Looking for volunteers to help test a new benchmark

[Trixter]

Without the exact system to debug on, I can't fix that, sorry. Luckily, the CPU tests are skippable with -s on the command-line.

 

[Cloudschatze]

Ah, well. Next time you're out west, feel free to drop by...

I have been skipping the CPU tests, all-the-same, and am eking out scores of 40 and 37 with both TGA (CGA) and VGA, respectively.

 

[Cimonvg]

hello
Just tested my IBM 5150 with MACH20 accelerator card. And it was slower then the Juko v20 I did test some time ago
..have to try my 286->486 cpu in the MACH20 card .. soon!

;PC5150revb_WITH_MACH20
[UID3B85189F]
MemoryTest=1263
OpcodeTest=1647
VidramTest=1406
MemEATest=2159
3DGameTest=1549
Score=13
CPU=Intel 80286
CPUspeed=8 MHz
BIOSinfo=1501476 COPR. IBM 1982 (10/27/82, rev. 0)
BIOSdate=19821027
BIOSCRC16=3B85
VideoSystem=VGA
VideoAdapter=VGA, Paradise PVGA1A, 8-Bit-Video, 8-Bit-ROMfixed-sync, 512kb Vi
Machine=PC5150revb_WITH_MACH20
Description=ERZO_DOS_320
Submitter=CIMONVG

Trixter - question. Where do we place new measurements ?
And, do you have any plans , regarding how many different computers the database should consist of ?
/cimonvg

 

[Compgeke]

I'm not sure if this is a bug or not, but I ran the test on my Sony PCV-LX800 (upgraded to a 1 GHz Pentium III) and it appears to be a lot faster than Tiberian Fiend's K6-2. I ran the test twice to check, going to try the stub now to see if there's a difference. The only modifications I made to the results are to change the CPU speed to 1 GHz (which the BIOS sees, not sure why the program detects it as 933 MHz), set the system model to PCV-LX800 and changed the video chipset from "unknown" to SIS630. To avoid confusion, all these scores are from the same machine.

;Sony PCV-LX800 (2)
[UIDF6DF3540]
MemoryTest=2
OpcodeTest=0
VidramTest=24
MemEATest=1
3DGameTest=0
Score=1538
CPU=Intel Pentium III E
CPUspeed=1000 MHz
BIOSinfo=. Awar,Copyright (C) 1984-98, Award Software, Inc. (09/20/00, rev. 0)
BIOSdate=20000920
BIOSCRC16=F6DF
VideoSystem=VGA
VideoAdapter=VGA, Tseng ET , DRAM, SIS630, VESA, 256kb Video Memory
Machine=Sony PCV-LX800 (2)
Description=
Submitter=

;Sony PCV-LX800
[UIDF6DF26AA]
MemoryTest=1
OpcodeTest=2
VidramTest=22
MemEATest=3
3DGameTest=3
Score=1541
CPU=Intel Pentium III E
CPUspeed=1000 MHz
BIOSinfo=. Awar,Copyright (C) 1984-98, Award Software, Inc. (09/20/00, rev. 0)
BIOSdate=20000920
BIOSCRC16=F6DF
VideoSystem=VGA
VideoAdapter=VGA, DRAM, SIS630, VESA, 256kb Video Memory
Machine=Sony PCV-LX800
Description=
Submitter=

Edit: Ran the stub twice and got very much similar results.

[UIDF6DF57DD]
MemoryTest=2
OpcodeTest=2
VidramTest=22
MemEATest=1
3DGameTest=1
Score=1548
CPU=Intel Pentium III E
CPUspeed=933 MHz
BIOSinfo=. Awar,Copyright (C) 1984-98, Award Software, Inc. (09/20/00, rev. 0)
MachineModel=01FC
BIOSdate=20000920
BIOSCRC16=F6DF
VideoSystem=VGA
VideoAdapter=VGA, Tseng ET , DRAM, unknown Chipset, VESA, 256kb Video Memory
Machine=AT clone

[UIDF6DF5B47]
MemoryTest=1
OpcodeTest=1
VidramTest=24
MemEATest=54924
3DGameTest=0
Score=1545
CPU=Intel Pentium III E
CPUspeed=966 MHz
BIOSinfo=. Awar,Copyright (C) 1984-98, Award Software, Inc. (09/20/00, rev. 0)
MachineModel=01FC
BIOSdate=20000920
BIOSCRC16=F6DF
VideoSystem=VGA
VideoAdapter=VGA, Tseng ET , DRAM, unknown Chipset, VESA, 256kb Video Memory
Machine=AT clone

 

[Trixter]

hello
Just tested my IBM 5150 with MACH20 accelerator card. And it was slower then the Juko v20 I did test some time ago
..have to try my 286->486 cpu in the MACH20 card .. soon!

Cool! I've added it to the database and will publish in a new release this weekend.

Trixter - question. Where do we place new measurements ?
And, do you have any plans , regarding how many different computers the database should consist of ?
/cimonvg

I'd prefer if they were emailed to me since that's easiest and doesn't clog up the forum, but either is fine. As for how many computers, I had no limit, but I don't accept results for artificially slow systems (ie. with turbo off or cache disabled) and I don't accept results for emulators either.

 

[Trixter]

I'm not sure if this is a bug or not, but I ran the test on my Sony PCV-LX800 (upgraded to a 1 GHz Pentium III) and it appears to be a lot faster than Tiberian Fiend's K6-2.

Not a bug, you now hold the record! I'm assuming this was a clean boot, right?

MemEATest=54924

Whoops, looks like I'll need to update the stub with my measurement fixes. I'll add that to the to-do list.

 

[Compgeke]

It was a mostly-clean boot, being the only thing loaded was the DLink ISA packet driver which failed to load (this system lacks a floppy drive so I used the 486's CF "hard drive").

 

[Trixter]

In any case, it will now warn you if you're in protected mode and refuse to run. I added a command-line switch to override in case it is a false positive (which would only happen in an emulator) or the CPU detection logic blows a gasket.

Other changes:

- Recompiled stub so it has the new measurement wrapper

- Enhanced /i so that it also spits out usec timings

- Added [letter] to jump to next system that starts with that letter in the browsing interface

- Added command-line option to work with a different database file.

- Added some small "special effects" if working on a color monitor (with command-line option to override)

Snag the new version at http://dosbenchmark.wordpress.com/downloads/

 

[Compgeke]

Re-ran on the Sony with your new update on a clean boot (MS-DOS 6.22 disk, nothing on it BUT topbench). The score was better, however I have noticed a bug, in that the program is seeing the 1 GHz Pentium III has a 933 MHz processor. The results I posted are ones I edited the CPU speed in to be correct for the Sony, however I have a feeling the problem is the program is seeing the speed as 133*7.5 for 997.5 MHz but then is reporting as 933 MHz as 997.5 is a odd result, but the 1 GHz Pentium III with a 133 MHz FSB isn't really 1 GHz.

Another thing I can think of is the BIOS doesn't have the CPUID for the processor, so it isn't showing the right speed to the OS even though it's seen right in the BIOS, thanks to this message:
<BIOS:> BIOS update data incorrect. CPUID=0000068A

And the BIOS's reported speed:
http://goput.it/b93.jpg

And the chip itself:
http://goput.it/szr.jpg

Yes, that must've been some really bad thermal paste to run like that.

Edit:
Updated the BIOS to what appears to be the latest (version 2004 but there's also a version 3002 from 2 years before...but I can't extract that one using anything and it won't run on my laptop to make the disk (wrong system)...it's not like that DOS updater could do anything to Windows 7 x64 anyways.

 

[Trixter]

I have noticed a bug, in that the program is seeing the 1 GHz Pentium III has a 933 MHz processor. The results I posted are ones I edited the CPU speed in to be correct for the Sony, however I have a feeling the problem is the program is seeing the speed as 133*7.5 for 997.5 MHz but then is reporting as 933 MHz as 997.5 is a odd result, but the 1 GHz Pentium III with a 133 MHz FSB isn't really 1 GHz.

The CPU and speed detection is a convenience for the user; it is not always accurate. I am glad you edited it to be correct before submitting the result.

I already munge some results to be more friendly; for example, I quantize results to 4.77 MHz or 7.16 MHz if it is very close to 5 or 7 respectively. I used to quantize 9.x to 9.54 MHz (appropriate for 8086 and NEC V30) but this screwed up legitimate 10 MHz results so I stopped doing that.

You've given me the idea to add a command-line switch to disable all MHz quantization, so I'll add that in the next release. That will return some interesting values (I got 12.8484 MHz from a 12-MHz 286 for example) but maybe it will help on the higher end.

 

[Trixter]

the program is seeing the 1 GHz Pentium III has a 933 MHz processor.

Just for you ;-) I added a command-line switch to not attempt any user-friendly normalization of the detected MHz -- see if it works better for you.

Full changelog:

• Added command-line option to do “real” MHz detection (raw result is returned without any type of user-friendly normalization). Warning, this has the potential to return some pretty wacky results since MHz detection is not an exact science; use only if TOPBENCH consistently mis-reports MHz by orders of magnitude or something. Besides, you should be fixing the value before committing it to the database anyway!)
• Added search-by-number jumps in the browser interface
• Bugfix: Importing numeric values from a text file with trailing spaces after the value resulted in a value of 0
• Restructured a few aspects of TOPBENCH so that there would be more free memory when debugging in the Turbo Pascal IDE
• Added TOPBENCH.TXT as super-quick documentation to go into the .zip file

 

[Compgeke]

I'll run that tomorrow to check it out, just doing my last right-before-bed forum checks now . I'll also see if it fixes the whacko Cyrix MediaGX speed (detects anywhere from 630 MHz-980 MHz) although I'm not worried about that as no one actually owns a MediaGX system.

 

[Cimonvg]

Cool! I've added it to the database and will publish in a new release this weekend.
....

Hello - strange. The 5150 motherboard is flawless with a v20 or the MACH20 (with original 80286). But with a Inbard386/pc (cpu:evergreen 20/60) or the MACH20 (with cpu "make it 486") the PC only starts with "beeep bep"
??
/regards cimonvg

 

[Cimonvg]

Hello - strange. The 5150 motherboard is flawless with a v20 or the MACH20 (with original 80286). But with a Inbard386/pc (cpu:evergreen 20/60) or the MACH20 (with cpu "make it 486") the PC only starts with "beeep bep"
??
/regards cimonvg

nothing much.
..Comment to my 1/2 year old remark (above) , the Inboard (with Evergreen) runs like a charm on the first PC 5150a. You just have to place the switches the right way
and today I did a benchmark on a IBM e220 tower-server (model 8646-42x, dual pentium III (not pro) cpu):

;Data collected by: TOPBENCH | Benchmark and detection stub | Version 0.97c
;This file contains fingerprinting information about your computer. Please
;email this file to trixter@oldskool.org with a subject line of "Benchmark" to
;help test these routines and seed the TOPBENCH database.

[UID26ED2D8A]
MemoryTest=54924 (some times reads "0" here)
OpcodeTest=0
VidramTest=158
MemEATest=0
3DGameTest=54924 (some times reads "0" here)
Score=320
CPU=Intel Pentium Pro (undistinguished)
CPUspeed=1229 MHz
BIOSinfo=COPR. IBM 1981, 1984 (12/07/01, rev. 2)
MachineModel=01FC
BIOSdate=20011207
BIOSCRC16=26ED
VideoSystem=VGA
VideoAdapter=VGA, Yamaha 6388, VESA, 256kb Video Memory (BIOS)
Machine=Compaq 286/386 or clone

 

[Trixter]

The stub is meant for systems that can't run the full tool (ie. have 128K RAM or don't have a IBM compatible BIOS), hence some of the wacky timings. Is there any way you could run the full TOPBENCH tool on your server and post the database section for that run?

 

[Cimonvg]

hello again
..not that high rating for this server ? CPU speed at 1.2 Ghz is kind of fast, perhaps I should try a better VGA card
TOPBENCH data:
[UID26EDBB20D]
MemoryTest=0
OpcodeTest=54924
VidramTest=139
MemEATest=0
3DGameTest=0
Score=334
CPU=Intel Pentium III x 2
CPUspeed=1143 MHz
BIOSinfo=COPR. IBM 1981, 1984 (12/07/01, rev. 2)
BIOSdate=20011207
BIOSCRC16=26ED
VideoSystem=VGA
VideoAdapter=VGA, Yamaha 6388, VESA, 256kb Video Memory (BIOS)
Machine=IBM 220 tower server 8646-42x
Description=
Submitter=cimonvg

 

[Trixter]

Keep in mind that TOPBENCH only tests 16-bit real-mode performance, and starting with the Pentium Pro Intel optimized microcode to favor 32-bit protected mode code over all others.

I'm going to release an update in a few weeks with many database additions and a bugfix; I'll reply back here when that's done.

 

[Wormetti]

http://www.youtube.com/watch?v=EXx2d1yvhRM

[UID495E12B381]
MemoryTest=618
OpcodeTest=239
VidramTest=1240
MemEATest=304
3DGameTest=253
Score=19
CPU=Intel 80286
CPUspeed=10 MHz
BIOSinfo=COPR. IBM 1981, 1987 (04/18/88, rev. 3)
BIOSdate=19880418
BIOSCRC16=495E
VideoSystem=VGA
VideoAdapter=VGA, unknown Chipset, 256kb Video Memory (BIOS)
Machine=IBM PS/2 Model 50Z 286-10
Description=
Submitter=wormetti@gmail.com twitter.com/wormetti

 

[Wormetti]

Compaq Portable III https://www.youtube.com/watch?v=zJ3k2QdJi1s

Why did this score a 15 in my video (using topbench -i) but 21 in the database.ini?

[UID1002A40B2]
MemoryTest=623
OpcodeTest=265
VidramTest=854
MemEATest=348
3DGameTest=264
Score=21
CPU=Intel 80286
CPUspeed=12 MHz
BIOSinfo=\JSF'PNA'CHL'5CRJ (C)Copyright COMPAQ Computer Corporation 1982,83,84,85 (01/29/87, rev. 0)
BIOSdate=19870129
BIOSCRC16=1002
VideoSystem=CGA
VideoAdapter=Compaq Portable III (Internal Plasma)
Machine=AT clone
Description=
Submitter=wormetti@gmail.com twitter.com/wormetti

;Commodore PC10-III
[UID95CB9133]
MemoryTest=3655
OpcodeTest=1773
VidramTest=2298
MemEATest=1966
3DGameTest=1889
Score=4
CPU=Intel 8088
CPUspeed=4.77 MHz
BIOSinfo=Copyright 1987,1988 Commodore Electronics Ltd. (09/08/88, rev. 0)
BIOSdate=19880908
BIOSCRC16=95CB
VideoSystem=CGA
VideoAdapter=CGA
Machine=Commodore PC10-III
Description=ctrl+alt+s (slow mode)
Submitter=wormetti@gmail.com twitter.com/wormetti

;Commodore PC10-III
[UID95CB4D0B]
MemoryTest=2469
OpcodeTest=1175
VidramTest=1744
MemEATest=1302
3DGameTest=1252
Score=6
CPU=Intel 8088
CPUspeed=7.16 MHz
BIOSinfo=Copyright 1987,1988 Commodore Electronics Ltd. (09/08/88, rev. 0)
BIOSdate=19880908
BIOSCRC16=95CB
VideoSystem=CGA
VideoAdapter=CGA
Machine=Commodore PC10-III
Description=ctrl+alt+t (turbo mode)
Submitter=wormetti@gmail.com twitter.com/wormetti

;Commodore PC10-III
[UID95CB1611]
MemoryTest=1851
OpcodeTest=882
VidramTest=1669
MemEATest=970
3DGameTest=926
Score=8
CPU=Intel 8088
CPUspeed=10 MHz
BIOSinfo=Copyright 1987,1988 Commodore Electronics Ltd. (09/08/88, rev. 0)
BIOSdate=19880908
BIOSCRC16=95CB
VideoSystem=VGA
VideoAdapter=VGAOAK OTI037C, 256kb Video Memory (BIOS)
Machine=Commodore PC10-III
Description=ctrl+alt+d (double speed mode)
Submitter=wormetti@gmail.com twitter.com/wormetti

 

[Trixter]

Your Compaq is showing a lower score because you're running it off of floppy, and the Compaq Portable III slows down when the floppy drive is active (this was an intentional hardware design decision on Compaq's part for compatibility with copy-protection schemes). If you run the full version and wait for floppy access to stop, you should get the correct score.

OR... download the new version (ack!) here: http://dosbenchmark.wordpress.com/downloads/ I added a continuous mode ("-l") that should show the score jump up when the floppy motor stops.

I should probably make an announcement about the new version, which I'll do in a new post.