Ian, I can hold my finger directly on the heatsink on both Epyc hosts. They are just barely warm to the touch.
Not even skin warmth. My heatsink on this daily driver C7H board is much warmer and and is uncomfortable to leave my finger on for very long. It is measuring at 64° C. right now via the asus-wmi-sensors report.
asuswmisensors-isa-0000
Adapter: ISA adapter
CPU Core Voltage: 1.23 V
CPU SOC Voltage: 1.08 V
DRAM Voltage: 1.41 V
VDDP Voltage: 665.00 mV
1.8V PLL Voltage: 1.81 V
+12V Voltage: 11.88 V
+5V Voltage: 4.99 V
3VSB Voltage: 3.33 V
VBAT Voltage: 3.18 V
AVCC3 Voltage: 3.33 V
SB 1.05V Voltage: 1.08 V
CPU Core Voltage: 1.25 V
CPU SOC Voltage: 1.09 V
DRAM Voltage: 1.46 V
CPU Fan: 1928 RPM
Chassis Fan 1: 0 RPM
Chassis Fan 2: 0 RPM
Chassis Fan 3: 0 RPM
HAMP Fan: 0 RPM
Water Pump: 0 RPM
CPU OPT: 0 RPM
Water Flow: 0 RPM
AIO Pump: 0 RPM
CPU Temperature: +83.0°C
CPU Socket Temperature: +59.0°C
Motherboard Temperature: +36.0°C
Chipset Temperature: +51.0°C
Tsensor 1 Temperature: +216.0°C CPU VRM Temperature: +64.0°C
Water In: +216.0°C
Water Out: +37.0°C
CPU VRM Output Current: 140.00 A
Based on my sense of heat via my finger, the heatsinks on the Epyc hosts feel half as hot as my Ryzen 5950X host. So that squares with my heat gun assessment.
Since I removed the side of the case to get the heat gun right on top of the heatsink, the temp was reading lower than when the side panel is put back on. After the case is back to 15 minutes of buttoned up equalization the temp reading is now at 41° C.
Still not anywhere near a temp that is uncomfortable to keep skin in contact with indefinitely.
I know that you were concerned with the temps of your VRM's on your SuperMicro board. But that is a bad design with the VRM heatsink right in the blast furnace output of the second cpu cooler. Also that heatsink has to dissipate 2 X 200W of heat for both cpus.
You have both the same EPYCD8 and ROMED8-2T mobos as I have. Why don't you put your finger on each of those mobo heatsinks and report what you feel. Unless you have no airflow across the motherboards, I think you will report the same feeling of just barely warm to the touch on your boards also.
I know that you were concerned with the temps of your VRM's on your SuperMicro board. But that is a bad design with the VRM heatsink right in the blast furnace output of the second cpu cooler. Also that heatsink has to dissipate 2 X 200W of heat for both cpus.
and the heatsink is more than 2x the size of the heatsink on the Asrock boards. but I've only been talking about the ROMED8-2T board in the last few messages.
Keith Myers wrote:
You have both the same EPYCD8 and ROMED8-2T mobos as I have. Why don't you put your finger on each of those mobo heatsinks and report what you feel. Unless you have no airflow across the motherboards, I think you will report the same feeling of just barely warm to the touch on your boards also.
my first comment about the heatsink temperature being burning hot to the touch WAS with the ROMED8-2T board (with the 7443P CPU @200W, same as yours). I never did a touch test with the dual EPYC board since the SM board has actual VRM temp sensors and reports it in the IPMI.
and about your ASUS board. tell me how it makes sense that a board with a 10+2phase beefy VRM setup, with much better VRM cooling, on only a ~185W CPU runs significantly HOTTER than the Asrock board with a weak VRM, tiny heatsink, and a 200W CPU? it just doesn't add up.
you may have better case airflow than I reducing the heatsink surface temp to tolerable levels, but I still don't think it's the right software reading on the board since mine is so far off between software and physical heatsink temp. and again the heatsink will not be as hot as the component itself, which is what the software reading will be showing.
Ian, I can hold my finger directly on the heatsink on both Epyc hosts. They are just barely warm to the touch.
Not even skin warmth. My heatsink on this daily driver C7H board is much warmer and and is uncomfortable to leave my finger on for very long. It is measuring at 64° C. right now via the asus-wmi-sensors report.
59C VRM, 79C Ryzen 3900XT.
69C VRM, 100C i5-8600K.
100! Oh, thanks for making me test that! I need a bigger heatsink when I use the internal graphics and all 6 cores at once. Looks like it's been thermally throttling, and maybe that's why it crashes once per 24 hours, why can't they throttle BEFORE the temperature at which they screw up?
Keith Myers wrote:
Based on my sense of heat via my finger,
I use my finger to test everything in computers. Things like disks should allow me to keep my finger on indefinitely. CPU/GPU heatsinks should be comfortable for a few seconds at least.
Actually, you could have a hot VRM and a heatsink that's not attached properly. My i5 I can keep my finger on the heatsink for several seconds before it's slightly uncomfortable, so it isn't getting anything like the 100C through it, no copper though, aluminium/steel basic stock Intel heatsink. You'd think they'd design a heatsink that worked with their own CPUs.
If this page takes an hour to load, reduce posts per page to 20 in your settings, then the tinpot 486 Einstein uses can handle it.
but I still don't think it's the right software reading on the board since mine is so far off between software and physical heatsink temp. and again the heatsink will not be as hot as the component itself, which is what the software reading will be showing.
Depending on the software, you can be looking at the wrong reading. There are so many sensors and the software sometimes displays the wrong one and labels it as something else. I too have never known any VRM be under 50-60C (MB reading) at full use, and they always burn my fingers.
If this page takes an hour to load, reduce posts per page to 20 in your settings, then the tinpot 486 Einstein uses can handle it.
and about your ASUS board. tell me how it makes sense that a board with a 10+2phase beefy VRM setup, with much better VRM cooling, on only a ~185W CPU runs significantly HOTTER than the Asrock board with a weak VRM, tiny heatsink, and a 200W CPU? it just doesn't add up.
you may have better case airflow than I reducing the heatsink surface temp to tolerable levels, but I still don't think it's the right software reading on the board since mine is so far off between software and physical heatsink temp. and again the heatsink will not be as hot as the component itself, which is what the software reading will be showing.
Ha! I would not call the POS heatsink on the C7H a proper heatsink. It is just a small block of aluminum. NO fins at all to dissipate heat. Just one single solid blade of only 1/4" depth cut into it. It is a joke of a heatsink. Pure eye candy of no functional use.
It occupies a position midway between the front fans and the rear fans up against the glass side panel. It gets pretty poor if any airflow.
Plus it is dissipating 205W of power as shown by Zenpower.
The Epyc mobos get much better airflow because of the dedicated fan mid-chassis blowing on the front of the boards.
I agree that the actual temp of the physical silicon is going to be hotter than the heatsink itself. The coils on the backside of the heatsink at the level of the motherboard and shaded by the heatsink from the cooling fan heatsink are much more toasty than the heatsink.
Without removing the heatsink and getting my K type thermocouple in direct contact with the VRM chips, we are not going to come to some definitive answer. All I can say is that the arbitrary sensor names shown in sensors do respond to cool spray or heat gun directed at the heat sink. Whether any of those are in fact actually a real telemetry reading from the VRM's, we are never going to know unless the mobo maker designers tell us the real story.
There is a real SIO chip on the mobo. It must be doing something even though the reported outputs make no sense because there is no published data sheets on the chip and no Linux distro provides the correct conf file to do proper scaling of the outputs.
I'll direct you to this series of posts at Level1Techs forum discussing concern over VRM temps. Nefastor is an electrical engineer. He shows FLIR images of the ROMED8-2T board under load with CyberPunk 2077 and also the stress burn-in test of Passmark.
He reports a max of 32° C. for the VRM heatsink. He mentions the BMC reports a "board temperature of 33° C. so he surmises that may be monitoring somewhere close to the VRM's.
Yeah I saw those posts. I was disappointed that most of his FLIR pics were from low load situations. Passmark is not a particularly heavy load. Nor is gaming and he even mentions it’s not full load. And he’s running a 7282, a 120W low power CPU.
also in that thread you’ll find another user with a 120W EPYC running on a Supermicro H12SSL-I board, that DOES properly report CPU VRM temps, this board also has a near identical VRM and heatsink setup to the AsRock boards. And he reported ~50C VRM temps under actual full load, at only 120W. That’s more in line with what I would expect. And of course higher for a more powerful CPU.
I still can't reconcile your observational differences between your finger touch test of your ROMED8-2T board with the same 7443P cpu under identical Universe cpu task 200W loads compared to mine.
Neither of my boards have heatsinks that are burning hot to the touch.
I can only assume you have complete "dead air" around yours. Pretty sure you are in mining chassis also so assume no closed case with poor airflow.
Ian, I can hold my finger
)
Ian, I can hold my finger directly on the heatsink on both Epyc hosts. They are just barely warm to the touch.
Not even skin warmth. My heatsink on this daily driver C7H board is much warmer and and is uncomfortable to leave my finger on for very long. It is measuring at 64° C. right now via the asus-wmi-sensors report.
asuswmisensors-isa-0000
Adapter: ISA adapter
CPU Core Voltage: 1.23 V
CPU SOC Voltage: 1.08 V
DRAM Voltage: 1.41 V
VDDP Voltage: 665.00 mV
1.8V PLL Voltage: 1.81 V
+12V Voltage: 11.88 V
+5V Voltage: 4.99 V
3VSB Voltage: 3.33 V
VBAT Voltage: 3.18 V
AVCC3 Voltage: 3.33 V
SB 1.05V Voltage: 1.08 V
CPU Core Voltage: 1.25 V
CPU SOC Voltage: 1.09 V
DRAM Voltage: 1.46 V
CPU Fan: 1928 RPM
Chassis Fan 1: 0 RPM
Chassis Fan 2: 0 RPM
Chassis Fan 3: 0 RPM
HAMP Fan: 0 RPM
Water Pump: 0 RPM
CPU OPT: 0 RPM
Water Flow: 0 RPM
AIO Pump: 0 RPM
CPU Temperature: +83.0°C
CPU Socket Temperature: +59.0°C
Motherboard Temperature: +36.0°C
Chipset Temperature: +51.0°C
Tsensor 1 Temperature: +216.0°C
CPU VRM Temperature: +64.0°C
Water In: +216.0°C
Water Out: +37.0°C
CPU VRM Output Current: 140.00 A
Based on my sense of heat via my finger, the heatsinks on the Epyc hosts feel half as hot as my Ryzen 5950X host. So that squares with my heat gun assessment.
Since I removed the side of
)
Since I removed the side of the case to get the heat gun right on top of the heatsink, the temp was reading lower than when the side panel is put back on. After the case is back to 15 minutes of buttoned up equalization the temp reading is now at 41° C.
Still not anywhere near a temp that is uncomfortable to keep skin in contact with indefinitely.
I know that you were
)
I know that you were concerned with the temps of your VRM's on your SuperMicro board. But that is a bad design with the VRM heatsink right in the blast furnace output of the second cpu cooler. Also that heatsink has to dissipate 2 X 200W of heat for both cpus.
You have both the same EPYCD8 and ROMED8-2T mobos as I have. Why don't you put your finger on each of those mobo heatsinks and report what you feel. Unless you have no airflow across the motherboards, I think you will report the same feeling of just barely warm to the touch on your boards also.
Keith Myers wrote: I know
)
and the heatsink is more than 2x the size of the heatsink on the Asrock boards. but I've only been talking about the ROMED8-2T board in the last few messages.
my first comment about the heatsink temperature being burning hot to the touch WAS with the ROMED8-2T board (with the 7443P CPU @200W, same as yours). I never did a touch test with the dual EPYC board since the SM board has actual VRM temp sensors and reports it in the IPMI.
and about your ASUS board. tell me how it makes sense that a board with a 10+2phase beefy VRM setup, with much better VRM cooling, on only a ~185W CPU runs significantly HOTTER than the Asrock board with a weak VRM, tiny heatsink, and a 200W CPU? it just doesn't add up.
you may have better case airflow than I reducing the heatsink surface temp to tolerable levels, but I still don't think it's the right software reading on the board since mine is so far off between software and physical heatsink temp. and again the heatsink will not be as hot as the component itself, which is what the software reading will be showing.
_________________________________________________________________________
Keith Myers wrote:Ian, I can
)
59C VRM, 79C Ryzen 3900XT.
69C VRM, 100C i5-8600K.
100! Oh, thanks for making me test that! I need a bigger heatsink when I use the internal graphics and all 6 cores at once. Looks like it's been thermally throttling, and maybe that's why it crashes once per 24 hours, why can't they throttle BEFORE the temperature at which they screw up?
I use my finger to test everything in computers. Things like disks should allow me to keep my finger on indefinitely. CPU/GPU heatsinks should be comfortable for a few seconds at least.
Actually, you could have a hot VRM and a heatsink that's not attached properly. My i5 I can keep my finger on the heatsink for several seconds before it's slightly uncomfortable, so it isn't getting anything like the 100C through it, no copper though, aluminium/steel basic stock Intel heatsink. You'd think they'd design a heatsink that worked with their own CPUs.
If this page takes an hour to load, reduce posts per page to 20 in your settings, then the tinpot 486 Einstein uses can handle it.
Ian&Steve C. wrote:but I
)
Depending on the software, you can be looking at the wrong reading. There are so many sensors and the software sometimes displays the wrong one and labels it as something else. I too have never known any VRM be under 50-60C (MB reading) at full use, and they always burn my fingers.
If this page takes an hour to load, reduce posts per page to 20 in your settings, then the tinpot 486 Einstein uses can handle it.
Ian&Steve C. wrote:and
)
Ha! I would not call the POS heatsink on the C7H a proper heatsink. It is just a small block of aluminum. NO fins at all to dissipate heat. Just one single solid blade of only 1/4" depth cut into it. It is a joke of a heatsink. Pure eye candy of no functional use.
It occupies a position midway between the front fans and the rear fans up against the glass side panel. It gets pretty poor if any airflow.
Plus it is dissipating 205W of power as shown by Zenpower.
The Epyc mobos get much better airflow because of the dedicated fan mid-chassis blowing on the front of the boards.
I agree that the actual temp of the physical silicon is going to be hotter than the heatsink itself. The coils on the backside of the heatsink at the level of the motherboard and shaded by the heatsink from the cooling fan heatsink are much more toasty than the heatsink.
Without removing the heatsink and getting my K type thermocouple in direct contact with the VRM chips, we are not going to come to some definitive answer. All I can say is that the arbitrary sensor names shown in sensors do respond to cool spray or heat gun directed at the heat sink. Whether any of those are in fact actually a real telemetry reading from the VRM's, we are never going to know unless the mobo maker designers tell us the real story.
There is a real SIO chip on the mobo. It must be doing something even though the reported outputs make no sense because there is no published data sheets on the chip and no Linux distro provides the correct conf file to do proper scaling of the outputs.
I'll direct you to this
)
I'll direct you to this series of posts at Level1Techs forum discussing concern over VRM temps. Nefastor is an electrical engineer. He shows FLIR images of the ROMED8-2T board under load with CyberPunk 2077 and also the stress burn-in test of Passmark.
AMD Epyc Milan Workstation Questions
He reports a max of 32° C. for the VRM heatsink. He mentions the BMC reports a "board temperature of 33° C. so he surmises that may be monitoring somewhere close to the VRM's.
ROMED8-2T VRM FLIR image temps
[Edit]
Just looked at the sensors via the BMC management page. "MB temperature" reports as 37° C.
Sensors output of AUXTIN2 matches 37° C. So that sensor label seems to match the BMC MB temp label.
Whether that is the VRM's is still a guess.
Yeah I saw those posts. I was
)
Yeah I saw those posts. I was disappointed that most of his FLIR pics were from low load situations. Passmark is not a particularly heavy load. Nor is gaming and he even mentions it’s not full load. And he’s running a 7282, a 120W low power CPU.
also in that thread you’ll find another user with a 120W EPYC running on a Supermicro H12SSL-I board, that DOES properly report CPU VRM temps, this board also has a near identical VRM and heatsink setup to the AsRock boards. And he reported ~50C VRM temps under actual full load, at only 120W. That’s more in line with what I would expect. And of course higher for a more powerful CPU.
big difference between a 120W CPU and a 200W CPU.
_________________________________________________________________________
I still can't reconcile your
)
I still can't reconcile your observational differences between your finger touch test of your ROMED8-2T board with the same 7443P cpu under identical Universe cpu task 200W loads compared to mine.
Neither of my boards have heatsinks that are burning hot to the touch.
I can only assume you have complete "dead air" around yours. Pretty sure you are in mining chassis also so assume no closed case with poor airflow.