What would be nice for these dual rail psu's is if the power connectors were labelled which rail they were on so you could do the load balancing yourself when you build the system...
But that might be convenient. And a problem if you're crap at maths, and don't really know anything about load ballencing.
You'd end up with an amateur using the wrong things in conjunction with each other, and causing more damage.
WhoBeDaPlaya
Diamond Member
- Sep 15, 2000
- 7,415
- 404
- 126
i know someone already pointed this one out but, and i quote:
mindless1
Diamond Member
- Aug 11, 2001
- 8,733
- 1,746
- 126
The article only mentions "couldn't play", hardly a sufficient description of what actually happened during those attempts.
Without more measurement of the systems in their failing state, rather than just swapping in another PSU, the problem is never isolated. That is a pity but if the report of regaining stability from adding more hard drives is true, it does appear to be a generalized rail-imbalance situation. To really know, they need to send these systems to a qualified engineer and have the appropriate PSU subsystem measurements taken.
Traditionally dual rail PSUs only have one common 12V rail, ie- one transformer winding, one (or a parallel pair of) rectifiers, one series of inductors and capacitors. Near the final output of the PSu the two rails are split with low-ohm resistors, or resistive wire elements.
As with any resistive element, higher current would drop the voltage. A PSu using such a design could be 2000W and it won't make a bit of difference, because it's not a lack of wattage bot a logical design problem in that it is not really appropriate for the dual SLI setup. More than anything, this signals that nVidia needs to either start segmenting their SLI approved list for different cards, or start weeding out those that aren't sufficient for new SLi'd products. ATI would do well to take the same philosophy.
You would have to open your PSu and reverse engineer it enougn to know for certain it has the rail-splitting implementation proposed above. If it does, simply tying the two rails together with another low-Ohm resistor might help. Directly bridging the two "might" as well but that completely circumvents the intented minor isolation between the two rails. In most exterme situations it might even help to simply replace the (deliberate) resistance with std., heavy copper wire. This is just thinking out loud, if after opening a psu this wasn't obvious you probably shouldn't be inside the psu in the first place.
That might be a stop gap measure rather than the ultimate solution as particular PSU designs may also have other subtly increasing losses, bottlenecks near their max output. Even so, maximum output may not be the problem at all, but due to Enermax's questionable rating system, I tend to agree with a prior poster that Enermax units should just be avoided altogether until someone has stripped one down and confirmed they have taken the higher road towards accurate sustainable ratings.
I've said it before and I'll say it again- it is not a valid test to load a psu with a static load for a day to determine if it's output is accurate. That it doesn't start smoking or shut off is not an indication it's fit for any particular use... and to nVidia/ATI/et al, if they provide a list they NEED to start being more stringent and specific who/what/where something is on that list.
Those of you who had thoughts about 2 PSU, that is actually a reasonable idea. Two $50 PSu will get you higher output than any one for $100 and significantly reduce the power density which has several positive effects in itself. Rail loading would be a signficant issue though, it seems fairly certain an additional load would need be placed on the 5V rail of one of the units and if enough load is necessary, for your system's cooling purposes it might be desirable to put that load outside the case rather than inside.
Of course there IS another option. Don't think "2nd ATX PSU", think 2nd PSu is a 12V-only unit. Not a wimpy power brick but a substantial unit. They're not cheap but if you have a few $K on a system that's still in line with a reasonable 10% system-budget-on-power plan.
Without more measurement of the systems in their failing state, rather than just swapping in another PSU, the problem is never isolated. That is a pity but if the report of regaining stability from adding more hard drives is true, it does appear to be a generalized rail-imbalance situation. To really know, they need to send these systems to a qualified engineer and have the appropriate PSU subsystem measurements taken.
Traditionally dual rail PSUs only have one common 12V rail, ie- one transformer winding, one (or a parallel pair of) rectifiers, one series of inductors and capacitors. Near the final output of the PSu the two rails are split with low-ohm resistors, or resistive wire elements.
As with any resistive element, higher current would drop the voltage. A PSu using such a design could be 2000W and it won't make a bit of difference, because it's not a lack of wattage bot a logical design problem in that it is not really appropriate for the dual SLI setup. More than anything, this signals that nVidia needs to either start segmenting their SLI approved list for different cards, or start weeding out those that aren't sufficient for new SLi'd products. ATI would do well to take the same philosophy.
You would have to open your PSu and reverse engineer it enougn to know for certain it has the rail-splitting implementation proposed above. If it does, simply tying the two rails together with another low-Ohm resistor might help. Directly bridging the two "might" as well but that completely circumvents the intented minor isolation between the two rails. In most exterme situations it might even help to simply replace the (deliberate) resistance with std., heavy copper wire. This is just thinking out loud, if after opening a psu this wasn't obvious you probably shouldn't be inside the psu in the first place.
That might be a stop gap measure rather than the ultimate solution as particular PSU designs may also have other subtly increasing losses, bottlenecks near their max output. Even so, maximum output may not be the problem at all, but due to Enermax's questionable rating system, I tend to agree with a prior poster that Enermax units should just be avoided altogether until someone has stripped one down and confirmed they have taken the higher road towards accurate sustainable ratings.
I've said it before and I'll say it again- it is not a valid test to load a psu with a static load for a day to determine if it's output is accurate. That it doesn't start smoking or shut off is not an indication it's fit for any particular use... and to nVidia/ATI/et al, if they provide a list they NEED to start being more stringent and specific who/what/where something is on that list.
Those of you who had thoughts about 2 PSU, that is actually a reasonable idea. Two $50 PSu will get you higher output than any one for $100 and significantly reduce the power density which has several positive effects in itself. Rail loading would be a signficant issue though, it seems fairly certain an additional load would need be placed on the 5V rail of one of the units and if enough load is necessary, for your system's cooling purposes it might be desirable to put that load outside the case rather than inside.
Of course there IS another option. Don't think "2nd ATX PSU", think 2nd PSu is a 12V-only unit. Not a wimpy power brick but a substantial unit. They're not cheap but if you have a few $K on a system that's still in line with a reasonable 10% system-budget-on-power plan.
flashbacck
Golden Member
- Aug 3, 2001
- 1,921
- 0
- 76
Oh man, if they do this, I just know they'll make a different power brick with weirdass proprietary plugs for every new card! Exactly what cell phone manufacturers do.
zephyrprime
Diamond Member
- Feb 18, 2001
- 7,512
- 2
- 81
I don't understand something. What exactly does putting more load on the 5V rail do to fix the problem? With my naive understanding of electronics, it seems like putting more load on the 5v line wouldn't do anything but I guess there's something that I don't know about?
Ronin
Diamond Member
Hell, even Motorola does it WITHIN their line. Talk about a pita if you have an older and a newer one.
Zebo
Elite Member
- Jul 29, 2001
- 39,398
- 19
- 81
Someone else had a problem with that SS Strider here in video - four rails none delivering power required what do you expect - I'm telling you guys a fat single 12V will run it fine. A 510W PCP&C will run it fine. That's what Alienware uses in thier decked out watercooled Sli boxen with RAID raptors.
It's been done before. Of course that was back in the days when SLI meant "Scan Line Interleave." Back to the future, baby.
Yup, that's how I still define SLI. I don't even remember the new words they came up with to fit the acronym. ;P
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 21K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter