keropi wrote:

Since I got some M3Ds I started testing them... tombraider/DOS was a nice surprise once Single Buffering was activated. Is there […]
Show full quote

Since I got some M3Ds I started testing them... tombraider/DOS was a nice surprise once Single Buffering was activated. Is there a list of "true" powervr accelerated games? Don't care about D3D ones... I searched but I can't get a definitive list 😐

Also I don't know if it's of interest, I have both games that Matrox gave with M3D cards: MDK Mission Laguna Beach and Hexen II Continent of Blackmarsh. Anyone looking for either?

edit:
well, I started uploading the images either way... 🤣

a. MDK Laguna Beach image (CCD, RAR5 compression)
<copyrighted software links removed by moderator>

b. Hexen II Blackmarsh (CCD, RAR5 compression)
<copyrighted software links removed by moderator>

c. Driver - patches - Ultimate Race (CCD, RAR5 compression)
<copyrighted software links removed by moderator>

all done! 😊

Sorry for bumping this old thread, but I was looking for Disk 3 to test my M3D and when I download these .rar files, the disk image that is extracted is unreadable by 7zip, cannot be mounted as a virtual CD and gives errors toward the end of the verification process after burning to an actual CD (I used ImgBurn). I believe there may be something wrong with the 5th part of the rar archive.

After burning to a CD I have 560MB of files on a disc. I can copy all of it off, but the image file is 768MB so something appears to be missing from the disc after burning.

Any possibility of getting it reuploaded, or even a single file download?

Thank you!

Yeah, lot's of cool stuff ! thx!
what CPU should I use for my M3D ? (currently quipped with PII-333)

a P4 or a P3.

PowerVR's not that great with slow non-intel CPUs.

It seems to run well with an Athlon T-Bird.

Anyone know anything about the CD issue I mentioned earlier with Disk 3? Extracted disc image is 768MB and cannot be opened with 7zip or mounted virtually, CD gives a verification error about 75% of the way through after burning, disk works but contains only 560MB of data. Is there something wrong with the disk or is this just some random quirk?

Well, I was able to get Ultimate Race running from disk 3. It works great! In the right game this little card is amazing! It has so few components and its so tiny... its really no wonder that this tech eventually made its way into the GPUs in the most popular mobile devices on the planet.

I don't have a direct comparison to a Glide game from this time period, but I don't remember games on my Voodoo Graphics cards from 1997 performing or looking as good.

Ozzuneoj wrote:

Anyone know anything about the CD issue I mentioned earlier with Disk 3? Extracted disc image is 768MB and cannot be opened with 7zip or mounted virtually, CD gives a verification error about 75% of the way through after burning, disk works but contains only 560MB of data. Is there something wrong with the disk or is this just some random quirk?

I suggest you start using other pieces of software , I downloaded the files from mediafire and nothing seems wrong

WinRAR and CloneCD were both used to create the files and now check them, Daemon Tools was used to mount the image

That's interesting...a PIII/P4/T-Bird have lot's of horsepower compared to a P133/P166MMX that was the target for the voodoo1 as the competitor.
I read that because of the "pci-bus-transfer" the main gpu-should be fast too and the onboard-gpu's cause/have "performance-problems" with the m3d.

dr.zeissler wrote:

That's interesting...a PIII/P4/T-Bird have lot's of horsepower compared to a P133/P166MMX that was the target for the voodoo1 as the competitor.
I read that because of the "pci-bus-transfer" the main gpu-should be fast too and the onboard-gpu's cause/have "performance-problems" with the m3d.

I don't think the PCI-bus transfer is an issue. That should be done with bus-mastering DMA transfers, and will be as fast as the PCI bus and host graphics card will allow, regardless of what CPU you use (and a P133 should be more than fast enough to saturate a PCI bus anyway).
I think the biggest issue is that the PowerVR is a tile-based deferred renderer, which means that an entire scene has to be batched up in the driver first, and the draw calls binned and sorted, before they can be processed by the PowerVR chip.
This also means that a powerful FPU is required.

The VooDoo card is an immediate mode renderer, so triangles can be fed directly to its rasterizer, and fast fixed-point arithmetic can be used.

There should be some kind of 'sweet spot' for the PowerVR where the CPU is fast enough to take the extra overhead, and the PowerVR's clever and efficient TBDR will deliver better quality rendering and higher framerates than a VooDoo.

Scali wrote:

There should be some kind of 'sweet spot' for the PowerVR where the CPU is fast enough to take the extra overhead, and the PowerVR's clever and efficient TBDR will deliver better quality rendering and higher framerates than a VooDoo.

I believe this sweet spot is post-3dfx glorious days though.

dr.zeissler wrote:

That's interesting...a PIII/P4/T-Bird have lot's of horsepower compared to a P133/P166MMX that was the target for the voodoo1 as the competitor.
I read that because of the "pci-bus-transfer" the main gpu-should be fast too and the onboard-gpu's cause/have "performance-problems" with the m3d.

Weird. How much data is on your original CD? Just seemed odd that the image was 200MB bigger than the contents of the disc after burning.

^ there is an AUDIO track on that cd, see the mixed-mode icon on the windows explorer on my shot (or the contents of the CCD/CUE files). That's where the extra megabytes are.

The data track of that image is indeed damaged, there are ECC errors.

-M1 ECC_Q error detected in sector offset 285964
-M1 ECC_Q error detected in sector offset 287645
-M1 ECC_Q error detected in sector offset 287726
-M1 ECC_Q error detected in sector offset 287807
-M1 ECC_Q error detected in sector offset 287848
-M1 ECC_Q error detected in sector offset 287990
-M1 ECC_Q error detected in sector offset 288192
-M1 ECC_Q error detected in sector offset 288335
-M1 ECC_Q error detected in sector offset 288598

Do a MD5 comparison of the "affected" files against a pressed CD and you'll see they are the same.
Anyways, anyone willing to upload another image go ahead, I can't really be bothered any more with this.

Yes, because that's what ECC is for. 😉 The bit-errors are corrected by the ECC on read. It's more of a hint for you that your CD is slowly dying.

But it's also the reason why Ozzuneoj get verification errors after burning.

konc wrote:

I believe this sweet spot is post-3dfx glorious days though.

Yea, else we'd be worshipping PowerVR instead of VooDoo now 😀

you'd find more powervr worshippers on revisionistnostalgia-fueled dreamcast fan forums though... the kind that believe the PVR2 is some sort of RSX

derSammler wrote:

Yes, because that's what ECC is for. 😉 The bit-errors are corrected by the ECC on read. It's more of a hint for you that your CD is slowly dying.

But it's also the reason why Ozzuneoj get verification errors after burning.

Just to clarify, it's a sign that which CD is slowly dying? Our new burned copies or the original?

I just want to be sure that all of the data is here. Sorry if I offended you Keropi. I greatly appreciate that you took the time to upload these, I was just trying to figure out why I had errors. I've never had an img file I couldn't open with 7zip and I've never had ECC errors after burning a CD, so it appeared that the image had problems. If the errors are part of the image but somehow don't matter due to error correction, then that's fine... I just wanted to know.

No offense, that's what my original cd is - can't do anything more tbh.
Maybe the disc is failing, maybe the weird mastering they did is to blame I have no idea. Even the volume label is weird so anything goes....

I've attempted a 16-bit dither shader trying to imitate the PowerVR 16-bit reduction i've noted. I've found that the Kyro II does the same and it's likely the Dreamcast and Neon250 does too.

What I did was start with a bit of red/blue brightness, then a typical ordered dither matrix subtracting over that, then a RGB454 reduction, and blend the differences of a RGB565 reduction addively slightly. I think Videologic did this to avoid the typical green of 16bpp imo (though what would be really more logical to use are 2 lookup tables to fidget the color values 7 indices higher every odd entry per 16/32 indexes per channel)

This is a PCem S3 Virge Battlezone 2 DEMO shot in 640x480x24 with the shader applied. (to go further with this would mean 32 tap bilinear interpolation with 4-bit alpha channel precision)