Tagan TG1100-U95 1100W Power Supply
- Thread starter Rob Williams
- Start date
Tech-Daddy
Tech Monkey
So... if I am reading this correctly, you would hook up PCI-e 1 and 3 to one card, and 2 and 4 to another card?
Is that correct?
How do you know "which" connector is on "which" rail?
Is that correct?
How do you know "which" connector is on "which" rail?
madmat
Soup Nazi
I have ways of making zem speak...1 & 3 are on one rail and 2 & 4 are on another so they're cool to hook up together. Don't get me wrong, a 6 is still a good score but it needs things like complete documentation, normal sleeving and a single big, quiet fan to get it pushed up to an 8.
Last edited:
P
pad6000
Guest
TG1100-U95 ripple
Hi. Just how bad was the ripple voltage with all four PCIe 12V combined? Did you get a measurement?
Hi. Just how bad was the ripple voltage with all four PCIe 12V combined? Did you get a measurement?
madmat
Soup Nazi
It was about 80mV at full load with a strangely oscillating waveform. If you're considering combining rails on one I'd really advise against it.
The ripple on each rail at 20A per rail was a rather clean 40-50mV which is well under spec and there was no oscillation in the waveform either.
What exactly are you wanting to do? Run an 8800GTX SLI setup? If so just run PCI-e 1 & 3 to one card and 2 & 4 to the other and that'll keep everything on its own rail.
The ripple on each rail at 20A per rail was a rather clean 40-50mV which is well under spec and there was no oscillation in the waveform either.
What exactly are you wanting to do? Run an 8800GTX SLI setup? If so just run PCI-e 1 & 3 to one card and 2 & 4 to the other and that'll keep everything on its own rail.
P
pad6000
Guest
Thanks for quick response!
Actually this is a non-PC application. Its to power a server blade which needs up to 80A. There are some single 12V rail PSUs, such as the Silverstone OP1000 and Power & Cooling 1KW-SR. I think I look at those.
Thanks
Actually this is a non-PC application. Its to power a server blade which needs up to 80A. There are some single 12V rail PSUs, such as the Silverstone OP1000 and Power & Cooling 1KW-SR. I think I look at those.
Thanks
Here is a question for Matt. We talked to Tagan at CES, and they showed us a similar model to this, except it had a switch on that back that could toggle between 5 separate +12v rails or 1 80a +12v rail.
Now generally, I thougt that there was no real such thing as "separate" rails, instead there is one +12v always, that just happens to split off into separate areas. If that's the case, then what good does this "switch" do?
I wish I could explain better, but I am just going by what he told me.
Now generally, I thougt that there was no real such thing as "separate" rails, instead there is one +12v always, that just happens to split off into separate areas. If that's the case, then what good does this "switch" do?
I wish I could explain better, but I am just going by what he told me.
Jakal
Tech Monkey
Sounds like they've incorporated a switchable transformer.
Transformers change AC to DC by the expanding and collapsing magnetic fields passing through a coil of wire. The "movement" of these magnetic fields produce a current upon the wire. Transformers are the most energy efficient devices man created, with a power factor of ~95% or better. If you were to take a transformer with a 1:10 ratio, 120V AC - 12V DC, you maintain the same voltage, but the current available has increased. If you're on a 30 Amp breaker then available DC current would be 10 times that, or 300 Amps. In that specific psu application you talked about, the transformer must have a maximum amperage rating of 80 Amps.
You can then go back and regulate the amperage rating for the 12V a few different ways. The simplest method would be a set of resistors that reduce the output current. Someone's using their head and applying Ohm's Law to the output of a power supply. I = V/R , or Current (I) = Voltage (V) / Resistance (R). Some simple Algebra leads to: V = IR, and R = V/I.
Resistors in parallel reduce current output, but not voltage. So, I'll demonstrate with a couple figures I've compiled, starting with a 40 Amp rating. Input will always be 12 volts at 80 amps, and using resistors is the simplest/cheapest method.
Reduction by 40A still maintains a 40A output.
12v / 40A = .3 ohms : Output of 12v @ 40A
12v / 30A = .4 : Output of 12v @ 50A
12v / 20A = .6 : Output of 12v @ 60A
12v / 10A = 1.2 : Output of 12v @ 70
12v / 0A = 0 or a direct output from rectifier/cap 12v @ 80A
As you can see, by including a switch to choose between output resistance, you can vary the output current. This can also be done with a single potentiometer, or rheostat. Similar to the voltage control on a dc fan, it varies the resistance to the load.
"So why don't companies produce higher amp power supplies?"
Well, to support the higher amperage ratings, it would require larger, higher rated, components, and heavier gauge wire. These components make for a larger psu, as well as, increasing the price.
Now that I've thought about how that could work, I'm slightly surprised this hasn't been introduced earlier. Use a larger transformer, and allow switchable amp ratings. I've seen the same application in power amps and such through school, but never applied it when thinking about power supplies. Quite nice indeed.
Transformers change AC to DC by the expanding and collapsing magnetic fields passing through a coil of wire. The "movement" of these magnetic fields produce a current upon the wire. Transformers are the most energy efficient devices man created, with a power factor of ~95% or better. If you were to take a transformer with a 1:10 ratio, 120V AC - 12V DC, you maintain the same voltage, but the current available has increased. If you're on a 30 Amp breaker then available DC current would be 10 times that, or 300 Amps. In that specific psu application you talked about, the transformer must have a maximum amperage rating of 80 Amps.
You can then go back and regulate the amperage rating for the 12V a few different ways. The simplest method would be a set of resistors that reduce the output current. Someone's using their head and applying Ohm's Law to the output of a power supply. I = V/R , or Current (I) = Voltage (V) / Resistance (R). Some simple Algebra leads to: V = IR, and R = V/I.
Resistors in parallel reduce current output, but not voltage. So, I'll demonstrate with a couple figures I've compiled, starting with a 40 Amp rating. Input will always be 12 volts at 80 amps, and using resistors is the simplest/cheapest method.
Reduction by 40A still maintains a 40A output.
12v / 40A = .3 ohms : Output of 12v @ 40A
12v / 30A = .4 : Output of 12v @ 50A
12v / 20A = .6 : Output of 12v @ 60A
12v / 10A = 1.2 : Output of 12v @ 70
12v / 0A = 0 or a direct output from rectifier/cap 12v @ 80A
As you can see, by including a switch to choose between output resistance, you can vary the output current. This can also be done with a single potentiometer, or rheostat. Similar to the voltage control on a dc fan, it varies the resistance to the load.
"So why don't companies produce higher amp power supplies?"
Well, to support the higher amperage ratings, it would require larger, higher rated, components, and heavier gauge wire. These components make for a larger psu, as well as, increasing the price.
Now that I've thought about how that could work, I'm slightly surprised this hasn't been introduced earlier. Use a larger transformer, and allow switchable amp ratings. I've seen the same application in power amps and such through school, but never applied it when thinking about power supplies. Quite nice indeed.
Last edited:
madmat
Soup Nazi
Most power supplies have one transformer that feeds out to the various rails via OCP (over current protection) per virtual rail, In the case of the 1100 I reviewed it has two transformers that fead two rails each which is why the ripple was so strange.
I'm guessing the switch turns the OCP on and off.
I'm guessing the switch turns the OCP on and off.
Thanks for the detailed explanation Jakal, that will require me to re-read when I'm not buzzed. Thanks also Matt, it makes more sense to me now. The Tagan rep was talking so fast, I didn't get to comprehend it that well, heh.
Hi I have a tagan 1100 watt . At first I also had my pciX cables connected to my 8800gtx sli like 1+2 and 3+4. I didn't notice anything strange, but ofcourse I have now changed it.
My question is if each rail has a max of 20amps will connecting two pciX cables on the same rail be any better than just connecting one? So if all I can get is 20amps on that rail, why do I need two cables? Will 20amps even be enough for one 8800gtx. It's more likely 16amps as no psu is 100% effecient. Why didn't they have one pciX cable on each of it's four rails. that makes more sense to me as then you can have more than 20amps going to each card.
I am very confused about this. Can someone explain all this to me please? Am I going to have enough amperage to have stable 8800gtx sli even if I overclock them? Is there anyway to use the Tagan so that I can get more thab 20amps on each card?
Thank you.
PS. What is this 25c thing all about? What will happen to my psu when it get to 50c? Will it crap out or something, or does it just mean that it will have alot less power at 50c, but how much less? Did you test that? What if I had a really good case that keeps things cool? Will i be able to stay below 50c then?
EDIT: I just quickly checke the spec of my 8800gtx and I have found that one of these bad boys needs a 450watt psu witha 30amp 12v current. SO for sli I would need 2x 450watt and 2x 30amps. That means I don't have enough power!!. My 12v rail is only 900watt and probably 750watt after the waste. My amps per rail is only 2x20amp or more like 16x2 after waste. What should I do, there is no psu out there tha even has 30amps per rail except the uber expensive PCP+C 1kw-sr or the Silverstoen Op1000. I can't get those where I am though. WHat do I do?
My question is if each rail has a max of 20amps will connecting two pciX cables on the same rail be any better than just connecting one? So if all I can get is 20amps on that rail, why do I need two cables? Will 20amps even be enough for one 8800gtx. It's more likely 16amps as no psu is 100% effecient. Why didn't they have one pciX cable on each of it's four rails. that makes more sense to me as then you can have more than 20amps going to each card.
I am very confused about this. Can someone explain all this to me please? Am I going to have enough amperage to have stable 8800gtx sli even if I overclock them? Is there anyway to use the Tagan so that I can get more thab 20amps on each card?
Thank you.
PS. What is this 25c thing all about? What will happen to my psu when it get to 50c? Will it crap out or something, or does it just mean that it will have alot less power at 50c, but how much less? Did you test that? What if I had a really good case that keeps things cool? Will i be able to stay below 50c then?
EDIT: I just quickly checke the spec of my 8800gtx and I have found that one of these bad boys needs a 450watt psu witha 30amp 12v current. SO for sli I would need 2x 450watt and 2x 30amps. That means I don't have enough power!!. My 12v rail is only 900watt and probably 750watt after the waste. My amps per rail is only 2x20amp or more like 16x2 after waste. What should I do, there is no psu out there tha even has 30amps per rail except the uber expensive PCP+C 1kw-sr or the Silverstoen Op1000. I can't get those where I am though. WHat do I do?
Last edited:
madmat
Soup Nazi
My god...calm down a bit. The ENTIRE SYSTEM requires a 450W PSU with 30A on the 12V rail. That's shared amongst the CPU, drives, fans, mobo, water pump if there is one and everything else plus the GPU so having 4-20A rails is more than enough. Take a Prozac and relax a bit. The 8800GTX uses at most 13A on DX9 and maybe, just possibly will require 18A on DX10 so quit panicking, you have more than enough power to run your system. You'll have 40A (20A X 2) shared amongst your 8800GTX SLI cards and another 40A for everything else. A $10,000 super duty home PC with dual dualcore CPU's and SLI 8800GTX's won't even touch that amount of power so don't freak out, you're well covered.
Yeah you're right, after rereading my post it was a bit OTT. Sorry about that. Thanks for clearing it all up though.
PS I think you mean Valium , not Prozac
U
Unregistered
Guest
How about the Tagan TG900 U95?
Great review.
I am in the process of building a 2 CPU Opteron workstation, with a SATA PCI-E RAID card and 8 HDs, plus floppy and 2 DVD drives. Will definitely use a XFX XXX 8800 GTX (factory overclocked). Will the 900 W PSU give enough power? How about if I decide to go SLI?
Thanks for your comments!
RRP
Great review.
I am in the process of building a 2 CPU Opteron workstation, with a SATA PCI-E RAID card and 8 HDs, plus floppy and 2 DVD drives. Will definitely use a XFX XXX 8800 GTX (factory overclocked). Will the 900 W PSU give enough power? How about if I decide to go SLI?
Thanks for your comments!
RRP
U
Unregistered
Guest
Thank you very much.
When would I need a higher than 900W PSU then?
I was looking at the the PSU calculator here http://extreme.outervision.com/psucalculator.jsp and total power required for my rig is between 892 W and 905 W. Isn't that too close?
BTW, I tried registering, but for some reason, never got the confirmation email. Thanks again for your help.
rrpalma
When would I need a higher than 900W PSU then?
I was looking at the the PSU calculator here http://extreme.outervision.com/psucalculator.jsp and total power required for my rig is between 892 W and 905 W. Isn't that too close?
BTW, I tried registering, but for some reason, never got the confirmation email. Thanks again for your help.
rrpalma
JacKz5o
Obliviot
You should be fine with 900W, but if you want to play it safe I would recommend 1KW or higher.
Also, I wouldn't recommend using PSU calculators like those, they're mostly inaccurate.
Also, I wouldn't recommend using PSU calculators like those, they're mostly inaccurate.
Yes, they are. To rrpalma, even 600W would probably handle that machine, but I'd go with at least 800W to be sure. People who buy 1KW machines have LOTS of components. But even then, I bet 90% of people who own them don't actually require them. It would almost be a challenge to purposely try to push a PSU of that size.