I wasn't implying that Rage Pro was a competitor with Voodoo II/3, rather that I switched to 3dfx, because ATi let me down.
The funny thing is that despite being launched in March 1997, I really don't remember the Rage Pro being available when I built a new system in August 1997.
It could be that it was available for AGP, but not PCI. (my system had no AGP slots). Too bad I tossed my collection of 90s Canadian newspaper ads, maybe I could have figured out the reason.

Meatball wrote on 2022-08-27, 15:23:

What are you using to overclock the Rage Pros?

I've used Powerstrip, version 3 something (probably last one), downloaded from page last year I think it is no longer updated.
HWinfo reports incorrectly clocks, (200 instead 75) but correctly memory.
But I was using it under Windows XP. With generic Windows XP drivers installed. (didn't use Ati one).

I had same problem as you, but only when I got "out of range" of what seems BIOS of the card allowing. 62-98 core works for me normally. I didnt try to push it further up or down. Memories works only between 68-116 range. If I try less or more, it randomly underclocks it or overclocks it. It's the only way, how I've reached 60 mhz memory value, but card doesn't work on this frequency. Z-buffer doesn't work, card is often crashing in 3D, with some kind of error. 800x600 resolution works, but 640x480 not. The least value, that is working for memory is 68 mhz.

My card is Ati Xpert@Work, SGR 8 MB with code "40200" (SGR memories has 8ns)
http://vgamuseum.info/index.php/technologies/ … e-pro-turbo-agp

Anonymous Coward wrote on 2022-08-27, 15:50:

The situation was somehow similiar here too (in 1997). In review of gaming magazines of first 3d accelerators in july 1997, (i've bought that magazine in that time), there is mentioned only only Virge , Monster 3D, some Fire proffesional card, Matrox Mystique and Ati Xpertvision or something like it with Xpert, but it's clearly Rage II chip. Rage Pro is there not mentioned. Maybe, it was released only in USA in march 1997? But in europe, it was shipped later? Not sure.

Anyway, I would say, after some drivers tweaking and after driver matured at beginning 1998, and Ati released Turbo version with new drivers, and 8MB version, it was better pick, than voodoo1. Driver support and compatibility was quite good even during 1999, which is suprise for me, as Ati driver sucked. It handled 32bit, and with 8MB memory, it could go often up to 800x600x32. In most of the games, it surpassed voodoo1, but many benchmarks are old (with original Rage Pro and old 1997 drivers... there voodoo1 seems better). But with latest drivers from june 1998 and later, it actually outperformed voodoo1, in direct3d for sure, and in opengl gradually situation became better (1997-early 1998 experience in opengl with rage pro had to be much worse). So card didn't get attention and value, it deserves. Definetely is on the Riva 128 level, with better image quality, better compatibility in DOS games (thanks to Mach64 2D core), CD/DVD/mpeg support from rageII+DVD version, and if someone picked TV out version, it had even nice bonus. With all-in-wonder pro card, even got TVtuner (but those cards was already out of "budget" league of course, as they were more expensive". When I've started doing little personal research around this card and playing with it , I am pleasantly suprised, because original "look" from few review pages is often underwhelming. I have feeling, it was all about Riva 128, voodoo1, Matrox Mystique.

Btw, I pinpoint a price check from march 1998, just for fun, it's from our country, so it's in czech crowns, but the "ratios", how much is something more expensive from other, can be seen here. about 4000 CZK should be around $100 in that time (meaning after import, so tax is paid, etc).. prices are there without DPH, which is +22% in that time for end user. Prices are from smaller markets, so as you find them in real shops, not in quantities of 100 or 1000. (note: average salary was around 10000sk/czk , so it was super super expensive for us, do not count it as 4000czk is 100$, like it's nothing. It was almost half monthly salary here in 1998)
Rage Pro is definately there at least in march 1998

W.x: it was interesting to see those prices, thankyou for sharing it. no wonder the virge dx and rage2 harvested the market at that time. In retrospect, i would have went with the rage2 instead of the virge tho (better compatibility, better scaling with faster cpu, and a little bit more tolerance for higher resolutions).

About the drivers: the drivers sucked both for the riva128 and voodoo1 on arrival as well. i remember getting the riva128 and voodoo1 with the early original driver discs. from 10 games, 8 will bsod, 2 will crash and so on. even nvidia and 3dfx required a few months to actually make their drivers remotely usable. in contrast, the rage2, virge, and virge dx came with more usable drivers on arrival (of course ati and s3 were already big names, and had a lot of experience with writing drivers, compared to the newcomers).

Ok, so new findings with rage pro. I've spent last 2 days playing with and testing Rage Pro Turbos.

The core starts to be unstable past 79mhz. Sometimes, it handles 81 or 82, but on both rage pro turbos, I get strange mirroring vertical distortion. In every game I tested, GLquake, GLhexen2, MDK, and Tomb Raider 2. Memory works up to 115 though. Cannot get past 116 mhz on memories, seems VGA BIOS resets its to 60 mhz. (so basically halves it).
Not sure, why it is so, if it is driver related, or can be somehow prevented, but if not, Rage Pro is then pretty much not overclockable on core. You can get like 5 mhz above 75mhz. (one of my Rage Pro Turbo is already unstable on 79mhz thought).
But memories at 100mhz are quite enough, so you won't get much performance by overclocking it to 115 (I often not get even 5% of performance in real game scenarios)

Even more interesting is fact, that my 4MB Rage Pro Turbo is much slower than 8 MB Rage Pro Turbo.
I was testing it like 2 hours, but it's still there. The differance is remarkable. In GLQuake 640x480x16, 4 MB Rage Pro Turbo gets 16.7 fps. But 8MB Rage Pro Turbo gets 21.8 fps. What's going on there? Both versions are same (SGR memory, same Ati-109- code , just one is Ati Xpert@Work (without TV out) and one is Ati Xpert@Play.

In GLHexen2 , differance between two were even more remarkable. Like 10.5 fps vs 19 fps. (almost double for 8MB version).
But 4MB memory should be enough for 640x480x16.
I've tried to lower resolution to 400x300x16, to eliminate memory full scenario, but it does also here.

Interestingly enoigh, when I've raise up resolution to 800x600x16 for GLhexen2, Rage Pro Turbo 8 MB started to have same abbysmal FPS as Rage Pro Turbo 4 MB on lower resolutions. 10.5 fps. I'm starting to thinking, that it's because it's using AGP texturing memory (system memory). Because it's same fast, which points up to memory limitation problem. But that would mean, that 4 MB Rage Pro cannot handle even 400x300x16 which is very weird. Because without memory limitation, both Rage Pro (Xpert@play and Xpert@work) should be same fast (in various benchmarks in historical magazines, they have same values... even all-in-wonder pro has same values). But how it is possible, than 400x300x16 already has lowered FPS for 4MB version. Because it should handle GLquake (one of the oldest accelerated games , and least demanding on memory... 4 MB should be enough).
The only explanation should be, that 4 MB is not enough, because it needs also memory for 2D part of VGA (drawing the screen). And from 4MB , too little stays for textures etc... so it needs to use AGP aperture memory.

I will test it on another 4 MB Rage Pro (original version, without Turbo, that came out in 1997). Unfortunately, those had only SDR memories, not SGR memories.
All cards had 75 core and 100 memory, so it's not because different clocks.

It seems similar to V1 or i810. Decent framerates only on 512x384 which looks very good on 15" TFT with 1024x768 native resolution.

As I am always in for multiboot to get most out of the system I keep always checking for native drivers for OS/2, Dos/Win3x, Win9x, Linux, Amithlon.
The Rage Pro has openGL support in Win9x...perhaps also working in (old) Linux, we will see how this turns out. (I always use Sarge 3.1)

And i took the ram expansion card from the rage pro, to implant it on the permedia2. It seems the rage pro will dislike this more than i thought.

Geri wrote on 2022-08-31, 10:39:

And i took the ram expansion card from the rage pro, to implant it on the permedia2. It seems the rage pro will dislike this more than i thought.

I am doing intensive testing right now, I will publish results later, when finished. I've included also Rage XL, and Rage LT Pro for comparsion, so 6 different "Rage Pro architecture" cards. As a little sneak peak, already tested SDR 8MB version of all-in-wonder pro and SGR 8MB Ati@Xpert , so SDR and SGR shows no differance at same clock speeds (75 for core and 100 for memory), 100% identical results. So it seems, that SGR brings no advantage for Ati Rage Pro from first sight. But again, I tested only 2 games, and 2 different resolutions. Maybe, there can be some scenario, eighter in game, or 3d display settings, that shows some difference, in advantage to SGR memory. Otherwise, it really doesn't matter, which version one have (before , I was focusing on getting SGR version, thinking, it will be maybe better somehow).
Also tested AGP 1x and AGP2x mode, also no differeance in FPS. Even AGP 1x handles it's performance without any bottleneck, so AGP 2x on Rage Pro architecture was only marketing gimmick.

Also tested AGP 1x and AGP2x mode, also no differeance in FPS. Even AGP 1x handles it's performance without any bottleneck, so AGP 2x on Rage Pro architecture was only marketing gimmick.

I've always thought that was the case and it's nice to have some hard evidence. Historical reviews always showed a noticable gap between AGP and PCI variants but I REALLY doubt its down to the bus bottlenecking but I've never found a satisfying explanation.

DrAnthony wrote on 2022-09-01, 00:45:

I've always thought that was the case and it's nice to have some hard evidence. Historical reviews always showed a noticable gap between AGP and PCI variants but I REALLY doubt its down to the bus bottlenecking but I've never found a satisfying explanation.

It heavily depends on what you are using the card in. For example, an anemic 166 mhz cyrix will not show too much difference between agp and pci, as it barely can even feed the pci bus with data. Once you have a stronger p2/p3 class cpu (including amd k6/2, or a very fast cyrix 6x86mx/m2) , the difference between agp and pci starts to open, up to like 20% typically. Of course there are no speed increase with some games, like, if the game doesn't uses the bus too much: for example, it uses immediate mode, but the poly count is too low.

Geri wrote on 2022-09-01, 10:30:

the difference between agp and pci starts to open, up to like 20% typically. Of course there are no speed increase with some games, like, if the game doesn't uses the bus too much: for example, it uses immediate mode, but the poly count is too low.

This.... I also have opinion, that there is realistic scenario, where AGP 2x mode can bring some little more fps over AGP 1x.
Anyway, I tested it with fast AXP 1600+ platform with DDR memory, where it could provide best chance for such scenario to occur, as everything through FSB and memory is opened up, but still, that slow Rage Pro cannot simply render so fast, that AGP 1x bandwidth would be used to full potential (to gain little more FPS , its enough, when it occurs like in 1/10 times over 1 second period... it would bring some increase in few frames), but it didn't occur in tested games.

Another scenario, where it could help is, when video card's memory is full, and it needs to use system memory for textures, which goes through AGP. In this case, when memory is very fast, it could teoreticaly help, that AGP 2x mode is up. AGP 1x should provide 266 MB/sec bandwidth, and definately, DDR memories are faster than this.
I've tried to test it on GLHexen2 800x600x32 bit, what should be enough for 4MB Rage Pro so it needs additional memory, but didn't experienced any difference in FPS betwen AGP 1x and AGP 2x mode. (set through motherboard BIOS, but verified both in HWinfo and Powerstrip, that it's actually really on AGP x2@x1 speed).
It's probably because core is so slow, that it doesn't need so much memory bandwidth to use anyway.

Anyway, I still hope, that there is some configuration or game, that takes advantage of AGP 2x over 1x, to gain at least 3% more performance. I still believe, that it can occur. 😀

The IT industry tends to prepare for potential bottlenecks before they happen. For example, if they see if a standard is not going to be enough within a few years, they move to avoid the problem, before the problems starting to add up. Thats why we already got AGP4x before we hit a wall with agp 2x. Most early agp 4x moherboards actually running in agp2x mode (i8xx chipset ftw), but no one is noticing it, so it doesn't really matters.

This also applies to pci-e, now we have gen5 or something already, but if someone compares the gaming performance with pci-e 2.0, 3.0, 4.0 and 5.0 modes, there will be no difference beyond 1-2% in mmost cases (people already did it https://www.youtube.com/watch?v=SvBovtT4Vf4). Only the first pci-e protocol is going to be somewhat troublesome mostly regarding graphics and ssd storage performance. Of course compatibility issues can occur in unfortunate cases. (my main pc still have pci-e 1.0 only, luckily everything booted up in it so far).

Once i had a P3 with buggy chipset (agp 2x) and wanted to use a GeForce2 in it (newer cards refused to boot up in it). The GeForce2 mx was crashing when agp 2x was enabled, if i limited it to agp1x, i lost about 10-15% performance in ut2003. So approx this gen could be the point when agp 1x vs 2x made significant difference.

MadMac_5 wrote on 2022-08-06, 21:44:

I had an ATI Rage Pro Turbo as my first 3D accelerator. It was on a PII-300 with 64 MB of RAM in 1998, and it was simultaneously a revelation and a source of pain. The 2D quality was spectacular, and a lot of 3D games looked... well, good enough. But it never could do alpha blending properly, no matter how many driver revisions happened (and I tried them ALL)

I also had an ATI Rage Pro AIW, and yes, alpha blending was the biggest weakness of the chip. Except that it worked quite ok, I could play even Quake 3. Alpha blending was fixed at some point, probably in the last available driver, it wasn't perfect, but it was much better. But with that driver I could not use the TV Tuner that was installed on the card 🙁

As for RageIIC, I didn't own one back in the days, only much later (in my collection) and it's good only for 2D, 3D is horrible.

I don't even consider the IIc a 3D accelerator 🤣

When considered on the time those video cards still existed on low end PCB brands, yes they were mostly 2D video cards for the lowest end market (I have different cards with 2000 date on them), but from a previous generation point of view it wasn't that bad with latest drivers. As a Virge/Trio3D serie alternative while never gaming oriented that late, it would have been (like many other products before and later) a "good" 3D accelerator in the 1996/97 times if driver were optimized before.
It's indeed a 3D accelerator considering the CPU of those times, such effects would have been impossible in a software rendering path.

Geri wrote on 2022-08-27, 23:39:

W.x: it was interesting to see those prices, thankyou for sharing it. no wonder the virge dx and rage2 harvested the market at that time. In retrospect, i would have went with the rage2 instead of the virge tho (better compatibility, better scaling with faster cpu, and a little bit more tolerance for higher resolutions).

Found another one, from october 1997, it is from different magazine, Computer. I was reading it, I have it physically, and didn't expect such cool price listing. So I've copied it from pdf. (it's without tax, which was around 22% at the time, so you need to add +22% to the price, if you are regular customer)

My opinion of the Rage Pro?

Well.
It was my first video card I put in a computer I built, back in 1998. I knew there were much better cards, but money was tight and I preferred to spend it on a good motherboard, CPU and memory (Abit BH6, C300A, 64MB Mushkin PC100).

So what happened next? Well I "liked" the card so much that my next ATI/AMD card was a Radeon HD7970. 😀 Was firmly a Nvidia guy in-between.

(they have redeemed themselves in the meantime IMHO, and my current one is a RX6900).

Recently, in my p2, I've been using a Rage Pro Turbo (agp) for 2D, and a Voodoo2 for 3d/glide.

As others have said, this Rage has pretty good 2d but leaves a lot to be desired otherwise.

W.x. wrote on 2022-08-30, 22:30:

The differance is remarkable. In GLQuake 640x480x16, 4 MB Rage Pro Turbo gets 16.7 fps. But 8MB Rage Pro Turbo gets 21.8 fps. What's going on there? Both versions are same (SGR memory, same Ati-109- code , just one is Ati Xpert@Work (without TV out) and one is Ati Xpert@Play.

With Rage3 (8MB) cards it is possbile to get better fps, more like 30-40fps with GLQuake @ 640x480x16 timedemo demo1 without overlock, use right OpenGL, best at the moment is 1094 or 1068 version for Quake1 and Quake2 1068 or 1076/1077 as i member and i recommend using Rage XL AGP pn 109-66700-01 or 109-66700-00 cards.

Rage3 OpenGL Win9x files for download:
https://www.upload.ee/files/14683268/Rage3-GL_Win9x.zip.html

If anybody has some other Win9x MiniGL/OpenGL versions for Rage3, let me know , Thank you!
ATi RagePro OpenGL files

MiniGL 4.10.1010, 4.10.1021, 4.10.1036

OpenGL 4.11.1051, 4.11.1058, 4.11.1060, 4.11.1065, 4.11.1068, 4.11.1076, 4.11.1077, 4.11.1084, 4.11.1085, 4.11.1091, 4.12.1091, 4.11.1094, 4.12.1094, 4.12.1095, 4.12.1096, 4.12.1103, 4.12.1104, 4.12.1106, 4.12.1107