Anand is back with another pre-launch quickbench session, this time with Nehalem instead of Conroe. http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3326
So far it's followed my performance expectations nicely, but one thing I completely got wrong was the power... their test sample ate a measly 15 watts more than a similar clocked Penryn. That doesn't appear to be a significantly hotter chip I was fully expecting to see.
Since all but the memory tests were done with single-channel Nehalem verses dual-channel Penryn, and I expect both platforms were limited by the basic DDR3-1067MHz memory and high memory timings, I'm betting launch day memory latency for Nehalem will be lower. Especially with real memory. I'm thinking DDR3-1333 with 6-6-6-18 timings (or better) should show some interesting results?
And around a 5-10% gain for having working triple channel memory for the other benchmarks they ran, although Anand quoted Gary Key as noting WinRAR showed 10% gains between single/triple channel use.
So far it's followed my performance expectations nicely, but one thing I completely got wrong was the power... their test sample ate a measly 15 watts more than a similar clocked Penryn. That doesn't appear to be a significantly hotter chip I was fully expecting to see.
Since all but the memory tests were done with single-channel Nehalem verses dual-channel Penryn, and I expect both platforms were limited by the basic DDR3-1067MHz memory and high memory timings, I'm betting launch day memory latency for Nehalem will be lower. Especially with real memory. I'm thinking DDR3-1333 with 6-6-6-18 timings (or better) should show some interesting results?
And around a 5-10% gain for having working triple channel memory for the other benchmarks they ran, although Anand quoted Gary Key as noting WinRAR showed 10% gains between single/triple channel use.