Link Copied
@Cris86 wrote:As soon as i can i will try to check if the screw are too tight and maybe cause some bending, who knows. Maybe is really a stupid thing like that which cause this extreme temp only why doing something a bit energy-demanding.
no no, that is not what i meant.
It's more like the water block mount, does not has enough pressure on the CPU IHS to make it straight/flat (if it's crooked).
And without a massive back plate, it can not deliver enough of that pressure, instead the motherboard bends.
Sorry, i am not so good at explainging stuff.
However, i did forget that i can use my backup hardware to illustrate what i mean by "bending the motherboard".
The example here is not perfect, the LGA2011 has a stock back plate with a thickness of 2,5mm.....there is no way that can bend.
And the stock CPU mount, has another plate (that is where those 4 bolts sit in) with a thickness of 1,5mm, nothing is going anywhere.
But, let's assume, that without additional back plate, the motherboard can indeed bend....
the setup on the right delivers 100% of the pressure to the CPU, as nothing else can bend, only the CPU can bend.
And you want to bend the CPU, if it's not crooked, then nothing will happen.
So a contact frame will do exactly that, it will force the CPU to the stock back plate, and the stock back plate would be obviously not crooked in any way.....i hope at least.
@n_scott_pearson wrote:Fact is, the processor can heat up faster than any cooling current solution can respond to.
Yes....with a tiny transistors, yes it would, except....
hmm, seeing the spikes on the temperature i would claim, that the heat transfer from the CPU to what sits on the CPU is not good enough.
I am currently running that CPU with a MO RA3 radiator and a big water reservoir, in total about 2,5L of water capacity.
On 50% rpm of the pump, it takes almost 3 minutes for the water to complete 1 cycle.
This means, for that time, the water temperature stays constant, i should be able to do literally anything i want, and the CPU temperature should not rise within these 3minutes.
When i run any unlimited stress test, the temperature spikes immidiately to over 80°C, there is barely any delay, no slow rising of the temperature, it jumps from 30°C to 95°C just like that, and few seconds later thermal throttling hits the CPU.
in the screen shot, the 'expected' part of the graph, actually happens bcoz i set TDP limit to 25W, then i turn off the water pump, and then i run a stress test, the water in the water block + the water block itself has the capacity to keep temperatures under 60°C for about 1~2 minutes, and with pump on, the temperature spikes down.
With pump turned on....the 'how?', is at TDP of 250W, it takes like 1 second of CPU memory stress test, for the CPU to reach 92°C, and when i stop the stress test, as seen above, it takes also like 1 second for the temperature to spike down - to exactly, same value as bevore the test.
Now if i set XTU logging monitor to 1000ms, the spikes will be logged with a duration of 1second, so from idle temp to 92°C in 1 second.
If i set it to 100ms, i get a duration of those spikes at about 1,2seconds (1200ms).
Raw monitoring log:
Raw as graph:
So the vertical value is the temperature (0°C to 100°C), the horizontal is "value x 100ms" (105 x100ms = 10500ms = 10,5seconds).
Same data but shown in XTU:
How can it jump within 1 or 1,2 seconds from 30°C to almost 100°C and then spike down, as if nothing happened?
That is the question here.
So obviously, the cooling perfomance musst be there, since it spikes down, in same, if not quicker time.
@Cris86 wrote:@KrissyG You've explained it right, i simply replied immediately without reading what i wrote which is the opposite of what my mind was thinking about at that time, sorry.
As soon as i can i will investigate more about everything in order to start solve some of my questions, thanks
@KrissyGabout your latest post which for me is way to advanced, i totally agree on the question about how is possible to pass from 30° to 100° and immediately back to 30° in fractions of seconds.
take your time, none of us is going anywhere here.
FYI, you can let the XTU save all the data in form of a text file for example, you just have to click on the 'file logging' button, and at some point click on it again to stop logging. I man you can let it log 24/7 too. And right next to that button if a 'folder' icon, if you click on that, it will show you the log files, you can open them in notepad.
The last thing you may need, is to set the logging intervals, which is in milliseconds.
so 100ms = 0,1 of a second
1000ms = 1 second.
So you can make it save the data even 10 times per second if you want to.
Unfortunately, the log files are sometimes skipping some values, as seen here "Package TDP" is missing from the first entry, or the interval for package TDP can not be set to 100ms and therefore it shows it only once per second:
This data allows me to take a closer look at the values, which XTU graph is not designed to do.
So this here is Excel which i fed with only the values of temperatures, and the duration of the rising temperatures can be seen as more or less 1 second (the space between 1sec and 2sec is the spike we see on XTU).
In fact, it is not just a spike, the CPU takes some power, apparently checks for the temperature, and then it does that in several steps again and again.....marked on the graph with red dots:
the graph for power usage looks similar, the ptoblem is as you agreed, that the temperatures should not rise as extremely as it does.
If you want to, you can make such log file, then post it here as attachement, and i will transform it into a graph with way higher resolution. Except, like i mentioned, some values can not be looged with an interval of 10 times per second, and some values are indeed missing sometimes anyway.
i think i found the problem, the actual chip is thinner, and we all know that thin metals....or silicone, heats up way quicker.
I would claim to have found the reason for the spikes as seen in this screenshot:
@chugzillafx wrote:the only thing that program doesn't allow us to do is set a loadline calibration, but you can undervolt there.
But what if GCC can?
There is however a problem, maybe only on my side?
The GCC returns the values to defualt each time i restart the PC.
HOWEVER! the default is nothing else but BIOS, so, once you change those settings in BIOS, GCC will return to those after restart.
And since BIOS has now a search box you can type in and search the setting, it is extremely easy to set it.
