For those whining about the fact that AMD was charging $1000 for their high end dual cores back in the early days of release...not to difficult to understand. As was noted they held the performance crown at the time, and thier desktop and server parts were interchangeable, they still charge $1000+ for their highend server chips.....more complex, better performance, lower power on higher clocks...extended instruction sets used only in workstation and server enviorments.
Where AMD is going out of their way to extend backwards compatibility, intel busts out a new socket with every new chip. Every 3 or 4 months they release a "new" highend chip that has 5-10% performance bump over older chips that cost 1/5th the price.
The cross licensing agreement between AMD and intel allows AMD to use x86 instruction set, and intel gets to use AMD's multicore archictecture, IMC design, and though they don't use it, i think hypertransport as well as x86-64....essentially they got the nehalem design from the agreement. With 32bit finally going extict....AMD will quickly be further down the losing end of that deal. But hey, intel trys to exploit a loophole with the AMD/foundry situation and say AMD breached the agreement....
Intel does many things well, instances where it includes cpu design are few and far between and rarely can be attributed to Intel's own innovation. Even when they have had the ambition to do something innovative, they have failed.
Itanium 64, joint venture between intel and IBM formed in 1992. They spent $2 BILLION...every year until it's launch in 2001. $18 billion dollars in R&D, they convinced microsoft to release an OS specifically for the IA64. They sold around 8000 machines. Why? Because thier 64 bit code....was not supported by ANY software. Users had the option to run it in 64bit mode, leaving them with no software to utilize, or they could set it to emulate 32bit mode....which left them with a very expensive machine that was 40-50% slower than mainstream 32bit chips.
The Timna, announced in 97 due to launch in late 98/early 99, was intels second attempt at a chip with an IMC, as well as a cpu/gpu in a single package. They went with RAMBUS memory for their IMC....upon doing a production run, they found a design flaw in the IMC, delayed the release reworked the design, did another production run...and the problem persisted. At this point they scraped Rambus and went with SDRAM...did a production run....and, surprise surprise the design flaw was still there. They scrapped the project in 2001 which was around the final delay of the...
Pentium 4 release! The chip that will scale to 10ghz. Too bad that it's performance gains took a nosedive clocked past 1.8, and it was being outperformed by AMD barton core chips that consumed something like 1/4th the power, ran at 2/3rds the clock and had double the memory performance of Intels DDR2 P4's....and then they overclocked with no effort to boot. Upon AMD launching the FX-55 it was discovered that intel would have to put out a 5.2ghz P4 (twice the FX-55 clock) to be on par in most area's of performance...and it would only require about 350w's compared to the FX55's 90w.
2001, intel's trifecta of contribution to the IT comunity
Intel doesn't work towards progress in performance and design, they work towards keeping the market tailored to what falls in their capabilities. Vista for example was designed with AMD/ATI hardware in mind as it was intended to be a 64bit OS exsclusively. The beefy UI overhead was supposed to be offloaded to the DX10 GPU...but wait, thats not fair cause intel's IGP's could barely render XP, and they had no prospect of a DX10 IGP....so that got dropped....tesselation, same reason....performance gains designed around an IMC....oops no no drop that too. 64bit oh no, intel doesn't have any 64bit chips to compete with AMD's. So they run a smear campaign on 64bit..5 or 6 years after they failed with IA64..and they ressurect a 32bit chip to improve on....P4? no no p3? well yeah bit's an pieces...p2? nah...ooo yes that's it.
The pentium pro. They had to go back to the best they had to offer in 1997 to release the Conroe in 2006. Which is a round about way of intel announcing "well, we discovered that between the pentium pro, and the p4...we really didn't do anything substantial enough to improve on....cause the rest was just crap" Instead of investing their substantal resources into developing a 64bit chip to compete with AMD, they put out an amazingly performing 32bit chip. Since AMD didn't really care about 32bit performance (shown by the fact that when comparing 32bit and 64bit versions of an app AMD sees a 17-25% performance gain in 64bit computing over 32) and this allowed software developers to stick with what they already had and knew well, it kind of hindered Vista's acceptance. People trying to run Vista ultimate with 2gigs of ram on a 32bit cpu are of course going to be dissapointed, it was never intended to be run under those conditions. With the C2/C2Q which included token 64bit performance, it's the same deal, small performance gains if any running 64bit over 32bit.
So here we are, with intel who bashed AMD's single die, multi core design, use of an IMC, hypertransport approach 6 years ago....using near identical architecture with nehalem.....after AMD told them how to do it. They don't have anything to trade now when AMD makes another innovative breakthrough, and they dont seem on the path to make any of their own. AMD is merely playing Intels game, and it's about damn time.
Facebook
Twitter
Instagram