I want a 720BE honestly. These expensive quad cores are for people with money
100% agreed. By my math, a 'three-core' P1 only needs to be clocked 20% higher to achieve similar results to a quad-core in most apps.
We've seen by the leaked Brit retailers that the 940 and 920 are going to cost ~$300 and ~$250. If we follow this trend down, the 2.6ghz variant will will be somewhere around $200+, and the 2.8ghz three-core 720BE (which will perform at stock similar to a 2.4ghz quad core that will not be released) will probably be $150-200.
Judging by the fact that the Phenom (1) three-cores OC similar to quad-cores, and use the same amount of power, the 95W (1.2v) Phenom 2 three-cores will probably do as well, if not better, than the 125W (1.3v) 920/940 that will be released essentially at the same time. Added to that fact is the X3 will be AM3 where-as the X4's will be AM2, allowing a greater lifespan and upgrade options.
1/2-2/3 the price for 4/5 the performance? YES PLEASE.
I predict the 720 will be the cheap bastard chip of choice until we see cheaper Nehalem derivatives/mobos.
I also wouldn't count out Propus, the X4 without the L3 cache. It's been reported it is only ~10% slower than with the L3 cache in desktop apps (probably because of all those cache misses in the L3), but that chip will probably only be 70% the size of Denab, allowing it to be priced very competitively if AMD chooses. While it may not LOOK attractive to not have any L3, I think it's very possible it will perform better than the X3's with the L3 at the same frequency. I think there's a reason beyond manufacturing priorities it's not coming out until months after Denab. If they are cheap (as the road map suggests by it being below the X3 and the 'quad core for the masses' moniker) and there is a BE, that could be another budget overclockers' dream.
Personally, I still don't see why AMD doesn't release a 3.2-3.4ghz chip at 125W (1.3v max) and even higher at 150w or less (1.4v max) considering there will be 95W quad-core chips @ 3ghz. In my mind, 16/17x (3.2/3.4ghz if at stock) multi is going to be the sweet spot as 16x250HTT = 4ghz (a system mem-synced with 1066 DDR2 at stock could achieve 4250mhz @266HTT...not an unreasonable feat considering the overclocks we've seen) It'd make more sense for AMD if the BE's used these multis to discourage buying a cheaper chip and obtaining the same OC. Nehalem uses 1.4v at stock and is rated at 130W, surely AMD wouldn't look terrible by releasing such parts...not any worse than the underwhelming 140W 9950BE...and it's not like it would take a massive stock heat sink to cool the thing at 1.4v as Intel has shown, while giving them competitive parts (at stock) if clocked closer to 4ghz. Maybe it'd be sneaky, but in my mind it would be a good move to drum up market/mind share. Surely a 3.8-4ghz speed-binned to hell Denab could take on the 3.2ghz top-o'-the-line Nehalem, considering nehalem is what, ~15% faster than Core2 and Phenom 2 is expected to have close to parity with those chips.
Another way to think of it is this: A Core i7 920 overclocks to 3.8-4ghz before being constrained by base clock/multi. These Phenom 2 940BE's have been shown to overclock to 4-4.4 on reasonable cooling (non-L2/Dice) If you take a 15% disparity between the two, they come out almost even while the pricing will be similar for the chips. When you take into account memory and motherboards though, the phenom setup would cost around $500 for chip,mobo, and 4gb DDR2 (or 8GB $50 more) while the nehalem setup w/ DDR3 would cost north of $800 with either a fast 4gb kit or a slower 6gb kit to have a tri-channel setup.
Personally I'll opt for the 720BE and 8GB on a decent board, which realistically could cost ~$400, half as much as a nehalem setup with less RAM, while surely being more than 1/2 as fast...I'm guessing ~75%-80.
This reminds me of the socket A days when the Athlon setups were slightly slower, but a hell of a lot cheaper. I miss my 2500+ MP on NF2...
Oh well...I digress. I better stop before I end up writing my own TPU article, if I didn't already.
