Trevor Wing, Group Marketing Director, VideoLogic.
GG: How long have you been with VideoLogic?
TW: Since December 1989, before that with Hewlet Packard for 17 years, leaving as a director of marketing for the data communications division.
GG: What’s your role?
TW: Overseeing all of the company’s marketing activities, and specifically responsible for the sales and marketing of our technology of our products under the NEC relationship.
GG: What’s the latest on the release of the PowerVRSG chip?
TW: We released the technology into the public domain in February this year and that covered the increased functionality and the chip level road-map for all the second generation chips, which includes five new chips this year and the individual chip announcements for those chips will be staggered throughout the year, the first one being the 2D/3D chip for the PC space.
GG: What are the latest benchmarks?
TW: We have the latest benchmarks, but we are not releasing any details apart from saying that we will deliver five times the performance of the current 0.35 micron chip called the PCX2 and that was bechmarked at 250,000 polygons a second. So the spec we put on the new chip is greater that 1.2 million polygons per second and above 120 mega-pixels fill rate.
GG: How confident are you that PowerVRSG is going to compete with Voodoo2?
TW: I feel very confident for two reasons. The first is that we know what Voodoo2 is capable of…the product is shipping. We have Voodoo2s, we benchmark them inside and out in great detail we know what our target is in terms of what we need to hit. We also have a test chip of second generation PowerVR running now which is on 0.35 micron, 66MHz, so we know what that is benchmarking against Voodoo2.
It’s looking very good…I can’t really quantify it except to say that it’s performing as per expectations and we will deliver over twice the performance of that with our 0.25 micron implementations, so those two factors lead us to assert confidently that we will be in a position to compete very, very well with Voodoo2 on performance and features set.
GG: Claims about chip performance in the past have been vastly over-estimated compared to what finally shipped. How can you re-assure customers that you can deliver what you say?
TW: It is true in the past that performance claims have been very, very variable. Most companies have claimed performance data that they can justify by possibly looking at one part of the pipeline and stacking the performance of that part of the pipeline.
For instance Riva 128 has consistently said five million polygons a second, and that probably is true for one part of the chip but it is certainly not true for the overall chip and it certainly isn’t true for the overall system. The Voodoo2 said initially 3 million polygons a second and that again was the throughput of the chip itself. It didn’t take into consideration the bus band-width supplying the chip or the memory bandwidth limiting the chip.
Click to continue
. . .
So what we’ve done is "let’s be absolutely reasonable about this" and with the second generation slide presentation (which we can send you) for the 0.25 micron, 100 MHz implementation the chip itself will do 4 million polygons a second. However, we then say that on a Pentium II 333 MHz with Windows running and a game running it is only going to achieve 1.2 million. That’s not a function of the chip’s rendering performance, it is a function of the CPU’s ability to supply polygons.
We have been criticized in saying that 1.2 million polygons isn’t very much with Riva claiming 5 million for the ZT, 8 million for TNT and Voodoo claiming 3 million for Voodoo2…what we’re saying is real, sustained throughput to deliver polygons to the screen for every frame in the game.
On the console-based systems it will be much better than that because the CPU in the console-based systems is a lot more powerful than pentium in floating point performance. I am confident we can deliver what we have said in terms of performance, I think in reality when all the chips are benchmarked together, the high performance chips no matter what I put in that category…the new Riva chips, the Voodoo2 chips and ourselves, we will achieve similar bechmarks in the popular benchmarks because we won’t be rendering limited we will be CPU limited.
GG: Have you distributed any of these boards to games manufacturers to test them out?
TW: Our policy is to send out development kits free of charge to qualified registered developers and we have done that. Our first generation developers kits are very widely spread. We’ve started feeding out the first of the second generation boards, we started doing that around November last year…so there are quite a few of those out there already however they’re based on the 0.35 micron technology running at 66MHz so they have got an idea of what the performance is going to be and so they can make the comparisons that we make.
GG: What’s been their reaction?
TW: Very good…and the demos if you ever get a chance to see them, we’d be happy to show you, we get them off the shelf…we have Incoming by RAGE, Forsaken and a few others and I think the performance speaks for itself…framerates in excess of 60fps and they look gorgeous and that is roughly half the performance of what we will deliver.
(Obviously at the moment we are not releasing any framerate figures, we are waiting for independent tests)
on Page 2