VOGONS


Reply 40 of 58, by bloodem

User metadata
Rank Oldbie
Rank
Oldbie

This is not a video card issue, not when he gets 17 FPS in Quake 2 Software renderer.

4 x Socket 3 / 4 x Socket 7 / 6 x Super Socket 7 / 5 x Slot 1 / 3 x Slot A / 5 x Socket 370
5 x Socket A / 1 x Socket 478 / 2 x Socket 754 / 3 x Socket 939 / 4 x LGA775 / 1 x LGA1155
Current rig: AM4 - Ryzen 5 3600X
Backup rig: LGA1151 - Core i7 7700k

Reply 41 of 58, by RockstarRunner

User metadata
Rank Newbie
Rank
Newbie

With the barebones system (no other cards than V3 and all non essential mobo stuff turned off) and 128MB, cpu running at 550mhz, I got it to hit 19fps in Q2 640x480 software.
So, it's a small improvement over the original... but highest 3dmark i've hit is 2490... still about 1000 points shy of where it should be.

Reply 42 of 58, by bloodem

User metadata
Rank Oldbie
Rank
Oldbie

Can you also test a normal K6-2 CPU and see if you notice any improvement?

4 x Socket 3 / 4 x Socket 7 / 6 x Super Socket 7 / 5 x Slot 1 / 3 x Slot A / 5 x Socket 370
5 x Socket A / 1 x Socket 478 / 2 x Socket 754 / 3 x Socket 939 / 4 x LGA775 / 1 x LGA1155
Current rig: AM4 - Ryzen 5 3600X
Backup rig: LGA1151 - Core i7 7700k

Reply 44 of 58, by texterted

User metadata
Rank Member
Rank
Member

If it's not playing the games you want to play, then it's time for a faster system!

Otherwise, play the games and not the benchies. 😁

Cheers

Ted

98se/W2K :- Asus A8v Dlx. A-64 3500+, 512 mb ddr, Radeon 9800 Pro, SB Live.
XP Pro:- Asus P5 Q SE Plus, C2D E8400, 4 Gig DDR2, Radeon HD4870, SB Audigy 2ZS.

Reply 45 of 58, by RockstarRunner

User metadata
Rank Newbie
Rank
Newbie
texterted wrote on 2020-10-08, 12:31:

If it's not playing the games you want to play, then it's time for a faster system!

Otherwise, play the games and not the benchies. 😁

What's wrong with wanting to solve a bottleneck? It's not about the benchmarks, they are just highlighting that there is a problem.

Reply 46 of 58, by bloodem

User metadata
Rank Oldbie
Rank
Oldbie

As promised, I did a few benchmarks with one of my SS7 boards.

SPECS:
MB: DFI P5BV3+(VIA MVP3 chipset)
CPU: AMD K6-2+ 550 MHz (no overclock)
VGA: 3dfx Voodoo 3 3000 AGP
Sound: Yamaha YMF 719E-S

3DMark99Max RESULTS:
- with 128 MB RAM (external cache enabled): 3279 points
- with 256 MB RAM (external cache enabled): 3084 points
- with 256 MB RAM (external cache disabled): 3043 points

* Note: I did not disable vsync for Direct3D. Since 3DMark99 does not disable vsync when running the benchmark, on a SS7 board scores will always be a bit lower with vsync ON (on newer platforms scores will be MUCH lower without disabling vsync)

Quake 2 SOFTWARE RENDERING 640 X 480 RESULTS:
- with 128 MB RAM (external cache enabled): 20.8 FPS
- with 256 MB RAM (external cache enabled): 19.9 FPS
- with 256 MB RAM (external cache disabled): 19.2 FPS

<thinking to myself>Hmm... that's odd, I clearly remember testing this board just a few months ago with a GeForce 2 MX and scoring 23 - 24 FPS in Quake 2 Software Mode. Hmm... could it be that GeForce has faster 2D speed compared to Voodoo 3?</thinking to myself>

Quake 2 SOFTWARE RENDERING 640 X 480 RESULTS (with a GeForce 2 MX):
- with 128 MB RAM (external cache enabled): 23.6 FPS
- with 256 MB RAM (external cache enabled): 23.1 FPS
- with 256 MB RAM (external cache disabled): 22.4 FPS

Yep, bingo!

*Note: Sound was enabled for the Quake 2 Software Rendering benchmarks, on High Quality / best performance.

So it seems that your Quake 2 Software Rendering scores were not as bad as I thought, but it's still unclear why your 3DMark99 scores are so low.
Still, as I said before, in a worst case scenario (no external cache & > 128 MB RAM), scores on a K6-2/3+ should be ~ 5 - 10% lower.

4 x Socket 3 / 4 x Socket 7 / 6 x Super Socket 7 / 5 x Slot 1 / 3 x Slot A / 5 x Socket 370
5 x Socket A / 1 x Socket 478 / 2 x Socket 754 / 3 x Socket 939 / 4 x LGA775 / 1 x LGA1155
Current rig: AM4 - Ryzen 5 3600X
Backup rig: LGA1151 - Core i7 7700k

Reply 47 of 58, by RockstarRunner

User metadata
Rank Newbie
Rank
Newbie

Thanks for doing all that, Bloodem, really appreciate the effort! I guess, at the end of the day, it's the motherboard having troubles with the cpu, meaning one of them will have to be replaced. I really wanted to use the + CPU, for the PowerNow features, so I'll have to try and find another mother board that will work well, that doesn't require messing around with. Preferably a nice small 10x10 Baby AT like the ga-5aa.

That aside, still trying a few more things, on the hope there is a solution to this boards issues.

Reply 48 of 58, by bloodem

User metadata
Rank Oldbie
Rank
Oldbie

You're welcome. Unfortunately I'm also out of ideas.
My best guess right now is that your board (or specific board revision) has issues with the K6-2+ CPUs. That being said, as I said before, I don't remember anyone complaining about bad performance on Gigabyte 5AA boards when running a K6-2+ CPU. So it's a real mystery.
Anyway, this is the reason why I love SS7 boards - each one behaves differently, having its own quirks and specific workarounds. So I kinda envy you, I wish I had this board so that I could debug the crap out of it 😀

4 x Socket 3 / 4 x Socket 7 / 6 x Super Socket 7 / 5 x Slot 1 / 3 x Slot A / 5 x Socket 370
5 x Socket A / 1 x Socket 478 / 2 x Socket 754 / 3 x Socket 939 / 4 x LGA775 / 1 x LGA1155
Current rig: AM4 - Ryzen 5 3600X
Backup rig: LGA1151 - Core i7 7700k

Reply 49 of 58, by Bancho

User metadata
Rank Member
Rank
Member

I've just finished putting my existing SS7 setup on a new board, so can give you something to compare too.

Iwill AX100 Plus Ali-V Motherboard
AMD K6-III+ 450 @ 550 (5x110)
128mb 133mhz SD-RAM
V3 3000 AGP
Ali AGP 1.54
Final Voodoo 3 Drivers

3Dmark 99 Score

3666 3DMarks
8233 CPU 3DMarks

I've Used CTU to set the enhancements for the CPU. My score does look quite a bit higher, I don't know if that's anything to do with the CPU being a K63+

Reply 50 of 58, by bloodem

User metadata
Rank Oldbie
Rank
Oldbie

In your case, the biggest boost most likely comes from the 110MHz FSB. 😀

4 x Socket 3 / 4 x Socket 7 / 6 x Super Socket 7 / 5 x Slot 1 / 3 x Slot A / 5 x Socket 370
5 x Socket A / 1 x Socket 478 / 2 x Socket 754 / 3 x Socket 939 / 4 x LGA775 / 1 x LGA1155
Current rig: AM4 - Ryzen 5 3600X
Backup rig: LGA1151 - Core i7 7700k

Reply 51 of 58, by RockstarRunner

User metadata
Rank Newbie
Rank
Newbie

Just to keep this up to date: I tried updating the v3 bios to 2.15.12, and tried setting Ali ago settings directly in the registry, with some settings I got that I thought would be correct. Score was just over 2500, so still not correct.
I actually managed to acquire a Asus P5A-B rev 1.4, so soon I'll be able to make a comparison between these boards with the same CPU/ram/gfx.

Reply 52 of 58, by RockstarRunner

User metadata
Rank Newbie
Rank
Newbie

Well then 😁
With nothing but a straight motherboard swap, from the Gigabyte GA-5AA to the Asus P5A-B, I'm now getting 3400 3dmarks! Same installation too, so all same drivers and settings (though the motherboard chips drivers did reinstall on first boot).
Does this not confirm that the Gigabyte board, although very stable, has issues with the k6+ and performance ?

Q2 Software timedemo 640x480
@550mhz 22.4fps
@600mhz 23.3fps

3dmark 99 @ 600mhz
3543 3dmarks
7926 cpu 3dmarks

Aida32 @550mhz
Read: 334 MB/s
Write: 125 MB/s

Attachments

  • IMG_20201027_202131.jpg
    Filename
    IMG_20201027_202131.jpg
    File size
    995.67 KiB
    Views
    343 views
    File license
    Fair use/fair dealing exception

Reply 53 of 58, by bloodem

User metadata
Rank Oldbie
Rank
Oldbie

Yeah, this was expected. It was clearly related to the motherboard or to its BIOS 😀
However, I'm not sure if it confirms that the Gigabyte board has issues with K6-2/3+ CPUs or not. In general, there are many things that work against you on SS7 platforms 😀

4 x Socket 3 / 4 x Socket 7 / 6 x Super Socket 7 / 5 x Slot 1 / 3 x Slot A / 5 x Socket 370
5 x Socket A / 1 x Socket 478 / 2 x Socket 754 / 3 x Socket 939 / 4 x LGA775 / 1 x LGA1155
Current rig: AM4 - Ryzen 5 3600X
Backup rig: LGA1151 - Core i7 7700k

Reply 54 of 58, by RockstarRunner

User metadata
Rank Newbie
Rank
Newbie

Bloodem, you were also correct about the USB issue. I enabled USB, to see if it would have an impact, and at 600mhz, 3dmarks dropped to 3000 & CPU 3dmarks to 7000, so that's quite a big hit. Needless to say, I disabled it afterwards, as I don't really need it anyway.

Reply 55 of 58, by bloodem

User metadata
Rank Oldbie
Rank
Oldbie

Yeah, I don't use USB on these old platforms either.
I find it much easier to quickly remove the HDD, plug it in my universal USB 3.0 rack and transfer everything I need. 😀

4 x Socket 3 / 4 x Socket 7 / 6 x Super Socket 7 / 5 x Slot 1 / 3 x Slot A / 5 x Socket 370
5 x Socket A / 1 x Socket 478 / 2 x Socket 754 / 3 x Socket 939 / 4 x LGA775 / 1 x LGA1155
Current rig: AM4 - Ryzen 5 3600X
Backup rig: LGA1151 - Core i7 7700k

Reply 56 of 58, by Tetrium

User metadata
Rank l33t++
Rank
l33t++
RockstarRunner wrote on 2020-11-03, 12:24:

Bloodem, you were also correct about the USB issue. I enabled USB, to see if it would have an impact, and at 600mhz, 3dmarks dropped to 3000 & CPU 3dmarks to 7000, so that's quite a big hit. Needless to say, I disabled it afterwards, as I don't really need it anyway.

I wonder if WinME would have the same performance drop from enabling USB btw (I presume you're using 98SE as mentioned in the OP).

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 58 of 58, by framebuffer

User metadata
Rank Newbie
Rank
Newbie

I've seen many bugs and inexplicable behaviors affecting 3DMark99 scores over the years.
Although is nice to run it for nostalgic reasons, I never trust it's scores and would highly discourage using it as only tool to test the 3Dperformance of a system.

For voodoo3 cards I'd use Q2/Q3/UT timedemo and eventually 3Dmark 2000

https://framebuffer.io