I was reading over the comments in the Soldam cube news article that Damage posted Saturday. Gignic, in post number nineteen, made reference to the fact VIAHardware.com had a similar story running. It being a slow Sunday, I decided to peruse what news Jasper and crew were reporting. Among the stories was a link to an article at Van's Hardware about the Intel 850 and 845 MCH chips having severe compatibility problems with AGP cards. These difficulties apparently were so severe as to result in the electrical death of the motherboard.
Van's Hardware, after some basic testing, came to the conclusion that Pentium 4 enthusiasts should look to other manufacturers for a reliable core logic chipset. I was very skeptical of this conclusion. After an update by VIA Hardware.com stated that the problem was a voltage specification issue, I decided to start digging into the Intel documentation.
I started with the specification updates for both the 845 and 850 MCH chips. If this issue really were a bug, those documents should specify it and detail why it happens and what Intel is going to do to fix it. Neither document contained any information about electrical difficulties with certain AGP cards. So I then moved to the design implementation guide and the design thermal and electrical implementation guide. I found the information I was hunting for in the design implementation guide, page twenty-one, section 220.127.116.11. There's a bulleted list of the features of the AGP implementation. Bullet number three reads:
AGP 1.5 V Connector support only. No support for 3.3 V or Universal AGP connectors.It doesn't get much clearer than that. The MCH supports only 1.5v AGP cards. Period. A follow-up article on Van's Hardware notes the above design limitation. The author's initial diffculties seem to have been related to some other failing, as he notes:
The only graphics boards on which we have seen a problem with i845 and i850 motherboards are some old SIS based graphic cards, on which the Vref generation signal for 1.5 volt was not properly implemented. It is not due to the SIS graphics chipset, but to a mistake in the board design made by some boards manufacturers, so it doesn't mean one must ban SIS based graphics boards, but only some improper boards manufacturing.This issue is only going to affect a handful of improperly designed cards. There will, of course, also be older AGP cards, like TNT or 3dfx cards, whose age likely rule them out of being compatible.
I wondered: Is only Intel limiting the AGP choices now? MSI now has their nForce board specification up for view. If you scroll down the the slots catgeory a important note pops up that is impossible to miss: "Note: AGP 2.0 2x/4x (1.5V only)".
It seems this is just another instance of the industry moving forward.
|Lenovo ThinkCentre and ThinkPad machines pack AMD PRO APUs||14|
|Nvidia licenses Rambus' DPA tech for side-channel data leak prevention||4|
|iOS 10.1 update includes portrait mode beta for iPhone 7 Plus||2|
|Biostar belatedly announces GTX 1060 graphics cards||12|
|HyperX Alloy keyboard gets lean and mean for FPS gaming||7|
|AMD drops prices on the Radeon RX 460 and RX 470||50|
|Reports: Radeon RX 470D is a budget Polaris card for China||9|
|Examining reports of slow write speeds on the 32GB iPhone 7||33|
|Cellular Insights dissects iPhone 7 Plus modem performance||11|
|A real "console monitor" would be 720p @ 30 Hz ;P||+63|