General advice is fit the largest radiator your case will allow.
Or in some of our cases, even if they don't fit in the case...they sit above them bahahahaha.....
Hi Zalster,
Believe it or not, I was tempted to do just that!:^)
I noticed on your #12789230 computer that you are running 2 x 3080 12GB GPUs. Do you notice any difference in running Einstein with a 12GB GPU vs a 10GB GPU? I'm curious... yeah, I know... Curious George!! (I had two bananas this morning)
I can barely wait until I get my new Meshify 2 XL case and water cooling 'stuff'. My intent is to put the setup I have now into the new case, and also the two 2070S's (all 3 GPUs water cooled & CPU too) put in there.
Don't skimp on the radiator capacity. If building a combined cpu-gpu loop, you should be using at least dual radiators or a very large 480mm radiator.
From EK's website.
The rule of thumb is to use at least one 120mm radiator (section) per each water cooled component plus one additional section. For example, if you're liquid-cooling a CPU and a single high-performance graphics card, we recommend using at least one 360mm (3x 120mm) radiator for the best results.
Since you are going to end up with 4 active components you can see that a single 360mm will not be sufficient.
Also another rule of thumb is a max 10-14° C. delta rise of the water temp over the ambient temp. Best is a 4-6° rise over ambient.
On my cpu-gpu loop with dual 1080 Ti's and the 5950X, I run a 37° C. water temp with a 25°C. ambient temp which gives me a 12° C. delta.
I would like that to be better but I had to compromise on a standard 30mm thick HWL GTS360 radiator in the roof to clear the top of the mobo and memory. I fitted a AlphaCool XT45 360mm rad in the front as another compromise to clear the gpus with the pump mounted to the front rad. I would normally have used my default preference of a XSPC RX360 but it wasn't available at the time I needed it and was building.
My preference would have been dual XSPC RX360 rads for the loop.
I noticed on your #12789230 computer that you are running 2 x 3080 12GB GPUs. Do you notice any difference in running Einstein with a 12GB GPU vs a 10GB GPU? I'm curious... yeah, I know... Curious George!! (I had two bananas this morning)
I have two 3080's in this host. Numbskull along with one of Ian's old 2080 Ti's.
The 12GB model runs about 10 seconds faster than the 10GB model both doing doubles. The 12GB model has 256 more CUDA cores than the 10GB model and slightly better memory bandwidth.
So yes, the 12GB models are faster running Einstein with our AIO app.
if this value is real, it could be the source of many of your ongoing issues, not just with the CPU processing but also with your constant GPU issues. A new PSU might be in order.
The PSU on the 3950x is a used EVGA 850 G3 one. The 24 and 8 pin cables came with it.
The 3950x box GPU is NOT giving me any trouble at all. But then I am running it under V0.92 :)
Just got done putting modular cable on there and am getting 12 volts under load (BOINC).
And when I shut down BOINC it still says 12 volts.
So I am guessing the 12 volt rail is not "the issue" with the CPU crashing on the 3950x. I am bringing down the CPU voltage stepping 1.39 to 1.38 etc. The displayed voltage under Asus sensors it a couple of hundredth below that.
Someplace under a displayed voltage of 1.31 volt I expect it to start crashing again.
Maybe I can get the CPU temp down some more.
I suspect I will have to jack up the CPU volts to 1.45 before I can get a semi-stable stab at 4.3 GHz. Which probably will make it too hot.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
I suspect I will have to jack up the CPU volts to 1.45 before I can get a semi-stable stab at 4.3 GHz. Which probably will make it too hot.
Tom M
Are you using air or water cooling? If water think of using a bigger radiator or more fans on the one you have or both. Assuming you are in a home or office you could even run a small tube from the a/c vent to near the fans cooling the radiator.
Just got done putting modular cable on there and am getting 12 volts under load (BOINC).
I'm guessing you meant you put on a Molex cable and measured the 12V with a multimeter and came up with 12V under load?
Yes.
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
General advice is fit the largest radiator your case will allow. Plenty of 360mm AIO kits out there now. Some even with full coverage Threadripper coldplates.
Just ordered a MasterLiquid ML360 Mirror ARGB for just under $100 from NewEgg
The 3950x is currently "screaming" along at 4.225 GHz. Any higher caused a CPU overheat shutdown. But it will boot and run till the overheating. Trying to see if 4.3 GHz is "really" a reasonable goal for production on this system. Clearly air cooling is not cutting it for this level of OC.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Keith Myers wrote: The 8 pin
)
This is what I meant to say. I'm sorry that I didn't explicitly state "how many gpus he has plugged into the mobo", I thought it was implied.
Proud member of the Old Farts Association
Keith Myers wrote: General
)
Or in some of our cases, even if they don't fit in the case...they sit above them bahahahaha.....
Zalster wrote:Keith Myers
)
Hi Zalster,
Believe it or not, I was tempted to do just that! :^)
I noticed on your #12789230 computer that you are running 2 x 3080 12GB GPUs. Do you notice any difference in running Einstein with a 12GB GPU vs a 10GB GPU? I'm curious... yeah, I know... Curious George!! (I had two bananas this morning)
Proud member of the Old Farts Association
GWGeorge007 wrote: I can
)
Don't skimp on the radiator capacity. If building a combined cpu-gpu loop, you should be using at least dual radiators or a very large 480mm radiator.
From EK's website.
The rule of thumb is to use at least one 120mm radiator (section) per each water cooled component plus one additional section. For example, if you're liquid-cooling a CPU and a single high-performance graphics card, we recommend using at least one 360mm (3x 120mm) radiator for the best results.
Since you are going to end up with 4 active components you can see that a single 360mm will not be sufficient.
Also another rule of thumb is a max 10-14° C. delta rise of the water temp over the ambient temp. Best is a 4-6° rise over ambient.
On my cpu-gpu loop with dual 1080 Ti's and the 5950X, I run a 37° C. water temp with a 25°C. ambient temp which gives me a 12° C. delta.
I would like that to be better but I had to compromise on a standard 30mm thick HWL GTS360 radiator in the roof to clear the top of the mobo and memory. I fitted a AlphaCool XT45 360mm rad in the front as another compromise to clear the gpus with the pump mounted to the front rad. I would normally have used my default preference of a XSPC RX360 but it wasn't available at the time I needed it and was building.
My preference would have been dual XSPC RX360 rads for the loop.
GWGeorge007 wrote: I noticed
)
I have two 3080's in this host. Numbskull along with one of Ian's old 2080 Ti's.
The 12GB model runs about 10 seconds faster than the 10GB model both doing doubles. The 12GB model has 256 more CUDA cores than the 10GB model and slightly better memory bandwidth.
So yes, the 12GB models are faster running Einstein with our AIO app.
Ian&Steve C. wrote: if this
)
The PSU on the 3950x is a used EVGA 850 G3 one. The 24 and 8 pin cables came with it.
The 3950x box GPU is NOT giving me any trouble at all. But then I am running it under V0.92 :)
Just got done putting modular cable on there and am getting 12 volts under load (BOINC).
And when I shut down BOINC it still says 12 volts.
So I am guessing the 12 volt rail is not "the issue" with the CPU crashing on the 3950x. I am bringing down the CPU voltage stepping 1.39 to 1.38 etc. The displayed voltage under Asus sensors it a couple of hundredth below that.
Someplace under a displayed voltage of 1.31 volt I expect it to start crashing again.
Maybe I can get the CPU temp down some more.
I suspect I will have to jack up the CPU volts to 1.45 before I can get a semi-stable stab at 4.3 GHz. Which probably will make it too hot.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Quote:Just got done putting
)
I'm guessing you meant you put on a Molex cable and measured the 12V with a multimeter and came up with 12V under load?
Tom M wrote: Maybe I can get
)
Are you using air or water cooling? If water think of using a bigger radiator or more fans on the one you have or both. Assuming you are in a home or office you could even run a small tube from the a/c vent to near the fans cooling the radiator.
Keith Myers
)
Yes.
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Keith Myers wrote:General
)
Just ordered a MasterLiquid ML360 Mirror ARGB for just under $100 from NewEgg
The 3950x is currently "screaming" along at 4.225 GHz. Any higher caused a CPU overheat shutdown. But it will boot and run till the overheating. Trying to see if 4.3 GHz is "really" a reasonable goal for production on this system. Clearly air cooling is not cutting it for this level of OC.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!