Part the fourth. The basic mistakes and recommendations.

This part has appeared essentially after first three and represents already concrete recommendations, which are based on year statistics of operation altered under Tualatin systems and experience of remaking not one hundreds (without exaggeration) motherboards / adapters.

First, the first, that would be desirable to say - now all previous theoretical calculations on "marketing" making evolution Socket370 completely are confirmed with numerous and long experience. All - is perfect worked, any real justifications of migration PPGA-> FCPGA-> FCPGA2 no and was not, all " basic changes " are result not flight of engineering idea, and just its "tamings". But here silly to be indignant - business is business.

Now the second, on "correctness" of remaking. In connection with change of quotations on SDRAM-and DDR-memory (when SDRAM became more dear) "advantage" of remaking under Tualatin has essentially decreased. However even in such situation she is meaningful is more often, since is thought over the adjusted system with Tualatin, dispersed up to 1.5-1.7GHz will not concede modern 1.7-2.1GHz-system (P4/Athlon, in case Celeron-478 - it be it is better). In general, personally (for myself) at comparison(on perfomance) with P4/Athlon-systems use the rating system similar AMD Athlon XP:

CelFreq+=1000 + (RealFreq - 1000)*1.5.

Accordingly, Cel-Tualatin, dispersed up to 1.6GHz:

CelFreq+=1000 + (1600-1000)*1.5=1.9GHz,

I.e. it can be positioned safely as real competitor P4-1.9GHz and AthlonXP-1900+. It is confident, that many will not agree with such approach is their problems (opinion) for my practice just and has led to to an empirical conclusion of the similar formula.

" Death after remaking ".

And now not about the most cheerful - about failure of the remaked systems. At me personally was not still while any (!) case when it would be possible to recognize fairly - has died (Tualatin) because of "remaking" (under Tualatin). Even at operation than them on 1.8V and is higher. Such systems because of other reason (directly connected with remaking) die: high frequency processors especially overclocked, especially with a raising of a voltage - "eat" much. Old power supply (and computer and on a parent motherboards) simply do not maintain and fail, creating thus impression of "abnormality" of remaking (and-or " impossibility of correctness " her in general). Therefore in case of desire to alter under Tualatin the next motherboards - do not forget will take care of a qualitative power.

First, power supply (computer) STRONGLY RECOMMENDED to replace o qualitative (in fact all the same it should be made later). Qualitative, quantity of Watt at anything (though it is recommended - 300Wt). Statements such as "... Yes five years worked property and no any problems " are not quoted, for I shall repeat the the same phrase - " eventually power units do not become better... " Constant practice of dying power supply in five-second-half-year in remaked under Tualatin to system tragically and regularly confirms it.

Second, the power supply (the processor, established on motherboard) too it is necessary TO SUPERVISE a power unit strictly. Since he usually has not been designed on such high currents that results to serious heat-up power transistors and filtering condensers. By the way, as a rule capacity "processor" power supply usually is the deterrent at dispersal.

In any case, do not forget "to feel" a power unit of the processor (if it is strongly heated - will take care of additional cooling), closely{attentively} to check up condensers on "pregnancy" (especially for motherboards from Abit, Acorp), at detection of those at once to replace (and for Abit, Acorp to supervise further).

If " money on new power unit no, and it would be desirable Tualatin" - take / remake on health, but in that case - the following recommendations:

1. Try to refuse overclock and to take the processor ~1.2GHz.

2. VERY MUCH an effective measure - downturn of a voltage of a power. Many processors fine function on native frequency at Vcc=1.3V. At such approach loading on power supply is very strongly reduced and a computer and the processor plus appears an opportunity to make silent computer - processor in such status is heated as somewhere 800MHz.

This point is especially basic for AT motherboards since it is usually "budgetary" models that is why on them, as a rule, stands the very low power of power unit (processor).

The separate recommendation for motherboards without support Vcc <1.8V. If at you power unit is AT or "JNC-shaped" (i.e. poor-quality, that it is possible to check up simply having weighed: qualitative weighs ~2kg, bad - ~1kg and it is less) ATX - do not risk to put on such system Tualatin, be limited 700-900MHz a Coppermine a maximum.

As be attentive happy owners of very much widespread payments Chaintech 6BTM - absolutely "gloomy" there stands power unit the processor.

The basic mistakes of remaking (under Tualatin).

The most important and most widespread mistake in a case non start the remakeded system - bad isolation AK4 (from the ground). Necessarily check it after remaking.

By the way, strongly recommended to get there (on AK4) Vtt (through the resistor 0.1-0.33kOm), instead of POWERGD (AK26). If isolation AK4 will be provided, but Vtt him "not will reach" (breakage, bad contact, etc.) the computer as a rule behaves very typically - is started, but hangs under loading (can hang only at start any applications / games, and can and not give to be loaded Windows).

If legs were broken out - check up, that the rests not shorted Therefore, besides, I personally recommend them unsolded - aesthetically and reliably. To do it it is necessary powerful enough, sharpened soldering iron, very warming up contact on the part of the soldering and " exhausting it inside " strong pressing, hooking thus from the return side any hook. By the way, that contact to extend, it can even be soldered simply accurately (to a copper wire).

Further, do not forget controlled contacts G35-G37, they should be necessarily short-circuited. In majority FCPGA of adapters so it also is, but non in all - check!
Necessarily supervise a voltage of a nucleus - very much often it turns out 1.3V (why - look in 3-rd part). Also do not overlook to ground VID4 (in some adapters / motherboards on the ground initially - check up) - I think, Vcc=2.3V to you to anything (why 2.3V - look in 3-rd part).

On a voltage. In some cases with it there are "glitches" - one of VID's neither from that nor from this can be exposed in "0" (though face value should be "1"). Similar it is noticed on Acorp-6VIA85P with VID3, that results to overestimated on 0.4V to a voltage. To solve it it is possible differently, but it is the easiest unsolde I correspond a leg directly on the PWM-CONTROLLER (RC5054/HIP6004BCB for this motherboard, VID3 is a leg 7).

Contact Y33 to touch usually does not. However in case of a Slot1-motherboard from Abit/Gigabyte is will be basic, without they be precisely not start. At what it is noticed, that for Abit the institution on this (Y33) Vtt through the resistor ~1-10kOm (similarly AK4) can suffice, and for Gigabyte, as a rule, it is necessary to make a "fair" divider (but it is not necessary from 3.3V or 2.5V, it is possible and from Vtt - resistors with the attitude ~ 1 to 6).

In case at you the interesting phenomenon of start of the processor on the lowered voltage is observed and non start on "native" - isolate mysterious contact Detect - AF36 (why - look in 3-rd part). It is noticed on some Slot1-motherboards (usually from Asus).

At remaking PPGA of adapters special attention give contact Z36: +2.5V for PPGA and RSVD for Tualatin. In some cases his removal to not do without. It is noticed on adapters MSI MS-6905 of the first versions (do not try to cut simply a thick track the leader to it - contact inside a layer, it is necessary to unsolded/break out).

In general, in case of remaking PPGA - be very vigilant, their remaking is very difficult (why 2.3V - look in 3-rd part). Necessarily supervise size Vtt voltage if instead of put 1.5V there 2-2.5V (it is usual ~2.1V) - check up, whether there is no contact of Vtt-legs with +2.5V (Z36, AB36).

BIOS remaking.

Tualatin (for example, almost all Gigabyte Slot1), the overwhelming majority simply incorrectly defines{determines} them, has no support of microcodes, etc. For elimination of it (in case AwardBIOS, for AMI yet has not made) it is possible to use program BIOS Patcher (except for addition of support Tualatin still more many that makes).

...

Article during spelling / updating.

...

Test laboratory ROM.by,
With Aleksejchik Vadim's help and Artiom's Borisionok,
With use of the numerous information from C.A.R.C.A.S.S.

Sevko Roman, 7.5.3.

Translate by victor

The used materials:

Datasheets:

Pentium® III Processor with 512KB L2 Cache at 1.13GHz

VRM 8.5 DC–DC Converter Design Guidelines

VRM 8.4 DC–DC Converter Design Guidelines

Pentium® II Processor Developer’s Manual

Intel® Celeron® Processor for the PGA370 Socket up to 1.40 GHz on 0.13 Micron Process

Intel® Celeron™ Processor up to 1.10 GHz

Intel® 815E Chipset Platform For Use with Universal Socket 370 Design Guide

Intel® 815 Chipset Platform Design Guide

...and many more another..

Part 1. Part 2. Part 3.