VOGONS


First post, by Smack2k

User metadata
Rank Oldbie
Rank
Oldbie

Hey all

I have a bunch of random PCI and early (1x / 2x) AGP Video cards (with a few 4x to) from various companies (ATI / S3 / Nvidia / others) that I'd like to benchmark just for the hell of it to see where they fall against each other. I've never gotten into benchmarking in the past so I wanted to know if there were any tips on setting up the system to do the testing? I have boards, procs, memory, etc around but didn't just want to throw anything together without getting some advice. Do I want a proc that is faster / newer than the video cards? Anything in particular about memory / HDD / motherboard? For the tests what are the recommended benchmark apps to test these mid 90s - early 2k cards? Looking to learn a lot more about the ins and outs of the cards along the way as well.

Appreciate the advice as I'd like to get this setup and start testing!

Reply 1 of 14, by cj_reha

User metadata
Rank Oldbie
Rank
Oldbie

I actually was doing the same thing with one of my Pentium 4 boards. It's the AOPEN AX45-533 U2 with the SiS 645DX chipset, and a "true" AGP Universal slot. It's accepted every old AGP card I've thrown at it even back to really early S3 Virge/ATi Rage 4 MB AGP 1x cards. I even got it to boot with a 1 MB PCI Trident 🤣

A lot of early cards had problems with 3dMark 99 under XP, weird framerates and artifacting, but I expected it. I was just doing it for fun 😜

Join the Retro PC Discord! - https://discord.gg/UKAFchB
My YouTube Channel - https://www.youtube.com/channel/UCDJYB_ZDsIzXGZz6J0txgCA

Reply 3 of 14, by Aideka

User metadata
Rank Member
Rank
Member

The faster the processor, the more it brings out the differences in the cards. I would use a fast Pentium 4 or Athlon system, with universal AGP slot. Sure, it may not be periord correct, but it is a good way to see the real difference between cards.

8zszli-6.png

Reply 4 of 14, by Smack2k

User metadata
Rank Oldbie
Rank
Oldbie

Perfect. That is what I wanted to know. Going to use a fast P4 as those differences are exactly what I want to see.

In terms of apps, outside 3dMark versions any other software you'd recommend to get some good results / see the differences in the mid - late 90s PCI and AGP cards?

Reply 5 of 14, by Ampera

User metadata
Rank Oldbie
Rank
Oldbie

Half-Life at high resolutions, as it was a high profile game.
If you can get a REALLY fast CPU, run a test of Doom, could be interesting to see how 2D performance has held up with these cards.
Stress 32-bit colours if possible. This was a thing that some cards couldn't even do in the late 90's, but extra points for good 32bpp colour performance.

Reply 6 of 14, by gandhig

User metadata
Rank Member
Rank
Member

Final Reality.

Dosbox SVN r4019 + savestates Build (Alpha)
1st thread & the only one related to the forum(?)...warning about modern-retro combo
Dead, but, Personal Favourite
Replacement for Candy Crush...Train the Brain

Reply 7 of 14, by Smack2k

User metadata
Rank Oldbie
Rank
Oldbie
Ampera wrote:

Half-Life at high resolutions, as it was a high profile game.
If you can get a REALLY fast CPU, run a test of Doom, could be interesting to see how 2D performance has held up with these cards.
Stress 32-bit colours if possible. This was a thing that some cards couldn't even do in the late 90's, but extra points for good 32bpp colour performance.

What would you consider a "really fast" CPU?

I am looking at a mid 2 GHZ Athlon XP or a 3.2 GZ P4 (If I can find the right heatsink for it) for the proc...

OS Wise, XP OK to run as the OS?

Reply 8 of 14, by clueless1

User metadata
Rank l33t
Rank
l33t
Smack2k wrote:

Perfect. That is what I wanted to know. Going to use a fast P4 as those differences are exactly what I want to see.

Just keep in mind that those results won't necessarily translate to a period correct system. So if you're ever building a system down the road, you still want to benchmark the cards you're deciding between in that system.

For example: On the fast system, card A is 30% faster than card B. But in a period correct system, their performance is nearly identical. Without that knowledge, you may pick card A for the period correct system. With that knowledge, you might decide to use card B in the period correct system and save card A for a faster system.

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 9 of 14, by Smack2k

User metadata
Rank Oldbie
Rank
Oldbie

My goal really is to benchmark a lot of lesser used / not as popular cards, PCI versions of more known AGP Cards, and other cards that you'd really only want to use if you didn't have a better option or weren't looking to spend more on the premium card but still get some decent gameplay with what you have. Seen so many benchmarks of more popular known high end cards but not as much for the "little guys". Just a side project I have been thinking to do for a while and now that I am out of room to put anymore retro machines in my retro cave, this was my next venture....so wanting to get the best overall testbed to use for mid 90s - early 2ks (01-02) cards of this nature for both 2D and 3D stats...

My big question would be what would be best apps to use? Would DOS benchmarking be better or Windows? This where my benchmarking noon comes out as I am not sure

For 2D benchmarking I can only currently find XP or newer apps and nothing older on Windows sids

Reply 10 of 14, by clueless1

User metadata
Rank l33t
Rank
l33t

DOS benchmarks for DOS era hardware; Windows benchmarks for Win era hardware. DOS benchmarks tend to reveal hardware-level performance while Win benchmarks can vary wildly by driver version. No drivers in DOS.

2D benchmarks for Win9x:
Ziff-Davis Winbench
Tom's 2D (tomshardware made a 2D windows benchmark back in the day)
PCMark (https://www.futuremark.com/benchmarks/legacy)
ATTO and Diskspeed (https://www.philscomputerlab.com/hdd-benchmarks.html)

Phil's DOS benchmark pack (https://www.philscomputerlab.com/dos-benchmark-pack.html)

EDIT: just found this post with a couple of nice links:
VOGONS Benchmark archive?

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 14, by Deksor

User metadata
Rank l33t
Rank
l33t

You can still increase DOS benchmark results on AGP cards to some degree with Slot 1 boards and Super socket 7 IIRC

Trying to identify old hardware ? Visit The retro web - Project's thread The Retro Web project - a stason.org/TH99 alternative

Reply 13 of 14, by kva

User metadata
Rank Newbie
Rank
Newbie

Just remember to do at least three runs of every benchmark on every card and present the average value.

Celeron Tualatin vs Celeron Conroe at equal clocks
Pentium Pro 256k vs Pentium Pro 1M vs Pentium II Overdrive!
VIA C3 vs VIA C7

My website all about old hardware

Reply 14 of 14, by Smack2k

User metadata
Rank Oldbie
Rank
Oldbie
Deksor wrote:

You can still increase DOS benchmark results on AGP cards to some degree with Slot 1 boards and Super socket 7 IIRC

Going to be running these all from one board as my base....with a P4 processor. Just trying to figure out what the best speed proc to use as I have a few different ones.

Once that is done, I can look into doing other boards and focus in on certain types. Just want to get a base down for all the cards from one source first.