|
Post by stormchaser on Jun 30, 2019 12:41:34 GMT -5
Nice NB frequency. Most FX chips arent too stable beyond 2.5ghz. You know, its funny you say that because I just resolved some reliability problems in reference to my NB speed.. I had the NB clocked at 2640mhz and the NB voltage at 1.35 and it was hanging up about half way through the AIDA64 Cache and Memory Benchmark. No matter what voltage, it would just lock the system up half way through the test. So after trying 1.4v I did the opposite and tried less voltage. Sure enough using LESS voltage did the trick! I don't know if this is just a lucky guess but it seems to have resolved my stability problems.
|
|
|
Post by ShrimpBrime on Jun 30, 2019 21:10:38 GMT -5
Sometimes less is more. Nice job.
|
|
|
Post by Vinster on Jul 1, 2019 12:22:26 GMT -5
like shrimp said... sometimes I'd bring the volts down to see what the system does. sometime you're hammering the volts harder that the chips can take and that adds to the instability
Vin
|
|
|
Post by stormchaser on Jul 2, 2019 14:42:20 GMT -5
So I got a little greedy and bumped up the clock speed to 4920Mhz @ ~ 1.5 vcore, its definitely not 100% stable but this is more of a web browsing machine anyway so I can ride that fine line of reliability.
|
|
|
Post by Bones on Jul 3, 2019 15:17:22 GMT -5
That's not bad at all.
|
|
|
Post by Vinster on Jul 3, 2019 18:41:37 GMT -5
what are you using to cool it?
Vin
|
|
|
Post by stormchaser on Jul 4, 2019 9:20:58 GMT -5
what are you using to cool it? Vin Im using a CoolerMaster Hyper 212, I also have a 140MM fan above and 120MM fan behind the cooler. The Biostar TA970 motherboard Im using has heatsinks on the VRM area, so as long as I don't overdue it with a game or a torture test, those VRMs stay pretty cool even at the higher clocks. EDIT: This particular Hyper 212 has been "modded" to say the least. I broke one of the studs off so its only held on with the three remaining studs, I had to bend the bracket to make it work and actually cool like it is supposed to. I will get pics at some point so you can see what Im talking about.
|
|
|
Post by Vinster on Jul 4, 2019 11:02:46 GMT -5
Ok, just make sure you get enough airflow around that cooler, that the CPU's TDP is only 61 degrees as per the AMD Spec, so keep an eye on the temps.
Vin
|
|
|
Post by stormchaser on Jul 4, 2019 11:14:15 GMT -5
Yeah I have core - temp running in my taskbar so I have real-time temp monitoring. I've heard the FX doesn't actually have a sensor, its just a mathematical calculation based on some CPU indications. I've heard, the higher the temp the more accurate it is. Because for now, it runs anywhere from 70*F to 95* at idle and 125-130*F under full load.
|
|
|
Post by ShrimpBrime on Jul 4, 2019 12:02:34 GMT -5
I've read that recently also. But when I was cooling sub zero, the temps looked pretty accurate to me.... -74 degree with DICE cooling seemed accurate for that cooling used.
70-95f seems accurate (enough) because I'd be willing to bet your ambient room temps are between those figures.
So what is doing the mathematical calculation? The Bios chip now can convert and calculate on it's own? I think not. There are Cpu core temps and board temps. These temps are given from a sensor, not a magical calculation actually. HOWEVER, the sensor may not be well calibrated at certain temperatures above or below a certain degree.
In short, you really only need to pay most attention to the 130F load temps.
|
|
|
Post by Mr.Scott on Jul 4, 2019 12:21:12 GMT -5
AMD processors have been like that for a long time, not just FX. There is no core sensor per-say, there is a die sensor. It is not accurate at all, but over 40c it is as close as it'll get. Sub ambient temps, you're better off looking at the socket temp. Neither will be very accurate. That is what K probes are for.
|
|
|
Post by ShrimpBrime on Jul 4, 2019 12:30:50 GMT -5
The core is located in the die. The sensor is simply not reading any core nor a single core. That's why we cannot call is a "core temp" and get away with it.
But then it doesn't make sense to say it's not accurate at all, but then becomes accurate suddenly at 40c, now this is your accurate "core temp" even though it's actually a die temp. At 40c the cores and die are closer together in temp to become accurate.
K probes are quite a bit more accurate. However doesn't change the location of the temperature sensor in the processor. Therefor the reading on the probes is also inaccurate.
Thus there is some degree of accuracy at idle and load or inaccurate at both states.
I can probe the DICE/LN2 pot and it should almost always differ from a core/board reading. I can view bios reading and view windows reading and always have a difference.
So then is the sensor inaccurate? The location of temp sensor inaccurate? Software vs hardware vs prove = accurate?
Meh, -148c. close enough. ++ 2.0v FX Phenom Ryzen ATHLON.....
|
|
|
Post by Vinster on Jul 4, 2019 22:04:52 GMT -5
Software vs hardware vs prove = accurate? Pick 1 and stick with it. if benching I'd integrate a probe in the block and use that all day.
Vin
|
|