I followed M2's development as close as I could, and even though I'm an outsider....I'm not a developer, I have no inside info on M2 or anything 3DO.... I do remember alot of things that I read over the years.
M2 was in no way anywhere near as powerful as Dreamcast.
However, it was 2 to 3 times more powerful than Nintendo 64 and MUCH more powerful than Saturn or PS1.
The Dreamcast was at bare minimum, 3 to 4 times more powerful than M2. That estimation came from Warp's Kenji Eno (D2 developer). At the time he said that, Dreamcast was new (mid 1998), had not been released yet, and the true performance of DC's PowerVR2DC graphics chip
was not known.
It's absolute fact that DC's SH-4 and PowerVR2DC combined power is much greater than that of the M2's twin 602 CPUs and BDA chips. M2 had 8 MB RAM for the console configuration, and IIRC maybe 16 MB for the arcade and industrial kiosk configs. Dreamcast had 26 MB RAM in its console config and around 56 MB RAM in its NAOMI arcade config.
The conservative spec for Dreamcast's performance was 3 million textured polyons/sec with lighting and every graphic rendering feature/effect turned on.
M2's performance with everything turned on was at MOST half a million polygons/sec. That means that Warp's statement of Dreamcast being 3-4 times more powerful would be wrong / or conservative itself.
M2 could supposedly push 700,000 textured polys/sec *without* effects, without lighting or gouraud shading, and over 1 million flat-shaded polys/sec.
Dreamcast's PowerVR2DC chip turned out to be significantly more powerful than Sega's conservative 3M polys/sec figure. Upto 7 million polygons/sec with all features could be rendered, but things like memory bandwidth and geometry storage data limits that to 5-6M polys/sec.
http://www.segatech.com/technical/polygons/index.html
The more stuff that's done on the non-graphics side like physics/A.I./ and other gameplay stuff, the fewer polygons get rendered, but most developers agreed (including Yu Suzuki) Dreamcast's poly performance was well over 3M, at least 4 million.
That's still a HUGE amount of performance compared to M2. Given that M2's in-game textured poly performance was likely 300,000 to 500,000 depending on the number of features turned on, it means Dreamcast is roughly TEN times more powerful, more or less.
Even if I am "half right" and we half DC's said polygon performance
(or double M2's), 5x more performance means that M2 is still not close to DC.
Of course I am talking about graphics, and the CPU side of things like MIPs and FLOPs would be a different comparison. Maybe that's where things are more like 3x in favor of Dreamcast ? Trying to remember, M2 CPUs could push... 70 MIPs each? DC's SH-4 could do 360 MIPS.
Floating point:
DC's SH-4: 1400 MFLOPs or 1.4 GFLOPs
M2's 602: 132 MFLOPs each = 264 MFLOPs.
Still, M2 was the most powerful console (and consumer) 3D hardware of its generation (the Saturn, PS1, N64 gen). The N64 could only manage 160,000 textured polys with all features on. PS1 could do only 180,000 textured polys with only gouraud shading & lighting, and no extra features at all, not even Z-buffering. The M2 was certainly the best home hardware of the day. Even the powerful 3Dfx Voodoo Graphics PC card of 1996 (which dominated PC graphics during 1997) was less powerful than M2, because realworld performance of Voodoo1 was only roughly 250,000 textured polys/sec with features on with the fastest CPUs. The M2 could push
at least 300,000 such polys/sec (remember EVERY feature on, in game) maybe somewhat more.
The MX hardware, I know somewhat less about it. I don't know if it was finished. In 1996 its performance was said to be double that of M2, with at least 1 million textured polys/sec with features on. In early 1998, a Next Generation Online article about the Nintendo/CagNet/Samsung MX deal said that MX performance was upto 4 million small triangles/sec. peak:
http://groups.google.com/group/rec.game ... ode=source
The MX chipset was a dramatically enhanced version of the M2 chipset sold to Panasonic and Matsushita, now capable of a 100 million pixel per second fillrate and utilizing two PowerPC 602 chips at its core.
(CagEnt's executives also boasted of a four million triangle per second peak draw rate, though the quality of those tiny triangles would of course have been limited).
I am going to believe/guess that meant flat-shaded polygons/sec similar to M2's 1M+ flat shaded poly performance. Making MX roughly 3-4 times stronger than M2, but still short of Dreamcast performance. However, there was a little known article by Intelligent Gamer magazine from 1996 that mentioned the *possible* use of on-chip graphics RAM with a version of the MX, pushing the performance to unbelievably high levels for the time, an unheard of 20 million polygons/sec. That was however, just a possible senario for MX, as it was still in development at the time. I don't know if there was any truth to that report, or if the embedded VRAM ever got implemented into the design. I am guessing it did not. Otherwise, Nintendo or someone would've wanted it really bad to use in a console. the first shipping console to use embedded RAM (eDRAM) for incredible performance was of course, the PlayStation2 with its Graphics Synthesizer chip that has 4 MB eDRAM.