According to Intel, in the Q2, 2002 the i845G chipset sold best of all among other chipsets of the company (and certainly, among Socket 478 chipsets). There is nothing strange: supporting a 533 MHz FSB, DDR200/266 memory (DDR333 unofficial) and working together with the ICH4 (which means almost free USB 2.0), the i845G has the richest set of functions. Besides, unlike the i845E, it has integrated graphics (entry-level, though), which is nearly free (the difference from the i845E is less than $5), can be used together with an external video card and provides decent 2D quality. It's quite a universal solution which doesn't match with a concept of an integrated chipset (cheap, simple, and a lot of integrated components), that is why boards based on the i845G are not of the mATX format (except Biostar and Chaintech in the today's competition), and a network controller (which comes in all office boards based on integrated chipsets) is mounted not on all of them. 

A couple of words on the DDR333 support. As compared with SiS and VIA which announced support of this memory type in their chipsets a long time ago (SiS also provides support for DDR400, though officially such memory is lacking), Intel doesn't hurry. No wonder that board makers and users want to use DDR333 ignoring Intel's recommendations. But they shouldn't do that. Our tests showed that just few boards were able to work correctly with quality modules of PC2700 (modules of this series are already certified by Intel for its future chipsets supporting DDR333). Some boards failed to start, some had bugs when we set the memory speed to 166 MHz - as a result, we had to estimate performance of the i845G systems only with DDR266, and we recommend you to do the same if you don't have an opportunity to test dozens of components for their full compatibility. 

As the JEDEC standard for PC2700 is lacking Intel couldn't provide correct support of this memory type. That is why we should follow the Intel's recommendations when choosing its chipsets as its specs are announced only when the company is completely sure of them: i845G (and other currently available on the market) are used with the DDR266, the DDR333 will arrive in a fall's lot of chipsets from Intel, where i845GE will replace i845G. Note that the memory is able to work at 166 MHz (DDR333) only when the FSB is set to 133 MHz (533 MHz Quad-Pumped) and the chipset supports only 1.5V video cards (old cards that require 3.3 V won't work). 

And now you can take a look at description of 7 new mainboards based on the i845G; the test results will be given for all of 9 boards including two models we tested before; and there is one more solution based on the VIA P4X266A chipset. 

Test conditions

• Processor: Intel Pentium 4 2.53 GHz (19x133 MHz), Socket 478 
• Mainboards: 
• Abit BG7 (BIOS 8R) 
• ASUS P4B533-V (BIOS 1004 beta 007) 
• Biostar P4TDG (BIOS 0522b) 
• Canyon 9IGA (BIOS of 17/04/2002) 
• Chaintech 9BIF0 (BIOS of 26/04/2002) 
• DFI NB76-EA (BIOS of 31/05/2002) 
• EPoX 4G4A+ (BIOS of 17/06/2002) 
• Jetway 845GDA (BIOS A02) 
• Leadtek P4I845G (BIOS of 09/07/2002) 
• Shuttle AV45GTR (BIOS AV45SI1S) on VIA P4X266A 
• Memory: 512 MB PC2700(DDR333) DDR SDRAM DIMM Samsung, CL 2 (used as DDR266 with CL 1.5 on all i845G boards and with  CL 2 on the VIA P4X266A) 
• Video card: Palit Daytona GeForce4 Ti4600 
• Hard drive: IBM IC35L040AVER07-0, 7200 rpm

• Windows XP Professional 
• DirectX 8.1 
• Intel Inf 4.00.1013 
• Intel Application Accelerator 2.2 
• VIA 4-in-1 4.40v(a)p3 
• NVIDIA Detonator XP 29.42 (VSync=Off) 

• VirtualDub 1.4.10 + DivX codec 5.02 Pro 
• WinAce 2.11 
• BAPCo & MadOnion SYSmark 2002 Internet Content Creation 
• BAPCo & MadOnion SYSmark 2002 Office Productivity 
• SPECviewperf 7.0 
• MadOnion 3DMark 2001 SE build 330 
• Gray Matter Studios & Nerve Software Return To Castle Wolfenstein v1.1 
• Croteam/GodGames Serious Sam: The Second Encounter v1.07 

Test results

The difference in the MPEG4 encoding speed hardly exceeds 2%. The board based on the second chipset goes on a par with the others. 

In the WinAce archiving the difference reached 4%, and the ASUS P4B533-V is leading here. The Shuttle on the P4X266A falls into the last position. 

In the SYSmark 2002 the gap is already 5%, with the EPoX and ASUS scoring the best marks. 

We have taken two tests from the SPECviewperf packet where the difference is greater. However, it's only 3%, that is why the victory of the ASUS's solution is quite formal. The VIA's board outshone all i845G models at the expense of different memory controllers and/or operation with AGP. 

The results of the 3DMark run in high resolutions revealed no difference. The same situation occurs in the games, where a speed in low resolutions could have shown a difference in operation with memory - but it didn't do that: we again have 3-4%, and the ASUSTeK's board turns out to be a leader again. 

Conclusion

But its performance is not much higher as compared with the other contestants, that is why you can save on it going with a bit less efficient solution which is also much cheaper. You can choose an appropriate system having looked through brief characteristics of the boards and also their descriptions. 
 
 
 
 

Article navigation: