Navigate:

Advice
Beginners
BIOS Guide
CPUs
Links
Mainboards
Memory
Network
Storage
Video/Sound Cards

Contact
Forum
SiteMap
Sponsors
WebNews
Home
. .

Prices:

Mainboards

ABIT
ASUS
Chaintech
Shuttle
Soyo
Tyan

CPU
Intel
P4 2.4C-800
P4 2.6C-800
P4 2.8C-800
P4 3.0-800
P4 3.2-800

AMD
AthlonXP
XP 1700+
XP 2000+
XP 2400+
XP 2500+
XP 2700+
XP 3000+
XP 3200+

Athlon64
Athlon64 3200+
Athlon64 FX-51

Opteron
Opteron 240
Opteron 242
Opteron 244
Opteron 246

Memory

Corsair
Crucial
Kingston
Mushkin
OCZ

Search Prices:


























































































































LOSTCIRCUITS

SHORTCUTS:
Dual Cores
Numbers and Test Configurations
Synthetic CPU and Memory Benchmarks
WorldBench5
Cinebench 2003
Lightwave [8], 3dsmax
TrueSpace and CPU power
3Dmark '05
Gaming Performance
Idle and max Power
Prime95 and Final Thoughts

Give Us Some Feedback on this Review

 Intel's P4 820D and 670
More Power to Duallies
(Review by MS September 3, 2005)
P4-670

Final Thoughts

Benchmarks are treacherous. Aside from obvious or less obvious optimizations like in the case of e.g. Abbyy - which we consequently discarded -, benchmarks are becoming poorer and poorer as indicators of real world performance. Benchmarks are easy to run in most cases, at least when it comes to single applications or single threads. Multithreading on the other hand, even though it is supported by essentially every current operating system, can turn into a can of worms, especially since quantitative assesments of the results is somewhat more dificult than in the case of a single application that is timed.

This is where we have some reservations about some of the multithreading approaches, as informative as they may be. The reason is that for example CPU utilization varies within the runtime of most applications, which means that the synchronization of the different workloads - or lack thereof - can change the outcome of the benchmark. Granted that the variations should be relatively minor, they can still alter the results to where they no longer accurately reflect the actual system performance.


Another possibility is that some applications simply take over and use all system resources and the result is that the overall runtime of two applications started at the same time is exactly the sum of the two applications, were they run consecutively. Either way, the vagaries are such that it is easy to favor one or the other architecture not on the basis of the applications chosen but simply by the sequence in which they are launched, which is the main reason why we are not showing particular combinations of benchmarks at this point and stick with the more conventional approach of running one benchmark and then the other. Finally, we try to think about what the hell we were actually doing other than burning energy. And experiencing the smoothness of dual cores.

Along these lines, the key message of this article is that the processor that came in dead last in most of the benchmarks we showed is, in fact, the best choice amongst the current fleet of Intel's offerings. The advantages of the multicore design are very obvious in any of the 3D rendering applications, the power consumption is borderline acceptable and the gaming performance falls into the same category. For those who were not paying attention, the above refers to the 820 D CPU that retails currently a little higher than $200.-.

On the other hand of Intel's processor gamut is the P4 670, a processor that appears to have been released "because Intel could release it" without very much incentive over any of the other 600 series CPUs. In short, performance is limited by the host bus interface that caps the memory bandwidth at 6.4 GB/sec and a utilization of about 75%. The result is a real bandwidth of 4.9 GB/s under optimized conditions which is in stark contrast to the bandwidth hype woven around DDR2. Let's just face it, without bumping up the host bus interface, it basically doesn't even matter whether the memory bus is running at DDR-333 or DDR2-667 or higher.

On the other hand, increasing the bus frequency is not possible because multipliers lower than 14 x are not supported by the Prescott core and, therefore, a 1066 MHz bus interface would bump the lowest speed to 3.73 HGz and the power consumption to some ungodly 160 W under load for the dual core design. In the case of the 670, the power consumption would stay about the same but there would be no possibility to implement the Enhanced Intel SpeedStep Technology. On second thought, that one did not seem to work anyway...

Bottomline is that of all current Netburst Architecture processors, the 820 D is the way to go since it offers the most reasonable compromise at a killer price. Those who bought a 925 / 915 chipset ... sorry pal, no go... you got screwed.....

"Drum pruefe, wer sich ewig bindet, ob sich nicht doch was besseres findet" (Thus, who is about to tie the eternal knot, better check whether there ain't anything better coming up).

Pentium4 820 D
(dual core)

next page: => More =>

All advice and educational articles on LostCircuits are free, but if you feel you can, please make a small donation to us!
Thank you!

General disclaimer: This page only reflects the author's personal opinion and assumes no responsibility whatsoever regarding any of the contents or any damages that may occur explicitly or implicitly from reading the contents of this site. All names and trademarks mentioned in this review are the exclusive property of the respective parent companies.
All contents of this site are protected by international copyright laws. Reproduction of the contents even in parts is not allowed except after written permission by the author and referral to this site.
Copyright 2002 - 2008 LostCircuits