VOGONS

Common searches


DOSBox 0.74 CPU Benchmark

Topic actions

First post, by Tertz

User metadata
Rank Oldbie
Rank
Oldbie

CPU is the main device responsible for performance in DOSBox. It's about how fast different CPUs may to be in DOSBox at gaming.
One of intentions is to compare the emulation performance of different generations of CPUs and to give base to compare the speed of different PC emulators.
Welcome to post the results of your modern or old PC for all tests' sections or some.

Preconfigured benchmark tests pack

The pack is useful for quick start. You may also read the description of the testing some below, to download and configure the tests yourself.
Contains DOSBox 0.74, Quake 1.06, Doom 1.9, Speed Test 1.14 and also Rufus 3.9: mediafire, yandex.
- tests1.bat - creates .txt files with results for Doom and Speedtst.
- tests2.bat - runs Quake demo1 in 800x600 and 320x200. After the game appears - press "Space" button to run a demo. Then record fps count (you may scroll the game's console by PgUp/PgDn buttons). To get average fps (as the variation for the same demo can be some %) from several demo's runs you may use F1 (for demo1) and F4, F5 for demo2, demo3 correspondingly. Then press "Esc" and choose to exit the game. The game will restart with the next screen resolution and you should repeat the actions.
Note: There should not be background apps and processes during benchmarks to do not reduce the results. For example, when you are using a CPU checking app during a test run to know the used CPU's frequency - those results are not valid and you'll need to take the results from another tests' run without any background apps.
It may be useful to turn off energy saving for CPU which may reduce the performance!

I. Common DOS games (non-FPU) performance based on Speed Test and Doom

Testing procedure:
1) Get DOSBox 0.74 [alternative], Speed Test 1.14, Doom 1.9.
2) Speed Test in DOS
- In case the PC has no DOS, it can be booted to DOS with Boot flash disk created by Rufus. This app has all needed system files. Choose there: FAT32 + FreeDOS or MS DOS and format a flash disk.
- Run Speed Test by the command: speedtst /L >res0.txt
res0.txt will have the string like "This computer 123 parrots, 456 XT". The needed number is near 'XT'.
3) Speed Test in DOSBox
- Make default emulator's config by running the shortcut in DOSBox section of Start button: DOSBox->Reset Options.
- Mount a folder as disk 'C', for example by writing the strings in [autoexec] section of the config:
mount c c:\dosdisk
c:
The testing have to go in windowed mode only. During the measurement DOSBox's window have to stay active (main) all the time, or changing window focus may reduce resources the emulator gets. No background applications like file managers, antiviruses or other as they take resources.
On P3 and earlier Speed Test inside DOSBox may be run without '/L'.
On P4 and higher it's better to make measurements inside DOSBox for 3 times to find an average.
3.1) In the config set: core=dynamic, scaler=none, cputype=486_slow, cycles=max
Run Speed Test by the command: speedtst /L >res1.txt
3.2) In the config set: core=normal, scaler=none, cputype=486_slow, cycles=max
Run Speed Test by the command: speedtst /L >res2.txt
3.3) Copy-paste 3 strings with results from these 3 text files or only the numbers near 'XT'.
4) Doom in DOSBox
- Install Doom in real OS or in DOSBox. Run its setup.exe to set music off and fx off, maximize ingame screen by '='. Also preinstalled and preconfigured Doom may be taken there from that theme.
4.1) In the config set: core=dynamic, scaler=none, cputype=486_slow, cycles=max
Run Doom by the command: doom -timedemo demo3 >dres1.txt
On P4 and higher it's better to make this measurement for at least 3 times to find an average.
4.2) In the config set: core=normal, scaler=none, cputype=486_slow, cycles=max
Run Doom by the command: doom -timedemo demo3 >dres2.txt
4.3) Copy-paste 2 numbers of realtics from these 2 text files or calculate fps by the formula below.
5) - Check actual (as max) frequency of CPU during the measurements by CPU-Z, as there are many CPUs with dynamic frequency (turbo mode, SpeedStep). Don't run CPU-Z during measuring of speed on the stages above! On this step you may use SuperPI to set CPU to max 1-core performance to get actual frequency in DOSBox, or to use DOSBox with a benchmark running inside. SuperPI or DOSBox window must stay active (main) during this taking the frequency.
- Report your CPU (model, actual frequency) and the measured results.

Results

                                       Speedtst, XT  || Doom, fps
------------------------------------------------------------------
AMD Ryzen 5 3600 4.2 GHz - 1038 232 || 291 80
Intel Core 2 Duo E8600 3.3 GHz 8435 718 184 || 167 63
Intel Pentium 2020M 2.4 GHz 6135 767 134 || 156 37
Intel Pentium E2180 2.0 GHz 4820 414 106 || 95 37
Intel Pentium 4 HT 531 3.0 GHz 2699 324 74 || 64 24
Intel Pentium 4 2.0A 2.0 GHz 2109 247 49 || 45 15
Intel Pentium III EB 1.0 GHz 2177 118 25 || 28 8.2
Intel Pentium III 750 MHz 1646 80 18 || 21 6.1
Intel Pentium III Mobile 600 MHz 1298 51 11 || 16 4.2
Intel Pentium III 500 MHz 1088 40 9.1 || 13 3.6
Intel Pentium II 400 MHz 803 14 3.2 || 7.6 2.1
Intel Celeron A 300 MHz 649 11 2.8 || 4.9 1.6
Intel Pentium Pro 200 MHz 430 2.3 0.8 || 2.9 1.0

Notes:
Numbers are IBM XT in Speed Test for: DOS, DOSBox (dynamic), DOSBox (normal).
Numbers after '||' are fps in Doom demo3 for: DOSBox (dynamic), DOSBox (normal).
Doom fps = gametics / realtics * 35 ; where gametics=2134 for demo3, realtics are measured.
- To run DOSBox you need Win98 or newer OS, also works in Win95 with IE 4.01SP2 (msie401sp2.exe, 52 Mb) and several .dll from Win98 (MSVCRT.DLL, OPENGL32.DLL, GLU32.DLL, WS2_32.DLL, WS2HELP.DLL)
- Doom performance testing of Android devices with Magic Dosbox.

Real machines for the comparision
1. Speed Test has its own chart.
2. Doom's fps are in Phil's testing. As examples for Doom: 386DX 33 MHz - 6 fps, 486DX2 66 MHz - 25 fps, Pentium 100 MHz - 55 fps, Pentium II 300 MHz - 100 fps.

II. Late DOS games performance based on Quake

Testing procedure:
1) - Get DOSBox 0.74
- Create a default dosbox-0.74.conf. You need to run DOSBox->Extras->Reset or to delete/rename your current dosbox-0.74.conf, then after you'll run DOSBox it will create a new default dosbox-0.74.conf. In Windows XP dosbox-0.74.conf is situated in Documents and Settings\<username>\Local Settings\Application Data\DOSBox.
- Input these settings in dosbox-0.74.conf:
scaler=none
core=dynamic
cputype=pentium_slow
cycles=max
All other settings leave as default.
//
The part for those who did not use DOSBox previously. To associate games folder with drive 'C' you need to input in DOSBox's .conf section [autoexec] 2 strings:
mount c c:\games\
c:
We assume that Quake will be installed in c:\games\quake. Then, after DOSBox will start, you may type to run it:
cd quake
quake
//
2) - Get Quake 1.06 Shareware. Use only this version of the game.
- Run DOSBox and install Quake in it
3) - Reboot your PC
- Close all unnecessary programs, including background ones (antivirus, hardware monitorings, etc.)
- Turn off your Internet connections and local network from Control Panel. In Windows XP: Control Panel->Network Connections->Local Area Network, right click->Disable.
4) - Run DOSBox and Quake inside of it. During demonstrations for fps measuring use DOSBox in "window mode" only and the focus must stay in the emulator's window (it has to stay active).
- After the game's start press ESC.
- Enter the game's menu by ESC and set video mode 800x600 (in 320x200 measurements are better to do after as they are quicker). All other settings leave as default.
- Open the console by ~. Input: viewsize 120. Then input: timedemo 0 (to stop demonstrations).
- Input: timedemo demo1. Close the console by ~ on time of every demonstration. (do so 3 times) [In the console, to call a previous command you may by pressing "up arrow"]
//
It's not obligatory part. For those who like a little of automation. This will set video mode 800x600, adjust viewsize, stop demonstrations and will allow to play demo1 by a single click of keyboard (you'll need to run demo1 3 times to get average fps in it): F1 for demo1. Other demonstrations: F4 - demo2, F5 - demo3. Insert 5 strings below in the end of file Quake\ID1\config.cfg, and then set config.cfg as "read only":
vid_mode 13
viewsize 120
timedemo 0
bind F1 "timedemo demo1"
bind F4 "timedemo demo2"
bind F5 "timedemo demo3"
//
- Open the console and look at fps results. You may list the text in the console up/down by PgUp/PgDn. Calculate average fps from 3 results there. Do same for 320x200 video mode.
- Then check actual (as max) frequency of CPU during demonstrations run by CPU-Z. As there are many CPUs with dynamic frequency (turbo mode, SpeedStep). The game's window must stay active (main) during taking the frequency. Don't run CPU-Z during measuring of fps on the stages above.
5) Report your: CPU (model, actual frequency), OS (with 32 or 64 bit), average fps for demo1. It's not obligatory, but additionally you may say average fps in demo2, demo3. Speed in different OS may differ.

Results, fps

Intel Core i7 7700K 5.0 GHz (Win10-64, DDR4 3600): 76, 86, 74 / 234, 238, 221
Intel Core i7 7700K 4.5 GHz (Win10-64, DDR4 3000): 68, 78, 68 / 228, 240, 221
Intel Core i7 4790K 4.5 GHz (Win10-64): 63, 71, 62 / 201, 205, 190
Intel Core i7 5820K 4.6 GHz (Win7-64): 62, 70, 62 / 204, 208, 191
Intel Core i7 4930K 4.6 GHz (Win7-64): 60, 68, 59
AMD Ryzen 5 3600 4.2 GHz (Win10-64): 59, 59, 59 / 184, 184, 183
Intel Core i7 4790K 4.4 GHz (Win8.1-64): 57, 64, 57
Intel Core i7 4770K 4.0 GHz (Win8.1-64): 55, 61, 54 / 174, 176, 163
Intel Core i5 4690 3.9 GHz (Win7-64): 53, 60, 53
Intel Core i5 3570K 4.2 GHz (Win8.1-64): 52
Intel Core i7 3930K 4.2 GHz (Win7-64): 47, 55, 49 / 159, 162, 150
AMD Ryzen 5 2600 3.9 GHz (Win10-64): 47 / 146
Intel Core i7 5700HQ 3.5 GHz (Win10-64): 46, 52, 45 / 149, 152, 141
Intel Core i5 2400 3.9 GHz (Win8.1-64): 44, 49, 44
Intel Core i7 2600 3.8 GHz (Win10-64): 43, 47, 42 / 134, 136, 126
Intel Core i5 4200M 3.1 GHz (Win7-64): 43
Intel Core i7 980X 4.1 GHz (Win8.1-64): 41, 46, 41
Intel Core i5 2500K 3.7 GHz (Win8.1-64): 40, 44, 39
Intel Core i5 4570 3.6(?) GHz (Win8.1-64): 39, 42, 39 / 142, 142, 132
Intel Core i3 550 4.0 GHz (WinXP): 39, 44, 39
Intel Core i7 930 4.0 GHz (Win10-64): 39, 43, 38 / 131, 134, 125
Intel Core i7 2760QM 3.5 GHz (Win7-64): 38 / 115
Intel Core i7 920 3.8 GHz (Win7-64): 38, 43, 38
AMD Phenom II X6 1100T 3.7 GHz (WinXP-64): 36, 39, 35
Intel Core 2 Duo E8600 4.0 GHz (Win7): 35 / 123
AMD FX 8320 4.0 GHz (Win8.1-64): 33
Intel Xeon L5408 3.8 GHz (Win7-64): 33, 37, 32
AMD Phenom II X6 1090T 3.2 GHz (Win8.1-64): 32
Intel Pentium 2020M 2.4 GHz (Win7): 31 / 97
AMD Opteron 185 3.1 GHz (WinXP): 27, 29, 26
AMD Phenom II X4 905e 2.5 Ghz (WinXP): 24, 27, 24 / 75, 77, 71
Intel Core 2 Quad Q9550 2.9 GHz (WinVis): 23, 26, 23
Intel Core 2 Quad Q9400 2.7 GHz (WinXP): 23, 26, 23
Intel Core 2 Quad Q6600 2.4 GHz (Win7-64): 20, 23, 20 / 71, 73, 67
AMD Athlon X2 5050e 2.6 GHz (WinXP): 20, 21, 18
Intel Pentium E2180 2.0 Ghz (Win7-64): 16, 18, 15 / 49, 50, 47
Intel Core 2 Duo T5600 1.8 GHz (Win7-64): 15, 18, 15
Intel Pentium 4 Mobile HT 532 3.1 GHz (WinXP): 13, 15, 12
Intel Pentium 4 HT 531 3.0 GHz (WinXP): 12 / 38
AMD Sempron 2800+ 1.6 Ghz (WinXP): 12, 13, 11 / 36, 37, 34
Intel Pentium D 930 3.0 GHz (WinXP): 12
AMD Athlon XP 2600+ 1.9 GHz (Win2000): 11, 12, 12 / 39, 40, 37
Intel Pentium 4 HT 3.0 3.0 GHz (WinXP): 9.6, 11, 9.3
Intel Pentium 4 2.0A 2.0 Ghz (WinXP): 6.2, 7.1, 6.0 / 22, 20, 19
Intel Pentium III 1400S 1.6 GHz (WinXP): 6.1, 7.1, 6.0
Intel Pentium III 1400S 1.4 GHz (WinXP): 5.0, 5.8, 4.9
Intel Celeron M 353 900 MHz (WinXP): 4.0, 4.7, 3.8
Intel Pentium III EB 1.0 GHz (Win98): 2.7 / 14
Intel Pentium III Mobile 866 MHz (WinNT4): 2.6 / 14
Intel Pentium III Mobile 650 MHz (WinXP): 1.6, 1.8, 1.5
Intel Pentium III Mobile 600 Mhz (WinME): 1.0, 1.2, 1.0 / 6.4, 6.6, 5.9
Intel Pentium MMX 200 MHz (Win98): 0.0 / 0.8

non-Windows
Intel Core i5 4200M 3.1 GHz (Ubuntu 14.10 x64): 18
AMD Athlon XP 2600+ 1.9 GHz (Linux Mint 17.2): 9.0, 10, 8.9 / 27, 27, 25
Broadcom BCM2836 900 Mhz (Raspbian): 0.2, 0.3, 0.2 / 1.3, 1.3, 1.2

Real machines for the comparision:
Intel Pentium III 1.4 GHz (DOS) [fastvid,nosound]: 64 / 257
Intel Pentium III EB 1.0 GHz (DOS) [fastvid,nosound]: 52 / 183
Intel Pentium III 600 MHz (DOS) [fastvid,nosound]: 36 / 127
Intel Pentium MMX 200 MHz (DOS) [nosound]: 9.7, 11, 9.5 / 38, 38, 35
Intel Pentium 100 MHz (DOS) [nosound]: - / 22
notes:
For AMD CPUs instead of fastvid may be used mtrrlfbe (run "mtrrlfbe lfb/vga wc")
The performance on real machines, besides CPU, may be affected by: chipset, video card, memory, etc.
- Also look at slighly different benchmark testing by: Thandor, Verstichelen, Phil.

Note:
Single number of fps is for demo1; where 3 numbers: demo1, demo2, demo3. Results after '/' are for 320x200 mode.
If a CPU has turbo mode, then its turbo frequency is placed. DOSBox uses a single core, so when there are no working concurrent programs the CPU most likely works in turbo mode, but not on the base frequency.

upd: Taking into account the existence of Phil's testing, mode 320x200 was added in Quake's testing. Added viewsize (120 instead of 100 by default) parameter in the procedure for the correspondence with Phil's results (though, later he've used default 100 in Quake, while still removing health/ammo bar in Doom). According to influence of new viewsize parameter, previous fps were remeasured or approximated by the division on experimentally gotten coefficients for demo1, demo2, demo3: 800x600 (1.041, 1.049, 1.051), 320x200 (1.133, 1.139, 1.138).

Last edited by Tertz on 2022-02-16, 11:13. Edited 238 times in total.

DOSBox CPU Benchmark
Yamaha YMF7x4 Guide

Reply 1 of 196, by leileilol

User metadata
Rank l33t++
Rank
l33t++

would be much easier to start Quake with this:

quake +vid_mode 13 +viewsize 100 +timedemo demo1

Disabling those other services, internet connections and programs shouldn't be necessary, unless your hardware really sucks, then it shouldn't have business running Quake in an emulator anyway

apsosig.png
long live PCem

Reply 4 of 196, by Tertz

User metadata
Rank Oldbie
Rank
Oldbie
leileilol wrote:

would be much easier to start Quake with this
quake +vid_mode 13 +viewsize 100 +timedemo demo1

Everything besides video mode must be as default. If default viewsize is not 100 then it's not appropriate, and in other case it's redundant.
Also, your offer is not significantly easier for a single game run for testing. Besides 10 sec to set video mode with game's menu, you'll need then to type manually in console timedemo to run the demonstration 2 times more.

Disabling those other services, internet connections and programs shouldn't be necessary

As programs may use CPU resources, then it's desirable. Network processes running in the background may take resources too.

unless your hardware really sucks

Your understanding of computers is very original. 😀 Anyway, any IBM PC is allowed for testing where DOSBox runs. If DOSBox may work in Windows 9x then even 386/486 may take part in the chart.
For example, on C2D I get 2 fps of loss during this testing when someone downloads/uploads data from me in my net. Besides net processes, we can't imagine all demanding miracles wich people may load on start-up and how that apps may interfere affecting system performance. So, the cleaner system memory - the more accurate results.

F2bnp wrote:

I get 33fps on my Phenom II x6 1090T

OS (with 32/64 bits)?
Is CPU frequency standard - 3.2 GHz?

I might get an FX 8320 for cheap this week, if so I will definitely try this out again 😀.

Would be nice.

Last edited by Tertz on 2015-05-11, 13:24. Edited 2 times in total.

DOSBox CPU Benchmark
Yamaha YMF7x4 Guide

Reply 6 of 196, by jk3one

User metadata
Rank Newbie
Rank
Newbie

CPU i7 2600 at stock (3,4 GHZ). Win 8.1 64bit
timedemo 1: 44,7 FPS
Very constant, by the way. I ran the demo 4 times and it came out at 44,7 every time. If something happens in the background (nothing big, just another program working or so), it may go down to 43 FPS or so, but basically DOSBox does a great job

Some more testing: in fullscreen, I get 45,8 FPS constantly (so its a little bit higher)

and timedemo 2: 49,7 FPS
timedemo 3: 44,5 FPS

Reply 9 of 196, by Tertz

User metadata
Rank Oldbie
Rank
Oldbie
ripa wrote:

You should also use fulldouble=false

It's by default and does not affect window mode.

SDL video output can be a bottleneck and affect Dosbox CPU cycle controller.

Mentioned "bottleneck" exists on every system, is minor and using frameskip would change the proportions between performances in not significant degree. While as this setting is not typicaly used, so the results just would become farther from practice.

Last edited by Tertz on 2015-05-11, 20:28. Edited 1 time in total.

DOSBox CPU Benchmark
Yamaha YMF7x4 Guide

Reply 10 of 196, by ripa

User metadata
Rank Oldbie
Rank
Oldbie

My results with Tertz's settings.

Windows XP 32-bit, AMD Athlon 64 X2 5050e (2.6 GHz):
timedemo1: 20.5fps
timedemo2: 22.4fps
timedemo3: 19.4fps

Fulldouble and frameskip actually had no effect on FPS.

PS. nice step-by-step instructions and benchmark.

Reply 11 of 196, by leileilol

User metadata
Rank l33t++
Rank
l33t++
Tertz wrote:

Also, your offer is not significantly easier for a single game run for testing.

Make it a batch file then.

Tertz wrote:

you'll need then to type manually in console timedemo to run the demonstration 2 times more.

bind f1 "timedemo demo1"

and just keep tapping F1. The function keys are bindable and work in a disconnected console

apsosig.png
long live PCem

Reply 12 of 196, by Tertz

User metadata
Rank Oldbie
Rank
Oldbie

leileilol

Your offers do not make the benchmark meaningfuly easier or more comfortable for a single tester.

upd:
I've added some automation commands in case someone will want to spare a couple of clickings and have additional messing with files.

Last edited by Tertz on 2015-05-27, 08:28. Edited 1 time in total.

DOSBox CPU Benchmark
Yamaha YMF7x4 Guide

Reply 13 of 196, by F2bnp

User metadata
Rank l33t
Rank
l33t

What a shameless showcase for Piledriver. I got an 8320 yesterday, by default it scores 31.5fps. Pushing it to 8350 levels (4GHz), it manages to score 34fps, which is still pretty lame 😵 .

Hopefully, it will speed up a few other tasks that I need it for 🤣 .

Reply 14 of 196, by Tertz

User metadata
Rank Oldbie
Rank
Oldbie
F2bnp wrote:

I got an 8320 yesterday, by default it scores 31.5fps. Pushing it to 8350 levels (4GHz), it manages to score 34fps

4 GHz is its turbo mode (current frequency should to show something like cpu-z). You've got 34 fps on overclocked (and hence maybe at turbo > 4 GHz) or on standard base frequency 3.5 GHz and hence on ordinary turbo = 4 GHz?

And wouldn't you be so kind to tell your OS and its bits?

DOSBox CPU Benchmark
Yamaha YMF7x4 Guide

Reply 16 of 196, by Tertz

User metadata
Rank Oldbie
Rank
Oldbie
F2bnp wrote:

No, I overclocked it to 4GHz and disabled Turbo. My OS is the same as above, Win 8.1 x64.

Thank you.

I wonder if the latest SVN builds improve perfomance somewhat, might be cool to give that a try.

Not a bad source about CPU for games is thg. According to a chart I saw for GTA 5, for additional $60 you may get i5 4590 wich is faster +25% there, compared to FX 8320 at 4 GHz. Similarly in DOSBox you'd get 34 fps +25% = ~43-46 fps. Some shops support return during several days, I'd think about this if some more speed in games is important.

For SVN I'd look on ykhwong's build from 2014-01-27.

DOSBox CPU Benchmark
Yamaha YMF7x4 Guide

Reply 17 of 196, by F2bnp

User metadata
Rank l33t
Rank
l33t

Tried the latest EmuCR build, managed to get a whopping 1 frame per second more 🤣. 35fps is about as good as it gets I guess, I imagine with an Intel CPU you might see more of a performance gain with this build, I suggest trying it out.

Thanks for the suggestions Tertz, but I'm not really interested in upgrading any further, this was a very inexpensive upgrade and I had been toying around with the idea for a while. I'd rather keep my money for when truly significant CPUs arrive and DDR4 becomes mainstream.

Reply 18 of 196, by Tertz

User metadata
Rank Oldbie
Rank
Oldbie
F2bnp wrote:

Tried the latest EmuCR build, managed to get a whopping 1 frame per second more 35fps

SVN 3914 gave me same fps. And ykhwong's gave same after set video options as 0.74 defaults.

I'd rather keep my money for when truly significant CPUs arrive and DDR4 becomes mainstream.

While I'm waiting for new 65w i5 and cheaper DX12 cards from NVidia.

DOSBox CPU Benchmark
Yamaha YMF7x4 Guide

Reply 19 of 196, by smeezekitty

User metadata
Rank Oldbie
Rank
Oldbie

I couldn't be bothered to shut down all the background processes or disable the network so it may be low.

Radeon 7850
Core 2 Quad Q9550 @ ~2.9 (FSB slightly bumped for strange reasons)

Windows Vista 32 bit
3 GB of usable RAM

Timedemo 1: 23.8
Timedemo 2: 36.3
Timedemo 3: 46.3
Average: 35.46