VOGONS


Voodoo 1 vs. Voodoo 2 on a 486

Topic actions

Reply 40 of 124, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Sounds like there isn't a way to measure frame rate in the Screamers, so no real way to add results to this thread.

If the Voodoo2 turns out to be slower at the same clock as the Voodoo1, that would be interesting, though I doubt the Voodoo1 can be clocked up to Voodoo2 speeds. I'll try to downclock the Voodoo2 and see what I get. I recall seeing in the OC software that the Voodoo2 can go down to 50 MHz.

mrau, sounds like you are interested in results using a moderately paced 486 - perhaps a DX4-100? I can run some results using an Am5x86-160 and an Intel DX4-100. How do you predict the percent difference between the Voodoo1 and Voodoo2 will change with these CPUs?

mrau, in the other thread I linked, the 27.x fps results were done at 800x600, if I recall correctly. There are probably more results in my World's fastest 486 link. I remember asking someone to compare the Voodoo2 vs Voodoo2 SLI in GLQuake using his AMD 5x86-160 system and the results were identical.

Plan your life wisely, you'll be dead before you know it.

Reply 41 of 124, by mrau

User metadata
Rank Oldbie
Rank
Oldbie
feipoa wrote:

Sounds like there isn't a way to measure frame rate in the Screamers, so no real way to add results to this thread.

bumma

feipoa wrote:

If the Voodoo2 turns out to be slower at the same clock as the Voodoo1, that would be interesting, though I doubt the Voodoo1 can be clocked up to Voodoo2 speeds. I'll try to downclock the Voodoo2 and see what I get. I recall seeing in the OC software that the Voodoo2 can go down to 50 MHz.

if downclocking alone cannot help then maybe the two chips can meet in the middle?

feipoa wrote:

mrau, sounds like you are interested in results using a moderately paced 486 - perhaps a DX4-100? I can run some results using an Am5x86-160 and an Intel DX4-100. How do you predict the percent difference between the Voodoo1 and Voodoo2 will change with these CPUs?

dx4100 sounds fine - percent difference should be small - the more powerful cpu will scale max 5% better than the slower one i presume on standard bus speed; with OC this might turn out to increase more but no more than 2-3% again;

feipoa wrote:

mrau, in the other thread I linked, the 27.x fps results were done at 800x600, if I recall correctly. There are probably more results in my World's fastest 486 link. I remember asking someone to compare the Voodoo2 vs Voodoo2 SLI in GLQuake using his AMD 5x86-160 system and the results were identical.

i remember playing this on a cyrix 100mhz in software mode - slideshow 😁

Reply 43 of 124, by mrau

User metadata
Rank Oldbie
Rank
Oldbie

i guess your first proposal was best - dx4.100 vx 5x86.160 for example; im not sure about cyrix - i believe areound 133 mhz they reach 5x86.160 performance; but then again a 100 mhz chip could be compared directly to the dx4.100 which is nice too 😀

Reply 45 of 124, by feipoa

User metadata
Rank l33t++
Rank
l33t++

My original plan was to run an updated ultimate 486 benchmark comparison utilising these 3D games and other software mode games, taking the results up to 133 MHz for any CPU (Pentium, RiSE, Winchip, etc). I didn't want to get too carried away with CPU combinations just yet. What I had in mind with this thread was to get an idea for how much faster a Voodoo2 might be on a fast socket 3 machine. Considering the results, I'm not sure if I should still use a Voodoo1 on such a comparison (because it is more period correct) or to use the higher performing Voodoo2. What do you think?

To more directly answer mwdmeyer's request, more CPUs will come along much later, but more immediately, I will try to get DX4-100 and Am5x86-160 results.

Plan your life wisely, you'll be dead before you know it.

Reply 47 of 124, by meljor

User metadata
Rank Oldbie
Rank
Oldbie

''period correct'' doesn't fit in any case here imho as both these cards really need pentium cpu's or higher to really shine. I think the voodoo1 came out with the p1-133 ? I think it really comes to life with 200mmx or even higher.

If the goal is to just get the absolute best out of a 486 based platform then simply use the voodoo2. To really play glide games build at least a Pentium(2) or something like that (but you know that).

asus tx97-e, 233mmx, voodoo1, s3 virge ,sb16
asus p5a, k6-3+ @ 550mhz, voodoo2 12mb sli, gf2 gts, awe32
asus p3b-f, p3-700, voodoo3 3500TV agp, awe64
asus tusl2-c, p3-S 1,4ghz, voodoo5 5500, live!
asus a7n8x DL, barton cpu, 6800ultra, Voodoo3 pci, audigy1

Reply 48 of 124, by Scali

User metadata
Rank l33t
Rank
l33t
meljor wrote:

''period correct'' doesn't fit in any case here imho as both these cards really need pentium cpu's or higher to really shine. I think the voodoo1 came out with the p1-133 ? I think it really comes to life with 200mmx or even higher.

Yea, one could argue that if you could afford a VooDoo, you could definitely afford a Pentium system as well (back in those days I had a P133, but could not afford a VooDoo. The first 3D accelerator I had in that machine was a Matrox Mystique. Later an Apocalypse 3Dx). So a fast 486 with a VooDoo wouldn't have been a common configuration.
It's a possible configuration, but you could argue the same for both the V1 and V2.

http://scalibq.wordpress.com/just-keeping-it- … ro-programming/

Reply 49 of 124, by leileilol

User metadata
Rank l33t++
Rank
l33t++
meljor wrote:

''period correct'' doesn't fit in any case here imho as both these cards really need pentium cpu's or higher to really shine.

Not true! A 6x86 and K5/K6 are okay too. 😀 Also i've known some who hung onto their iDX4/AM5x86s until 1999 and yes this is america and the "Try VoodooGet Voodoo." marketing game was still going strong.

it's certainly more period correct than the usual "retro" PIIs with 512mb PC133 ram, cf flash, and a Geforce from the XP era.

"You need pentium for real true voodoo experience!" arguments should be left out of the thread. It's not like it's a FPU-demanding 3d card like any PowerVR something. I can't think of any lighter and more spec appropriate 3d card where it'd be beneficial.

apsosig.png
long live PCem

Reply 50 of 124, by feipoa

User metadata
Rank l33t++
Rank
l33t++

There will always be someone who demands period correctness, someone who demands it to fit what they consider practical, and those who don't one way or the other. Oh so many opinions. In the end, we all will do what interests us.

I bought a 486 new in Jan 1997. It was, without a doubt, a budget system, but to me, it was fast. I used it for 2 years solid. During these two years, I upgraded the RAM and graphics. I bought a Mystique 220, thought I could easily have bought a Voodoo1, or even a Voodoo2. In my mind, I consider both of these options as "period good enough" (albiet a waste of horsepower). In this thread, I'm really just wanting to see how a Voodoo1 and Voodoo2 perform on the socket 3 platform, not caring about reemerging debates on what constitutes a 486 and what doesn't, what is practical, what isn't, etc. It is intended to be a thread of curiosity, not so much a system build to aspire towards.

If someone else wants to re-run the comparison using a Pentium MMX or PII, I encourage them to do so. It would be interesting to see how the percent difference between V1 and V2 changes and what the slope is like. It might also be interesting to see if the differences goes to 0 with a sufficiently slow CPU.

Plan your life wisely, you'll be dead before you know it.

Reply 51 of 124, by swaaye

User metadata
Rank l33t++
Rank
l33t++

I actually did something like this myself maybe 10 years ago. I had a MSI 4144 motherboard and Am5x86, Cx5x86 and a POD. The POD really boosts what you can get out of a 486 board with a Voodoo. Especially the POD that I have which seems stable at 100 MHz (40x2.5).

Unfortunately for the POD, compatibility is a problem, and it caused the motherboard cache to not work. It showed enabled but didn't register in any cache benchmarks. This really hurts the POD in other applications. It could be felt just operating Windows 95.

Reply 52 of 124, by dirkmirk

User metadata
Rank Oldbie
Rank
Oldbie

I've noticed in your testing feiopa that you don't test the Cyrix 5x86-120?

Is an important CPU because its one of the easiest options for a fast 486 system, buy a IBM 5x86C an appropriate motherboard and away you go, I agree the AM5x86-160 is also a good cpu for the same reason.

Whilst its interesting to see what a 66mhz/133mhz cyrix can do its not really a viable option for most people.

Reply 53 of 124, by feipoa

User metadata
Rank l33t++
Rank
l33t++
swaaye wrote:

I actually did something like this myself maybe 10 years ago. I had a MSI 4144 motherboard and Am5x86, Cx5x86 and a POD. The POD really boosts what you can get out of a 486 board with a Voodoo. Especially the POD that I have which seems stable at 100 MHz (40x2.5).

Unfortunately for the POD, compatibility is a problem, and it caused the motherboard cache to not work. It showed enabled but didn't register in any cache benchmarks. This really hurts the POD in other applications. It could be felt just operating Windows 95.

Indeed, for 3D games, the POD shines bright with accelerated graphics, but I too noticed it a bit sluggish in Windows GUI in comparison to the Cyrix 120 or AMD 160. I find the POD acceptable, though, in boards which properly use the L1 cache in write-back mode and when the chip is run at 100 MHz.

dirkmirk wrote:

I've noticed in your testing feiopa that you don't test the Cyrix 5x86-120?

Is an important CPU because its one of the easiest options for a fast 486 system, buy a IBM 5x86C an appropriate motherboard and away you go, I agree the AM5x86-160 is also a good cpu for the same reason.

Whilst its interesting to see what a 66mhz/133mhz cyrix can do its not really a viable option for most people.

Yes, apparently I'm not finished! Seems that people are mostly interested in absolute frame rates rather than a generalised Voodoo2 vs. Voodoo1 run. The Am5x86-160 and Cx5x86-120 aren't going to be much different than what was presented. The Cx5x86-120 and Am5x86-160 have the advantage of a 40 MHz FSB and no cache/RAM wait states. The 66 MHz FSB is really just a means to get a Cyrix running at 133 MHz.

I don't mind running extra tests. Looks like we have DX4-100, Cx5x86-120, and Am5x86-160 on the list. I wouldn't be surprised if someone asks for a POD100. I have had some things I'm attending to now, so it will probably be a week before I can get to work on further tests.

Plan your life wisely, you'll be dead before you know it.

Reply 54 of 124, by dirkmirk

User metadata
Rank Oldbie
Rank
Oldbie

Don't mean to be a pain, personally I don't see the point of the Intel dx4-100 for these tests as you would go straight to a am5x86-33/160.

IMO the mainstream "performance" CPUs are:

AM5x86-133 @ 160mhz
IBM5x86C-100 @ 120mhz
POD 83mhz stock

Overclocking past that is generally a specialist area.

I guess I was more interested in the actual peformance difference between the cyrix at 120/133mhz, very greatful for your work you've done over the years but I feel like it was a big omission to not extensively test the cyrix 5x86-120 vs the 133 version in your major 486 benchmark project, to determine whether the $1,0000-$1500 133 version is much better than the $50 IBM chip.

Edit: don't worry about testing the 120, I agree that for the purpose of this testing it's quite pointless as your simply showing the difference between the voodoo cards.

If you ever wanted to revisit the extensive 486 benchmark thats more my interest being an optimised 5x86-120.

Edit: Just re-read through your ultimate 486 benchmark, I thought I read somewhere that you didn't set the register enhancements for the 120mhz, I must've imagined something.

Reply 55 of 124, by feipoa

User metadata
Rank l33t++
Rank
l33t++

I completely understand the reasoning for wanting the Cyrix 5x86-120 and Am5x86-160. Both are easily obtainable. There will always be someone who is not happy with the results though, e.g. complains that I tested the Cyrix with LSSER and wants the setting reversed because their motherboard doesn't support LSSER. Stuff like that.

Keep in mind that I do intend to remake the ultimate 486 benchmark thread with by using more games, e.g. early 3D and whatever else I have in my folder waiting for testing. I would like to use the MB-8433UUD instead of the crippled M919 due to its PCI bus issue at 40 MHz. Currently, the 3D data for 40 MHz CPUs is deflated becuase the PCI bus is run at 27 MHz. I also intend to take all non-socket3 chips up to 133 MHz so that the results bleed more into next generation data.

Do you own a S1R3 Cyrix 5x86? This chip will work with branch prediction in Windows.
So it might be interesting to include the Cyrix 5x86-120 with branch prediction and the Am5x86-160 for this Voodoo comparison thread.

I also treat the POD83 as almost mainstream. Seems that several people have ones which run at 100 MHz stock. And of the 2 chips I have modified to run at 4 V, they both work at 100 MHz.

Plan your life wisely, you'll be dead before you know it.

Reply 56 of 124, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Here is a little bit of a teaser for the ultimate "486" benchmark version 2.

NexGen Nx586 PF110 running GLQuake in Windows 95c. Notice how the GLQuake results are the same as the IBM 5x86c-133 (28.5 fps)? It is only running at 102 MHz, though I somehow was hoping for better results. There are two screen shots because one is at 640x480, the other at 800x600. Results were the same, minus 0.1 fps.

I may also have my hands on the venerable Am5x86 which runs at 200 MHz for version 2 of the ultimate comparison.

NexGen_Setup.jpg
Filename
NexGen_Setup.jpg
File size
502.26 KiB
Views
1618 views
File license
Fair use/fair dealing exception
NexGen_POST.jpg
Filename
NexGen_POST.jpg
File size
105.18 KiB
Views
1618 views
File license
Fair use/fair dealing exception
Desktop_NexGen.JPG
Filename
Desktop_NexGen.JPG
File size
202.31 KiB
Views
1589 views
File license
Fair use/fair dealing exception
NexGen_PF110_GLQuake_640x480.jpg
Filename
NexGen_PF110_GLQuake_640x480.jpg
File size
407.68 KiB
Views
1618 views
File license
Fair use/fair dealing exception
NexGen_PF110_GLQuake_800x600.jpg
Filename
NexGen_PF110_GLQuake_800x600.jpg
File size
415.38 KiB
Views
1618 views
File license
Fair use/fair dealing exception
Last edited by feipoa on 2018-08-18, 18:14. Edited 2 times in total.

Plan your life wisely, you'll be dead before you know it.

Reply 58 of 124, by willow

User metadata
Rank Member
Rank
Member
feipoa wrote:
meljor wrote:

Hold on now! WHERE ARE THE SCREAMER RESULTS?? 🤣

Good job and thanks, very appreciated! Pretty good fps in some games i must say, especially for a 486 system.

I was reading on Wiki that Screamer is a DOS-based software mode game. Is there a 3dfx patch for it and does the game have a means to run either a timedemo or instantaneous frame rates?

Do Screamer 2 or Screamer Rally provide a means to run in 3dfx glide mode? And is there a means to measure frame rate via a counter or timedemo?

Screamer 1 don't have 3dfx version, only software and s3 virge accelerated version with bilinear filtering
Screamer 2 and rally have 3dfx version.

Reply 59 of 124, by feipoa

User metadata
Rank l33t++
Rank
l33t++

I'm about 1/3 of the way finished with these extra CPUs. I'm testing Am5x86-160, Cyrix 5x86-120 S0R5, Cyrix 5x86-120 S1R3 (for BTB), Cyrix 5x86-133/4x, and Intel DX4. For the individual who wanted the Intel DX4, do you prefer it tested at 100 MHz or 120 MHz?

Plan your life wisely, you'll be dead before you know it.