OC Review: The Gigabyte GA-990FX-Gaming

The buying of a new 990FX motherboard for gaming right now is the height of stupidity. So while this board does offer bells and whistles like Killer Ethernet, M.2, USB 3.1 and an upgraded audio section I will ignore them since these features change absolutely nothing about the fact that 990FX is terrible platform for everything except benching AM3+ CPUs. While this is a "gamer" focused board is does also offer some overclocker goodies like post code LEDs and onboard buttons. As such I'm hoping for it to be a strong alternative as a benching board to the 990FX Sabertooth R2.0 and Crosshair V Formula-Z both of which actually lack some key overclocking features that this board offers. It also helps that Gigabyte's GA-990FX-Gaming comes in at a significantly lower price than the Sabertooth and the Fomula.

VRM

CPU VCore

- International Rectifier 3564 4+1 phase controller

- 4 phases using IR doublers to make 8

- low side Renesas K0393 rated at 40A @ 25C°(ambient)
- high side Renesas K03B7 rated at 30A @ 25C°(ambient)

- 680nH Inductors
- 12 5K rated 6.3V 560uF capacitors

- 200KHz to 2MHz 
- large VRM + NB + PCH heatsink with 2 heatpipes

CPU NB

- 1 phase using a doubler to make 2

- Renesas K03B7 rated at 30A @ 25C°(ambient)

- 200KHz to 2MHz 
- shares main heatsink

RAM

- 2 true phases

- Same MOSFETs as CPU core - No heatsink

- Unknown minimum to unknown maximum KHz

Verdict: 7/10
The VRM on the 990FX-Gaming is strong offering 320A to the core and 80A to the CPUNB as long as you keep the VRM cool. On LN2 that is not a problem as the LN2 pot will be cooling down the entire motherboard. However on air and watercooling the VRM will need some extra cooling because during my short testing with minimal airflow the VRM heatsink was reaching into the 90s with core voltage at 1.55V. Direct airflow from a fan quickly got the temperatures under control however without it I'd expect the IR 3564's OTP to kick and either throttle the CPU or shutdown the VRM. So if you plan to bench this board with an air or water cooler I recommend you stick a fan on the VRM heatsink(which recommend for most AM3+ boards anyway). You might also want consider attaching some small heatsinks on the back side of the PCB in this area:

BIOS

Maximum Voltages: 

- VCC: 2.3V

- CPUNB: 2.1V

- VDDR: 2.1V
- HTlink: 1.8V
- PCIe: 2.435V
- PLL: 3.135V
- NB: 1.735V 
- 5 Vcore LLC settings

CPU:

- NB clock from 400 to 6300

- 1 core per CU mode as well as 2 4 and 6 core mode

DRAM:

- XMP Support

- All timings and settings available

- Supports per channel timing settings
- maximum ratio of 10.66X for 2133mhz at 200mhz FSB
 Extras:
- core unlocking on Phenom II and Athlon II CPUs 
- 8 onboard BIOS profiles and option to store profiles on USB stick or SATA drive

Verdict: 9/10

The BIOS uses the standard Gigabyte layout. All the voltage limits are high enough that even with LN2 cooling you can kill anything and everything you could install on this motherboard. The VRM settings however are a little lack luster. There are no switching frequency, response time or current limit settings like you would find on an ASUS board and you only have LLC settings for the core voltage. The core voltage LLC settings are also pretty bad with Everything but the medium and extreme setting dropping far too much voltage to be useful. The medium setting is perfect for daily usage and even benchmarking since it idles at the voltage you set and drops only 20mv going under load. The extreme LLC setting on the other hand provides a huge amount of voltage boost going under load and as such is useless for daily use and I wouldn't use it even when benchmarking just because of the 80mv voltage spread it has. Memory overclocking on the 990FX-Gaming is on the tedious side especially when pushing above 2133mhz as the board does not have a 12x ratio. So in order to get any memory clock above 2133mhz you need to start messing with the FSB. That's not really a major issue if the board showed you the impact of FSB on the NB and HT link speeds but it doesn't so regardless of your FSB setting you will only ever see the NB ratio * 200. Of course you need to run your NB above your memory clock so you just need to use the 11x NB ratio when using the 10.66X memory ratio but it would still be nice to see what the actual NB speed is going to be. Other than that memory overclocking on the board works perfectly well and the board recovers from crashes quickly without trashing your settings so you don't even have to save overclock profiles every time you change something even though I do recommend doing that just in case the board doesn't manage to recover properly from an extra bad mix of settings that require a BIOS wipe. Overall the BIOS on the 990FX-Gaming has everything I want and need when overclocking and is easy to navigate after you get used to Gigabyte's layout.

Air Testing Results:

Max RAM speed (FX 9590 + 1 4GB stick): DDR3-2603 13-16-16-36-2T 1.75V
PSC best settings(FX 9590 + 2 2GB sticks): DDR3-2500 8-12-8-28-2T 1.83V
1600mhz low timings(Phenom II X2 555BE): 1600 5-7-5-18 1.9V
Max FX 9590 validation clock 2 cores with 1.65V: 5222mhz
Max CB11.5 core clock with 1.55V: 4950mhz
Max FSB: 320mhz
Verdict: 9/10

Gigabyte may only rate the board to do 2000mhz on the RAM and the highest RAM multiplier is only 10.66x however with a bit of tweaking I got the RAM as high as 2603mhz. More is probably possible however with AM3+ compatible CPUs your NB clock needs to be equal or greater than the RAM clock and as such maximum frequency ended up limited by the North bridge which needed to be on a 11X ratio to stay ahead of the RAM speed. This is not ideal as it forces your NB clock to be a few 10s of mhz higher than the RAM speed which makes you run into NB stability problems that much sooner.
My FX9590's Norhtbridge showed just how bad it is when I did my PSC testing. The kit is capable of doing 2600 8-12-8-28-1T on a little over 1.86V. My FX 9590 on the other hand is absolutely incapable of dealing with much more than 2400mhz with such low timings without giving the north bridge a lot of voltage. In the end I managed to get the system to just barely finish SuperPi 32M with 2500 8-12-8-28-2T with 1.83V on the RAM and 1.45V(too much to be safe on air or water) on the North Bridge. Since it's my CPU that's holding the RAM back I'm gonna say that the board handles high performance RAM just fine you just need a capable CPU.
Now some of you maybe wondering about why I do a 1600mhz low timing test. The test exists because most Phenom II and Athlon II CPUs can't go much beyond 1600mhz RAM clock so to get good memory performance it is important to be able to get really low timings at speeds between 1450 and 1650mhz. I'm happy to say that the 990FX-Gaming didn't have any problems doing 5-7-5-18 timings however the board can't go bellow 5 so if you have a kit of RAM that can pull of 1600 at CL4 you're out of luck.
The 990FX-Gaming manages the same maximum core clock at 1.65V as my Sabertooth did. However the software for the Gaming is a lot more painful to work with. Gigabyte's Easy Tune require's Gigabyte's App Center software which requires .NET 4.5 and really didn't like my stripped down win 7 install. So I had to go and make a new install of windows to get Easy Tune to work. Considering that I remember having similar issues with AIsuite so it's not like Gigabyte's obnoxious OC software is anymore obnoxious than ASUS's I'd still prefer a stripped down utility however I do understand that since no one really cares about AM3+ any longer that there isn't a slim OC utility for it.
For cinebench clocks the 990FX-Gaming matches the Sabertooth exactly so again no complaints from me.
Maximum FSB varies from motherboard to motherboard and maxing it out completely is very time intensive. A high maximum FSB is very important when benchmarking CPUs with locked multipliers like the Athlon II X3 460 or a Sempron 145. The Gaming getting an easy 320mhz is inline with what the Sabertooth does which is inline with most 990FX boards. They generally start having issues going above 330mhz on the FSB and as such the Gaming is not inferior to any of them in this department either.
Unfortunately even with the great overclocking results I can't give the 990FX-Gaming a 10/10 here because I've never tested any of the boards that are priced above the Sabertooth and they may deliver better results than this board.

Conclusion: 25/30 Points, 0.179pts/$(based on Amazon price at the time of writing)
Gigabyte's 990FX-Gaming makes for a great AM3+ overclocking board. Personally I prefer it to my deceased 990FX Sabertooth R2.0 and I'm not saying that just because the Saberooth decided to die on me. The fact is that Gigabyte's BIOS offers more features than the Sabertooth's. All the voltages you need are accessible and can be pushed higher than on the Saber. This is especially true of the Gigbayte's Vcore setting which goes all the way to 2.3V compared to 2.075V on the Saber. That's enough voltage to quickly kill FX chips even with LN2. The Gigabyte board also offers core unlocking on Phenom II and Athlon II chips which the Sabertooth R2.0 nor Formula-Z offer. You also get dual BIOS as well as on-board buttons both features that the Sabertooth lacks. In testing the Gaming managed the same results as the Sabertooth and recovered much more smoothly from bad settings than the Saber ever did. This board is however not perfect. The VRM is slightly weaker than the Saber's and while still plenty power full for water or LN2 cooling it requires direct airflow at high voltages and loads which the Sabertooth doesn't need as long as you just have decent airflow in your case. Overall I like the Gigabyte better than the Sabertooth between the buttons, the dual BIOS, the core unlocking and the lower price the Gigabyte is both less hassle to work with, better featured and cheaper than the Sabertooth and from my aircooled testing you don't seem to lose any performance for any of these benefits. Other than that the 990FX-Gaming also offers a bunch of "gamer" features that the Saber also lacks. Really if this board had been available when I was buying my Sabertooth and I weren't a massive ASUS mobo fan I would have probably taken this instead because it's just better and cheaper. So if you're looking for a relatively cheap motherboard for benchmarking ALL(you can't convert X2s to X3s on the Saber or Formula-Z!) AM3+ CPUs the 990FX-Gaming is one of the best out there. Oh and you can use it to play games I guess... if that's your thing.

As soon as I get my LN2 Dewar topped up again I will do another live-streamed 8+Ghz attempt with the 990FX-Gaming and after that it will be featured for any AMD CPU benching I do until AM4 boards launch.

If you enjoy AHOC's content and would like to see more of it please donate using the button in the top left of the page.

The cheapest LN2 GPU pot ever.

GPU pots are expensive. Like really really expensive. Slims go for 100-200$ and can't handle any of the big GPUs high power GPUs. Fat pots are even more expensive ranging from 200-400$. So basically that means I can't run LN2 on GPUs............
......
.......or does it?

Ladies and gentlemen I present you the Tapester GPU LN2 pot.

It's not pretty but it doesn't leak LN2 and it should be enough for some hilarity on a live stream. So then what is the Tapester? It's a reference HD 4870 heatsink wrapped in a liberal amount of electrical tape. It will mount on 4800 and 5800 cards. It should also fit several other AMD cards but I don't have them so I neither care nor know.


Here's some pics of the first iteration with a few hundred ml of LN2 in it.































The surprisingly leak proof bottom


Pouring LN2 into this first version was a very wasteful affair since I didn't give the top much of a funnel into which to pour.

Also during the whole cooling process the pot was making some horrible cracking noises as the tape shrunk from the cold. One of the cracks even shook the whole pot


I let the pot defrost and dry before I did an inspection of the damage that the LN2 did to it.



This is the only place where I found a visible crack in the tape. I think this is the one that shook the pot when it cracked. I'm also pretty sure that the tape cracked because I stretched it too much when putting it on. Either way I'll just add more tape.

The bottom of the HD 4870 has a layer of plastic. The tape didn't stay stuck to this when I added LN2 so this are got an extra piece of tape to hold it together better.

This wasn't caused by the LN2 but by lazy tape placement. one thing I noticed is that the tape held best when it didn't have any loose edges so this had to go.

Here too the tape started coming off. The solution is obviously more tape.

































I added some extra tape to the top for a proper funnel. After all if I can't get LN2 in the pot it isn't a good pot.

Here's the fixed bottom. The tape is just barely low enough to clear the center shim.

Here's a full frontal shot of the abomination.

And here it is fully insulated with 5mm armaflex and ready to roll. To test it I plan to just run some Unigine Heaven Basic preset with the 4790K on air cooling. If this works I'll have a few more HWpoints to add to my total. But much more importantly I will a working LN2 pot for 2 of my GPUs.Unfortunately I won't do much benching with this for a while because to get a good amount of points I'd also need LN2 on the CPU and I currently don't have the equipment needed to LN2 both a GPU and a CPU.

If you enjoy the content you see on Actually Hardcore Overclocking and would like to see more please share it or donate using the donate button in the top left of the site. All donations will be used to make more content even if just helps buy more themal paste. Right now though all donations will go towards a capture card.

Saturday Fury X Benchmarking Stream!

I'm going to be live streaming again. This time it will be 3 Fury Xs vs a whole lot of 3Dmark Firestrike and Firestrike Extreme.

I've already set a score of 33970 points with HWbot rules that got me the 18th spot foir 3GPUs in Firestrike on HWbot. However it has come to my attention that I'm not particularly efficient so here's the plan for the stream:

Beat this score

and this one

And this one too

Improve my current Firestrike score on HWbot.

Since bench marking Firestrike on Fury Xs isn't a super involved process I will be doing my best to answer questions during the stream.

I will be starting at 7GMT and basically going until I beat all the score or drop dead trying.

As always the stream will be up on my twitch channel

Possibly the last batch of custom Fiji BIOS

And here is the most likely final batch of custom BIOS. Now including a BIOS for the Fury Nitro and a newer version of the Fury X BIOS which should be better than the old version.

The BIOSs offer raised power limit and a 3D load voltage of 1.3V. This leads to better overclocking capabilities and better performance than the stock BIOSs running similar amounts of voltage in Trixx/Afterburner/GPUTweak. The power limits on all the BIOSs are 400W and 350A you should not need to raise the power limits but can if you wish to do so. The newer Fury X BIOS is based of the latest AMD provided BIOS and seems to clock better than the last one I made but I haven't fully tested it yet(it is safe to use).

Here is  a download link for them: https://www.dropbox.com/s/v18py3rhbef8rmx/ProFijiBIOSs.zip?dl=0

In other news I've done a quick test run of Firestrike with 2 cards and the results are looking promising. The cards aren't fully finished but at this rate I should be able to break 40K graphics some time soon. With some work on the 5960X I can probably get into the top 20 for 3Dmark Firestrike on HWbot. Currently I'm number 65 on the Firestrike rankings for 2 cards. Here's a link to the score: http://hwbot.org/submission/3185463_ 

Mind you I got that score with 0 artifacts so there was still some room in the cards left to push it but it was 3AM and I got the results I wanted to see. Also the 5960X was only 100mhz above my daily OC since I wanted to test the GPUs not mess around with CPU stability issues. Once I have everything working 100% I will be live streaming the overclocking sessions. I'm coming for you 3 way TITAN-Xs and once I'm done they will rename Firestrike to Furystrike.

MFW I get 38.8K GPU score in Firestrike

Also for those of you who watch the youtube channel do not worry. I have not forgot about you, there will be content soon I promise.


If you enjoy reading AHOC and would like to see more content please make a donation using the button in the top left corner. All donations will go towards producing more content. If there is something specific you would like to see you can leave it in a message with your donation or in the comments bellow.

More Custom Fiji BIOSs

So I noticed that I forgot to include the 3584 core BIOSs in the last BIOS pack. Those are now included. I also got my hands on a Windforce Fury BIOS so those of you who have can now enjoy the magic of overpowered BIOSs. Also apparently the 3776 core and 3840 core BIOS for all Fury cards were name swapped this should now be fixed. Enjoy!

For those who do not know what the custom BIOSs are

Here's the dropbox link

I'd like to make a "pro" BIOS for the Nano however I do not want to set peoples GPUs on fire so I need a Nano to test with. Even as much as I like to think that I'm a famous overclocker. I'm not and so while I can ask AMD/Sapphire/Powercolor.... bla bla bla for samples I don't get them. I think you get the idea of where this is headed. I need money. There's a donate button in the top left of the page.

BTW if you're wondering about my voltmoding progress with the Fury X. Then I got good and bad news.
The Fury X I started with still works. The bad news is that it's currently stuck at +20% Vcore and I can't really bench the card till I fix that. I do know how to fix that because I already did the same thing to my 290X. The problem is that the fix isn't very tidy and I'd like it to look good. It also doesn't help that the fix makes the Volt mod less sensitive. So I'm currently thinking about how to fix the card the best way.

Here's a silly Fury X for the time being

Also I'm considering throwing the Fury Xs on Dry ice if things go well on the stock cooler. Should be pretty awesome based from what Joshy Ocuk told me at The Big Freeze. Apparently HBM absolutely loves cold and at around -70C I won't even have to do much to keep the card happy since the cold bugs are bellow -100C°.

Fury X testing results

Optional intro

I hate power management algorithms. Sure they save you some power but they make overclocking a nightmare. The smarter they try to be the worse they are. Back when this blog first started I wrote a really long and terrible rant about Nvidia's boost algorithm and AMD's beginnings of what is now know as power tune. The TL;DR of that article can be: Nvidia's is dumb because you can't set core clock and AMD's is dumb because it's broken. Well Nvidia's hasn't changed and AMD's is possibly more broken now then ever before.

I promised a Fury/ Fury X overclocking guide some time ago. I also promised to whoop the ass of Nvidia GTX 980 Ti SLI setups in 3Dmark. If you follow me anywhere where I can be followed you are aware that I haven't delivered either and the facebook followers should know that I've complained that I can't even beat my old Fury score because AMD's drivers are garbage.

Well today I will show you what the problem is.

End of optional intro

TESTING METHOD
I used free cam mode in Unigine to park next to a wall so I could quickly track FPS changes. I was running Unigine Heaven at very low settings so that I was getting about 274FPS at stock clocks. I was running Unigine windowed and had Trixx open so I could change settings and immediately see the FPS changes. After every batch of readigns I checked that the FPS at stock clocks was still the same.

Here are the results in all their glory:

The core clock and HBM clock scaling in Unigine at 1440x900p is unsurprisingly garbage however what is surprising is the voltage scaling. You can see that even though I did not touch core clock or HBM clock, FPS goes down as I raised core voltage. Also note how power limit did nothing to prevent this drop in FPS. Basically AMD's power tune is as usual completely freaking broken.

Now I would like to see if this is the case for all Fury Xs so if you have one could you do me a favor and just check what happens to your FPS if you're at stock clocks and set core voltage to +96mv and to +198mv. you can mod Trixx to do that using this guide and post your findings in the anonymous comments bellow this post. You will need to run at least 55% fan speed to prevent the VRM from overheating.

For me this perfectly explains why running a 1225mhz core clock with +175mv scores lower than stock clocks in 3Dmark.

Luckily I can just hard mod the card for more power however I'd rather not do that and so I started doing a little investigating. First thing I found is that my PowerColor Fury Xs have 2 100% identical BIOSs. This goes against what some reviewers said about the Fury X coming with a 350W and 250W BIOS. So I've started searching for a 350W BIOS in hopes of getting rid of at least some of the negative scaling. I've also tried to contact AMD about providing a custom 600W BIOS but I doubt I will get a response much less an actual 600W BIOS. So I'll probably end up soldering 600ohm resistors in place of the 300 ohm resistors used for current sensing.

BTW for all those wondering about Fury X VRM temps. They are directly related to core temperature(duuurrr they share a heatsink) so if you're running stock setting and the core is at 70C the VRM is between 90-100C. If the core is at 40C the VRM will be at 60-70C.

Update:
So I've tried several custom BIOS with power limits as high as 768W and got no benefit. I also tried the clock blocker program also no benefit. I'm going try a 65000W BIOS but at this point I don't think it will make a difference. 

DON'T READ THIS IF YOU DON'T LIKE PEOPLE COMPLAINING
As you've probably noticed the table is missing 2 of my favorite data sets. Real world core voltage and power draw are both missing because someone stole my parcel which contained the equipment for measuring those. I still can't believe that the UK doesn't offer pay on delivery. I'm also in limbo about ordering anything else because the idiots running deliveries here would probably leave that in front of my door too.

R9 380 Nitro Volt mod guide

As you all know I wanted to make an Sapphire R9 380 Nitro(review of the card) volt mod guide. However me being me I royal messed up my R9 380 and had to leave for the UK before I fixed it. However a kind gentleman on Facebook has provided PCB pictures. They aren't exactly what I need however they are good enough to do the guide with. Lets get to it.

DISCLAIMER: Everything past this point will void your warranty! If the GPU stops working due to you doing any of the things described bellow I am at no fault what so ever.

So first lets breakdown that PCB into what's important.

There isn't much to modding the card. Hook some sensing wires to the memory and core voltage VRMs and pull a GND from where ever you want to. I highlighted some good candidates in brown.

 

First up core voltage and core current limits. The current limit is simple to disable just cut the 2 traces highlighted in green. Now your card will think it's pulling 0W regardless of what the powerdraw actually is. So power related throttling is history. However it does make it easy to blow up the VRM by running Furmark since there is nothing to stop the card from pulling too much power. This should only be a risk when going over 1.45V core voltage. To cut the traces I suggest digging into them with a needle. This picture also shows how you could try modding core voltage however I recommend that you avoid soldering onto a pin or trace because it is really hard to not screw up. 

Instead solder a 22K(+/-5K) ohm variable resistor from the red point to ground. This will give you free control over core voltage. I recommend not going over 1.35V on air cooling for daily usage. For benching 1.45V is my ceiling. More than that seems to rapidly degrade 28nm GPUs.

The VRAM is an easy mod. All you do is hook a 220K(+/-50K) ohm variable resistor from the pin highlighted in orange to ground and you'll have VRAM voltage on manual control. There is no power limit for the VRAM so you don't have to worry about that. I don't rember the stock voltage of the VRAM on the 380 but don't go over 1.7V for daily usage. On sub zero 1.85-1.95V should be optimal.

Practical Tips
Use a fine chisel tip for all the soldering. A needle tip will make everything much harder because the tip of the pen won't actually melt anything.

If you have a temperature regulated iron set it to 240-250C for everything.

When you finish soldering one of the connections to an SMD component I recommend using plastidip to hold it in place. Hot glue is harder to apply without breaking a weak soldering joint and is very bulky which might prevent you from fitting the heatsink back on the card.


Please donate if you want to see more guides! I can't do mods on GPUs I don't have and unfortunately I'm not important enough to get free samples yet...
If just 200 of you were to donate 5$ I would be able to do almost any GPU you're interested in.

Overclocked Power Draw Calculator

Back in 2012 when I first started this blog(there aren't any posts left from then because it was literally just me ranting about everything in existence) I figured out how to calculate overclocked power draw. Over the years I have tested the equation and tried to perfect it. I failed to perfect it. However I have found all it's flaws and so I feel safe in releasing a program that does the calculation for you. The usage should be self explanatory. However I recommend you read the README.txt to understand the flaws. Like everything else I make it's still written in Java. Java 7 might run it but you will most likely need Java 8.

Download Link

Pic of it in action:

Donate if you want to see the app get better! With enough funds I can make this take operating temperature and architecture into account and possibly extend the functionality to include some degradation rate estimates. I could also make it a whole system power draw calculator but I think that's unnecessary because there are plenty of those on the net already.

Something Quickly: How to calculate overclocked current and power draw for CPUs and GPUs.

This is what you use to calculate power draw:

(MHzOC/MHzstock)* (VOC2*(TDP/Vstock2))

Just plug in the numbers and you'll get the power draw of the CPU or GPU you're overclocking once it's overclocked. It does have a few issues it only works for specific temperatures and TDP doesn't always give accurate values it's best if you measure the initial power draw rather than relying on the TPD to be accurate. However if you do you power draw and stock voltage measurement when under load and your temperature when overclocked isn't different this will be accurate. A power draw value isn't all that useful when you're trying not to kill a VRM. To get current draw you just use this version of it:

(MHzOC/MHzstock)* (VOC*(TDP/Vstock2))

The power draw value is good for when you're designing a chiller or trying to choose coolers the current draw value is for not killing a VRM. The equation has a +/- 10% margin of error so keep that in mind when using it(AFAIK it does not work with Nvidia Maxwell GPUs at all). It can be much more accurate if I could do more research into it but that requires that I can get my hands on a bunch of CPUs to test it with and several liters of LN2 so I can keep very specific temperatures.

Just for the sake of doing a quick proof: 

 My 3960X measured 353W at 5Ghz using 1.56V running IBT. It has a stock voltage of 1.18V, TDP of 130W and stock clock of 3.3Ghz. Lets plug them in:

(5/3.3)*(1.562(130/1.182))= 344W

So it a little bit of but hey better than nothing.