P4 Powersupplies

[Floppyman]

Hi all,

Is it possible to buy to buy an adapter for a standard powersupply so it will work with a P4, or do you need to have a special P4 powersupply? Thanks in Advance.

[HAL9000]

You have to research the board that you intend to use. Some can use a standard power supply and will just take an extra power connector from the power supply for the added juice (these are typically, but not always, the SDRAM and DDR equiped boards). Others require the special power supply with an additional 12v connector and another that may look familiar to half of the old AT power supply connectors (Typically, but again, not always, the RDRAM boards). No, there is no adapter.

[Toaster]

Hello Flopster,
Such power adaptors exist but are not recommended.
The P4 (RDRAM and many others system boards) requires ample 12 volt power that Intel decided required more power then the traces on a system board could handle reliably. Using the appropriate supply, the needed 12volt power is available directly to the system board and close to the regulators.
I mesured 3.5 amps drawn via the 4 pin molex connector for the P4 on the recent build I had just done. Otherwise, this power would have to be derived from a device power connector via an adaptor which you are considering.
Many power supplies have minimal wiring for such a connection at being about 20-18 guage wire. Now you have 2 possible connections to become problematic as opposed to one.
I have seen users building P4s without using the additional connector and simply relying on the system board traces to route 12volt power to the system board regulators (VRM). I recommend against this practice as the traces arn't designed for this kind of duty.

[HAL9000]

Whoops, OK, I've never seen, nor heard of one, but who would know better than the power supply guru

[Floppyman]

Hi guys,

Thanks for the replies. The board I'm talking about is the Asus P4T-E. My powersupply has that connector that looks like half of an AT powersupply motherboard connector, would this work with that board? The powersupply itself is a 300 Watt inside a Supermicro SC-760A case. Thanks again.

[HAL9000]

I've heard some people run without the 12v connector, but the way I look at it is if you're going to do a job, do it right, which will mean getting the proper power supply for that ASUS board.

[Floppyman]

Thanks Hal,

Question is, does the Asus board I mentioned use that connector my powersupply has, because if it does I wouldn't have to buy a new powersupply to use with a P4. Thanks again.

[Toaster]

Hello again Flopster,
The ASUS P4T-E (the system board I used is an EE but I suggest the E is the same, correct me if I'm wrong) uses a molex style 4 pin connector for the CPU VRM. This connector is a squareish, 4 cavity "female" connector. There may be variants so look closely on the system board.
The system board I used did NOT use the auxilliary "AT" style connector so it was unused. The auxilliary connector is +5volts and +3,3volts I believe.(I didn't use it so I'm not 100% sure)
The Intel specific P4 boards use the "aux" connector (some). Most Intel specific boards use the 4 pin molex connector to supply +12volts to the VRM.
It seems other system boards are using variations of the VRM connector and I have seem implementations of a simple "device" power connector on the system board for VRM power.
I had considered "converting" an ATX supply (Aztec) for P4 use as the molex connector simply supplies 12volts and ground. I didn't have a per spec supply on hand and thus the reason I considered a conversion. Alas a supply was made available and the conversion wasn't necessary. However, I am considering another P4 using the system board you mention and the "typical" cases use power supplies I don't much care for.
While an adaptor might be stable if of desent quality, many power supplies are now built to P4 spec. I contacted Aztec about P4 specific supplies but as yet no reply. I'm hoping to locate a retail location as Aztec supplies are mostly OEM.

[Floppyman]

Hi Toaster,

So you suggest I should probably keep my eyes open for a P4 powersupply then, correct? Funny thing is, this case I have came in two versions, a P4 ready one, and another with a regular P/S. I bought the one with the regular P/S. Maybe I should email supermicro to see if there is an adapter available or if there's another way for it to supply support a P4.

[Toaster]

Hello Flopster,
Yup, I'd opt for the correct supply were it me.
The case you plan to use is designed for the P4?
The case system board floor has the 4 mounting holes for the heatsink and the like. I found the heatsink/fan to be heavy enough that these holes should be used for the added support.
Look at it this way Flopster, the P4 is hard to "break" and an adaptor might be okay for the short term but I wouldn't do so. The amount of current is high enough that you could have a voltage drop sufficient enough to cause problems.
Would they exchange the supply for a modest fee? At any rate, I'm sure you could make use of a handy dandy supply kicking around...yes?
One minor note, that particular system board supports memory interleaving.
Install your memory in alike sets of 4 for optimum performance.
You won't need your CRIMMs but don't toss 'em. Obviously, make sure your board supports the memory types and number of modules. There are some boards that the CRIMMs are mandatory but this may be only to Intel specific boards.

[HAL9000]

That's interesting Toaster. I'm looking at upgrading to the P4T-E board in the future, and if I understood you right, if I want 512Mb of RAM, I'm better off installing 4 x 128Mb modules instead of 2 x 256Mb modules and the CRIMM's?

[Toaster]

Howdy again,
Me and a buddy when we built with the ASUS board, we tried several memory configs.
The local computer stores here were very co-operative as I assisted them on numerous occaision. They were very enthusiastic about the P4 and its capabilities.
They all scampered to supply most anything we desired for testing purposes and simply asked for the return of unused gear and a full "report".
Me and my buddy both knew about memory interleaving but were unsure to what extent performance would increase after enabling the feature.
We tried 512MB in a dual DIMM and quad DIMM configs. The dual config was 2 256MB RDRAM-800 modules and the quad was 4 128MB RDRAM-800 modules. (we returned their modules after the experiment and bought PC-1200 modules )
The dual set-up ran fine but the memory moves were a full 25% below that of the quad config. When we installed our PC1200 RDRAM modules, the percentage jumped to over 30%. This we are still investigating and with what data we have suggests lower memory latencies.
We then removed 2 of the PC1200 modules and installed the CRIMMs. Performance dropped by about 25% as far as memory moves.
The result would compare such like a BX class board running at CAS2/133mhz and CAS/3 at 100mhz. Quite a difference IMHO and well worth it for the intial build.
While reconfiguring would't be worthwhile to many, if one is buying the goodies to build then by all means buy identical pairs.
Micron was my choice for memory as they would match the set of 4.
However, after "reporting" back to the shops, they now will match modules.
It seems that all 850 class RDRAM board support this feature to some extent as we also tried a true Intel board. The Intel board was an ultra stable system board but lacked overclocking abilities.
Over the coming week or so, the system will be entrusted to me for safekeeping.
During this time, I am going to explore the registers of both the chipset and the P4. There may be some changes instore if my suspicions pan out.
It seems the ASUS board is tuned for SCSI as well for reasons I'm unsure about as yet.
Rumors suggest a PCI64 board by mid 2002.
So far as the P4 CPU, the 473pin 1.6-1.8 appear (unproven yet) to be the "sweet spot" .
Again, Intel is using SMD resistor paks on the reverse of the CPU for what seems to be configuration. I'll be exploring this when the cost of the CPUs plummit in the coming 2 months when I feel it might be worth "popping" a CPU. (there seems to be more risk here then with the PIII/Cel-mine) More on that as it becomes available.

[HAL9000]

Thanks Toaster. When I spring for my upgrade, I'll probably go with the 4 module set then as I'm not really worried about filling my RAM slots, 512Mb should be plenty.

[Toaster]

Woops, a disclaimer might be in order here.
Be advised that we used the ASUS P4T-EE board. This is an "eval" board.
I suspect the "E" class board would perform about the same. The "EE" board had additional BIOS enhancements for overclocking. Unfortunately, an "E" class board was not available at the time of our build. I removed the heatsink/fan from the chipset to see "EVAL". I suggest the chipset to be a minor mod as the chipset driver was the same as for the "E" class board. The boards look identical however in almost every respect. I suspect that the chipset was "optimized" but I cannot confirm this until I explore the registers. I expect to see memory optimization and register configs. These are my suspicions but as yet unfounded.
Give me about a week or so for the full "poop".

[Floppyman]

Thanks for all the info guys. I had this board in mind for my next upgrade, and this info already proven to be quite interesting.

[HAL9000]

Toaster, I e-mailed my ASUS rep and he tells me that the P4T-EE board is an evaluation board as you mentioned, that is all the second "E" stands for and that it is in fact the same board. He figures that you should have had the same overclocking results had you had the P4T-E board available to you.

[Toaster]

Howdy Hal,
Did SAUS go into any detail about the "EE" board?
From the register poking I've done so far, it suggests a "pro" board.
The BIOS date of the board is 12/06/01 so this suggests a revision beyond the "E" board rather then prior. We had thought that "maybe" this board was an eval of the "E" board (we were a tad worried about this) but the BIOS date and system board julian date suggest otherwise as well as register diffs.
We recieved a "software" manual (PDF) and a bulk packed system board.
It's interesting that your rep suggested an eval board which suggests production has probably begun.

[HAL9000]

No, he didn't have too much detail other than "To the best of his knowledge" that "that is what the second E usually designates".

[Toaster]

Howdy Hal,
After a bit more poking, I found a date within the chipset that suggests a revision of some sort. The chipset date is 11/13/01.
I'm still poking and pushing and Iv'e found other registers that I can not as yet identify fully. I have found an "Interleaving" option thats currently disabled that suggests the system board can utilize faster memory.
I also was able to disable the EIDE ports fully. (this had GREAT pucker value because I thoght I boogered something with a bad poke).
If the ports are disabled via BIOS, the ports still existed within BSD but were considered "disabled". When i poked the register and shut down the ports, BSD did not "see" the ports at all. This option isn't available via BIOS but seems handy for recovering IRQs because the system now can utilize these resources.
Odd...still more stuff hidden within...

[HAL9000]

You never cease to amaze me Toasty.

[Werewolfdaddy]

Everyone on this particular thread never ceases to amaze me, Especially Hal. DVD...

[Redo40]

What do you all think of the ASUS P4B266 mobo? It has a EZ Plug that lets you use your existing power supply instead of buying the ATX 12V. Of course it is based on the 845 chipset and DDR memory, so it would be useless if you wanted RDRAM.