In honor of March 22nd which marks the 20th year anniversary of Intel releasing the first Pentium processor (The socket 4 Pentium 50) I have a RAR.
Lets say by act of ROB the Intel of 1993 suddenly has a duplicate headquarters building from 2013 appear next to there 1993 HQ, inside is everything from 2013 related to the 20 year history of Intel from it's design documents to production reports to everything they need to produce everything from the Pentium 133 to the first Pentium 4 to the Core set of processors and the low power atom mobile designs.
In other words if Intel has written it down either on paper or electronically it's available for the 1993 Intel team to peruse. They don't get any copies of current 2013 hardware except what Intel produced itself sans two midrange 2013 computers which contain copies of all of the software documentation. The network wiring and servers are missing it's just all of the documentation and a few engineering samples of the various processors (Non-working but still identical physically to the real thing)
How far would Intel with it's 2013 knowledge be able to race ahead of Moore's law?
RAR Intel from 2013 to 1993
Moderator: Edi
RAR Intel from 2013 to 1993
"A cult is a religion with no political power." -Tom Wolfe
Pardon me for sounding like a dick, but I'm playing the tiniest violin in the world right now-Dalton
Re: RAR Intel from 2013 to 1993
Most of the info would be more useful to Intel from a business standpoint than a technological one - in other words, they'd know not to use RDRAM, not to design the Pentium 4 or anything even vaguely resembling it, to update x86 to include 64-bit extensions rather than producing a totally new VLIW design, to take integrated graphics a bit more seriously, and that AMD would ultimately turn out to be their main competitors rather than Cyrix.
Without the technology to actually manufacture any of the chip designs on their hands, they'd be limited to using it to gradually improve their existing designs. And bearing in mind the usual chip design and manufacturing timescales, it probably wouldn't start to make any difference to their products until the introduction of the Pentium Pro in 1995. Ultimately though, Intel wouldn't really have any reason to start making uber-powerful chips on the order of Ivy Bridge or Haswell a whole lot sooner than they actually did/are going to, so all we'd see is probably just moderately better and more sensibly designed versions of the products that Intel have produced until now, along with Intel not getting their backside kicked by AMD for most of the years from 2000 to 2006.
Without the technology to actually manufacture any of the chip designs on their hands, they'd be limited to using it to gradually improve their existing designs. And bearing in mind the usual chip design and manufacturing timescales, it probably wouldn't start to make any difference to their products until the introduction of the Pentium Pro in 1995. Ultimately though, Intel wouldn't really have any reason to start making uber-powerful chips on the order of Ivy Bridge or Haswell a whole lot sooner than they actually did/are going to, so all we'd see is probably just moderately better and more sensibly designed versions of the products that Intel have produced until now, along with Intel not getting their backside kicked by AMD for most of the years from 2000 to 2006.
Re: RAR Intel from 2013 to 1993
The entire semiconductor industry hit a power wall at 90nm; this sort of advance knowledge is priceless. There's a lot of non-technical useful information, like how AMD got so good (DEC and NexGen talent, mostly). The way the industry will end up going will help, too.
In the end not much will change; producing the technology to produce the fabs will advance roughly with Moore's Law. You can only validate a design so quickly. The worst that could happen is that Intel gets lazy with such an advantage.
In the end not much will change; producing the technology to produce the fabs will advance roughly with Moore's Law. You can only validate a design so quickly. The worst that could happen is that Intel gets lazy with such an advantage.
- Starglider
- Miles Dyson
- Posts: 8709
- Joined: 2007-04-05 09:44pm
- Location: Isle of Dogs
- Contact:
Re: RAR Intel from 2013 to 1993
No, it will go much faster. Getting new litho processes to work acceptably is a difficult, extended process of experimentation, tweaking, tuning and outright trial-and-error. This is why it is much harder to set up the first fab for a given process than the second and subsequent fabs. Having an exact description of all the chemical and mechanical steps could potentially halve the interval between new processes being available, or allow skipping every other process in the historical sequence, although it may not make sense to actually do that for commercial reasons (why go faster than necessary simply to beat all your competition).phongn wrote:In the end not much will change; producing the technology to produce the fabs will advance roughly with Moore's Law. You can only validate a design so quickly. The worst that could happen is that Intel gets lazy with such an advantage.
1993 was right when Itanium was transitioning from a research project into a full-blown development effort. Althought the obvious thing to do would be to cancel it, the amount of support / management investment into the project would probably mean that it just gets redesigned. If it got more future technology than the main Pentium line (and effectively unlimited funding from Intel's complete market dominance) it would have a good shot of taking over from x86 as the main ISA. Knowledge of the future importance of GPUs and a reasonable sequence of working GPU designs from Intel's various integrated graphics efforts makes a huge difference. Intel could and almost certainly would get an early dominance of that market, kill ATi and prevent Nvidia from ever existing. Similarly knowledge of the future importance of smartphones is huge, Intel will definitely buy or crush ARM and use its massive process advantage to field Atom or something similar early, to ensure that it has a complete monopoly in all segments. On a minor note working designs for cheap gigabit and ten gigabit NICs will make those standards viable earlier and probably accelerate adoption by a couple of years. Similarly earlier arival of multi-core processors will pull forward various parallel programming technologies by a few years (that existed in theory since the early 80s but didn't have widespread applicability).