970M Pro 3 Problems

PetrolHead wrote: From what I understand, Akirareturns _has_ tried to set the timings manually. In any case the CR value has nothing to do with his issues. The timings are a lot looser than stock and even at stock settings the modules (as well as the IMC) can probably handle 1T. If not, that would result in instability, not changed timing values or lower-than-normal FSB clock speed.

Thanks. I'm fully aware of that. I've been using AMD solely for over a decade now.

What say you on the possibility that the board is throttling itself, thus keeping his changes from becoming static?

Also, if it was throttling, then the motherboard would definitely be faulty. The motherboard is supposed to support that processor, so at stock speeds there should be absolutely no issues - especially since Akirareturns is using the stock cooler that blows air downwards towards the motherboard, thus cooling also the socket and providing air flow over the VRM heat sinks. There should also be some headroom for overclocking even without those two 120 mm fans and despite that 4+1 power phase design.

All that's wrong? Let's see...

-"The motherboard is supposed to support that processor, so at stock speeds there should be absolutely no issues"

Officially supported? Check. (http://www.asrock.com/mb/AMD/970M%20Pro3/?cat=CPU) Also, if a processor is officially supported, the motherboard has to be able to cope with running the processor at stock speeds and with the stock fan. Otherwise ASRock is guilty of false advertising (it already kind of is, since the correct HT Link speed is not supported).

-"Akirareturns is using the stock cooler that blows air downwards towards the motherboard"

Correct and "required" for the processor he has by ASRock ("For cooling the CPU and its surrounding components, please install a CPU cooler with a top-down blowing design.").

-"thus cooling also the socket and providing air flow over the VRM heat sinks"

Correct, although the air flow over the VRM heat sinks is not huge.

-"There should also be some headroom for overclocking even without those two 120 mm fans and despite that 4+1 power phase design."

Correct based on my experience with this motherboard and my own processor, which has the same TDP as the FX-8150. The only thing holding me back on overclocking my processor is the socket temp, which gets to around 72 C during Prime95 when I overclock to 3.5 GHz from 3.2 GHz, but this is only because I don't have a top-down blowing CPU cooler at the moment. Also, this:

I wouldn't say I recommend overclocking either due to the power phase design. However, since overclocking at stock voltages causes less power drain and heat load than increasing the voltage, and since the board needs to have some sort of safety margins to officially support a 125W TDP processors, I'd still say there is some headroom for those who are willing to venture into those safety margins. Much naturally depends on the CPU. If the CPU can't be pushed further on stock voltages, then the headroom's pretty much gone.

Btw, a small note: None of the FX-8xxx processors are officially rated as 140W processors (TDP), only 125W and 95W. If TDP wattage is what you meant, that is.

Regarding the cooler, I don't think only AMD can be blamed here. Sure, the stock cooler is not hugely efficient and due to the small fan it's also quite noisy. I've also read some people having temperature issues, even though the stock cooler should in principle be guaranteed to be enough. Maybe the stock cooler wasn't properly installed or the case was poorly ventilated, who knows. What I know of my own experience is that a stock cooler from an FX-8xxx series kept my CPU in practice as cool as my current aftermarket CPU cooler does, and my current cooler is recommended for 130W TDP. Thus, a 125W TDP shouldn't be a problem, but maybe the FX-8xxxs aren't really 125W TDP?

The way I see it is this: Regardless of what AMD says the TDP is and what sort of stock cooler it provides with its processors, ASRock must have tested all the processors listed in the supported processors list and those tests must have been performed with stock coolers at the very least. If ASRock just says a certain CPU is supported, then it should work using the stock cooler. If an aftermarket cooler is required, then this should be stated very clearly. If no such requirements are stated and a supported CPU runs into problems using the stock cooler, then either ASRock's testing procedure is flawed or they just don't care if components get cooked. Either way, it's on them, not AMD.

I'm about to go off on a tangent here, so apologies in advance.

It's interesting to note that ASRock doesn't list a top-down CPU cooler as a requirement for my current processor, but it is required for the FX-8150. Both processors have a 125W TDP and both processors actually have identical stock coolers (or so I've been told - I got mine used and it came with a newer CPU cooler from an FX-8xxx processor, which I've since "upgraded"). Although the socket temperature stays below the "safe" limit of 70C during stress testing at stock speeds, it still maxes out somewhere above 60C (69C with CPU-NB/HT @ 2200 MHz and RAM @ 1600 MHz), which is less than ideal. Also, I worry about the VRMs, so I've been planning on switching my current aftermarket cooler to a top-down model. The problem is that there aren't many around that are designed for efficient processors. Last time I checked the best I could find (that fits my mATX case) was Noctua NH-C14S, which has a maximum recommended TDP value of 140W and only a low OC potential according to Noctua. Thus, as far as CPU cooling goes, it's not a big improvement over my current - or the FX-8xxx series stock - cooler. The socket, RAM and VRMs are hopefully another matter.


Edited by PetrolHead - 26 Nov 2015 at 9:09am

PetrolHead wrote: I checked my BIOS settings and the board doesn't indeed seem to support HT speeds beyond 2400. The plus side is that this shouldn't affect performance noticeably, as memory speed doesn't really matter apart from synthetic tests and some specific tasks.

I have to correct myself a bit, as I've made an error here: HT Link speed doesn't affect memory performance, CPU-NB speed does. HT Link speed is the speed with which the CPU communicates with the NB chip (not the CPU-NB) and thus affects things like PCI-e performance, but as 2600 MHz should provide enough bandwidth even for a multi-GPU setup, a speed of 2400 MHz should not be a bottleneck in most systems.