I have my own question...I bought a Tesla K80 gpu, 2 of them actually, and one online video says to use the plugs marked 'cpu' from the power supply to power the K80. My question is isn't that cable needed to make the pc run like it should or is it not needed if I plug it into the gpu.
I did buy 2 adapters to go from 2 standard 8 pin gpu plugs to the 8 pin K80 plug but the 'cpu' plug would be alot easier cable routing wise.
I have my own question...I bought a Tesla K80 gpu, 2 of them actually, and one online video says to use the plugs marked 'cpu' from the power supply to power the K80. My question is isn't that cable needed to make the pc run like it should or is it not needed if I plug it into the gpu.
I did buy 2 adapters to go from 2 standard 8 pin gpu plugs to the 8 pin K80 plug but the 'cpu' plug would be alot easier cable routing wise.
mikey
you need a PSU with more than one CPU 8-pin connector. do not take the CPU plug away from your motherboard.
Tesla/Datacenter GPUs tend to use the "EPS 8-pin" plug standard, which is the same as the CPU 8-pin plug. this is just the sandard uses in servers which the Tesla is designed for. most real rack mount servers don't have PCIe/VGA style power connectors. that's mostly only a thing on consumer and workstation systems. the 8-pin PCIe cable is only rated for 150W, but the EPS/CPU 8-pin plug is rated for 300W.
the power cable setup is 180deg flipped on between the EPS and VGA formats, they are opposite on the 12v and GND connections, so do not try to plug a PCIe/VGA cable directly into the K80 without an adapter cable. you can get adapters that will take 2x PCIe/VGA 8-pin and convert it to a single EPS/CPU 8-pin
How did you get 5*6pin power connector from these ancient work station's PSU?
Those are called PSU Y-cables:
What you actually get with them is the use of your Primary (1st, original) PSU only for motherboard & CPU, while the activation wirings also goes to Secondary (2nd, additional) PSU which can then power up the rest of the GPUs. Hope you like it!
I have my own question...I bought a Tesla K80 gpu, 2 of them actually, and one online video says to use the plugs marked 'cpu' from the power supply to power the K80. My question is isn't that cable needed to make the pc run like it should or is it not needed if I plug it into the gpu.
I did buy 2 adapters to go from 2 standard 8 pin gpu plugs to the 8 pin K80 plug but the 'cpu' plug would be alot easier cable routing wise.
mikey
you need a PSU with more than one CPU 8-pin connector. do not take the CPU plug away from your motherboard.
Tesla/Datacenter GPUs tend to use the "EPS 8-pin" plug standard, which is the same as the CPU 8-pin plug. this is just the sandard uses in servers which the Tesla is designed for. most real rack mount servers don't have PCIe/VGA style power connectors. that's mostly only a thing on consumer and workstation systems. the 8-pin PCIe cable is only rated for 150W, but the EPS/CPU 8-pin plug is rated for 300W.
the power cable setup is 180deg flipped on between the EPS and VGA formats, they are opposite on the 12v and GND connections, so do not try to plug a PCIe/VGA cable directly into the K80 without an adapter cable. you can get adapters that will take 2x PCIe/VGA 8-pin and convert it to a single EPS/CPU 8-pin
+1 thing, as I used to run K20Xm (which is basically ~1/2 of K80) - it is not worth the electric bill. Like, it would not give you so much of the power you would expect. Link: https://www.techpowerup.com/gpu-specs/tesla-k80.c2616
What you essentially get is the 2x GeForce GTX 670 on one card...& you have to keep it cool, really well for those 300W to dissipate, so put a good duct & a big fan(s)!
Also note that ~200% of that K80 is the GTX 1660 card, which spends TDP 120W & is pretty cool down with double slot design & you can get for ~50€! Not to mention that Tesla M4 has the same speed as GTX 670, while spending around TDP 50W for same amount of ~50€!
& I speak from experience, as I currently run Teslas P4 in 2 systems!
So the HP z840 has a beefy power supply, rated for 1125W minimum, with that increasing with available wall voltage. It also has special 6pin cables that are rated higher than PCIe standards, up to 216W each. So I simply got some 1x 6pin PCIe to dual 6pin PCIe spliter cables, giving each GPU 108W to work with. Since the P4000 pulls 105W TDP, this worked out beautifully for me.
I have my own question...I bought a Tesla K80 gpu, 2 of them actually, and one online video says to use the plugs marked 'cpu' from the power supply to power the K80. My question is isn't that cable needed to make the pc run like it should or is it not needed if I plug it into the gpu.
I did buy 2 adapters to go from 2 standard 8 pin gpu plugs to the 8 pin K80 plug but the 'cpu' plug would be alot easier cable routing wise.
mikey
you need a PSU with more than one CPU 8-pin connector. do not take the CPU plug away from your motherboard.
Tesla/Datacenter GPUs tend to use the "EPS 8-pin" plug standard, which is the same as the CPU 8-pin plug. this is just the sandard uses in servers which the Tesla is designed for. most real rack mount servers don't have PCIe/VGA style power connectors. that's mostly only a thing on consumer and workstation systems. the 8-pin PCIe cable is only rated for 150W, but the EPS/CPU 8-pin plug is rated for 300W.
the power cable setup is 180deg flipped on between the EPS and VGA formats, they are opposite on the 12v and GND connections, so do not try to plug a PCIe/VGA cable directly into the K80 without an adapter cable. you can get adapters that will take 2x PCIe/VGA 8-pin and convert it to a single EPS/CPU 8-pin
Thank you VERY much, that helps alot!! I will be using the adapters then so the only thing left is changing the video in the bios to "above4k" IF it has it. I found it on one pc, but not on another pc, this is becoming a harder test than I at first thought. I found some software to tell me which MB is in each pc in Windows but not in Linux yet, I'm still looking though.
So the HP z840 has a beefy power supply, rated for 1125W minimum, with that increasing with available wall voltage. It also has special 6pin cables that are rated higher than PCIe standards, up to 216W each. So I simply got some 1x 6pin PCIe to dual 6pin PCIe spliter cables, giving each GPU 108W to work with. Since the P4000 pulls 105W TDP, this worked out beautifully for me.
+1
A Proud member of the O.F.A. (Old Farts Association).
So the HP z840 has a beefy power supply, rated for 1125W minimum, with that increasing with available wall voltage. It also has special 6pin cables that are rated higher than PCIe standards, up to 216W each. So I simply got some 1x 6pin PCIe to dual 6pin PCIe spliter cables, giving each GPU 108W to work with. Since the P4000 pulls 105W TDP, this worked out beautifully for me.
+1, when you have such a data, yes you can make it this way. Especially as that 105W is actually pulled usually ~75W from motherboard & rest (~40W) from 6pin adapter. All that can be shown with freeware GPU-Z program, while under stress (in BOINC). B)
Reason for knowing this, is that I need GPU-Z for temp.monitoring on my 2nd GPU (Tesla), as Tthrottle can by itself only monitor primary card. :/
I have my own question...I
)
I have my own question...I bought a Tesla K80 gpu, 2 of them actually, and one online video says to use the plugs marked 'cpu' from the power supply to power the K80. My question is isn't that cable needed to make the pc run like it should or is it not needed if I plug it into the gpu.
I did buy 2 adapters to go from 2 standard 8 pin gpu plugs to the 8 pin K80 plug but the 'cpu' plug would be alot easier cable routing wise.
mikey
mikey wrote: I have my own
)
you need a PSU with more than one CPU 8-pin connector. do not take the CPU plug away from your motherboard.
Tesla/Datacenter GPUs tend to use the "EPS 8-pin" plug standard, which is the same as the CPU 8-pin plug. this is just the sandard uses in servers which the Tesla is designed for. most real rack mount servers don't have PCIe/VGA style power connectors. that's mostly only a thing on consumer and workstation systems. the 8-pin PCIe cable is only rated for 150W, but the EPS/CPU 8-pin plug is rated for 300W.
the power cable setup is 180deg flipped on between the EPS and VGA formats, they are opposite on the 12v and GND connections, so do not try to plug a PCIe/VGA cable directly into the K80 without an adapter cable. you can get adapters that will take 2x PCIe/VGA 8-pin and convert it to a single EPS/CPU 8-pin
_________________________________________________________________________
chihwei wrote: How did you
)
Those are called PSU Y-cables:
What you actually get with them is the use of your Primary (1st, original) PSU only for motherboard & CPU, while the activation wirings also goes to Secondary (2nd, additional) PSU which can then power up the rest of the GPUs. Hope you like it!
non-profit org. Play4Life in Zagreb, Croatia, EU
Ian&Steve C. wrote: mikey
)
+1 thing, as I used to run K20Xm (which is basically ~1/2 of K80) - it is not worth the electric bill. Like, it would not give you so much of the power you would expect. Link: https://www.techpowerup.com/gpu-specs/tesla-k80.c2616
What you essentially get is the 2x GeForce GTX 670 on one card...& you have to keep it cool, really well for those 300W to dissipate, so put a good duct & a big fan(s)!
Also note that ~200% of that K80 is the GTX 1660 card, which spends TDP 120W & is pretty cool down with double slot design & you can get for ~50€! Not to mention that Tesla M4 has the same speed as GTX 670, while spending around TDP 50W for same amount of ~50€!
& I speak from experience, as I currently run Teslas P4 in 2 systems!
non-profit org. Play4Life in Zagreb, Croatia, EU
So the HP z840 has a beefy
)
So the HP z840 has a beefy power supply, rated for 1125W minimum, with that increasing with available wall voltage. It also has special 6pin cables that are rated higher than PCIe standards, up to 216W each. So I simply got some 1x 6pin PCIe to dual 6pin PCIe spliter cables, giving each GPU 108W to work with. Since the P4000 pulls 105W TDP, this worked out beautifully for me.
Ian&Steve C. wrote: mikey
)
Thank you VERY much, that helps alot!! I will be using the adapters then so the only thing left is changing the video in the bios to "above4k" IF it has it. I found it on one pc, but not on another pc, this is becoming a harder test than I at first thought. I found some software to tell me which MB is in each pc in Windows but not in Linux yet, I'm still looking though.
taketwicedailey wrote: So
)
+1
A Proud member of the O.F.A. (Old Farts Association).
mikey wrote: ... I found
)
For linux: inxi -b
taketwicedailey wrote: So
)
+1, when you have such a data, yes you can make it this way. Especially as that 105W is actually pulled usually ~75W from motherboard & rest (~40W) from 6pin adapter. All that can be shown with freeware GPU-Z program, while under stress (in BOINC). B)
Reason for knowing this, is that I need GPU-Z for temp.monitoring on my 2nd GPU (Tesla), as Tthrottle can by itself only monitor primary card. :/
non-profit org. Play4Life in Zagreb, Croatia, EU
mountkidd wrote: mikey
)
Thanks Mountkidd!! I didn't know this until now.
Proud member of the Old Farts Association