9 Posts
0
3865
April 15th, 2021 20:00
Area-51 R4, replace PSU with 3rd party one?
The issue is when I use a non-OEM video card such as the EVGA 2080 TI, it draws more power than the stock OEM 1080 TI, which causes the computer to shut immediately when the EVGA 2080 TI goes under load. I have a 1500-watt power supply that is showing an overvoltage of 12.62 volts on pin 24. This is .02 volts out of specifications and the computer is not even attached to the power supply during this test, so the power supply is not under load. It should be enough for Dell to replace the power supply, but they will not because the computer passes all the diagnostics using the OEM card. Dell will not fix my issue, even though I was told when I bought the $5000 computer that it was fully upgradable and could easily support a two-way SLI due to the 1500-watt power supply. So much for a spending $600 for a 5 year service plan.
I have the service manual, but it does not give me the specifications of the type of power cables that connect to the motherboard and other peripherals.
Has anyone successfully replaced the OEM Alienware power supply with a third-party
version?


Cass-Ole
8 Professor
•
1.9K Posts
•
9.7K Points
0
April 21st, 2021 01:00
The x3 red 6pin ports on the PSU run power to the CPU 8pin (up behind the heatsink, top middle of board) &
in the lower right hand corner is the GPU/PCIE_PWR 8pin, so you have two solid 8pin (EPS/CPU) cables (CPU is long, PCIE is short) sending power to the mthrbrd, make sure PCIE_PWR 8pin is connected to mthrbrd & to PSU
Search 800GY on eBay for a replacement
If you want an aftermarket PSU, if it's 1200watts or more it will come with two CPU cables & two CPU 8pin ports to plug into the front panel, one will go to your CPU 8pin header (behind heatsink), one goes to PCIE_PWR 8pin (lower right)
If another 800GY, or an aftermarket PSU won't cure this, you can sell it afterwards, after testing, since they can retain resale value (though u may lose a little money back relative to what u paid). Or try Amazon/MicroCenter/Newegg & return it if it won't cure this
Then suspect the motherboard
*Your 24pin 3.3v/5v/12volt readings all look normal, in-spec
*Try the GRFX card in different PCIE slots, if u haven't already
*We can be sure that dual-2080Ti was an option (on 51 R5 at least, which is about identical to R4), I can't say what your issue is, or the why of it, but when in doubt, post at the EVGA forum also, someone w/2080Ti may have heard of this happening & have advice
redxps630
11 Legend
•
16.1K Posts
•
82.1K Points
0
April 15th, 2021 22:00
a little perplexed by this report. Pin 24 of psu 24 pin ATX should be Gnd according to ATX standard. Which pin are you testing voltage?
are you saying the stock psu is slightly inferior in quality that it is not matching ATX pinout spec exactly on some pins, or it is inadequate to supply enough current on 12V rail to power 2080ti under load? Not clear if the two issues are related.
One 2080ti needs two 8 pin each of which could come from a pair of 6 pin. The two pairs of 6 pin should come from different 12V rails if the psu has multiple 12V rails.
RSRAM
9 Posts
0
April 15th, 2021 22:00
Sorry, I meant pin 14 was 12.62 volts which is out of the 5% tolerance for a power supply.
Cass-Ole
8 Professor
•
1.9K Posts
•
9.7K Points
0
April 16th, 2021 00:00
Yes, u can swap out to an aftermarket type (Note: Corsair Hx1200i, Ax1200i, Ax1500i will not work)
Pin14 (blue), in the 24pin end, is called -12v (minus 12volts) ... although the blue wire is installed into the cable harness, -12v on the motherboard header pin14 is not used, so it's a dead wire, dead pin, unused wire, a meaningless wire + meaningless voltage reading ... take the positive +12v voltage reading from pins 10 & 11 (relative to ground)
*Make sure the solid 8pin PCI-E_PWR connector is connected to motherboard
*The device-end / hardware-end of the vid-cable is an 8pin (6+2), make sure they're both seated
*As a test, on your PSU black vid-cable ports (black 6pins, R1-R6), insert a cable into black port R1, then cable #2 into black port R2, see if shutdown occurs. If so, move cable #2 from R2 to R3 & retest; continue more tests if needed in ports R4, R5, R6 until all ports are tested. If shutdown occurs on all (R1 + R2...R6), REPORT BACK
RSRAM
9 Posts
0
April 16th, 2021 00:00
I'm confused with the instructions in the last paragraph. My video cables are not black. Shorting out pins 1 and 2 on the VGA connector would connect both 12v powered pins together?
These are the voltages that I measured:
VGA-1, PCI 8 Power Connector:
1- 12.17v 5- 0v (GND)
2- 12.17v 6- 0v (GND)
3- 12.17v 7- 0v (GND)
4- 0v (GND) 8- 0v (GND)
-------------------------------------------
VGA-2, PCI 8 Power Connector:
1- 12.17v 5- 0v (GND)
2- 12.17v 6- 0v (GND)
3- 12.17v 7- 0v (GND)
4- 0v (GND) 8- 0v (GND)
-------------------------------------------
CPU Power Connector:
1- 0v (GND) 5- 12.16v
2- 0v (GND) 6- 12.16v
3- 0v (GND 7- 12.16v
4- 0v (GND) 8- 12.16v
------------------------------------------
25 Pin Motherboard Connector:
1- 3.333v 13- 3.333v
2- 3.333v 14- 12.62v
3- 0v (GND) 15- 0v (GND)
4- 5.108v 16- PS on
5- 0v (GND) 17- 0v (GND)
6- 5.108v 18- 0v (GND)
7- 0v (GND) 19- 0v (GND)
8- 4.134v 20- NC
9- 5.137v 21- 5.108v
10- 12.17v 22- 5.108v
11- 12.17v 23- 5.108v
12- 3.333v 24- 0v (GND)
speedstep
11 Legend
•
47K Posts
1
April 16th, 2021 06:00
Number of pins DOES NOT indicate what the plug is for.
CPU EPS12v connections ARE NOT GPU PCI-E connections.
CPU are 4 + 4
GPU are 6 + 2
RSRAM
9 Posts
0
April 16th, 2021 08:00
You didn't answer my question. I used the same diagrams you posted to get the voltage readings.
Vanadiel
8 Professor
•
7.1K Posts
•
29.7K Points
0
April 16th, 2021 10:00
Forgot to add that unless you have a calibrated volt meter, you cannot be sure about the 0.02 Volt over spec. That could be the result of the error of margin on your volt meter.
To be sure you would need a calibrated volt meter that is calibrated to 0.01 volt precision.
redxps630
11 Legend
•
16.1K Posts
•
82.1K Points
1
April 16th, 2021 10:00
Re: VGA-1, PCI 8 Power Connector:
1- 12.17v 5- 0v (GND)
2- 12.17v 6- 0v (GND)
3- 12.17v 7- 0v (GND)
4- 0v (GND) 8- 0v (GND)
-------------------------------------------
VGA-2, PCI 8 Power Connector:
1- 12.17v 5- 0v (GND)
2- 12.17v 6- 0v (GND)
3- 12.17v 7- 0v (GND)
4- 0v (GND) 8- 0v (GND)
these are correct 8 pin pcie connector voltages. Dell 1500w psu 800GY etc. has two 12V rails supporting 48A on each rail. It is designed for multi gpu support. But combining two 8 pin connectors should give you up to 300w in those connectors, plus 75w from PCIex16 slot. For some reason your report seems to suggest your 2080 Ti is not getting enough power from 12V rails when under load. If your two 8 pin connectors are from different 12V rails each capable of 48A, you should not have problem. Since you have 6 black PCIe sockets R1-6 on psu panel, I would experiment different combinations of two 8 pin to see if crash issue can be resolved.
note the 1500w and 850w Dell psu have identical psu panel layout, but 1500w psu label spec 12V A, B, C, D = 32,48,48,16A, 850w psu label spec 12V A, B, D = 32,48,16A.
Vanadiel
8 Professor
•
7.1K Posts
•
29.7K Points
0
April 16th, 2021 10:00
There are 6 and 8 pin PCI Express power cables to power up video cards.
You should never connect 12 volt rails together on them, because each rail is supposed to carry a certain current. If you use 1 wire to feed 2 separate pins, all current for each pin goes over the common wire and if the wire size (AWG) is not the correct size for that current it can fail, even causing an electrical fire.
It is possible on a 6 pin that pin 2 is not connected. I have never seen it myself, but it is in the PCI Express spec.
6 pin: 3 x 12 volt rails, total of 2.083 A per rail, total of 75 watts per connector.
8 pin: 3 x 12 Volt rails, total of 4.167 A per rail, total of 150 Watts per connector.
Dell does not always keep itself to the specs, and sometimes uses custom pin layouts. I ran into that with my R10 recently. Keep that in mind when trying to replace this PSU.
Also, I am not sure why the 1500 Watt PSU should not be able to power up your system. Maybe I read it wrong, but it's normal for larger size PSU's to slightly be over voltage spec when not loaded. This will not pose an issue and it should drop down within spec once it becomes loaded up. It's common for over clockers to manually increase voltages to get better results.
http://www.playtool.com/pages/psuconnectors/connectors.html#pciexpress
redxps630
11 Legend
•
16.1K Posts
•
82.1K Points
1
April 16th, 2021 10:00
Re: Make sure the solid 8pin PCI-E_PWR connector is connected to motherboard
From service manual A51 R4 motherboard does not appear to have Dell proprietary 8 pin PCIe/gpu pwr socket, which is a feature of Aurora R5-8 motherboard.
Area 51 R4
Aurora R5 (#21: GPU power)
RSRAM
9 Posts
0
April 16th, 2021 12:00
Is there any other proprietary connectors going to the motherboard? Or is all of the power connectors standard ATX?
Vanadiel
8 Professor
•
7.1K Posts
•
29.7K Points
1
April 16th, 2021 12:00
If it would be out of tolerance on the low end I would agree with you.
If it is out of tolerance on the high end, I would not be concerned about it. I would actually prefer it to be on the high end, as that lowers the current draw for the load. Especially since we are talking here about 0.02 Volts, 2 hundreds of a volt.
When a power supply fails it is typically due to age of the capacitors. It is possible your aging power supply is buckling under load. To check for that you should monitor the voltages with HWINFO64 or similar program and look for min/max recorded values over time when the system is being under load.
I have had this happen myself, and my 12 volt common rail would drop under load to below 11.4 volts and cause instability. It would be above 12 volts when not under load.
I don't know how the internals of that PSU are, if it has a common 12 volt rail or multiple rails.
But I would check that first.
You also said you tried a separate power supply to power up just the video card, with the same results. In that case the power supply to the video card would be correct and not the issue, but the PCIe power might still be the issue.
I would use HWINFO, log the voltages and see how much they fluctuate over time to get a base understanding on how the voltages behave.
RSRAM
9 Posts
0
April 16th, 2021 12:00
I tried using different VGA connectors. I have two each labelled VGA1 through VGA 4. I'm assuming that the VGA cables with the same name are from the same modular connector. The EGVA support already had me using different combinations of VGA1 through VGA 4. That is I did the following sequence:
VGA1 - VGA2
VGA1 - VGA3
VGA1 - VGA4
VGA2 - VGA3
VGA2 - VGA4
VGA3 - VGA4
Note that I also tried using the VGA connectors from an external 750 watt power supply. I shorted out pins 3 and 4 of the 25 pin motherboard connector to power it up. I got the same results of severe artifacts on the monitor at the boot prompt.
This is the third replacement card that EVGA has sent me, and they are not going to send me another one. I have had this issue of intermittent abrupt shutdowns for a year now. In each case, when the EGVA card is under load when playing games such as DCS world, the power supply shuts off. I just started having issues where my monitor will not turn on if I have my Valve Index plugged in at boot up even when using the OEM 1080 TI. So EVGA support thinks the power draw from either the power supply or on the PCI-E bus is fluctuation enough to cause the power supply to shut off.
RSRAM
9 Posts
0
April 16th, 2021 12:00
I get that, but my issue is that with no load, I shouldn't be getting a voltage so close to being out of specifications when all of the other voltages like 3% or less within tolerances.