Random Phrase: Home sweet home...


Read Me First!

This is the Spode's Abode archive. The old articles and forum have been kept here purely for historical purposes and are no longer updated.

Not all portions of this archive may work as expected.


Please visit the new site.



Intel D865PERL Motherboard - Canterwood vs Springdale
Written by Peter Barnard (23/July/03)
Page 3 of 6
Supplied By: Intel

Untitled Document

Overclocking

In a word, no. Intel don't take kindly to their creations being fiddled with, and you wont find any overclocking options on this board. You do have a full range of memory timing options of course, and there is a burn in mode to up the FSB by 1, 2,3, and 4%. You wont find any voltage adjustments though, or even any FSB options, the board limits you to what the chip identifies itself as needing. If you want to overclock, look elsewhere.

Performance and Benchmark Results.

And now for the numbers. Here I shall be comparing the D865PERL against the Canterwood sample that we reviewed a few months ago, and clocking it down to 1.6gig to compare it to a similar speed AMD chip, a palomino running at 1750mhz, on a 166FSB, with Dual DDR 333, using an nForce 2 motherboard. Unfortunately this is neither quite the same clock rate as the p4, nor is 166mhz AMDs most recent FSB standard. It cant really be called a fair comparison to an AMD system, but I have included it anyway to give some perspective to the figures.

Benchmarks used were 3dmark 2001se build 330, SYSmark 2002, and Sciencemark 2.0 membench. All tests were run with a Gigabyte of Dual DDR 400, in 2 modules, kindly supplied by Overclock.co.uk. Graphics cards used were Both ATI Radeons, an 8500 and an 9700 pro, both running at stock speeds.

3d Mark 2001 SE build 330

 
Radeon 8500
Radeon 9700 Pro
 
1024 x 768
1280 x 1024
1024 x 768
1280 x 1024
D865PERL @ 1.6ghz
8486
7013
10478
9626
nForce 2 XP @ 1.75ghz
9198
-
13676
12132
D865PERL @ 3ghz
10470
7924
15390
13398
Canterwood sample @ 3ghz
10941
8033
-
-
D865PERL @ 3.2ghz
10479
7726
16063
13765
D865PERL @3.2ghz w/o HT
10601
7943
16096
13538
D865PERL @3.2ghz w/o Dual DDR
10421
7905
15477
13522

Compare the numbers for the Canterwood sample @3Ghz and the D865PERL @3Ghz, and you will see just what a difference there is between the Canterwood and Springdale chipsets. The only difference performance wise between the two chipsets is the Canterwood includes PAT, which adjusts your RAM timings on the fly. Here we see just how effective it is, giving scores that would require a CPU running perhaps 300mhz faster to achieve this with the Springdale. As the figures show, Dual DDR is only giving marginal benefits with the slower card, but the much faster 9700 is making better use of the extra bandwidth. Turning off Hyper-Threading gives a slight improvement with the 8500, but the 9700 seems to be able to make much better use of the Hyper-Threading. For more information about Hyper-Threading, see my article on the 3.2Ghz P4. It is hard to make any firm conclusions from comparing the AMD platform to the Intel one, but it seems that AMD is still leading the way in power per MHz performance.

ScienceMark Memory Bandwidth
D865PERL
Dual DDR400
3553 MB/s

D865PERL

Single Bank DDR 400
2480 MB/s
Canterwood Sample
Dual DDR 400
4035 MB/s
nForce 2
Dual DDR 400
2295 MB/s

Here you can quite clearly see the benefit of Dual DDR, giving almost a gigabit a second more bandwidth. Strangely, this did not affect the 3dmark scores as much as you might think, whilst the Canterwood, which gains another 500 MB/s or so from the PAT technology, scores an awful lot higher. I suspect this is due to the fact that PAT reduces latencies, whereas dual DDR merely gives more data per cycle, without speeding up the response times of the RAM. Only the faster 9700 could put the extra bandwidth to good use. In terms of raw data throughput, both Intel chipsets beat the nForce 2 by a long way. Whilst the nForce 2 doesn't have PAT obviously, most of the gap can be attributed to the fact that Dual DDR is effective on the Intel platform, due to the quad pumped FSB eliminating the FSB bottleneck. If you cannot afford dual DDR, or it just isn't a convenient option, then you will find little difference in memory performance between the nForce 2 and the Springdale.

SYSmark 2002

 
D865PERL @3.2
D865PERL @1.6
nForce 2 @ 1.75
Internet content creation
423
236
226
Office productivity
173
121
143

The SYSmark results are rather interesting, as the Internet content creation benchmark shows just how much a 3.2ghz P4 can reduce processing times for complex things like image and video editing. If you compare the AMD system with the 1.6Ghz P4, you will see the p4 doing better on the Internet content creation, most likely due to the better memory bandwidth speeding up the manipulation of large files, since the Internet content creation benchmark involves a lot of multitasking with some very large files. The AMD system does much better on the Office Productivity benchmark however, probably because of the CPU advantage the Athlon architecture gives, and the fact that the smaller file sizes would not find the lower memory bandwidth a bottleneck. If you look at the difference between all the Office productivity scores, you will see how little the speed of your CPU matters if all you use is Office type applications.

To summarize the performance of the Intel D865PERL motherboard, it is certainly very powerful in terms of power and bandwidth, but in terms of latency and speed, it is not as good as the Canterwood, or indeed the nForce 2 when you take into account the differences between the p4 and the Athlon. This means the board is a good option for high end workstations, or even servers, where dealing with large amounts of data is important. It isn't so good for the gamer, who needs quicker response times for more frames per second, which can make the difference between fragging and being fragged.


<< Back | 1 | | 2 | | 3 | | 4 | | 5 | | 6 | Next >>



Copyright Andrew Miller
Please read our disclaimer

Search the site:

Random

Athlon 64 Chip Guide (Sempron, Newcastle, ClawHammer)