RAM,Opti,32K cache,WDC90C30 512K.Conner CF170A.
IDT Winchip C6 200MHz,DFi P5BTX/L,Number Nine S3 ViRGE 2MB,SoundBlasterCT2260,1.2Gb Fujitsu.
User avatar
BSA Starfire
Oldbie
 
Posts: 796
Joined: 2014-3-22 @ 05:20
Location: UK

Re: Let's benchmark our systems with caches disabled

Postby clueless1 » 2016-4-27 @ 21:46

BSA Starfire wrote:Results added for a AMD Duron 650 MHz on Dfi AK74 motherboard. 512 MB RAM & Nvidia TNT 16 MB STB Velocity 4400 graphics card.

Sorry I have been a bit slow on contributions, been really busy this last 2 weeks.
Best,
Chris

Thanks! No worries, Chris. I appreciate any results people submit. :) I'd love to see some 486 results, if anyone reading this has one they'd like to benchmark.

Cheers.
The more I learn, the more I realize how much I don't know.
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks
User avatar
clueless1
l33t
 
Posts: 3651
Joined: 2015-12-22 @ 17:43
Location: Midwest US

Re: Let's benchmark our systems with caches disabled

Postby MMaximus » 2016-5-07 @ 17:29

I updated the spreadsheet with results from my 233MMX build http://www.vogons.org/viewtopic.php?f=25&t=42572

I benchmarked 8 different combinations of CPU speed and caches but I'm just gonna post the highest and lowest results here for simplicity:

P233MMX with L1 and L2 enabled
3Dbench: 165.1
PCPbench: 55.8
DOOM realticks 903 (82.71 fps)
Speedsys 176.23

P233MMX @ 100mhz with L1 and L2 disabled
3Dbench: 10.8
PCPbench: 2.7
DOOM realticks 17304 (4.32 fps)
Speedsys 10.45
User avatar
MMaximus
Member
 
Posts: 448
Joined: 2014-3-30 @ 20:28
Location: EU

Re: Let's benchmark our systems with caches disabled

Postby clueless1 » 2016-5-08 @ 01:02

Thanks, MMaximum. I see too that tyuper is adding an Intel 486DX4-100. :) Very cool. Regarding your results, here's some equivalents that you are hitting:
P233MMX with L1 disabled performs like a 486DX-33
P233MMX @ 100Mhz with L1 disabled performs like a 486SX-25
P233MMX with L1+L2 disabled performs like a 386DX-33
P233MMX @ 100Mhz with L1+L2 disabled performs like a 386DX-25
The more I learn, the more I realize how much I don't know.
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks
User avatar
clueless1
l33t
 
Posts: 3651
Joined: 2015-12-22 @ 17:43
Location: Midwest US

Re: Let's benchmark our systems with caches disabled

Postby tyuper » 2016-5-08 @ 01:53

Nice to hear that :) The lowest I could get without touching jumpers:
3DBENCH: 3.4
PCPBench: 0.4
SpeedSys: 3.5
Doom: 57998 ~ 1.28 FPS
Slower are only 286-8, that's what TOPBENCH stated :lol:
tyuper
Newbie
 
Posts: 64
Joined: 2013-8-01 @ 15:25

Re: Let's benchmark our systems with caches disabled

Postby clueless1 » 2016-5-12 @ 20:24

tyuper wrote:Nice to hear that :) The lowest I could get without touching jumpers:
3DBENCH: 3.4
PCPBench: 0.4
SpeedSys: 3.5
Doom: 57998 ~ 1.28 FPS
Slower are only 286-8, that's what TOPBENCH stated :lol:


How did you manage this? :)
The more I learn, the more I realize how much I don't know.
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks
User avatar
clueless1
l33t
 
Posts: 3651
Joined: 2015-12-22 @ 17:43
Location: Midwest US

Re: Let's benchmark our systems with caches disabled

Postby DX7_EP » 2016-5-14 @ 02:02

I went and added my results from the vintage laptop onto the spreadsheet. It is a bit silly how my system has a bit of gap between <i386DX-16 and ~AMDX4-100 levels of performance.

It is interesting this system does allow a user to switch CPU cache write policies between default write-back and slower write-through modes in the BIOS. As a result, I've specified which mode I was using in each result set.
CM-64, FB-01, SC-55ST, SC-8850, SD-20, SD-90, MU1000-EX + PLG-150VL
User avatar
DX7_EP
Newbie
 
Posts: 45
Joined: 2014-5-31 @ 00:57
Location: medic_ihavenoidea.wav

Re: Let's benchmark our systems with caches disabled

Postby clueless1 » 2016-5-14 @ 17:34

Just a quick note: Now that we have a decent database of results, I picked some common/popular combinations of CPU/chipset/graphics that represent their platforms well and isolated their results on their own spreadsheet tab. Also updated the OP to reflect this. Link to spreadsheet:
https://docs.google.com/spreadsheets/d/ ... sp=sharing
The more I learn, the more I realize how much I don't know.
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks
User avatar
clueless1
l33t
 
Posts: 3651
Joined: 2015-12-22 @ 17:43
Location: Midwest US

Re: Let's benchmark our systems with caches disabled

Postby melbar » 2016-5-18 @ 18:18

By the way, i have replaced the GF2-MX400 with a RivaTNT2 i've bought last week. It works quite well after deinstall and reinstall the same drivers.

A first test with the clock-speed which i use mostly and with all caches:

As expected, the RivaTNT2 is as fast as a GF2-MX400 (which performs with a fast CPU, let me say 1,5-2,0Ghz, between GeForce256 and GeForce256-DDR). My CPU with 170MHz is definitely too slow and also with the max. 500MHz the RivaTNT2 should be enough...
With a RivaTNT2 i have compared to GF2-MX400:

3DBench2: -0,40%
PCPBench: -0,36%
Speedsys: -0,44%
DOOM (realticks): -0,85%
DOOM (fps): -0,88%

So, no need to update the speadsheet...
User avatar
melbar
Member
 
Posts: 340
Joined: 2016-1-31 @ 13:38

Re: Let's benchmark our systems with caches disabled

Postby clueless1 » 2016-5-18 @ 18:57

I found that any differences between graphic cards goes completely away when you disable caches. Even comparing onboard Cirrus Logic to TNT2, there is a huge difference at full speed, but nearly identical scores after caches are disabled.
The more I learn, the more I realize how much I don't know.
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks
User avatar
clueless1
l33t
 
Posts: 3651
Joined: 2015-12-22 @ 17:43
Location: Midwest US

Re: Let's benchmark our systems with caches disabled

Postby tyuper » 2016-5-19 @ 00:28

clueless1 wrote:How did you manage this? :)

Look for "Boot Up System Speed" option in your's mainboard BIOS setup. :happy:
tyuper
Newbie
 
Posts: 64
Joined: 2013-8-01 @ 15:25

Re: Let's benchmark our systems with caches disabled

Postby BSA Starfire » 2016-6-02 @ 13:54

Results added for VIA/Cyrix III 600 MHz, 3dfx Voodoo Banshee.
386SX 33MHz,4MB Ram,ALi M1217 motherboard,OAK OTI077 1MB. Seagate ST1144A.
386DX 25MHz,IIT 387,8MB RAM,Opti,32K cache,WDC90C30 512K.Conner CF170A.
IDT Winchip C6 200MHz,DFi P5BTX/L,Number Nine S3 ViRGE 2MB,SoundBlasterCT2260,1.2Gb Fujitsu.
User avatar
BSA Starfire
Oldbie
 
Posts: 796
Joined: 2014-3-22 @ 05:20
Location: UK

Re: Let's benchmark our systems with caches disabled

Postby melbar » 2016-6-02 @ 15:47

Interesting results with your Via-Cyrix III (600Mhz). With 3Dbench & PCPbench, you'll get a 486DX25-33, and with Doom you match to a 486DLC-33.

That's quite good.

When i compare my results of PIII, i've made in the past (only 3Dbench & PCPbench, with a slow RageXL), i've got these strage results:
PIII 650: 3Dbench 353.7, PCPbench 136.1
PIII 433: 3Dbench 235.7, PCPbench 105.9
PIII 433 (L1,L2): 3Dbench 58.5, PCPbench 1.4
User avatar
melbar
Member
 
Posts: 340
Joined: 2016-1-31 @ 13:38

Re: Let's benchmark our systems with caches disabled

Postby BSA Starfire » 2016-6-02 @ 15:53

Well the VIA/Cyrix III has no level 2 to start with, so I guess that is a seriuos limiting factor. You can find my scores with a Celly Mendocino and same board on the spreadsheet too.

EDIT, I just realised those were with a a Chromatic Mpact2, so not really comparable, sorry.
386SX 33MHz,4MB Ram,ALi M1217 motherboard,OAK OTI077 1MB. Seagate ST1144A.
386DX 25MHz,IIT 387,8MB RAM,Opti,32K cache,WDC90C30 512K.Conner CF170A.
IDT Winchip C6 200MHz,DFi P5BTX/L,Number Nine S3 ViRGE 2MB,SoundBlasterCT2260,1.2Gb Fujitsu.
User avatar
BSA Starfire
Oldbie
 
Posts: 796
Joined: 2014-3-22 @ 05:20
Location: UK

Re: Let's benchmark our systems with caches disabled

Postby j^aws » 2016-6-05 @ 22:40

Added VIA Cyrix III: Ezra-T @ 1000MHz, 1200MHz and 200MHz, using onboard VGA (C&T F69000), and on a 440BX board.
j^aws
Oldbie
 
Posts: 644
Joined: 2013-2-02 @ 19:39
Location: UK

Re: Let's benchmark our systems with caches disabled

Postby clueless1 » 2016-6-23 @ 14:08

Thanks for the results, I resorted. :) I appreciate everyone's input!
The more I learn, the more I realize how much I don't know.
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks
User avatar
clueless1
l33t
 
Posts: 3651
Joined: 2015-12-22 @ 17:43
Location: Midwest US

PreviousNext

Return to General Old Hardware

Who is online

Users browsing this forum: B.Gumble, BastlerMike, hard1k, Munx, realoldguy23, Thallanor and 16 guests