VOGONS


First post, by Hanamichi

User metadata
Rank Member
Rank
Member

Quick and dirty test of these cards, as they often come up in DOS PCI card conversations. Might be useful for some folks.

Other people have done much better testing but maybe not including all 3 on the same system.

Benchmarks taken from Phils DOS benchmark.


The cards:
Jaton Video Magic 128 (ET6000) 2MB MDRAM
Apocalypse 5D (ET6100, PCX2, DEC PCI Bridge) 4MB MDRAM
Legend 64 (ARK2000) 2MB VRAM


Test System:

AOpen AX3S Pro-U (815e B stepping chipset)
Pentium III-S 1400
2x 256MB Infineon PC133 2225 modules
Athlon XP S462 cooler, nothing special
No sound card installed, onboard sound disabled


Tualatin 1400 @ 1400MHz/1.45v/10.5 x 133.6MHz Bus/133MHz 2,2,2,5 SDRAM/33MHz PCI

Codeblock version

                                        ET6000             ET6100               ARK2000
No graphics booster
Quake VGA 243.8 245.0 236.0
Quake 640x480 72.4 71.9 (Need UniVBE)

MTRRLFBE
Quake VGA 252.9 252.7 Not compatible with MTRRLFBE without UniVBE
Quake 640x480 72.2 71.9 "

Fastvid
Quake VGA 252.9 254.9 231.9 (Doesn't work properly on 815E?)
Quake 640x480 72.3 72.2 (Need UniVBE)

UniVBE + MTRRLFBE
Quake VGA not tested not tested 231.2
Quake 640x480 not tested not tested 78.6

Table screenshot

PIII1400 ET Tests.PNG
Filename
PIII1400 ET Tests.PNG
File size
64.49 KiB
Views
3397 views
File license
Fair use/fair dealing exception

Tualatin 1400 @ 1749MHz/1.625v/10.5 x 166.6MHz Bus/166MHz 3,3,3,7 SDRAM/42MHz PCI

Codeblock version

                                        ET6000             ET6100               ARK2000
No graphics booster
Quake VGA 283.6 288.6 267.3
Quake 640x480 83.6 84 (Need UniVBE)

MTRRLFBE
Quake VGA 291.3 297.4 Not compatible with MTRRLFBE without UniVBE
Quake 640x480 84.3 83.6 "

UniVBE + MTRRLFBE
Quake VGA 294.6 297.4 231.2
Quake 640x480 84.3 84.2 87.7
Quake 800x600 43.1 57.7 58.3

Table screenshot

PIII1750 ET Tests.PNG
Filename
PIII1750 ET Tests.PNG
File size
56.08 KiB
Views
3397 views
File license
Fair use/fair dealing exception

Notes

Quake VGA = 320x200 resolution
Repeated benchmarks at 640x480 vary, for example 83.6-84.3 so they can be deemed equal and will be CPU bound. some 3D cards tested higher so must be a Tseng design limitation.
Due to the Apocalypse card having a PCI bridge, the mb defaults to turn on integrated graphics at boot no matter what is set. So this was left on for all tests.
RAMDAC comparisons differ by vendor for the ET6000 and ARK2000
RAMDAC comparisons done on a Sony Trintron CRT with <3000 hrs usage


Performance conclusions

ET6000 vs ET6100 performance is negligible unless the PCI bridge is holding back the ET6100. edit - ET6000 struggles at SVGA, possible due to 2MB althought the ARK does well also with 2MB.
The ET6100 does pull away just a small amount when the CPU is overclocked, checked several times with reboots.
The ET6100 didn't seem to under perform with the PCI bridge like some people were indicating, the Apocalypse 5D sonic might be different?
ARK2000 comes close in non enhanced performance but has some VESA 2.0 and fastvid issues that may or may not be resolved with more time UniVBE fixes this.
This combo of 815E chipset and Tualatin 1400 benefits only a little from DOS graphics boosters and has good PCI performance for these limited tests. (My 440GX system shows much bigger gains from using boosters)


RAMDAC comparisons

Subjectively:

Legend 64 - Not great, serious vertical banding issue visible in Windows, doesn't show so much in DOS gaming but it's there (could be my card). Colours a bit dull especially whites. 60% rating
Jaton Video Magic 128 - Fine, no real issues, quite sharp at low res but not tested at high resolution. Colours look good, a bit less vibrant then the ET6100. 85% rating
Apocalypse 5D - Really nice, very sharp at low res and vibrant colours on par with Matrox at low resolutions. Should be the best at higher resolutions out of the three. 100% rating

I would place both the Tseng RAMDACs above your bog stand NVIDIA RIVA - GF3 partner boards and my ARK card a bit below.


Takeaways

I really wanted the ARK to do better it's still faster then most PCI cards and at 320x200 line doubled the lower dac quality is not that noticeable.
Get the ET6000 out of the three compared for PCI only DOS systems unless you want the rarity and just a slightly better visual quality of the ET6100 at a much higher cost.
Other PCI cards to be tested in the future 😀

Last edited by Hanamichi on 2020-10-04, 21:39. Edited 8 times in total.

Reply 2 of 20, by Hanamichi

User metadata
Rank Member
Rank
Member

RAMDAC quality photos below (hard to capture on a CRT).
Nokia 808 used.
You can (just) make out the vertical lines and a bit of green/grey overtone on the ARK.

ARK2000 WIN

ARK2000 RAMDAC WIN.jpg
Filename
ARK2000 RAMDAC WIN.jpg
File size
968.72 KiB
Views
3422 views
File license
Fair use/fair dealing exception

ARK2000 DOOM

ARK2000 RAMDAC DOOM.jpg
Filename
ARK2000 RAMDAC DOOM.jpg
File size
1.1 MiB
Views
3422 views
File license
Fair use/fair dealing exception

ET6000 WIN

ET6000 RAMDAC WIN.jpg
Filename
ET6000 RAMDAC WIN.jpg
File size
1.31 MiB
Views
3422 views
File license
Fair use/fair dealing exception

ET6000 DOOM

ET6000 RAMDAC DOOM.jpg
Filename
ET6000 RAMDAC DOOM.jpg
File size
1.09 MiB
Views
3422 views
File license
Fair use/fair dealing exception

ET6100 WIN

ET6100 RAMDAC WIN.jpg
Filename
ET6100 RAMDAC WIN.jpg
File size
1.58 MiB
Views
3422 views
File license
Fair use/fair dealing exception
Last edited by Hanamichi on 2020-10-04, 19:57. Edited 2 times in total.

Reply 3 of 20, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

You need SciTech Display Doctor aka Univbe for ARK2000. It won't hurt for Tseng cards too. You may also want to bench Quake with vid_nopageflip "0" or "1" command, which may improve performance. Also vid_wait should be set to 0.

I would place both the Tseng RAMDACs above your bog stand NVIDIA RIVA

Any Riva card has superior DAC - 230 Mhz or more.

Last edited by Stiletto on 2020-10-03, 22:32. Edited 1 time in total.

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

Reply 4 of 20, by Hanamichi

User metadata
Rank Member
Rank
Member

Thanks, I'll try Univbe and add results. I'm not sure what Phil has set as defaults in his benchmarks but I will try them out.

I know the Nvidia DACs have higher MHz/bandwidth for higher resolutions but on some boards the quality is not as good as the Tsengs at low resolutions (subjectively).
This might not be much of an issue when converted to a digital image on an LCD.

I guess it's the quality of the analogue circuit design that makes the difference at least for CRTs when bandwidth is not the issue.
The sure way round that for Nvidia cards is to get a Canopus 😀

Reply 6 of 20, by Hanamichi

User metadata
Rank Member
Rank
Member
Doornkaat wrote on 2020-10-04, 06:25:

Thank yu for your tests. Could I ask you to test wether the Apocalypse 5D will send its 3D signal to an AGP card when using the AGP card as primary video adaptor?

Tested with GF2 AGP.

Seems to work with GLQuake (SGL wrapper) and Ultimate race. The menu in Ultimate race had some weird lag but that might be the usb mouse i was using. The actual game ran like expected.

It would be a good topic to test performance between host 2D cards later 👍

Reply 8 of 20, by Doornkaat

User metadata
Rank l33t
Rank
l33t
Hanamichi wrote on 2020-10-04, 13:35:
Tested with GF2 AGP. […]
Show full quote
Doornkaat wrote on 2020-10-04, 06:25:

Thank yu for your tests. Could I ask you to test wether the Apocalypse 5D will send its 3D signal to an AGP card when using the AGP card as primary video adaptor?

Tested with GF2 AGP.

Seems to work with GLQuake (SGL wrapper) and Ultimate race. The menu in Ultimate race had some weird lag but that might be the usb mouse i was using. The actual game ran like expected.

It would be a good topic to test performance between host 2D cards later 👍

Hey, thanks for trying this for me.😃
I have an Apocalypse 5D somewhere and never thought about this possibility before. I never really used the card because it only has the two soldered MDRAM chips onboard limiting the output to 640*480 if I'm not mistaken. Being able to use another 2D card along with it changes everything.😄
You wouldn't know a source for the MDRAM chips though? I only ever found large lots and I can't spend that kind of money.

Reply 9 of 20, by Hanamichi

User metadata
Rank Member
Rank
Member

No worries, I was curious if it would work.

Apocalypse 5D is a special card, here you go, I'm guessing these modules work alongside 908 ones... I might get some for the ET6000

https://www.ebay.co.uk/itm/MPN-MD909SJ-5-100- … 68/113186689235

Reply 10 of 20, by clueless1

User metadata
Rank l33t
Rank
l33t
Hanamichi wrote on 2020-10-04, 13:36:

Tested Univbe, which allowed me to use graphics boosters on the ARK2000 but on this platform it was slower then without them 🙁

But what is the Quake 640x480 score for the ARK2000? Univbe should allow you to run it, which you couldn't in your original tests.

Thanks for the numbers!

edit: btw I think with the ARK, you need to go into Quake settings and manually pick 640x480. Phil's setup may not be mapping to the correct resolution for all cards in the 640x480 test.

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks

Reply 12 of 20, by clueless1

User metadata
Rank l33t
Rank
l33t
Hanamichi wrote on 2020-10-04, 20:02:

Thanks for the tip on the resolution for the ARK.

Alright tests will be done, looks like the ARK was the winner your comparison, nice. Still hoping it can pull something out the bag!

I've since gotten an ARK1000 PCI card and it's ever so slightly faster than the ARK2000 in many benchmarks. It could just be a difference in BIOS/DRAM timings, but it's there. And I'd like to point out that since I've done all these benchmarks, I've gone back to the onboard Cirrus Logic (the slowest in my tests) for my DOS gaming. While it's fun to see the performance differences, at no point have I ever come across a game where the difference in speed between VGAs has made the difference between a game being smooth, and a game being jerky. Generally, VGA games will ALL be smooth on a 200Mhz Pentium (what I have) no matter which PCI graphics card, and all SVGA games will be jerky no matter which PCI graphics card. 🤣. I'd still love to get an ET6100, ET6000, and ET4000 and re-run all my tests just for fun.

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks

Reply 14 of 20, by Hanamichi

User metadata
Rank Member
Rank
Member
clueless1 wrote on 2020-10-04, 20:17:
Hanamichi wrote on 2020-10-04, 20:02:

Thanks for the tip on the resolution for the ARK.

Alright tests will be done, looks like the ARK was the winner your comparison, nice. Still hoping it can pull something out the bag!

I've since gotten an ARK1000 PCI card and it's ever so slightly faster than the ARK2000 in many benchmarks. It could just be a difference in BIOS/DRAM timings, but it's there. And I'd like to point out that since I've done all these benchmarks, I've gone back to the onboard Cirrus Logic (the slowest in my tests) for my DOS gaming. While it's fun to see the performance differences, at no point have I ever come across a game where the difference in speed between VGAs has made the difference between a game being smooth, and a game being jerky. Generally, VGA games will ALL be smooth on a 200Mhz Pentium (what I have) no matter which PCI graphics card, and all SVGA games will be jerky no matter which PCI graphics card. 🤣. I'd still love to get an ET6100, ET6000, and ET4000 and re-run all my tests just for fun.

Just updated the results in the table screenshots, 640x480, 800x600 OC victories for the ARK with UniVBE and MTRRLFBE!
ET6000 left trailing at the end.

It seems age doesn't matter with these ARKs 😀 maybe I will keep an eye out for a 1000 series or a different brand 2000 series.
I understand completely, I am trying scratch the PC upgrade itch that never ends, except with PCI cards it has too once you have benchmarked them all 🤣
I actually grew up using Quadro Macs and Power Macs where mostly you just care about the Sonnet CPU upgrades until 3D APIs had taken off.

Reply 15 of 20, by Hanamichi

User metadata
Rank Member
Rank
Member
jasa1063 wrote on 2020-10-04, 20:40:

Thanks for doing the testing. This makes me think my next video card acquisition will probably be an ARK2000 PCI as I never owned one in the past.

Welcome! I might add more later on and hope you enjoy the card whichever you pick.

Reply 16 of 20, by Doornkaat

User metadata
Rank l33t
Rank
l33t
Hanamichi wrote on 2020-10-04, 19:03:

No worries, I was curious if it would work.

Apocalypse 5D is a special card, here you go, I'm guessing these modules work alongside 908 ones... I might get some for the ET6000

https://www.ebay.co.uk/itm/MPN-MD909SJ-5-100- … 68/113186689235

Thank you very much for the link but can you confirm those work on the Apocalypse 5D? I have read on multiple occasions the ET6100 needs MD908SJ-5-120. Those are also what's soldered on my Apocalypse 5D.

Reply 17 of 20, by Hanamichi

User metadata
Rank Member
Rank
Member
Doornkaat wrote on 2020-10-04, 21:20:
Hanamichi wrote on 2020-10-04, 19:03:

No worries, I was curious if it would work.

Apocalypse 5D is a special card, here you go, I'm guessing these modules work alongside 908 ones... I might get some for the ET6000

https://www.ebay.co.uk/itm/MPN-MD909SJ-5-100- … 68/113186689235

Thank you very much for the link but can you confirm those work on the Apocalypse 5D? I have read on multiple occasions the ET6100 needs MD908SJ-5-120. Those are also what's soldered on my Apocalypse 5D.

Not sure, some other normal ET6100 cards have the 909s. I would guess they are compatible but maybe someone else here knows for sure.

I found these:
https://www.xfree86.org/3.3.6/tseng12.html
How much memory is it on this ET6000? ( I could be wrong )

Reply 18 of 20, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

909 chips are uncommon occurrence on normal Tseng cards. I think they were mostly used on shorter PCB variations without additional sockets for extra RAM. They are pretty much useless in 4 chips config.

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