What Makes a PSU SLI Compatible? Got Answers Here!

[GuitarDaddy]

I have read this entire thread and all the linked information, and I'm still not convinced. Until someone comes up with a dual rail PS with 25a on one of the rails, I won't be trusting my SLI rig to a dual rail PS.

Although Ive read tons of technical mumbo jumbo, the issue is pretty simple. Can the amps on two or more seperate rails be added together to meet the requirements of the motherboard? I say No, and I haven't seen any convincing arquments to the contrary. The concept of the power rails "sharing" amps doesn't make sense to me, How is a rail with 18a going to share amps with another rail with 18a?

Ignoring the technical aspects for a moment, and focusing on first hand experiences. I post regularly here at Anandtech and on three other forums, and I have been watching postings closely regarding the A8N-SLI since I received mine 12/13/04 (it has become somewhat of an obssesion according to my wife.).

And what I have seen is a high percentage of the people having power related issues on this board are using dual rail PS's. And I have read several posts of people who RMA'd their dual rail, replaced it with a strong single rail and their problems went away.

And if your waiting on "Professional Reviewers" to do an in depth analysis of single rail versus dual rail on SLI boards, don't hold your breath. These guys are not much more than paid advertizers and they will steer clear of a sticky issue like this that could jepordize their relationship with any of the major manufacturers.

 

[Algere]

Although Ive read tons of technical mumbo jumbo, the issue is pretty simple. Can the amps on two or more seperate rails be added together to meet the requirements of the motherboard? I say No, and I haven't seen any convincing arquments to the contrary.

Think of it this way, if you have a dual rail PSU with one rail that is dedicated to powering the motherboard & another rail that powers everything else. You pretty much have 18A more or less usable to the motherboard & to the motherboard alone. The reason you may believe 18A is not enough to power a motherboard may be because other (older perhaps) dual rail implementations would have one rail dedicated not only to the motherboard but to alot of other components as well. Depends on the configuration of a dual rail PSU.

The concept of the power rails "sharing" amps doesn't make sense to me, How is a rail with 18a going to share amps with another rail with 18a?

The 18A/18A rating e.g. for a shared amp PSU isn't what they have, it's what they can have @ most/max.


Here's an analogy:

You have two 18 ounce pitchers but only 35 ounces of lemonade to pour into those pitchers. Now one pitcher belongs to Bobbie, Cindy, & Greg while the other pitcher belongs to Albert, Sandy, & Mark. Now imagine each pitcher gets filled with 17.5 ounces of lemonade each (35 ounces total). Let's say Albert is a greedy sneak & steals half an ounce from Bobbie, Cindy, & Greg's pitcher & puts it into his. One pitcher now has 17 ounces while the other has 18 ounces.


Now replace...

Pitcher = Rails
Ounce = Amps
Lemonade = Available amps
Bobbie, Cindy, & Greg = CPU, motherboard, & SATA
Albert, Sandy, & Mark = PCIe, 4-pin drives, etc.

And what I have seen is a high percentage of the people having power related issues on this board are using dual rail PS's.

Most likely older dual rail PSUs designs like the one implemented into Antec's Neopower & if not, user error & such.

And if your waiting on "Professional Reviewers" to do an in depth analysis of single rail versus dual rail on SLI boards, don't hold your breath. These guys are not much more than paid advertizers and they will steer clear of a sticky issue like this that could jepordize their relationship with any of the major manufacturers.

I usually put my faith in general consensus & in belief that not all reviewers are in for the take. If that's foolish, then foolish I am .

 

[Insomniak]

Originally posted by: Algere

Here's an analogy:

You have two 18 ounce pitchers but only 35 ounces of lemonade to pour into those pitchers. Now one pitcher belongs to Bobbie, Cindy, & Greg while the other pitcher belongs to Albert, Sandy, & Mark. Now imagine each pitcher gets filled with 17.5 ounces of lemonade each (35 ounces total). Let's say Albert is a greedy sneak & steals half an ounce from Bobbie, Cindy, & Greg's pitcher & puts it into his. One pitcher now has 17 ounces while the other has 18 ounces.


Now replace...

Pitcher = Rails
Ounce = Amps
Lemonade = Available amps
Bobbie, Cindy, & Greg = CPU, motherboard, & SATA
Albert, Sandy, & Mark = PCIe, 4-pin drives, etc.

The problem with this analogy is it doesn't match real world use. It's more like 2 people on one pitcher, and 5 on the other - and the pitchers are equally filled. One side is going to come up short.

Even if you dedicate a single 12v rail to the Mobo/GPU, and power everything else with the other rail. would you really trust multiple HDDs, ODDs, Fans, and whatever else to a 12v rail with only 18A? I wouldn't.

A dual rail PSU can only work, as I see it, if the COMBINED DRAW of all the devices on EACH RAIL does not exceed that rail's max amperage. In other words, you have to split your people equally between pitchers OR ELSE.

Is it possible? Yes. Is it convenient? No. Does it leave you as much room for error as a single rail PSU with a strong (26+) amperage? No.

In light of this, is there a compelling reason to use dual rail PSUs? Not that I can see, but to each his own. When they get to the point where they have 24+ amps on each 12v rail, they'll probably be able to cut it - until then, I think everyone is better off sticking with a single, beefy 12v rail.

 

[ImJacksAmygdala]

This is the first hardwarw review site I have seen that even talks about the PSU in relation to using SLI.

http://www.hardwareanalysis.com/content/topic/38623/

The A8N-SLI is rather picky about what powersupply is used with it, and that?s partly due to the fact it is designed as an ATX 12V v2.0 motherboard, featuring a 24-pin power connecter and dual 12-volt rails coming from the powersupply. Although this recommendation goes for all systems, the A8N-SLI demands you have a good powersupply, the minimum requirement is 400-watts. It doesn?t have to be an ATX 12V v2.0 powersupply per se, nor feature a 24-pin plug, but it needs to be able to supply at least 18-amps of current or 200-watts on the 12-volt rail. Of course there?s a difference between running a single PCIe video card with this motherboard and running two in SLI. Up until two GeForce 6600GTs you can get by with the following powersupplies:

Power supplies tested with a single GeForce 6x00/GT/Ultra or up to 6600GT SLI

- Zalman ZM400B-APS, 400-watts
- Antec TruePower 430, 430-watts
- Tagan TG480-U01, 480-watts
- Enermax Coolergiant EG435AX, 430-watts
- Antec NeoPower 480, 480-watts

Power supplies tested with GeForce 6800GT and 6800 Ultra in SLI

- Tagan TG480-U01
- Antec NeoPower 480
- Fortron Source FSP550-60PLN*
- Antec TruePower 550 EPS12V*

The power supplies designated with a * are needed if you want to run the A8N-SLI with an Athlon 64 FX 53 or 55 processor, two GeForce 6800 Ultras and two or more high capacity Serial-Ata drives. However they need to be modified in order to convert the 8-pin 12-volt connector, coming off of the 2nd 12-volt rail to two 6-pin PCIe power connectors, this is a simple procedure that means extracting the pins from the original connector and splitting them out to two 6-pin PCIe connectors. These will be connected to the GeForce 6800 Ultras directly, hence powering them off of their own 12-volt rail within the power supply. From our testing this is the only way to guarantee absolute stability. Take notice that when running SLI the EZ-Plug must also be connected to guarantee stable operation.

Its interesting to note that Sander Sassen, CEO of Hardware Analysis, says that the minimum requirement for the PSU in SLI needs to be able to supply at least 18-amps of current or 200-watts on the 12-volt rail. All of the power supplies I have seen with dual 12volt rails that advertise "Supports SLI" only meet the minimum at 18A on both 12volt rails. We now know that the rail1 and rail2 current is not additive no matter what resellers advertise so if this is true I'd rather just go with a high current single 12volt rail PSU with enough power to meet my overall system requirements.

Thats a nice Asus A8N-SLI FAQ and I can't wait for them to add to it, but we really need a review dedicated to this topic. I'm emailing sites and dropping hints in forums. Are you?

Xbit labs just released a PSU review and it totally skated around this topic just to save you 12 pages of reading.

 

[Insomniak]

It says in the A8N-SLI Deluxe manual that it needs a minimum of 18A on the 12v rail. How all these people have been missing it is beyond me - although I think most people probably think they can add the amps of the two rails together - not the case.

 

[ImJacksAmygdala]

There in lies the problem Insomniak. You get a cookie!

Another shinning example of why we need to have a proper SLI PSU review. I think you hit the nail on the head.

 

[FastEddie]

Originally posted by: ImJacksAmygdala
This is the first hardwarw review site I have seen that even talks about the PSU in relation to using SLI.

http://www.hardwareanalysis.com/content/topic/38623/

The A8N-SLI is rather picky about what powersupply is used with it, and that?s partly due to the fact it is designed as an ATX 12V v2.0 motherboard, featuring a 24-pin power connecter and dual 12-volt rails coming from the powersupply. Although this recommendation goes for all systems, the A8N-SLI demands you have a good powersupply, the minimum requirement is 400-watts. It doesn?t have to be an ATX 12V v2.0 powersupply per se, nor feature a 24-pin plug, but it needs to be able to supply at least 18-amps of current or 200-watts on the 12-volt rail. Of course there?s a difference between running a single PCIe video card with this motherboard and running two in SLI. Up until two GeForce 6600GTs you can get by with the following powersupplies:

Power supplies tested with a single GeForce 6x00/GT/Ultra or up to 6600GT SLI

- Zalman ZM400B-APS, 400-watts
- Antec TruePower 430, 430-watts
- Tagan TG480-U01, 480-watts
- Enermax Coolergiant EG435AX, 430-watts
- Antec NeoPower 480, 480-watts

Power supplies tested with GeForce 6800GT and 6800 Ultra in SLI

- Tagan TG480-U01
- Antec NeoPower 480
- Fortron Source FSP550-60PLN*
- Antec TruePower 550 EPS12V*

The power supplies designated with a * are needed if you want to run the A8N-SLI with an Athlon 64 FX 53 or 55 processor, two GeForce 6800 Ultras and two or more high capacity Serial-Ata drives. However they need to be modified in order to convert the 8-pin 12-volt connector, coming off of the 2nd 12-volt rail to two 6-pin PCIe power connectors, this is a simple procedure that means extracting the pins from the original connector and splitting them out to two 6-pin PCIe connectors. These will be connected to the GeForce 6800 Ultras directly, hence powering them off of their own 12-volt rail within the power supply. From our testing this is the only way to guarantee absolute stability. Take notice that when running SLI the EZ-Plug must also be connected to guarantee stable operation.

Its interesting to note that Sander Sassen, CEO of Hardware Analysis, says that the minimum requirement for the PSU in SLI needs to be able to supply at least 18-amps of current or 200-watts on the 12-volt rail. All of the power supplies I have seen with dual 12volt rails that advertise "Supports SLI" only meet the minimum at 18A on both 12volt rails. We now know that the rail1 and rail2 current is not additive no matter what resellers advertise so if this is true I'd rather just go with a high current single 12volt rail PSU with enough power to meet my overall system requirements.

Thats a nice Asus A8N-SLI FAQ and I can't wait for them to add to it, but we really need a review dedicated to this topic. I'm emailing sites and dropping hints in forums. Are you?

Xbit labs just released a PSU review and it totally skated around this topic just to save you 12 pages of reading.

Oh, we have a two hundred post thread regarding what Sander Sassen, CEO of Hardware Analysis, says.

:laugh:

 

[Algere]

The problem with this analogy is it doesn't match real world use. It's more like 2 people on one pitcher, and 5 on the other - and the pitchers are equally filled. One side is going to come up short.

The analogy was a specific example of how amp sharing works, not actual usage. In real world usage, the standard SLI rig wouldn't suck bone dry the amps available on the +12V rails of the 600W Enermax - which the analogy was modeled after. If I intended to describe power consumption under today's real world conditions within that analogy I could've said that Bobbie, Cindy, & Greg (CPU, MB, & SATA) and/or Albert, Sandy, & Mark (PCIe, 4-pin drives, etc.) were full & couldn't drink all of the lemonade.

If you do however wanna go into real world usage then ok. In a previous post, you said the A8N-SLI needs @ least 15A.

if any of your devices need more than 15A, like an A8N-SLI Deluxe, then you're boned.

Jdogg's experience is one example of a dual rail PSU powering an A8N-SLI with 2 6800 GT's. Since Jdogg is using a 600W dual rail Enermax, we know that it has 18A max on the rail. That one rail is in charge of powering up the MB, CPU, & SATA drives per specs. Right there we can get an idea of the power distribution from that one rail. So we have 18A available, if the A8N-SLI does indeed need 15A for the motherboard itself as you've said, that would only leave 3A or even less if SATA drives are plugged into the PSU's native SATA connectors. 3A sounds short for a current generation CPU.

Now if the power consumption of an A64 is indeed around 7A as I've heard (speculation?). That would mean on a 18A rail, the A8N-SLI doesn't have 15A to use but around 11A or even less with SATA drives attached. So does your power consumption figure of the A8N-SLI sound realistic, not really IMO. But hey! I could be surprised & find out that an Athlon 64 really does consume only around 3A of 12V power.

Even if you dedicate a single 12v rail to the Mobo/GPU, and power everything else with the other rail. would you really trust multiple HDDs, ODDs, Fans, and whatever else to a 12v rail with only 18A? I wouldn't.

Perhaps you overestimate the power consumption of all of those components.

Maximum power consumption AFAIK (not typical)
WD Cavier SE (See electrical) = 530mA
92mm medium speed Panaflo = 150mA
NEC ND-1300A DVD-RW = 1150mA or if I'm correct 1.15A

4 x WD Caviar SE = 2.12A
6 x 92mm medium speed Panaflo = 900mA or 0.9A
3 x NEC ND-1300A = 3.45A

Total = 6.47A

That leaves around 11A left for other things.

In light of this, is there a compelling reason to use dual rail PSUs? Not that I can see, but to each his own.

I could see a possible reason one might want a dual rail PSU. Cleaner power to components that are otherwise connected to other "noisy" components like on a single rail PSU. Wouldn't someone who's overclocking want to minimize noise to other components to achieve a higher OC (e.g., OC'ing a video card while not worrying about noise from a OC'd CPU affecting stability)? Also although I can't prove or disprove due to experience, could it be possible that with a cleaner source of power being supplied, one could expect components to last longer instead of dying prematurely?

A dual rail PSU can only work, as I see it, if the COMBINED DRAW of all the devices on EACH RAIL does not exceed that rail's max amperage. In other words, you have to split your people equally between pitchers OR ELSE.

Is it possible? Yes. Is it convenient? No.

Hence shared amps. Shared amps however is more beneficial to lower wattage PSUs (450-500W) than higher wattage PSUs since they can maneuver/balance the shared amps more freely between the 2 rails. Although that doesn't do much for higher wattage PSU's like a 600W Enermax, there's also the addition of improved distribution of power to components due to rearrangement from previous models. If you look @ the differences between the dual rail Enermax 470W & the 485W version - 485W model replaced 470W model. You'd notice that the configuration of the 470W model is exactly like that of the Neopower. One rail powers CPU, other rail powers everything else that's +12V dependant. While the 485W model follows that of the 600W version (specs above).

 

[shira]

Originally posted by: Insomniak

Originally posted by: McArra
TT purepower 680W is also supposed to be for SLI setups and has 3 12V rails, 2 of them with 15A and another one with 8A, for a total of 38A. This is the PSU I will receive next week, is supposed to be really really powerful. I'm not using SLI though, I'm an AGP owner.

Amps still aren't additive - if any of your devices need more than 15A, like an A8N-SLI Deluxe, then you're boned.

The basic problem is that if you have THREE power-hungry components (A. Video Card 1 B. Video Card 2 C. Everything else), then on a 2-rail PSU one of the rails has to power 2 of these 3 things. So if the PSU's total power is split evenly or nearly evenly between the two rails (as seems to be the case with most of these 2-rail PSUs), you have problems. And you lose pretty much the only true benefit of separate rails: power isolation (since the power-draw variations of the video card on the common power circuit with "everything else" will affect those other components).

Really, if you want the advantage of a power isolation, you need a three-rail design: Two of the rails each need to be powerful enough to cover the needs of the hungriest video card (let's say 150W and 10A on each of rail, that ought to be "future proof"). And the third rail then needs sufficient power to easily drive the rest of the system (let's say 20A and 300W). Better yet would be an adjustment that allows you to set the power levels of each of the three 12V rails. Of course, MBs would need to be redesigned to accept three independent sources of power.

 

[Algere]

Originally posted by: ImJacksAmygdala
Its interesting to note that Sander Sassen, CEO of Hardware Analysis, says that the minimum requirement for the PSU in SLI needs to be able to supply at least 18-amps of current or 200-watts on the 12-volt rail. All of the power supplies I have seen with dual 12volt rails that advertise "Supports SLI" only meet the minimum at 18A on both 12volt rails.

From what I've absorbed unless I missed something, he said 18A. So as far as anyone knows, he just means 18A total from a PSU's entire output, not just one rail.

We now know that the rail1 and rail2 current is not additive no matter what resellers advertise

Wattage/Power is additive for PCIe video cards, if motherboard & PCIe 6-pin are on seperate rails. It's as simple as understanding that you get power from the PCIe 6-pin plug & PCIe slot.

Wattage/Power is the product of amp*voltage (current*amount).

 

[shira]

Originally posted by: Algere

Wattage/Power is additive for PCIe video cards, if motherboard & PCIe 6-pin are on seperate rails. It's as simple as understanding that you get power from the PCIe 6-pin plug & PCIe slot.

Wattage/Power is the product of amp*voltage (current*amount).

Power is NOT additive if what you mean is that amps from rail 1 can be shunted over to rail 2. If that were the case, there would be no point in having separate rails, since there would be no power isolation between the rails.

Now, are you saying that the two video cards are on the same rail, separate from the rest of the MB? My understanding is that only one of the video cards is on the second rail. If what you are saying is correct, then a 2-rail system could be very attractive, since it would provide true power isolation. But in order to work, one rail would need to be sufficiently powerful to completely meet the power needs of the two video cards. And the other rail would need to be sufficiently powerful to completely meet the needs of the rest of the system. Again, there's no power-shunting.

 

[Algere]

Power is NOT additive if what you mean is that amps from rail 1 can be shunted over to rail 2. If that were the case, there would be no point in having separate rails, since there would be no power isolation between the rails.

Dual rail was originally intended to seperate the CPU from everything else from what I've understand to date. Not the motherboard from drives & esp. PCIe power plug. For some reason (SLI?) PSU companies started to branch away from the original design which sprung new designs, e.g. triple rail for one among other changes. As for the question IDK exactly, shot in the dark but maybe perhaps the PCIe plug charges some capacitors while other capacitors are charged by the PCIe slot. If you thought I meant power from one rail joins another and then both combine go into the video card, that's not what I mean.

P.S. Could explain why there are 2 groups of capacitors on my Radeon 9800 instead of them being grouped altogether within a single area. One group of capacitors near the 4-pin molex connector, other group closer to the card's external metal bracket.

Now, are you saying that the two video cards are on the same rail, separate from the rest of the MB?

Under the old design they were on the same rail. To compromise (for SLI?) they reintegrated some components with the CPU - sacrificing some rail seperation but while keeping others intact. That's one design, triple rail likely seperated components even further.

My understanding is that only one of the video cards is on the second rail.

Curious where you got that from. Not saying your wrong cause these new "dual rails" seem different & unique between various PSU manufacturers. These new dual rails arent created under a unified standard, afterall if it was we wouldn't have one PSU manufacturer making triple rail PSUs while others are changing which devices are & aren't shared.

 

[ribbon13]

Multiple rails are a useless buzzword.

When I encounter multiple 12v rails I do this

Rail1 -- diode
Rail2 -- diode -- 1a 12v load resistor -- 500mf 12v cap -- feed ->
Rail3 -- diode /

Also for multiple power supplys.

 

[FastEddie]

Originally posted by: Algere

Curious where you got that from. Not saying your wrong cause these new "dual rails" seem different & unique between various PSU manufacturers. These new dual rails arent created under a unified standard, afterall if it was we wouldn't have one PSU manufacturer making triple rail PSUs while others are changing which devices are & aren't shared.

I did a quick post last night casue it was way past my bedtime. See I'm old.

What is curious is that new multi-rail psu's are popping up left and right without taking power consumption into consideration. That is the crux of my argument, where marketing is playing a bigger role than the electrical dynamics of the new SLI technology. A good source for power quidelines comes from AMD--->
Builders Guide for Opteron Server Platform
Here the use of multiple rail psu's (which is what they are designed for. ie. Multi-Processor Systems) clearly defines the power requirements in the manufacturer and use of dual rail psu's. However, the single processor A64 based system is a completely different animal, with completely different power needs.

This will give you an accurate assessment of power requirements in K8 A64 platforms---> Processor Power & Thermal Data Sheet

 

[ribbon13]

Excellent info man!

In server enviroments having multiple rails happens a lot more often because of multiple hotswap PSUs.

 

[Algere]

I did a quick post last night casue it was way past my bedtime. See I'm old.

Won't hold that against ya

What is curious is that new multi-rail psu's are popping up left and right without taking power consumption into consideration. That is the crux of my argument, where marketing is playing a bigger role than the electrical dynamics of the new SLI technology. A good source for power quidelines comes from AMD--->
Builders Guide for Opteron Server Platform
Here the use of multiple rail psu's (which is what they are designed for. ie. Multi-Processor Systems) clearly defines the power requirements in the manufacturer and use of dual rail psu's. However, the single processor A64 based system is a completely different animal, with completely different power needs.

Interesting info from what I can tell from my brief skim, I'll have to read a bit more into it later on. For now the question still remains, what PSU (if such exist) powers one gfx card on one rail & another gfx card on another.

This will give you an accurate assessment of power requirements in K8 A64 platforms---> Processor Power & Thermal Data Sheet

I wish that data sheet explained if all that 89W of power derived from the 12V line alone or a combination of 12V, 5V and/or 3.3V. AFAIK if it is powered by 12V only then that could explain the 7A figure I hear about (89W/12V = 7.42A).

 

[Algere]

Originally posted by: ribbon13
Excellent info man!

In server enviroments having multiple rails happens a lot more often because of multiple hotswap PSUs.

Wonder if that's the main difference between multiple rails on desktop & multiple rails for servers. One is meant to isolate noise shared between components while the other is meant for PSU swapping. Does that mean multiple rail PSUs for servers have double or more voltage lines, i.e. 2*12V,2*3.3V, etc. etc.? If so could it be used as a means of some kind of RAID 1 backup solution but for PSUs?

 

[ribbon13]

I mod every PSU i get so it feed into a custom power box I made. Diodes on all positive sides, feeing into 1a load resistors for each voltage, then a quarter farad stiffening cap for each voltage. works great. draws from all rails equally, and the stiffening cap makes all multiple-rail advantages moot.

 

[FastEddie]

Originally posted by: Algere

Interesting info from what I can tell from my brief skim, I'll have to read a bit more into it later on. For now the question still remains, what PSU (if such exist) powers one gfx card on one rail & another gfx card on another.

Therein lay the crux of the problem. Both PCI-E ports are being powered by the motherboard. Additional connectors to individual card(s) do an assist. They do not take over 100% of fullfilling the power reqirements to these cards for the motherboard. Where one dual rail psu seperates the rails via the 4-pin square 12v AUX connector (most use this method) as one rail, and all else use the second rail. That puts the burden on the same source, and gives you the result that one rail is getting hammered, with a constant demand at or near PEAK values for that rail. You can't convince me that such an intent was an engineered strategy, because psu's are not designed to run perpetually at PEAK.

This is in current terms. What happens when dual core becomes a functional reality on these boards. We already know dual core runs out of the box on the A8N SLI. What happens now? Just slam the other rail as well, until either the psu (and/or the attached components) takes a powder?

An engineered dual rail strategy would provide 24a or better on the main rail, and 12a on the processor/memory rail. Now you have a balance. The system would never run at a constant value so close to PEAK on either of the rails. Slap a couple of PCI-E power connectors to run with the main rail and YES by all means call this psu SLI Ready.

 

[shira]

Originally posted by: FastEddie

An engineered dual rail strategy would provide 24a or better on the main rail, and 12a on the processor/memory rail. Now you have a balance. The system would never run at a constant value so close to PEAK on either of the rails. Slap a couple of PCI-E power connectors to run with the main rail and YES by all means call this psu SLI Ready.

Taking this a step further: An even more useful implementation would be to allow the user to specify how he wanted some total Amperage specification to be split up between the main and CPU/memory rails. The GFU makers could assist in the set-up by documenting recommended amperage levels for their products.

 

[FastEddie]

Originally posted by: shira

Originally posted by: FastEddie

An engineered dual rail strategy would provide 24a or better on the main rail, and 12a on the processor/memory rail. Now you have a balance. The system would never run at a constant value so close to PEAK on either of the rails. Slap a couple of PCI-E power connectors to run with the main rail and YES by all means call this psu SLI Ready.

Taking this a step further: An even more useful implementation would be to allow the user to specify how he wanted some total Amperage specification to be split up between the main and CPU/memory rails. The GFU makers could assist in the set-up by documenting recommended amperage levels for their products.

I mod every PSU i get so it feed into a custom power box I made. Diodes on all positive sides, feeing into 1a load resistors for each voltage, then a quarter farad stiffening cap for each voltage. works great. draws from all rails equally, and the stiffening cap makes all multiple-rail advantages moot.

That's what ribbon13 is basically doing.

 

[GuitarDaddy]

I have worked with 2 guys on two seperate threads lately in the CPU/overclocking forum using the Enermax 600 with 18a and 17a rails. In both cases their boards run fine with a single midrange GPU at stock speeds, but they can't raise HTT over 230-235 even with the CPU and RAM underclocked without hard crashing.

That tells me the rail suppling the mobo is running at max capacity, and when you try and increase HTT, the increased power required by the chipset overloads the power supply

 

[liquidboss]

So what has been determined here is that there are no power supplies out yet that have 24+ amps on a 12v rail and the 6-pin connectors for PCI-e cards? That is unfortunate...

 

[ribbon13]

If you run an socket 940 here you go

http://www.pcpowerandcooling.c...ut/whatnew_850_ETX.htm

 

[Algere]

Therein lay the crux of the problem. Both PCI-E ports are being powered by the motherboard. Additional connectors to individual card(s) do an assist.

I concur & that's pretty much established. I'd still like to know if the statement from the original quote is true or not about a PSU that has one 6-pin/4-pin molex on one rail & another 6-pin/4-pin molex on another rail for video cards. AFAIK 4-pin & 6-pin power plugs have always been on the same rail - on a multi-rail PSU.

They do not take over 100% of fullfilling the power reqirements to these cards for the motherboard.

I assume that's the case since the motherboard is the only source of 3.3V for a video card.

Where one dual rail psu seperates the rails via the 4-pin square 12v AUX connector (most use this method) as one rail, and all else use the second rail.

That's likely if not true that most PSUs use that method since that was the first/original (standard?) design for ATX dual rail which PSU manufacturers adopted.

You can't convince me that such an intent was an engineered strategy, because psu's are not designed to run perpetually at PEAK.

As a counter-argument you could assume PSUs weren't designed to hold so many amps (or wattage once calculated) on one rail either while @ the same time being cost-effective, capable, or a combination of both. As an example would you want to connect the power requirements necessary for a 3000 person LAN party via one 2-socket wall outlet (rail) & a bunch of daisy-chained power strips or more wall outlets (rails)? Then there's power problems that could affect all components on the same rail.

Also I'd agree if the discussion was limited to early & to an extent, some (not all) existing dual rail implementations. However since "more than one 12V rail" is listed in the new ATX12V 2.0 spec, I don't expect it to go away. I actually expect PSU makers to improve on the original existing design as some have done. Even highly-acclaimed PSU seller PCP&C is implementing multi-rail into their future PSU line-up.

An engineered dual rail strategy would provide 24a or better on the main rail, and 12a on the processor/memory rail. Now you have a balance. The system would never run at a constant value so close to PEAK on either of the rails. Slap a couple of PCI-E power connectors to run with the main rail and YES by all means call this psu SLI Ready.

What I'd like to see for a desktop aimed multi-railed PSU would be a higher maximum amp limit per rail, shared amp capability, & if necessary/plausible, the more rails the better.

So what has been determined here is that there are no power supplies out yet that have 24+ amps on a 12v rail and the 6-pin connectors for PCI-e cards? That is unfortunate...

Not on one rail of a dual rail PSU (yet) but there's enough single rail PSUs out there that meet & exceed those goals.