VOGONS

Search

Common searches

Voodoo 5 5500 PCI, expected performance?

Topic actions

First post, by paradigital

Posted on 
User metadata
Rank Oldbie
Rank
Oldbie

So, I purchased a "for parts" Voodoo 5 5500 PCI from eBay around 3 weeks ago, it arrived today (well done eBay Global Shipping Program, good job). It was known to be missing at least one capacitor and have another knocked for six. After taking a good look around the card it was apparent that there were 3 missing cylindical caps (C41, C59, C85), as well as the two SMD caps near the Molex connector (C74, C89). There were also a couple of resistors "knocked" half off (one side still attached) around the aforementioned C85. There was also no cooling at all (no heatsinks or fans, I assume harvested for parts to fix another card).

So, replaced the missing components, and added some 40x40x16mm pure copper heatsinks, along with brand new 40mm fans. It boots, it installs drivers and it appears to run a-ok, however performance seems a bit... lacklustre.

I've never owned a Voodoo 5 5500 before so don't have a frame of reference for expected performance, but paired with my Abit BH6, 100fsb Slot-1 PIII 850 (running at 112*8.5, 952MHz), and running Windows 98 SE, I seem to be limited to bang on 70fps (give or take a few decimal points) when testing with Q3A 1.30's demo "four".

I'm running the Evolution Drivers for V4/5 (v1.00.09) from the FalconFly 3dfx archive.

Enabling or disabling single chip mode under antialiasing seems to make no difference to the performance, it's 70fps at 640x480x16 with both chips or a single chip, it's 70fps at 800x600x16 with either both or a single chip., 1024x768x16 drops a bit to 61.4fps in single-chip mode and 69.2fps with both chips active. This suggests to me that both VSA-100 chips are working (temperature of both using the highly-accurate "finger on heatsink" or "finger on rear of chip" methods appear consistent with each other when in dual-chip mode), and that I'm CPU bound? Do I need to pair this card with a faster PC (I've got options, I just had this machine to hand as I'd been testing other video cards in it, like my newly acquired GeForce 256 DDR)? I've looked at period benchmarks for the V5 5500 (albeit AGP), and an Athlon 750 seems to beat my framerates by a noticable 5-10%.

Anything I should try?

Reply 1 of 8, by paradigital

Posted on 
User metadata
Rank Oldbie
Rank
Oldbie

So I'm now pretty convinced something else other than CPU bottleneck is locking me to 70fps. 3DMark2000 is also hitting only 70fps on the helicopter tests, which seems far too coincidental.

Reply 2 of 8, by PC-Engineer

Posted on 
User metadata
Rank Member
Rank
Member

Sounds that V-sync is active. Look for an option in the driver to disable it.

And it seems that both chips are working correct.

Epox 7KXA Slot A / Athlon 950MHz / Voodoo 5 5500 / PowerVR / 512 MB / AWE32 / SCSI - Windows 98SE

Reply 4 of 8, by auron

Posted on 
User metadata
Rank Oldbie
Rank
Oldbie

is 70hz your monitor refresh rate, coindicentally? i don't know how these 3rd party drivers work, but i recall that on the official drivers, you had to download a seperate utility from 3dfx to toggle vsync for OGL/D3D, respectively, and rather confusingly checking the boxes would actually turn on vsync.

also i believe V5 AGP can be 5-10% faster in period titles and a lot faster than PCI in newer titles that massively texture thrash (V5 has effectively just 32 megabyte of texture memory). also since you are using newer 3rd party drivers here and p3 and athlon can perform differently depending on program, your comparison wouldn't be apples to apples anyway.

Reply 5 of 8, by paradigital

Posted on 
User metadata
Rank Oldbie
Rank
Oldbie

Ironically, no, the monitor is an old 4:3 60hz tft panel.

3dfx Tools has the option to disable vsync, both in the Advanced tab (one setting for each api, d3d and ogl/glide), and also as tick boxes in the Overclocking tab. Both options for each API are set to disable vsync.

I’ve gone back and removed all video drivers and started again with the latest official 3dfx drivers, and now I can break 70fps, but not my much. 3DMark2000 can max out the FPS slider (99 fps) on the low detail helicopter test, and Quake 3 now gets to 76fps depending on resolution (colour depth makes no difference)

Quake 3 Arena 32-bit (32-bit):

1024x768 73.9fps
800x600 76.3fps
640x480 76.3fps

As this is no longer a round figure, I’m thinking it must be CPU bottlenecked.

I’ll try it with either my Athlon XP 2800+ machine or P4 2.8GHz Shuttle tomorrow.

Reply 6 of 8, by cyclone3d

Posted on 
User metadata
Rank l33t++
Rank
l33t++

Overclocking the PCI bus should help as well. You can also try adjusting the PCI latency in your BIOS.

I did some testing on PCI vs AGP with Voodoo 3 2000 cards quite a while ago using the same motherboard.

For DOS games that were not 3D accelerated, the AGP card was way faster than the PCI card. I was only testing out the DOS 2D speeds and didn't actually run any 3D accelerated games.

Yamaha modified setupds and drivers
Yamaha XG repository
YMF7x4 Guide
Aopen AW744L II SB-LINK

Reply 7 of 8, by The Serpent Rider

Posted on 
User metadata
Rank l33t++
Rank
l33t++

For DOS games that were not 3D accelerated, the AGP card was way faster than the PCI card

Switching to AGP will give about 25% to 33% increase in VESA modes (the only thing that matters) and that's with much faster CPU. And no, Doom timedemo result doesn't matter.

As this is no longer a round figure, I’m thinking it must be CPU bottlenecked.

No, that's both PCI and FSB bottlenecks combined.

Last edited by The Serpent Rider on 2020-07-27, 18:53. Edited 1 time in total.

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

Reply 8 of 8, by paradigital

Posted on 
User metadata
Rank Oldbie
Rank
Oldbie

So the PCI bus is currently at 37MHz due to the 112MHz FSB. I’ve can’t get up to 133fsb for 44MHz PCI as the P3-850 just won’t do it unfortunately.

I’ll play with the PCI Latency settings 😀

Go to top of page Go to top of page
Back to Video
Close window
Close window