The Core 2 was pretty decent, but it looks a whole lot better compared to an Athlon 64 rather than a Phenom. If AMD had been faster in getting the Phenoms to market than Intel was with the Core 2 Quads, the Core 2 Quads would still have looked solid but wouldn't have completely walked all over AMD's product line like it did with the aging K8 dual cores. Intel would have had to have pulled out something like Nehalem to compare with the Phenom to get the same effect as a Conroe vs. a K8. That's the whole feet-to-the-fire thing I was talking about.
I know I7 initial release was disapointing but i do believe whilst the economic pressures will prevent more frequent releases what it will provide when we need its the time to develope the next step in processing power.. I still feel that more can be done with i7 appart from apps and sli setups...
The i7 release was disappointing? I didn't think so. It would have been disappointing if the i7 didn't perform better than the Core 2s, but it certainly did, and by an okay margin too. The NetBurst -> Core transition was a real anomaly as the Pentium Ds were pretty uncompetitive at the time and priced far too high, while the Core 2s were very good and in all actuality, priced too low (this started a price war that led to Intel losing a bunch of money for a while.) Intel was much more conservative this time in keeping the i7 prices high enough to not kill the market for their existing Core 2-based products. I think the only real disappointment was that the initial prices of an i7 motherboard + three sticks of DDR3 + the i7 were much higher than we'd all been used to paying. The i7s also run hot, but the TDPs are similar to the Phenom X4s and the faster Kentsfields.
Also, the real spot where the i7 shines is in the Xeon form. One look at the architecture of the i7 should tell you that it was designed as a server chip first as it has massive memory bandwidth with a three-channel IMC and the QPI links, plus the ability to handle a bunch of threads with HyperThreading. Almost all other server CPUs out there have some form of multithreading to help with handling highly-threaded server applications (UltraSPARC T1 and T2, IBM POWER6 and POWER7.). The FSB wasn't completely tapped out on the desktop and Intel could have gone a bit farther with 1600 MHz FSB non-Extreme Edition CPUs and been fine, plus many desktop apps couldn't use more than four cores and HT wasn't really needed. But those features were needed in the server space.
Only a Octocore in my mind will fit the bill with loads for cache and a dose hyper threading for good measure..
That's a Xeon MP 75xx "Beckton," which should be coming out sometime here. It has eight cores, HyperThreading, and 24 MB of L3 cache, which I think qualifies as a "load." However, it will probably be one of those "if you have to ask the price..." kind of units. I think we mere mortals will probably have to wait a couple of years for Sandy Bridge to see eight cores at a reasonable price from Intel.
Then the software will have to play catchup with true multi threaded apps to take advantage of the hardware.
I hope Windows 7 utilises multithreading more - with a dose of DX 11 help... A multi threaded capable API... mmm thats a thought - wonder if a version of PHSYX will be in DX11... as standard..
Some software already is playing catch-up with multithreading as tests of the Core i7s with HyperThreading on vs. off have shown. Windows apparently keeps improving multithreaded support, but it's mostly the client applications that need to be improved. IIRC even Windows XP can handle 32 cores, although it isn't that great at scheduling them. I don't know what DX11 will have, but I doubt that Microsoft would make NVIDIA's proprietary PhysX part of it. My money is on MS making their own physics engine and putting it in a future DX revision. They like to keep control over their graphics API and don't want to be at the mercy of a third party. I could only see PhysX being in DX sometime if Microsoft bought it from NVIDIA, and then we'd see PhysX made to work on at least AMD/ATi GPUs if not also Intel's IGPs. I think Intel's IGPs probably can handle some GPGPU type of work like physics since they're DX10-compatible and I think DX10 requires enough FP programmability to make GPGPU work possible.