Intel i5 4570, Windows 8.1 x64, 8gb ram, no process in background

320 x 200 demo 1 969 frames - fps 141, 143, 142.141
320 x 200 demo 2 985 frames - fps 144, 141.6, 139.7
320 x 200 demo 3 1090 frames - fps 134, 132, 130.2

800 x 600 demo 1 960 frames - fps 39.5, 39.4, 39.4
800 x 600 demo 2 985 frames - fps 43, 42, 41.5
800 x 600 demo 3 1090 frames - fps 39.2, 39, 39

Core speed always at same speed for each test

Thank you, Arklay.
Your results are 15% lower than should to be compared to other i5/i7 3xxx-4xxx with same frequency. Some with 4-core CPUs reported that presence of background applications gave higher results to them (+10%). Maybe similar situation, or something other like a concurrent application.

As requested by Tertz (via PM), here are the fps results for both my P3 Slot 1 and P4 Northwood PCs.

Pentium III 600MHz Katmai Slot 1 (no overclocking), Asus P3B-F (440BX), 512mb PC133 ram, DOS 7 clean boot (Windows 98 SE), Nvidia GeForce 2 GTS 32mb (fastvid enabled), sound disabled

320x200=143.2

800x600=37.2

Pentium 4 Northwood 2400MHz Socket 478 (no overclocking), Soyo P4I-845PE, 2gb DDR333 ram, DOS 7 clean boot (Windows XP SP3), Nvidia GeForce 6600 256mb (fastvid enabled), sound disabled

320x200=233.7

800x600=72.8

BIOS settings for both PCs were set at defaults. As for DOSBox, I'll just skip for now.

EDIT - To answer some questions asked by Tertz (via PM again):

- Pentium III Katmai runs at 100MHz FSB.
- Pentium 4 Northwood runs at 533MHz FSB and is non-HT.
- I had to get sneaky with the Pentium 4/Windows XP PC to boot DOS 7. Created a separate 4gb FAT32 partition, booted with a modified Win98SE floppy (has an option for clean boot ^^ ), installed Quake 1.06 in it and ran the benchmark as well. Used the same boot floppy with the Pentium III/Windows 98 SE PC as well.
- Viewsize for both tests was 120. I could rerun them again to verify, but have to wait for some more free time. Also, neither PC has working sound in DOS - both have PCI sound cards and DOS sound emulation is not enabled, no joysticks/gamepads or network cards are installed, and no CD/DVD-ROM drivers are loaded either.

EDIT 2 - Got some time to rerun the benchmarks again ,making sure fastvid was enabled and viewsize was set at 120.

Pentium III 600MHz Katmai (100MHz FSB, fastvid 111 32 D8000000):

320x200=126.7

800x600=35.9

Pentium 4 Northwood 2400MHz non-HT (533MHz FSB, fastvid 111 256 D0000000)

320x200=219.3

800x600=71.3

Once again sound was disabled - these PCs are Windows only builds and I don't have the time or need to enable and debug DOS sound emulation from PCI sound cards just for a benchmark. And if these results are still not good enough then all I can say is I'm sorry, you won't get anything more from me. Just take it as is or forget about it.

Thanks, JayCeeBee64, for results and specializings.

> Pentium III 600MHz Katmai 320x200=126.7 800x600=35.9

This looks closer to expected and seems as viewsize was 100 previously.
Approximation coefficients for converting viewsize 100->120 wich were gotten experimentally for demo1, demo2, demo3 are: 800x600 (1.041, 1.049, 1.051), 320x200 (1.133, 1.139, 1.138). For example: 143.2/1.133 = 126.4

> Pentium 4 Northwood 2400MHz 320x200=219.3 800x600=71.3

Results on your same CPU+MB are lower than in Phil's testing on 26% (219/295=74%), - it's significant. I'm sorry but it's better to decline them. Maybe other video card matters, - Nvidia FX5500. If you'll make retest with fps closer to real possibilities of your machine then it will be resonable to input them.
The main interest for work in DOS is on machines from DOS era - the last typical were P3. So results for P4 are not very important in DOS, but in DOSBox are more meaningful.

these PCs are Windows only builds and I don't have the time or need to enable and debug DOS sound emulation from PCI sound cards just for a benchmark

I doubt sound off vs SB ISA on P3 (maybe even P2) and higher affects fps much in Quake, so it's acceptable.

Intel Core i7 4770K 4.0 GHz (Win8.1-64): 55, 61, 54

Ran it on my Win98 Machine in Win98 DOS NO AUDIO Geforce Ultra 5950 for fun:

P4 3.4Ghz (Win98 SE): 127, 141, 127

Shagittarius wrote:

Intel Core i7 4770K 4.0 GHz (Win8.1-64): 55, 61, 54

Thanks. And for 320x200 mode fps is needed too (with viewsize 120, etc). At least demo1.

If you'd wish to input data for P4 3.4 Ghz it should to be either in DOSBox, or in DOS with fastvid (Win9x has DOS-boot option). Concrete P4 model may be checked with cpu-z and that list.

Core i7 4770K 4.0 Ghz (Win8.1-64) @ 320x200 : 174,176,163

Shagittarius
Thanks for data.

Hi, thanks for the benchmark invite. Per your instructions, here are my results.

Core2Quad Q6600 2.4ghz stock (Win 7 x64)

800x600:
demo1 - 20 fps
demo2 - 23 fps
demo3 - 20 fps

320x200:
demo1 - 71 fps
demo2 - 73 fps
demo3 - 67 fps

fyy
Thank you.

Tertz wrote:

fyy
Thank you.

What is interesting is that Core2Duo E8600 @ 4 ghz is 66% higher clocked than my Q6600 (4 ghz vs 2.4 ghz). And it's score is also 75% better in first test and 73% better in 320x200. It seems to scale very close to linearly based on clock (for same architecture) and is not taking into account additional cores

fyy wrote:

What is interesting is that Core2Duo E8600 @ 4 ghz is 66% higher clocked than my Q6600 (4 ghz vs 2.4 ghz). And it's score is also 75% better in first test and 73% better in 320x200. It seems to scale very close to linearly based on clock (for same architecture) and is not taking into account additional cores

DOSBox is a single core application, DOS games too.

Intel Core i7-3930K 4.2 GHz (win7-64): 47, 55, 49 // 159, 162, 150

spazz

Thank you.

Cool idea.

Intel Core i7-5820K @ 4.6 GHz (Win7 x64): 62, 70, 62 / 204, 208, 191

I knew I built this rig for some reason, and I've finally found it 😜

galneon wrote:

Intel Core i7-5820K @ 4.6 GHz

30% overclocking. How in good old times, if it's on air, without decap and is stable in Linx.
Thank you.

I knew I built this rig for some reason, and I've finally found it

While Unreal with Voodoo (software emu) still gives ~10 fps (640x480) on today top CPUs. Multithreading support is needing to improve this, or Intel will remember about the progress instead of only raking money by a shovel.

Tertz wrote:

30% overclocking. How in good old times, if it's on air, without decap and is stable in Linx.
Thank you.

Yeah, it's kind of ridiculous but it's stable with linpack (via OCCT) on air at that clock. Individual 5820Ks typically range from 4.4-4.6 for max overclock. Honestly, when I saw the heat produced by latest version of Prime95, I cut that test short. I just couldn't handle it--like nuking my GPU with Furmark.

While Unreal with Voodoo (software emu) still gives ~10 fps (640x480) on today top CPUs. Multithreading support is needing to improve this, or Intel will remember about the progress instead of only raking money by a shovel.

Then I suppose Unreal is what my next build (in ~5 years) should be about 😜

galneon wrote:

Then I suppose Unreal is what my next build (in ~5 years) should be about 😜

Things are worse. 30 fps is what comfortable 3D gaming needs. There is no significant GHz gain during 10 years - P4 had 3.8 GHz in 2004, so after 5 years we may have same ~4 GHz. Q6600 was produced in the beginning of 2007 and it's difference of single core speed with today top CPUs like i7 5820 (produced at 2014 end) on the same frequency is only 1.5 times, - this is after 7.5 years. If such situation will stay, 3 times rise of speed would need: log (3 on 1.5) * 7.5 = 2.7 * 7.5 = 20 years.

Just testing this on Windows 10 retail (did the update, not a clean install). Again very consistent - and slower:
CPU i7 2600 at stock (3,4 GHZ). Win 8.1 64bit

timedemo 1: 44,7 FPS on Win8.1 but around 42.5 FPS on Win10
timedemo 2: 49,7 FPS (W8.1) vs. 47 FPS (W10)
timedemo 3: 44,5 FPS vs 42 FPS

Well, MS should polish it more, I think.

jk3one wrote:

timedemo 1: 44,7 FPS on Win8.1 but around 42.5 FPS on Win10
timedemo 2: 49,7 FPS (W8.1) vs. 47 FPS (W10)
timedemo 3: 44,5 FPS vs 42 FPS

I made changes in the procedure by "viewsize 120".
Your previous results for Win 8.1: 44.7, 49.7, 44.5 were for default "viewsize 100".
With viewsize 120 they should be: 43, 47, 42
So Win10 has results identical to Win8.1 if that Win10 results were gotten with the current viewsize.

I'll enter the results for Win10 if you'll also test in 320x200 wich exists as obligate in the current procedure. Frequency should be checked by recommended sequence to be sure CPU works in full turbo.