386SX wrote on 2024-03-15, 10:47:

Trashbytes wrote on 2024-03-15, 07:47:

Intel branded 740 cards are always solid choices, IIRC hercules and ELSA also made versions as did Creative. […]
Show full quote

386SX wrote on 2024-03-15, 06:04:

Which was the better built and/or fastest board using the i740?

Intel branded 740 cards are always solid choices, IIRC hercules and ELSA also made versions as did Creative.

I have the creative model and its a solid i740 but really they are all pretty much the same.

If you want something a little special you can also look for a Real3D Starfighter 8/12Mb or 16/24Mb i740 PCI, be warned these are expensive models to pursue.

I didn't remember such PCI versions at all. Interesting. This is a video chip/card I wish I had back in those days. I think to remember Intel was pushing quite a lot the whole AGP thing as the next generation feature.

There is a 24Mb PCI Real3D Starfighter up on evilbay right now for an eye watering 2700ish USD, no idea why its at that price as most other Starfighter cards are ~100 - 200USD for AGP. (Its Chinese seller so no idea if its legit or not)

386SX wrote on 2024-03-15, 06:04:

The Serpent Rider wrote on 2024-03-14, 21:06:

For what it's worth, i740 was a decent card. Very good rendering quality and speed, compared to all garbage that was released only one year prior. It also had pretty decent market penetration.

Which was the better built and/or fastest board using the i740?

The Real3D version has special drivers from them (they are the original creators of the chip), and can still use Intel drivers if needed. I remember that the Real3D driver CD is packed with nice goodies too, with even Windows themes. More importantly, they allowed general settings that override the ones in games (maybe there was that with the late Intel drivers too, I can't remember).
To overclock, it's done by swaping the oscillator, no matter the card. No software OC.
I only owned AGP versions of the card.

JustJulião wrote on 2024-03-15, 12:26:

The Real3D version has special drivers from them (they are the original creators of the chip), and can still use Intel drivers i […]
Show full quote

386SX wrote on 2024-03-15, 06:04:

The Serpent Rider wrote on 2024-03-14, 21:06:

For what it's worth, i740 was a decent card. Very good rendering quality and speed, compared to all garbage that was released only one year prior. It also had pretty decent market penetration.

Which was the better built and/or fastest board using the i740?

The Real3D version has special drivers from them (they are the original creators of the chip), and can still use Intel drivers if needed. I remember that the Real3D driver CD is packed with nice goodies too, with even Windows themes. More importantly, they allowed general settings that override the ones in games (maybe there was that with the late Intel drivers too, I can't remember).
To overclock, it's done by swaping the oscillator, no matter the card. No software OC.
I only owned AGP versions of the card.

Impressive! I suppose one chip is a PCI to AGP bridge and the other the early i740. How much less did it perform?

386SX wrote on 2024-03-15, 12:42:

JustJulião wrote on 2024-03-15, 12:26:

The Real3D version has special drivers from them (they are the original creators of the chip), and can still use Intel drivers i […]
Show full quote

386SX wrote on 2024-03-15, 06:04:

Which was the better built and/or fastest board using the i740?

The Real3D version has special drivers from them (they are the original creators of the chip), and can still use Intel drivers if needed. I remember that the Real3D driver CD is packed with nice goodies too, with even Windows themes. More importantly, they allowed general settings that override the ones in games (maybe there was that with the late Intel drivers too, I can't remember).
To overclock, it's done by swaping the oscillator, no matter the card. No software OC.
I only owned AGP versions of the card.

Impressive! I suppose one chip is a PCI to AGP bridge and the other the early i740. How much less did it perform?

About 15% slower for the 12mb version (compared to the 8mb AGP).
The AGP version is definitely more interesting.
Paired with with a fast FSB and fast RAM, it was doing noticeably better. I wouldn't bother with the PCI version.

At the begining of this video you can see some of the goodies I got from the Real3D CD (wallpaper, cursor pack) even if it's not very relevant.
This is a system I built around the Starfighter on which I used the wallpaper and the cursor pack, but this particular video was during a testing session of the PowerVR PCX2. I temporarily swaped the Starfighter for an ATi card because it gives better results with the PowerVR PCX2.
https://www.youtube.com/watch?v=RoMuxZ3EmCI

JustJulião wrote on 2024-03-15, 12:57:

About 15% slower for the 12mb version (compared to the 8mb AGP). The AGP version is definitely more interesting. Paired with wi […]
Show full quote

386SX wrote on 2024-03-15, 12:42:

JustJulião wrote on 2024-03-15, 12:26:

The Real3D version has special drivers from them (they are the original creators of the chip), and can still use Intel drivers if needed. I remember that the Real3D driver CD is packed with nice goodies too, with even Windows themes. More importantly, they allowed general settings that override the ones in games (maybe there was that with the late Intel drivers too, I can't remember).
To overclock, it's done by swaping the oscillator, no matter the card. No software OC.
I only owned AGP versions of the card.

Impressive! I suppose one chip is a PCI to AGP bridge and the other the early i740. How much less did it perform?

About 15% slower for the 12mb version (compared to the 8mb AGP).
The AGP version is definitely more interesting.
Paired with with a fast FSB and fast RAM, it was doing noticeably better. I wouldn't bother with the PCI version.

At the begining of this video you can see some of the goodies I got from the Real3D CD (wallpaper, cursor pack) even if it's not very relevant.
This is a system I built around the Starfighter, but this particular video was during a testing session of the PowerVR PCX2. I temporarily swaped the Starfighter for an ATi card because it gives better results with the PowerVR PCX2.
https://www.youtube.com/watch?v=RoMuxZ3EmCI

Very interesting, I'll look for the AGP one, too bad the later version of this chip were not released (I suppose some prototype boards might be around somewhere). About the chip revisions/clocks, any major differences?

386SX wrote on 2024-03-16, 10:32:

JustJulião wrote on 2024-03-15, 12:57:

About 15% slower for the 12mb version (compared to the 8mb AGP). The AGP version is definitely more interesting. Paired with wi […]
Show full quote

386SX wrote on 2024-03-15, 12:42:

Impressive! I suppose one chip is a PCI to AGP bridge and the other the early i740. How much less did it perform?

About 15% slower for the 12mb version (compared to the 8mb AGP).
The AGP version is definitely more interesting.
Paired with with a fast FSB and fast RAM, it was doing noticeably better. I wouldn't bother with the PCI version.

At the begining of this video you can see some of the goodies I got from the Real3D CD (wallpaper, cursor pack) even if it's not very relevant.
This is a system I built around the Starfighter, but this particular video was during a testing session of the PowerVR PCX2. I temporarily swaped the Starfighter for an ATi card because it gives better results with the PowerVR PCX2.
https://www.youtube.com/watch?v=RoMuxZ3EmCI

Very interesting, I'll look for the AGP one, too bad the later version of this chip were not released (I suppose some prototype boards might be around somewhere). About the chip revisions/clocks, any major differences?

I don't know about the different revisions of these cards.
There are some i752 cards AFAIK, but it was integrated in i810 and some i815 chipset variants. It was, again, pretty good for what it was, even if it has its usual crowd of haters because better dedicated PCI and AGP cards existed.

JustJulião wrote on 2024-03-16, 11:17:

386SX wrote on 2024-03-16, 10:32:

JustJulião wrote on 2024-03-15, 12:57:

About 15% slower for the 12mb version (compared to the 8mb AGP). The AGP version is definitely more interesting. Paired with wi […]
Show full quote

About 15% slower for the 12mb version (compared to the 8mb AGP).
The AGP version is definitely more interesting.
Paired with with a fast FSB and fast RAM, it was doing noticeably better. I wouldn't bother with the PCI version.

At the begining of this video you can see some of the goodies I got from the Real3D CD (wallpaper, cursor pack) even if it's not very relevant.
This is a system I built around the Starfighter, but this particular video was during a testing session of the PowerVR PCX2. I temporarily swaped the Starfighter for an ATi card because it gives better results with the PowerVR PCX2.
https://www.youtube.com/watch?v=RoMuxZ3EmCI

Very interesting, I'll look for the AGP one, too bad the later version of this chip were not released (I suppose some prototype boards might be around somewhere). About the chip revisions/clocks, any major differences?

I don't know about the different revisions of these cards.
There are some i752 cards AFAIK, but it was integrated in i810 and some i815 chipset variants. It was, again, pretty good for what it was, even if it has its usual crowd of haters because better dedicated PCI and AGP cards existed.

Did any real AGP boards even existed with the i752 in prototypes? EDIT: I've read now that some existed for reviews, it'd be great to find those boards.

JustJulião wrote on 2024-03-16, 11:17:

There are some i752 cards AFAIK, but it was integrated in i810 and some i815 chipset variants. It was, again, pretty good for what it was, even if it has its usual crowd of haters because better dedicated PCI and AGP cards existed.

Has anybody seen/done comparison tests between i740 and integrated i810? Iv seen rumors stating i752 didnt offer any performance gains, is the same true for i810?

Well don't trust the 3Dmark scores because the i810 has a driver cheat which will slideshow the game tests and then claim it got 25fps or so.

rasz_pl wrote on 2024-03-16, 12:49:

JustJulião wrote on 2024-03-16, 11:17:

There are some i752 cards AFAIK, but it was integrated in i810 and some i815 chipset variants. It was, again, pretty good for what it was, even if it has its usual crowd of haters because better dedicated PCI and AGP cards existed.

Has anybody seen/done comparison tests between i740 and integrated i810? Iv seen rumors stating i752 didnt offer any performance gains, is the same true for i810?

I've only found the vintage3d tests about it and seems interesting.

What about one of the Number Nine cards? I remember the "Ticket to Ride 3D Revolution" was meant to be quite bad.

While I am deciding which of these cards to buy (they are all expensive as usual these days, I am going to buy one today), yesterday I was testing the 3D hw rendering of the Mystique 220 4MB SGRAM and it's interesting that beside no textures seems rendered in more modern benchmarks/games and lack many needed features, it's actually rendering in hw and fast beside the speed coming from the broken graphic I suppose.
Then again installed the Rage Pro Turbo AGP and always impressed that ran so well even in Quake2 and partially even Quake3. Ok that I am testing these with a 500Mhz P-III to push these to their speed limits but still surprise me every times. It's almost "too good" for the level of cards I am looking for.

Finally, you are getting more specific. Yes, Mystiques were fast. To get slower, aim for Laguna3D. If it should be more broken, try Trident 3DImage.

386SX wrote on 2024-03-17, 06:35:

Then again installed the Rage Pro Turbo AGP and always impressed that ran so well even in Quake2 and partially even Quake3. Ok that I am testing these with a 500Mhz P-III to push these to their speed limits but still surprise me every times. It's almost "too good" for the level of cards I am looking for.

But the Rage Pro is unusual because it received driver updates for so many years. Did any of its 1997 competitors get significant driver updates until 2002?

When I had a Rage Pro in 1997/8 it was next to useless for games of the time. There was no OpenGL at all, the D3D performance was comparatively poor, and the CIF API was useless. There's a contemporary review of it here: https://www.tomshardware.com/reviews/3d-accel … step,51-26.html.

But the chip remained on the market way into the 2000's, and I suspect driver updates made use of massively more powerful CPU's as time went on. I recently went back to my original Rage Pro in a Pentium III system with the new drivers and was amazed at how much it could do, and how well it was doing it. Games that I had to play in software mode at the time had become playable and even enjoyable using the Rage Pro.
So my suspicion is that the Rage Pro is in effect 'cheating' with its reputation here, compared to the cards it was originally competing with, due to the sheer length of time it was on the market.

Putas wrote on 2024-03-17, 11:49:

Finally, you are getting more specific. Yes, Mystiques were fast. To get slower, aim for Laguna3D. If it should be more broken, try Trident 3DImage.

I think I'll buy them "all" if I can, I really like this generation of chips that feel like they were almost going to make it through those difficult rendering times before the much better but boring later cards. The Mystique really seems a fast engine maybe also thanks the far too fast cpu and lack of many rendering features. Final Reality miss many of that check list points.

Interesting that I'm also testing at the same time a generic Virge/DX 4MB (late 1999 built, only smd components) end up much better in graphic compatibility while much slower of course with so much to render. Powerstrip read the Virge DX rams @ 40Mhz (?) and after what may be 60Mhz begin to have problems. Actually running 3DMark on this board still seems like incredible. I was also trying to make Thief game running with some textures but as seen in reviews, no textures but really fast.

Spark wrote on 2024-03-17, 12:03:

But the Rage Pro is unusual because it received driver updates for so many years. Did any of its 1997 competitors get significan […]
Show full quote

386SX wrote on 2024-03-17, 06:35:

Then again installed the Rage Pro Turbo AGP and always impressed that ran so well even in Quake2 and partially even Quake3. Ok that I am testing these with a 500Mhz P-III to push these to their speed limits but still surprise me every times. It's almost "too good" for the level of cards I am looking for.

But the Rage Pro is unusual because it received driver updates for so many years. Did any of its 1997 competitors get significant driver updates until 2002?

When I had a Rage Pro in 1997/8 it was next to useless for games of the time. There was no OpenGL at all, the D3D performance was comparatively poor, and the CIF API was useless. There's a contemporary review of it here: https://www.tomshardware.com/reviews/3d-accel … step,51-26.html.

But the chip remained on the market way into the 2000's, and I suspect driver updates made use of massively more powerful CPU's as time went on. I recently went back to my original Rage Pro in a Pentium III system with the new drivers and was amazed at how much it could do, and how well it was doing it. Games that I had to play in software mode at the time had become playable and even enjoyable using the Rage Pro.
So my suspicion is that the Rage Pro is in effect 'cheating' with its reputation here, compared to the cards it was originally competing with, due to the sheer length of time it was on the market.

ATi really did a good drivers job later on those old cores to be at least compatible with those far too heavy games. Last time I checked (I'll try it soon) even the weak Rage IIC end up being an interesting slow chip but very compatible. The Rage Pro I was testing run Unreal Tournament @ 640x480 quite well, beside the textures seems to be low res until you look directly to them and they are like reloaded in higher details. I remember the Voodoo Graphic running this game with Glide API and being already impressed by it, honestly the Rage Pro seems to have done all it was possible, even if at the end.
With all their missing features, other chips probably only needed much faster and better drivers development to get to acceptable levels. Even Trident actually seems did a nice job for its chips software support (Blade3D but also from reviews the 3Dìmage serie drivers).

386SX wrote on 2024-03-17, 13:17:

Interesting that I'm also testing at the same time a generic Virge/DX 4MB (late 1999 built, only smd components) end up much better in graphic compatibility while much slower of course with so much to render. Powerstrip read the Virge DX rams @ 40Mhz (?) and after what may be 60Mhz begin to have problems.

I have a generic Virge DX which has a stock clock of 50 MHz, but can be overclocked to 70 MHz without producing any artifacts. The chip gets very hot at that frequency though, so I added a small heatsink using thermal adhesive, which helped a bit. Ideally, you also want some kind of active cooling blowing fresh air over the chip (a good 120mm case fan should work) to make it even more stable.

The best Virge DX cards from reputable manufacturers (e.g. Diamond) had a 72 MHz stock clock (or higher) and thus performed much better out of the box.

Joseph_Joestar wrote on 2024-03-17, 13:48:

386SX wrote on 2024-03-17, 13:17:

Interesting that I'm also testing at the same time a generic Virge/DX 4MB (late 1999 built, only smd components) end up much better in graphic compatibility while much slower of course with so much to render. Powerstrip read the Virge DX rams @ 40Mhz (?) and after what may be 60Mhz begin to have problems.

I have a generic Virge DX which has a stock clock of 50 MHz, but can be overclocked to 70 MHz without producing any artifacts. The chip gets very hot at that frequency though, so I added a small heatsink using thermal adhesive, which helped a bit. Ideally, you also want some kind of active cooling blowing fresh air over the chip (a good 120mm case fan should work) to make it even more stable.

The best Virge DX cards from reputable manufacturers (e.g. Diamond) had a 72 MHz stock clock (or higher) and thus performed much better out of the box.

I wonder why Powerstrip seems to read ram clock (40Mhz) but not the chip clock at all and of course you can't increase that. Need to check memory timings as well.

386SX wrote on 2024-03-17, 13:52:

I wonder why Powerstrip seems to read ram clock (40Mhz) but not the chip clock at all and of course you can't increase that. Need to check memory timings as well.

I always used MCLK.EXE for overclocking my Virge DX. It will also correctly read the stock clocks.

That's a DOS utility though, and I'm not sure if it can be used in a Win9x environment.

Joseph_Joestar wrote on 2024-03-17, 13:54:

386SX wrote on 2024-03-17, 13:52:

I wonder why Powerstrip seems to read ram clock (40Mhz) but not the chip clock at all and of course you can't increase that. Need to check memory timings as well.

I always used MCLK.EXE for overclocking my Virge DX. It will also correctly read the stock clocks.

That's a DOS utility though, and I'm not sure if it can be used in a Win9x environment.

I'll try that. Anyway DRAMs are 35ns. The card is called VC963C-3D, 86C375 chip built in 23 week of 1998.