WHAT HAPPENS WHEN you get two very smart people, Tim Sweeney from Epic Games, and Andrew Richards from Codeplay, with diametrically opposed views together in the same room to talk about the future of graphics hardware? Easy, a debate. At GDC this year, that is exactly what SemiAccurate did, and through the magic of video cameras, you too can enjoy the spectacle.
It started out as any normal conversation at GDC would, talking to someone smarter than yourself about topics that they know in great depth. This time, I was talking to Andrew Richards about what new architectures he has his compilers working on. Codeplay makes compilers for chips that mere coding on drives many insane. The compilers autoparallelize C/C++ code, hiding architectural complexity from coders, theoretically making their job easy.
The talk turned to the architecture that was on the minds of most during GDC, Larrabee. Andrew was adamant that it would not work, or at least would never scale. The conversation went something like this:
Charlie Demerjian: You really think Larrabee doesn’t work?
Andrew Richards: Yes.
CD: But it does, Intel has working chips, and has demo’d it.
AR: It will never scale.
CD: Tim Sweeney thinks it will.
AR: Tim Sweeney is wrong.
CD: Would you say that to his face?
(Note: At this point, I started thinking this could get fun.)
CD: On camera?
CD: OK, I’ll see what Tim has to say.
Off to the epic Epic booth. The conversation with Tim Sweeney, Unreal Engine coder extrodinaire went something like this:
Charlie Demerjian: Andrew Richards says that Larrabee will never work right. Do you think it will?
Tim Sweeney: Yes.
CD: He says it will never scale, and has no future. What do you think?
TS: Andrew Richards is wrong.
CD: Would you say that to his face?
CD: On camera?
(Note: Game on)
We agreed to meet a few hours later, and the gods of timing and meeting rooms conspired to have all of the rooms in the epic Epic booth filled with people doing real work. The three of us wandered around GDC until we found an area that was relatively quiet and curtained off from wandering developers. Other than the odd caterer and janitor noisily wheeling something by, it was quite desolate, so we sat on the concrete floor and fired up the camera.
Our lovely venue/concrete janitorial area
I laid out the plastic silverware to use when the debate got heated, and fired up the camera. As a first test of my new camera’s ability to take movies, it was an abject failure. Debugging much later on showed that the memory speed of the SD card did not quite live up to it’s rated speed. The result were several sub-30 second clips of the author introducing himself, Tim Sweeney and Andrew Richards, then stopping abruptly. Without a video camera to film the debate, we called it a day and agreed to continue the following afternoon. I also felt like somewhat of an unprepared clod for wasting their time.
For the remainder of that day, and the following one, the overarching task was to find a working video camera that could be borrowed for about an hour. In a convention full of geeks with literally thousands of journalists, this should be a piece of cake, right? Wrong. No one had anything usable, but several offered their cell phones as a replacement. How can a hundred PR people, journalists, geeks, and PR agencies not have a video camera? Oh yeah, karma.
At the very last minute I ran into a friend, Robert Lupo, who offered to film it with his flip camera. All four of us, and and audience of two others, came together in a meeting room in the epic Epic booth for The Debate: Attempt II. The weapons were readied for the debaters, and I did my best to keep things moving along with a minimum of umms and ahs.
The true hero of this debate however was Robert Lupo. He went well above the call of duty here, literally on his knees filming Tim Sweeney and Andrew Roberts debate for over 45 minutes. Since we didn’t have a tripod, he was also holding the camera for the entire time.
Because of this, the video quality is a bit shaky, and the audio has ups and downs. We did our best to clean it up, but there is only so much you can do. Given the circumstances, the choices were to do it this way or not at all. Since the topic was and is quite pertinent, and the two debaters don’t get within sporking distance of each other more than two or three times a year, we went for it.
The debate ended up being just under 45 minutes long, and was chopped up into Youtube compliant lengths, we opted for picture quality instead of length. Installments will be posted until we run out, and if we ever get a host that has the bandwidth to support it all, we will post the whole thing as a single video.
With that, enjoy part 1 of the first SemiAccurate debate/argument/food fight over the future of graphics hardware.S|A
Latest posts by Charlie Demerjian (see all)
- SemiAccurate has Skylake-SP die shots, sizes, and more - Jul 21, 2017
- Skylake-SP has a diverged core - Jul 19, 2017
- Intel’s Purley platform architecture is a step forward - Jul 13, 2017
- Intel launches Purley aka metal Xeons - Jul 11, 2017
- AMD’s Epyc has lots of connections - Jun 26, 2017