Thoughts

Firstly, I will run 64-bit benchmarks using Windows XP 64-bit as soon as it's released. Some early testing with RC2 shows weirdness. While that weirdness may well persist in the final version, I'd like to wait and see.

And it's that 64-bit ability of Prescott-2M that's its main selling point, rather than any performance to be gained by the double-sized L2 cache, compared to Prescott-1M. I missed out gaming performance since I honestly couldn't find a game that really enjoyed the 2MiB that the 6-series or 3.73GHz XE had. While I could show gaming scaling with the XE, that was due to bus speed increases and the 3.46GHz Gallatin-2M was often faster. The working set of most games simply doesn't care for much more than 1MiB or so of on-CPU memory.

So while it's nice that it's there, it generally doesn't do much. Factor the bigger L2 out of the equation while comparing to the 5-series chips and you're swapping 200MHz core frequency for EIST and other power and heat management schemes, the enabled 64-bit ability and the NX bit capabilities.

EIST and the new C1 ACPI state generally help in terms of power consumption (Damage will show you explicitly, here) but since the J stepping 5-series supports the new C1 state, you can get the large majority of the benefit across the 5-series range now, if you're buying a brand new CPU. While there are some mainboards that don't program the ACPI table correctly for the C1 state, some don't support EIST entirely (at least for the moment).

So we factor that out of the equation too and it really is the trade off of 200MHz core frequency for 64-bit and the NX bit. Hmm. With the NX bit protection cicumvented in Windows XP SP2 recently, well, you can see where I'm going.

The obvious statement to then make is "if you're considering a processor just for 64-bit computing reasons, why are you considering the 6-series P4?", since the Model 4000+ is almost always faster than the 570J or 660 it competes with on price. A 3400+ Athlon 64 will have a lot of fun smacking a 630 around too (and even the faster 540J). Extrapolate anywhere between 3400+ and 4000+ as you compare to P4 and you can see what I mean.

The Extreme Edition CPUs remain a pointless exercise for Intel. A 570J will best a 3.73GHz Extreme Edition in more than a few tests, just on that 66MHz. For $999, does anyone really care any more? Get an FX-55 if you're that rich, and give the change from a grand to me as thanks.

I'm all for the technology enhancements that enable the reduction of heat and power consumption, like EIST. But Intel's competition has being implementing it since the very first days of Opteron and it's not a compelling reason to go out and purchase a 6-series Pentium 4.

Infact, the only reason I can think of to go with any of the new processors is if you have an irrational adversity to shopping anywhere that's not Dell.com. 64-bit is done much better elsewhere, the bigger L2 cache is largely ignored by mainstream software, EIST buys you little if anything and C1 support is sketchy on many mainboards at the moment. Intel needs to drop the 5-series completely and offer the 6-series at the 5-series pricing, matching clocks, to give it any sniff of a chance with people that know what makes a PC tick.

That it's the reason for Microsoft's holdup of 64-bit Windows XP is a bitter pill to swallow, especially when the pill ends up giving you nothing but the technological squits.

Prescott-2M feels like a stepping stone to nothing much, as the consumer processor industry steps sideways, rather than forwards. We'll see what happens with dual-core.

[Pages]
   Last Page
   [1]· Introduction
   [2]· Processor Examination
   [3]· System Setup and Notes
   [4]· Memory Tests
   [5]· ScienceMark 2.0 Primordia, We Crypto, Realstorm 2004
   [6]· Kribibench, MP3 Encoding, Cinebench 2003, PicColor v3.0
   [7]· Thoughts



	TheThirdMedia Hardware → CPU Guide → CPU Article > Intel's Prescott-2M: Pentium 4 660 and Pentium 4 Extreme Edition 3 73GHz