Why not let PowerVR build RSX?

Understanding of PowerVR's real-world benefits isn't relevant for device makers who pick their GPUs based upon brand. Console makers, like Sony who's the topic of this thread, are selling mostly on the strength of their own brand and would be more interested in the performance of the GPU.
Remember that I replied to this post.

If you're talking about console makers and not the PC space, why'd you mention Kyro? You mention MBX vs. PSP, but what benchmarks have you run on the two? None you say? How much money did Sony save by developing it in-house for a one-time cost instead of paying a royalty on every unit? Say what you want about their interest in performance, but there's no competition here. PSP wouldn't sell any more units with AA. A thread here was even talking about how its best games aren't even 3D. Other games are based on PS2 architecture, so it makes sense to make it vaguely similar in design.

As for the CLX2, what are you comparing it with? And where did you get comparison numbers to justify the claims you're making? 90% of your argument is nothing more than conjecture. Even if you did have data, it's all outdated because today's 3D workloads have no resemblance to those of the past.

So can you lay off this ignorance crap? MBX is being licensed where it makes sense.
Kyro 2's performance was so favorable that, of the two benchmarks being used to argue against it, the first one, MDK2, actually shows victory for PowerVR over its most direct competition (and a near tie with higher cost competition with the highest, yet still frame rate relevant, detail setting) and the other one was 3DMark, not a relevant application in itself and not even indicative in this case of the games which budget cards, like Kyro 2, ever ran.
You don't have to convince me. I'd buy a Kyro II instead of a Geforce MX or Voodoo5 in a hearbeat, and I don't care about 640x480 benchmarks at >100fps. I'm talking about common Joe perceptions. Benchmarks like that are enough to show that T&L matters, that's all, especially for value consumers who didn't have an Athlon 1GHz. 3DMark2001 was all the rage back then, too. Anyway, now that you tell me you're not talking about video card manufacturers, this Kyro talk is meaningless.
 
Any number of reasons might explain why Sega Sammy, a company who's been familiar with PowerVR's benefits before, turned it down recently for Lindbergh, including the possibility that its TBDR might not be optimal anymore for modern workloads as proposed.

One of the primary reasons to suspect for companies who missed adopting PowerVR back when its performance was so clearly dominant, however, is obviously a lack of understanding. All companies wouldn't suddenly have become so tech savvy, though the example PowerVR has made definitely appears to have had effect as the majority of the major semiconductor companies now use it for, at least, Dreamcast/Kyro class workloads.
 
What if they switched to nvidia cause they needed better sw support? (compilers, tools, libraries, etc..)
 
In addition to PowerVR's suitability for various market spaces, CLX2, Kyro 2, and MBX technologies represent various time periods, at least, when PowerVR compared favorably.

Since the pricing of SoCs like Freescale's i.MX31 family shows that MBX adds several bucks to the cost at most, Sony wouldn't have saved a lot by going in-house for PSP, and they didn't attempt to reorient PSP technology for their Ericsson phones, and instead use MBX, after all that work despite their company mandate for sourcing semiconductors in-house whenever possible.

The CLX2 is compared favorably to Voodoo3 and TNT2 by developers with experience using each yet released at least six months ahead of both.
 
In addition to PowerVR's suitability for various market spaces, CLX2, Kyro 2, and MBX technologies represent various time periods, at least, when PowerVR compared favorably.

Since the pricing of SoCs like Freescale's i.MX31 family shows that MBX adds several bucks to the cost at most, Sony wouldn't have saved a lot by going in-house for PSP, and they didn't attempt to reorient PSP technology for their Ericsson phones, and instead use MBX, after all that work despite their company mandate for sourcing semiconductors in-house whenever possible.

The CLX2 is compared favorably to Voodoo3 and TNT2 by developers with experience using each yet released at least six months ahead of both.

Benchmarks definitely showed the CLX2 with some major losses (could have been drivers, but I believe it was linked to a lower polygon rate and some sort of memory limitation from what I remember) against the voodoo3, tnt2, and matrox g400, though I do think it beat the tnt and voodoo 2. If it released 6 months before tnt2, that makes it sort of a mid generation card, sort of like the dremacast was a mid generation console. BTW, was it CLX or CLX2?
Kyro 2 was a very attractive geforce 2 mx competitor, but once again was released mid gen shortly before the geforce 3 came out. Not that the kyro 2 was in the same price bracket, but software support for cards without hardware t&l was already dwindling quickly at that point. I must admit though, I was very tempted to upgrade to the kyro 2 from my voodoo 3, but the fact that it wasn't a high end card put me off (was waiting for the kyro 2 se or whatever), and I eventually got a geforce 3 ti 200.

MBX...why I've got a pda running on an MBX Lite. And not 1 single piece of software to use it, besides a video player and a snes emulator. No idea why MBX has seen so little market support, especially when an ARM9 with integrated MBX Lite would have been a dream for the nintendo ds.
 
CLX2 (since it was Imgtec's only console part, it could be called just "CLX" without confusion, though the "2" at the end specifies that it was part of PowerVR's 2nd generation) had substantially heftier 3D than its PC cousin, the Neon 250, and I'm not sure which benchmarks would've directly compared Dreamcast's CLX2 to competitors' PC processors.
 
CLX2 (since it was Imgtec's only console part, it could be called just "CLX" without confusion, though the "2" at the end specifies that it was part of PowerVR's 2nd generation) had substantially heftier 3D than its PC cousin, the Neon 250, and I'm not sure which benchmarks would've directly compared Dreamcast's CLX2 to competitors' PC processors.

Ah, alright, yeah, I was referring to the Neon 250.

How was the CLX2 significantly more powerful? Wasn't it even a lower clock rate?

As for benchmarks comparing the CLX2 to PC stuff...best I could say is console ports to PC games, but that's not a direct comparison due to (arguably) beefier cpus on the pc side and way more memory. Quake 3 and Unreal Tournament didn't run all that great on the Dreamcast though, nor did 4x4 Evo.
 
Various silicon saving reductions ranged the performance of the Neon 250 from three-quarters to one-half that of the CLX2, despite the PC part's higher clock rate.

The number of Z comparators/modifier volume calculators was apparently reduced from 32 to 16 for a smaller tile size on Neon 250 of 32x16 compared to CLX2's 32x32. That should've directly scaled down the peak fill rate for opaque overdraw and modifier volumes from 3.2 Gpix/sec to 2 Gpix/sec.

Several of the other subtle changes in aspects like texturing and general efficiency are described in a previous thread:

http://www.beyond3d.com/forum/showthread.php?t=14660
 
Any number of reasons might explain why Sega Sammy, a company who's been familiar with PowerVR's benefits before, turned it down recently for Lindbergh, including the possibility that its TBDR might not be optimal anymore for modern workloads as proposed.

not only that, but Lindbergh is linux based, isn't it? nVidia has had a long history of having some of the most linux friendly GPUs, and their openGL support is also top notch. regardless, i'm pretty sure sega wasn't basing the choice completly on performance, considering they picked the GF6 series when they could have had the 7s.
 
not only that, but Lindbergh is linux based, isn't it? nVidia has had a long history of having some of the most linux friendly GPUs, and their openGL support is also top notch. regardless, i'm pretty sure sega wasn't basing the choice completly on performance, considering they picked the GF6 series when they could have had the 7s.

Cost was likely also an issue.
Linux based huh? Surprised we haven't seen a pc port of the game already if it was that abstracted from the hardware.
 
CLX2 (since it was Imgtec's only console part, it could be called just "CLX" without confusion, though the "2" at the end specifies that it was part of PowerVR's 2nd generation)
Actually, CLX2 was an upgraded version of CLX. I can't recall if the latter was actually made nor what the additions were.
 
personally, I wouldn't let powerVR make RSX, but i'll be sure use their assets to make one :) , (figuratively speaking :) )
 
Back
Top