Fox5 said:
nAo said:
PC-Engine..you will like this one
Hmm, how many NURBs were taken?
Perhaps MS also told the "truth" with 125 mpps for XBox......
Maybe as a complete system, but I don't think NV2A could do that alone.(wasn't it downgraded to 100mpps anyhow?)
working at a gamestop through out highschool and still holding a job there for the discount i can honestly say in 8 years i have never heard anyone ask that .
You work at a gamestop for the games discount?
I'm guessing your in college and this isn't your life career then. I don't think I would have ever pictured you in retail.
Oh and I thought babbages was gone ? I know the only one over here closed down .
Didn't Gamestop eat Babbages and Funcoland? BTW, I always wondered if Babbage's was named after Charles Babbage.
But 'til now and Allard's claims of 1 teraflop performance for Xenon
Is it possibly that the entire Xenon system could even have 1 teraflop of performance, or is it just something that was said because it doesn't have to mean anything right now since specs aren't final? Or maybe it's of the "some other hardware would require 1 teraflop of performance to achieve the same results as a dedicated processor we're using" variety. Like maybe an 68000 would have to be capable of 1 teraflop of performance to equal xenon.
I am fairly certain that Microsoft expects the entire Xenon to provide over "teraflop" of computing performance when all the processing performance of the CPUs and GPU are added together. even though they probably have not locked down the final clockspeeds, they know the minimum they will get, and therefore, what they can claim performance wise.
the GPU will have the lionshare of this floating point performance. its shader units will all be floating point even IF they are not unified shaders (current pixel shaders are fp based) but anyway the GPU shaders are expected to be unified, so they ALL could be used to crunch geometry / lighting / vertex shading data. this has to be how they can claim more than a "teraflop".
The original Xbox GPU alone was first rated at 120 to 140 GFLOPs and then downgraded to 80 GFLOPs. the CPU is rated at about 3 GFLOPs (regardless of what they can actually do, and actually sustain in realworld)
Xbox's polygon spec was and is as follows (AFAIK and IIRC)
original 300 MHz GPU:
300 million micro polygons per second
150 million transformed or T&L'd polygons/s
over 100 million textured polygons/s with all effects on, sustained
first downgrade to 250 MHz
125 million transformed or T&L'd polygons/s
100 million textured polygons/s with all effects on, sustained
second downgrade to 233 MHz
116 million transformed or T&L'd polygons/s
100 million textured polygons/s with all effects on, sustained
AFAIK Microsoft never gave their "micro polygon" performance figures when XBox GPU was downgraded to 250 MHz and then again to 233MHz.
at 250 MHz, the micro polygon count should've been listed at 250 million micro polygons/s and at 233 MHz, it should've been 233 million/s.
because of that, people thought the polygon downgrade was alot larger than it was: they thought 'oh that sucks, look, Xbox downgraded from 300 million polygons/s to 125 million (or 116 million). when in fact they were comparing the micro polygon count before downgrade (thus the HIGHEST figure) to the transformed / T&L'd polygon count after the downgrades (the smaller figures) it was Microsoft's fault for never clearing up that bit of confusion. I suppose they should have, it could've been to their (slight) benefit.
the actual drop was from 150 down to 116 million.
(or 300 million micro polys down to 233 million)
notice that the "sustained polygons/s with all effects on" figure basicly never changed with the downgrades. that was kinda curious.
obviously ALL of these figures are stretched, best-case polygon performance figures, if everything is superefficient and perfect, no bottlenecks, perfect programming, theoretical etc. Xbox games only push a fraction of Microsoft's smallest figure (100 million sustained), I don't know of any Xbox game that reaches even 30 million polygons/sec, the highest ones are all probably 15-25 million or less.
