Asrock X370 Fatal1ty Gaming K4 Review

That is some good numbers right there alright, if only there was a way to set a faster memory divider than the XMP of 3200MHz I'd love to see if this kit of F4-3200C15D-16GVK has 3400 and even 3600MHz in it and what performance would be like.
 




Don't let the chart fool you, the difference in Cinebench between 2133MHz and 3200MHz memory speed is not significant, but there is a small bump in this test when switching to faster memory Cinebench is just not bandwidth limited but all about raw CPU speed which is why the faster memory only has a small impact here.





That is one hell of a difference in Cloudgate simply switching to 3200MHz memory speed you really see how sensitive to bandwidth and latency Zen is here. The CCX just loves the extra speed it can run at with faster memory.



 

The Stalker Call of Pripyat benchmark is an excellent test of a system as a whole, using X-Ray 1.6 the engine places every bit as much stress on the GPU as it does the CPU and memory subsystem, if there is a bottleneck this benchmark will find it. Above are the settings that were used.



 

This is some interesting results really highlighting how you can potentially cripple Zen gaming performance if you use slow memory. Average performance changes very little but those minimum frame rates increase by 47.1% when increasing to 3200MHz memory speed! Now that is what you call a night and day difference and makes me wish for dividers above the XMP maximum even more to see what else Zen has to give.


Test your Metal!

I haven't done this in a while, and every time I do it always comes back with a slightly different name where I've not settled on one but I think this is a good one, it's time to "test your metal!". The purpose in this section is to test thermodynamic efficiency of any heatsinks.

For those that don't know the most efficient type of heatsink is always one that uses high grade thermal interface material, the thinner the better, good contact between heatsink and IC (a good, solid imprint of the heatsinked ICs is a sure indication of optimal contact), solid way of mounting the heatsinks, and the heatsinks themselves will have a lot of fins to maximise cooling capacity and take advantage of any natural airflow with the base of the heatsink being free of any milling marks that would hinder thermal efficiency.

For testing I will be using a DT8380 infrared temperature gun, accurate to +/- 2%.

With that all said, let's take a look at the K4s heatsinks.





I'm pleased to see Asrock have not changed with their approach to heatsinks, the base is nicely finished with proper standoffs to prevent board warping and solid contact between IC and TIM is not a problem in the slightest. For consistency with other X370 reviews I've done I'll replace the Mosfet thermal pads with some Laird 6W/mK pads - the very same NASA used to use on the space shuttle. This is to ensure that the ground is as equal as possible and any differences are related to the thermal design and not the thermal interface material.

The one area of concern is the lack of fins on the Mosfet heatsinks which mean any natural airflow is going to be wasted. You see Mosfets get hot, very hot, particularly when a system is under load and even more so when overclocking. For the latter case particularly this is going to lead to hotspots and you need an effective heatsink. As for the southbridge heatsink the lack of fins should not pose much of a issue as the southbridge doesn't create much heat but why have any component getting hotter than it needs to, right?

To testing then!





So what can we take away from these results? The Mosfets stay well within their operating temperature range but the amount of heat being generated at 1.37v+ just cannot be effectively dealt with by natural airflow and even active airflow directly over the heatsinks will not be as effective as it could be due to the lack of fins on the Mosfet heatsinks. Under prolonged load duration this leads to a lot of heat inside the case where it cannot be effectively pulled away from the heatsinks with airflow. These heatsinks are quite effective, but the lack of fins prevents them from reaching maximum effectiveness. If the Mosfet heatsinks had been finned all the way across and a heatpipe been present in my test system at least the 1.37v side Mosfet temps would have been 5-7c lower which in terms of preventing heat build-up inside a case is quite substantial.

As expected the southbridge faired absolutely fine, a finned design would improve temps for it but there is nothing to worry about, I'd just like to see a southbridge heatsink that is smaller.

Overall the Mosfet and southbridge heatsinks perform decently, but there is room for improvement particularly for the Mosfet heatsinks a finned design with a heatpipe to spread the thermal load would have seen them doing a fair bit better.


BIOS / UEFI

The Asrock Fatal1ty Gaming K4 is simply filled to the brim with options in the UEFI so I have tried to limit the screens to what you will be most interested in while also giving a peek at just how many options this board has elsewhere in the UEFI. All screenshots have been taken from UEFI 2.2.

 





Just look at all those options and we are still on the same UEFI page! There is a very real chance Asrock have more options on their single OC Tweaker page than the AX370 Aorus gaming 5 has in the entirety of it's UEFI. Are you paying attention Gigabyte? You should be because this is how you make a good, no great, UEFI.




Let's take note of the ability to save UEFI settings to removable media such as a USB drive, very handy for quickly backing up settings instead of jotting them all down prior to a update or useful for sharing with friends or fellow overclockers who have the same board. I sense a very beneficial ecosystem could develop from this simple feature.



 

 

 



That brings us to the end of what I think you all would be most interested in, the Fatal1ty UEFI is simply superb, well laid out and enough options to make your head spin. There's a few niggles but they are very minor ones, simply astoundingly good work by Asrock, showing the Gigabyte AX370 Aorus Gaming 5 how it should be done. Fatality!


Overclocking

Before we get in to this it's worth reminding everybody that you play the silicon lottery when overclocking not all CPUs and memory are created equal, as such any overclocking results you see should be taken as a rough guideline only of what you may or may not be able to achieve.

So with that said, what can you realistically expect to get from a Zen CPU? On the whole 3.7 - 3.9GHz you can consider to be fairly average for most Zen chips, with more exotic cooling and higher voltages you might be able to attain 4.0 - 4.1GHz, and do remember that with SMT on it will play a role in your maximum OC as will it possibly play a role in some applications performance - not all software benefits from SMT and some applications will actually lose performance with SMT on. For these overclocking tests I'll favour absolute core speed which means I'll be seeing how far I can go with SMT OFF.

Let's get to some overclocking then, which turned out to be very good like many other aspects of the K4. It is also worth mentioning that vcore stability is (As)rock solid with barely any variation and the G.Skill RipjawsV 3200MHz kit performs well in both sets of DIMM banks, the love between Asrock boards and G.Skill continues I'm happy to say.





Oh me, oh my, oh golly! Would you look at that! A mite over 4GHz for around the 1.3v mark anyone? I'll take that thank you kindly, what's that? You'll run 16GB RipjawsV at 3200MHz with 14-14-14-25 timings and give me over 60FPS minimum on the CoP sunshafts benchmark too? Well I'm not going to say no to that!

These results are simply awesome. If we cast our minds back a bit we will recall that with the R7 1700 at 3.7GHz with 2133MHz memory the CoP benchmark had a minimum FPS in the sunshaft test of 36.7, it then jumped to 54.1 when changing memory speed to 3200MHz, and now we see the 300MHz OC has netted another 6.6FPS on top of that. These results well and truly answer that gaming is great on AMD when paired with a good memory kit. What I find more interesting though is how memory frequency makes far more difference to Zen than overclocking it does, that'll be due to that pesky latency and CCX speed. My advice; Zen for gaming is awesome as long as you pair it with a 3000MHz or 3200MHz memory kit.

I also attempted overclocking with SMT ON, the resu

Edited by ket - 19 Jul 2017 at 2:10am

XHellAngelX wrote: Hello, how can you reach 3200 MHz RAM on this board ?

I reached 3200MHz with a lot of coaxing and fine tuning. With the newest AGESA reaching 3200 should be much easier than when I wrote this review things have matured substantially over the last year or so.