Nehalem: The End Of Mainstream Overclocking?! - Page 2

Last message on previous page:

You are correct about the average user not overclocking their pocs but the people that do OC are the people that determine the supremecy of one chip over another. Now here is what I mean. When your neighbor / mother / aunt wants a new computer who do they go to. They go to you, me, pretty much the kind of people that hang out in here. What are you going to tell them? You are going to tell them to make sure to get a computer with a AMD proc not because they want to OC but because in your mind it's the best proc. It's us that determines the direction of products coming out of the manufacturers. Just let Intel know and believe me they will pay attention even though we are the minority.

Some of you guys are ranting and raving about this as if it's fact. It's only a Fudzilla rumor. You know, the same sort of guys that said Phenom FX would come in at 2.8GHz last year, and reverse hyperthreading, the 3GHz Phenom scoring 30K in 3dmark 06, 2.9Ghz Barcelona 'dancing in the aisles!' etc.

Even if were true, a stock Nehalem will blow anything AMD has away even while overclocked. All AMD has against Nehalem is a 45nm shrink of Phenom (even then, it's on a substantially inferior 45nm process, with no HK/MG) with more L3 cache and a few minor tweaks. Hell, all AMD has until the second half of 2010 is a 45nm K10.

The quad-graphics support would be for ATI cards only(and larabee I suppose, dont know if it will support multi-GPU or not), nVidia wont share the nessisary info to allow SLi to work on anything other than an nVidia chipset.

Larrabee is supposed to be a multi-core GPU. I am not sure if this means 2 or 4 or maybe more. All I know is it seems interesting and I am anxious to see what Intels R&D can really do. We already know they can make good CPUs when need be but a GPU is a whole other matter.

'Multicore' makes little sense when applied to GPUs, since they've been using multiple hardware pipelines for years. The last GPU I worked on had something like two vertex pipelines, four texture pipelines and sixty-four pixel pipelines... if you want twice the performance, you just double the number of pipelines and increase the memory bandwidth to match; unless you have bad bottlenecks in your design there's little reason to put a second independent core on there instead.

The reasons are costs and efficiency. Itll cost less eliminating the north bridge and putting it on die. Itll be more efficient on die, or faster. Itll be more power efficient also. But at this point, unless Intel releases the PLL info to the mobo makers, which I doubt they will, therell be no ocing on the Lynnfields. There will be on the Bloomfield package. Its going to get more costly ocing Intels new arch, thats for sure

The only caution I throw out here is this, remember all those enthusioast EEs theyve made? Ever see em go down in price? Even after the C2Ds were out the P4 EEs were 900$ or more. If they decide thats how theyre going to handle this enthusiast solution the same way, that may not happen. But Im hoping it will, so it opens the doors for the bang for the buck guys, and the people with little or no moneys