ASRock 990FX Extreme 4 Motherboard Review

ASRock has given us an opportunity to review one of their midrange AMD boards, the 990FX Extreme4. Although the word ‘Extreme’ is in the name, the number (4) shows us that it falls in the midrange between the Extreme3 and Extreme7 from their 990FX line. As with a lot of midrange offerings, they do what you need them to do, but may not necessarily give you everything you want. Let’s take a closer look at this board and see if it falls into line with that thinking.

Taking a look at the board

I was really anxious when I saw that UPS dropped off the package at my house. What I really wanted to do was put this board on the test bed and not turn it off unless it was dead or I ran my power bill too high for the summer. Yet, once I opened up the package, I realized that I had to do the photo shoot before trying to torture this board to oblivion.

The ASRock box is pretty generic, and is completely geared toward the gaming community. When you look at the box, and open up the flap to look at all the specs, it screams out “Hey, you gamer, look at all these cool things I can do!” The ASRock box has two compartments: the first one holds the motherboard and the other one holds the accessories. You can see the accessories in the following section.

When you finally take a look at the board, after being enthralled by the amazing box, you realize that there is not much to its appearance. In a lot of ways, it’s a bit boring. Now don’t get me wrong: I love the color scheme, it reminds me of the old Asus socket 939 boards. I grew up on overclocking with those boards. Yet, if you are in the market for components to build a budget gaming computer, this board will come up and when you look at it and compare it to the Gigabyte and Asus boards that are priced close to the ASRock Extreme4, these looks are not going to win.

For the CPU, the board is packed with a 8+2 power phase. What that means is that 8 phases are dedicated to the CPU power plane, and 2 are dedicated to the CPU-NB power plane. This is pretty typical for AMD motherboards. This power will be efficient for any voltage you want to pack into your Phenom II or Bulldozer CPU. The power plane uses an old L6717A voltage regulator by STMicroelectronics. Attached to each phase line are two MOSFETs with the cleaning inductor and capacitor for cleaning the signal. As for the DDR power plane, you can see that it is nicely packed down by the southbridge. This keeps the socket clean and away from the CPU fan.

When looking at the CPU power plane, you probably noticed something that should not be there. Yes, that is the BIOS battery. The BIOS battery is horribly placed on this board and will be covered by the heatsink – you cannot access it for those “oh crap” moments when you really mess up the settings. Next to the battery is the CLR CMOS jumper, another horrible choice of placement, although there is a CLR CMOS button on the back plane of the board, which somewhat makes up for the bad placement. It will be interesting to see what happens when I try LN2 or DICE with this board. I am afraid that the battery may freeze over and cause a short if it starts to melt ice around it.

Moving over to the backplane, we see everything is typical. There are eight USB slots, two of which are USB3. There is 7.1 channel audio controlled by a Realtek controller, a Broadcom 10/100/1000 MBps network port, SPDIF and optical, along with your classic PS2 keyboard and mouse ports. Firewire and eSATA did manage to get placed on this board as well.

The 990FX and SB950 are the controllers for this board. The 990FX takes care of PCI-E lanes to the CPU while the Southbridge takes care of everything else, controlling the SATA and IDE ports along with the BIOS. The SB950 southbridge only controls six out of the eight SATA3 ports found on this board: a Marvell chip was added on to take care of two other ports. The six SATA3 ports controlled by the southbridge support RAID in various flavors.

I really would not say that the design of this board is unique, or the color scheme for that matter. I have seen the blue and white on brown PCB for quite some time. What I have not seen is the gold caps being used by motherboard manufacturers. They are a bit difficult to implement into a board color scheme. They fit in pretty well for this motherboard, and as you can tell, I fell in love taking photos of them. The caps are pretty much the only part of the color scheme that makes the board stand out.

Specifications

Accessories

The accessories are pretty typical. You have your power cables, PATA and SATA cables, and your USB brackets for the rear. A nice plus is the front USB 3 bracket. If you have some extra room and your case does not support USB3 yet, this could be a nice add-on.

The one accessory I am going to talk most about is the optional fan you can put onto your board. First off,  I don’t know why is this not already on the board when it is shipped. Most people would use it, and could use it depending on the airflow in their case. I do understand why you would want to swap between a fan or a fanless setup: some people like to keep their boxes quiet and it is nice to have that option. Although, there is one more problem: the screws that are used to hold these two parts down are a bit rubbish, or it could be the mounts themselves. It required the use of the small screwdriver that they give you and one of my own to get each screw out. ASrock could have done better with this. Yet, truth be told, how many times are you going change this thing out? Reviewer Note*: After some more use with the removal of the heatsink and the fan (during the temp testing) I found I did not need the screw drivers anymore. I only need simply lift and the heatsink or fan will come off with no problem.

As an engineering student, I do like to test everything that can be tested. For my amusement, I decided to test how much the fan actually helps with the temperature on the MOSFETs. Using a K-Type temperature probe and my thermometer, I placed the probe at the center underneath the heatsink to get the best temperature read and swings during load and idle times. I decided to go with using my 24/7 build with stock hardware settings. Ambient temperature in my room (as read by thermostat) is 75 °F (23.8 °C). All fans were kept on high with no power savings.

The results are not terribly surprising but it does make you question a couple things. First off, the fan, with system idle, only had the MOSFETs 1 °C lower in temperature than the heatsink alone. I would have expected to see a bit more in the difference, but since this is an open case test platform I can only assume that the heatsink may actually have a higher idle if it is in a case. Second, the difference in the incline, if you follow the logarithmic line you will notice that that the heatsink has a faster incline. This shows that the fan really does help keep the temperatures lower for a longer period of time, while the heatsink heats up much faster. Third, the end result. You may notice that the test without the heatsink lasted longer than the one with the fan. This is because the fan leveled off after some time. The heatsink kept rising, and the spreadsheet can only hold so many data points. If I could have gone on longer, I would not have been surprised to see the heatsink temps rise higher.

In the end, all of this is not too surprising. A fan outperforms a heatsink (if you can call it that, its more like a metal lid). If you were to put this system in a case, I could only imagine that the results would be much more dramatic. The fan would probably show a bigger lead. I do have to make one last comment about this. I would like to see how the two would perform when the board is under LN2. One thing Extreme Overclockers have to worry about is frost build up on the board, preventing this from happening can allow us to bench longer. Fans usually help with moving the condensed air away from the board. I would like to see if the fan or the heatsink would be better suited for the task, but that is a test for another time.

UEFI BIOS

Now, time for the UEFI BIOS. This is the first time that I have had a board with the UEFI BIOS and I must say I like it, but there are some things that could be done to make it a lot better. Maybe it’s ASRock’s particular UEFI BIOS, but I found some bugs and a couple items that should be in the UEFI BIOS, but are missing. Overal, I like it, and I cannot wait to see what comes out of these UEFI BIOS’s in the future.

First up is the main screen. It looks pretty normal, but there is something missing. Neither the main page, nor any other page, has the current UEFI BIOS version listed anywhere. Normally this is not a big deal, and in a way really does not matter. But the truth is, I would like to know my BIOS version before I spend the time looking up a newer version. It’s a time saver more or less.

Next up is the OC Tweaker page, with the default settings displayed. I really like how ASRock sectioned all the categories into appropriate groups. Each sub menu is very clean. You are given three save slots with this board, and as you can tell I have already used that feature. I also like where they are located. Most of the time this option is stored somewhere else in the BIOS.

The OC Mode has 2 different settings: Manual Mode, and CPU OC Mode. The CPU OC Mode is basically a quick overclocking tool for your CPU. It will overclock your CPU by a percentage. This is a nice feature for those that really do not know what settings to play with and want a bit of an extra boost. How stable each of these settings may be, I do not know. From my experience in the past with other motherboards, they are hit and miss.

The first picture just gives you a picture of how the OC Tweak page looks when you first enter into full Manual mode, and the next two pictures, I show one of my overclocking settings during testing. All of the frequencies can be typed in, but the multipliers and voltages are a pull down menu. I wish that voltages were not a pull down menu, and that I could increment the voltages at in smaller increments, not 0.25 v. Well, the 0.25 v increment is only for the CPUv and CPU-NBv, the HTTv and VDDA’s have their own settings and way too high increments. I do like how the multipliers will auto update their frequency as you increase the HTT bus. The problem is that the CPU will not update past 18x multiplier. A reboot in the BIOS will auto correct this though.

ASRock’s DDR Overclocking page, for this motherboard, is one of the worst I have ever played in, and needs to be redone. Everything works, and there is no problem for when I overclock my RAM. It is how the page works that needs to be redone. As you can see with the two screenshots, each menu has a sub-menu that you can key in your value. What they should have done was insert a drop down menu or a key in section, not both. It is really frustrating to enter in the values and move around this menu. Please ASRock, change this.

The Advanced tab is pretty typical. You can access power saving features for your CPU, Northbridge and Southbridge options and work with your storage. Built onto this board is a BIOS debug hex screen, power on and reset buttons. You can configure how to keep the LED back lights on or off during sleep, or while the computer is on. These options are called: Good Night LED and Onboard Debug Port LED. Unfortunately, the default to these buttons/hex screens are off when your computer is ON. Having these off is suppose to help with LED brightness in a dark room or while your asleep, but I really do not believe they are bright enough to bother someone. Furthermore, the power buttons and reset buttons back lights do not stay on when the computer is off, but with active power. I would like to see all three of these features stay on, when power is running through the computer even if the computer is off.

Last there are the Hardware Monitor, Boot, and Security pages. As you could expect, they are typical. The temperatures in the Hardware Monitor seem to be spot on. The voltages are pretty accurate, but the CPU voltage is a bit off. As I was working the benchmark results for this board, I used my USB memory stick to install OS’s. At one point or another I deleted one of the boot priority options. You cannot get this option back. I did not notice this until I had deleted two of the three options I started with. After that, I tried to figure out how to get it back. A BIOS update brought back the options.

Overall the BIOS is alright. It needs a lot of updates and fixes in my opinion. I did write down a couple more bugs that I found and hopefully Asrock will fix them soon. Having a bad BIOS with a board labeled Extreme leaves a bad feeling with the user.

AXTU

ASRock Extreme Tuning Utility reminds me of the very first Asus overclocking tool. It has nothing to do with the looks, even though they do use somewhat the same color scheme. It is more about the simplicity and the options that are given. I always thought those days were done and gone, although I should not make it sound that bad. Right now, manufacturer-supplied overclocking utilities are very advanced. TurboV and Easy Tune from Asus and Gigabyte offer tons of features, and these tools can be used on almost any of their motherboards. Now, granted, most of the settings will not be used, unless you’re someone like me. So, something like AXTU is completely reasonable for those that are not like me. Since ASRock does list this as one of their higher end boards, I cannot imagine not having some sort of software that allows me to work in the OS as if I was still in the BIOS.

The hardware monitor page and fan controller are pretty typical. The temperature readings on these pages are pretty accurate. During my 24/7 testing I solely used this software to watch my temperatures. CoreTemp and AIDO64 were no use for me while using this board and CPU. The voltages were pretty close to accurate, but later on you will see just how accurate the readings were.

Now this is what I’m talking about: the overclocking sections of this tool. As I said before, AXTU is a pretty simple tool. You can easily adjust some parts of the CPU and work with most of the voltages. You pretty much have enough to work with if you were to overclock in a benching OS. However, there is one item I would love to see in this program: controlling the different frequencies of each core of the CPU. This is one thing that I see in AOD and Asus’ software. The feature is very helpful for when I really want to maximize the system during competitions.

OC DNA is a bit of an interesting section, and it is not something you see a lot with overclocking software. You are able to switch between BIOS settings that you have saved in the BIOS. This is pretty cool and could come in handy for those overclocking gamers. Let’s say you wanted to put some more power down on your computer for playing a heavy game like Witcher 2, but when you’re done, you do not want your system to be using this same power while your viewing your favorite website: Reddit.

IES or Intelligent Energy Saver is an attempt to control the number of power phases that are turned ON for the CPU. Now that motherboard manufacturers have to cater to more than just those that want a motherboard that looks like nice and overclocks well, new energy saving technologies are being implemented. Controlling the power planes is not a bad idea, and from some playing around it does a pretty good job of it. When my CPU demanded performance, the power plane gave all 8 to the CPU. At idle, the CPU sat with 2 planes turned on.

I did manage to find some bugs with the software as well. One of the most troublesome is that if you increase the CPU voltage, the LLC will reset back to Auto. Even if you have it set to something like 1/4, it will ignore that setting. For this test, I played with LLC through the UEFI BIOS first to see how LLC affected the CPU voltage through Prime95 tests. After the Prime95 test I would adjust the voltage in AXTU and bring it back down to where I started. Then, I launched Prime95 once more and recorded the results. The voltage for the CPU was set at 1.45 v in both the UEFI BIOS and AXTU.

As you can tell, AXTU would simply ignore the setting set in the UEFI BIOS. This is most likely a bug in the software and a quick update will certainly fix this. In the end, I can see through the little faults that the software has, and see what it could do. Later in this review, I will get down to the good stuff of pushing the system to the limit, and I relied mostly on this software.

Test Systems

• Windows 7 64-bit Operating System: Fresh install with updates, and few programs installed
• AMD Phenom II 1090T CPU @ 3.0 GHz
• OCZ Black Edition 1600 Cas 8 (2×2 GB) 4 GB RAM
• Asus Matrix 5870 GPU
• Drivers: Cat 11.8
• All BIOS settings were set to default, only the RAM and DRAM voltages were changed.
• Water cooled on an open-air benchmarking station

As with all tests, reviewers have a structure we adhere to. For motherboard reviews, we like to test things at the most default settings. This gives the most general and non-biased view point of  how well the motherboard works for most users. Since this is my first time doing a motherboard review, I do not have any data to compare with so this time around we will just have to look at the scores and just say “Uhm… Quite”.

CPU Bench

For testing, I worked only with the CPU and memory. Motherboards tend to yield the same scores given enough samples in the data set. The benchmarks chosen were SuperPi, Wprime, Cinebench, 7zip and the AIDA 64 package. These benchmarks are all designed to either focus on the CPU, or the CPU and Memory. All benchmarks were run three times in a row and an average was calculated.

Everything is pretty normal. If I have my calculations right, most of the people that open this review are going to look through results and probably not read anything, so hopefully you read this. Benchmarks are nothing more than somewhat useless for motherboard reviews. The only reason that we do them is to hopefully catch memory or bus latency. Given a big enough data set, the score for all boards will average over time. Instead of looking at the results, look at what it can do for overclocking, and read about how the BIOS behaves and the software packages. That is where a motherboard can be different from its competitors.

Overclocking

Overclocking with this board was a bit of a hassle at first, or at least for me. As an extreme overclocker, I like to set very specific overclocking settings in the BIOS and then do a couple tweaks in the OS once I know settings are stable. It helps for when I am trying to find the absolute maximum overclock, not just the boot maximum. This kind of style helps extreme overclockers, or overclockers that like to benchmark competitively. Since this board does not give true readings from the software or BIOS, I was worried about what voltages I was actually setting and using while benching.

I was able to request some information from the engineers over at ASRock. I asked for onboard contact points so that I could wire my digital multimeter up to the board and get live voltage readings. They gave me the CPUv, CPU-NBv, NBv, and DDRv. The CPUv and CPU-NBv became very helpful. I found that there was quite a bit of a difference from what various software was reading and to what my multimeter was reading. To regular users and overclockers, these differences are not that big and you would probably figure out the offset without even knowing it. Yet, for competitive overclocking, knowing these offsets can help me tune my system for specific start-ups and benchmarks.

After wiring up these read-outs, I started to benchmark with stock settings. I learned quickly that what I was given in any of my software was nothing near to what I was getting on my multimeter. In a previous section, I showed the difference with the CPU voltage. I will not go over that again. The NBv and the DDRv were right on, and since I do not play too much with those voltages, these contact points were hardly touched. What I kept my eye on most of the time was the CPU-NB voltage. Stock voltage of the CPU-NB is 1.125 V; what is on my multimeter is 1.138 V idle. That is a 0.013 V difference and it’s pretty big. The difference will continue to increase as the system is being pushed. Most likely this is due to the CPU-NBv LLC. On a lot of the new high-end AMD boards, you are now able to control the LLC on the CPU-NBv line. An option like that for this board may help with the gap that I’m seeing.

24/7 Overclocking

For 24/7 overclocking, it was pretty simple to get up to the normal speed of 4.0 GHz. 4.1 GHz took only a small push on the CPU voltage, and after that, I was able to find my absolute boot-up maximum and stable overclocks. These settings are only based on the safe range of voltages that I list in my Phenom II Overclocking Guide. That is to say, that voltages in the range 1.45 – 1.50 V are safe for 24/7 use, but I prefer 1.45 V.

I really did not see a difference in this board compared to my M4A89GTD-Pro (with the 890GX chipset). I have to be honest, since I push my hardware so much during benching sessions, I like to keep my 24/7 settings at stock. This helps limit degradation of the chip. What will really show the difference with this board is how far I can push the hardware with benchmarking.

Pushing The Limit

Now it’s time to push the system to its maximum. At this point, I changed my operating system to a custom-built benchmarking OS. These are specially designed for those that want the best results from benchmarks, and are no-where close to being suitable for 24/7 systems. Who cares about those 24/7 overclocks: I like to see how far I can push this board with my favorite CPU in the world, the 1090T.

For 2D benches, I have a custom built Windows XP 32 bit. Most of the services are disabled and only necessary drivers are installed; video drivers are not even installed. This allows the CPU and memory to avoid any use by any small services that may otherwise run in the background. This then gives me the best chance to push the system, and for the hardware to be dedicated completely to the benchmarks.

The results speak for themselves – they are not bad. Compared to what I used to get on 890GX and 890FX boards, I find the results an improvement. These results were done on an water-cooling system. I had a lot of fun pushing the SuperPi 1M times lower and lower. The CPU kept taking the increase in speed each time, allowing me to run the bench each time. I only stopped at 4.5 GHz because I did not want to increase the voltage any further. This board definitely has some tolerance.

I really like 2D benches and Heaven, working with 3DMark Vantage and 3DMark 06 is not really my thing, but since these two tend to put a lot of strain on the CPU (Heaven does nothing with the CPU), I gritted my teeth and pushed hard. For 3DMark 06, I was very very surprised to see how far I got with my CPU. The CPU tests are not particularly stressful to modern CPUs, but it does require a lot of stability on the user end. All I can say is that it was a breeze to get to 4.4 GHz and have it pass. After that, it was more of a CPU voltage limit, but I really did not want to go beyond what I had already set since I am only on water cooling. As for Mark Vantage, I could only hit a maximum of 4.2 GHz. Vantage really does put a strain on the system and it really likes having high frequency memory with low timings. Since my memory was pretty much at maximum settings I called it quits there. All in all though, I am very impressed in how far I was able to push these benchmarks on this board.

Last, but not least, here are the maximum pushes. Both of these results were me just pushing the hardware as hard as I could. It’s not surprising to see the bus speed up at 323 MHz. I know that it can give a lot more, but for the life of me I could not get the memory to behave at the lowest ratio. You will have difficulty booting up at 300 MHz into the OS, but once in there, your system will find the 300 MHz to be stable.

As for the maximum CPU run, that is a pretty sad run. I normally push only one core and will see easily 5 GHz with the voltage that I supplied. The reason why I didn’t get 5 GHz is because the software that ASRock gives you does not support individual CPU ratio increments. I could have disabled more CPU cores too, and maybe got a higher score, but I do not know which core is the best to push. With the right software, I could see which core is the strongest, and show how far I could push it. ASRock: if you really want this board to make a statement, please include individual CPU ratio overclocking in your software.

Conclusion

I am going to give this board a Overclockers Approved stamp, but only barely. To my standards, which I admit are pretty high, the board does not come close to what I would call Extreme. In fact I would call this board a generic, typical, motherboard. ASRock did a good job with the board, but left a lot to be desired; there are a lot of changes that I believe should be made before this board can really compete against the big dogs.

You may be wondering why would I give an approval for a motherboard that I believe has a lot of problems. If I were to place this board in my gaming computer, give it a slight overclock for everyday use and close the case up, I would never be able to tell the difference between this board or any other AM3 or AM3+ motherboard. This is both good and bad. It is good that I can trust this board, and know that when I am using it my system will run perfectly fine everyday that I turn it on. It is bad because it does not stand out: it does not allow me to go that extra step and take my hardware onto the next level of overclocking. This is why I give it an approval, because the majority of you that are looking for a motherboard want to put something in a case and forget about it. Maybe tweak here and there. This board will do just that, but if you want something to take your breath away you will have to look elsewhere.

– Dolk