• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Gigabyte 990FXA UD5 R3.0

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Annunaki

New Member
Joined
Feb 26, 2019
Hey guys been lingering around the scene for sometime now not really been into over/under clocking until just recently with the AMD FX9590 power sucker! (yes i know), Anyway I've had this CPU for about 3 years now and basically ran it with optimised defaults with Turbo Boost switched off as this caused system hangs, So since then its been running at 4.7ghz albeit unbeknown to me unstable which has recently been showing its instability in the recent Windows 10 creators insider releases, TBH I have been using my PC more recently which is probably why these issues are popping up.

So I've garnered the info from various threads here and there with regards to down clocking this beast as its plainly obvious my Scythe tower cooler with a push and pull fan setup just cant tame this beast at 5ghz or 4.7ghz without going into a thermo nuclear meltdown..... Anyway I seem to have found a happy stableish kind of place at 4.5ghz which runs prime small FFT for 2 hours (i got bored after this time) without killing off any workers or burning my house down but the blend test is another story well it ran for 7ish hours before worker 8 died twice over 2 days...

So snooping in the bios again and I find LLC... Which seems to fluctuate any set voltage I give to the vcore.. I will start with 1 question and reply when I can to any counter question.

How can I have the vcore stay put? So like when I set it for 1.404 it will stay at 1.404 and not venture +/- off at about 0.025v each way..

Thanks in advance for any help..
 
We'll need to know what motherboard it's in, for a couple reasons. That chip is a beast when it comes to power draw, not many motherboards can keep up. A lot fewer than claimed to, that's for sure. Having said that, your cooler isn't up to the task. What do your temps look like when Prime fails?
 
Right in the title AL Gigabyte 990FXA UD5 R3.0

Annunaki welcome to the forums, you're correct in thinking that your cooler may not be up to the task, the 9590 really needs a triple rad to run properly. I would back the multiplier down to 4.5 and start with 1.4 V test it out with the stress test of your choice, personally I use P95 blend. Watch the temps and the socket temps try to keep the core below 60C and the socket below 70C you can use hwmonitor it has a freee version and will show both temps.
If it crashes at 1,4V then bump it up a bit, if it doesn't then try raising the multi add voltahe as necessary until you reach your temp limits or stability keep failing.
 
:facepalm:
Yeah, that board should run a 9590 at stock speeds. Johan45 has it nailed on the cooler, too. In fact, I think AMD recommended liquid cooling for those chips under all circumstances. HWMonitor is handy because it puts the max temps right on the same line as the min. and current temps. Makes it very easy get the necessary info without jumping through hoops.
 
Hey guys thanks for the reply and i have noted about the 3 fan aio but just to add I have started at the lowest clock and lowest voltage I could then worked up bit by bit increasing the multiplier then the voltage to compensate. My issue at the moment is getting the voltage stable at 4.5ghz which is all over the place due to the LLC option in the bios..

There are several setting auto, regular, normal, high, ultra high and extreme and none of them seem to stabilise the voltage. If anything I'd say High is the most stable but it still moves +/- by atleast 0.025mv which obviously has a huge effect on temps on the + side and instability issues on the - side.

Basically I want to keep my voltage at 1.416v which seem 4.5ghz is stable at but LLC keeps increasing or lowering which in effect make the clock unstable when it drops below 1.416. I have all c states disabled and also cool and quiet off I did find a setting apm which I have enabled and I notice a new function in hwmonitor package power which seem to be the cpus TDP draw.

At the moment funds are short (kids, car, mortgage etc) so I'm not bothered too much about running it at its default clock of 4.7ghz. I can run it at 4.5ghz with a temp margin of 12c or so shown in aod but like in my OP after 7 hours of prime95 blend test worker 8 died so I'm guessing this is due to the vdroop LLC crap.

Today well from last night I've kept all bios settings the same but dropped the multiplier to 20 (vcore 1.404ish) which gave me a clock of 4ghz and a margin on 13ish and when I woke up it was still chugging away with all workers.

I'm at work now so hopefully when I get home all 8 workers will still be active and this should hopefully rule out and memory (1886) instability, Also just to add my psu is a Corsair 1000 so I'm sure that is plenty for the 9590.

Thanks again for any input.
 
At 4.5 GHz, if I read this right you still had a 12° margin and failed after 7 hours? If so, I would just bump the voltage up a smidge and call it a day. Unless you're doing some serious work with this machine and absolute stability is necessary 7 hours of P95 is pretty good and most likely your PC will never work that hard under typical use.
 
At 4.5 GHz, if I read this right you still had a 12° margin and failed after 7 hours? If so, I would just bump the voltage up a smidge and call it a day. Unless you're doing some serious work with this machine and absolute stability is necessary 7 hours of P95 is pretty good and most likely your PC will never work that hard under typical use.

Hi yeah that's the issue though if I bump up the voltage a bit more when LLC kicks in voltage goes a bit higher than my cooler can handle and then I end up like -2c below the margin which is obviously not acceptable but like you said my usage would probably never hit what prime95 stresses at but OCD much...

Today prime ran for over 12 hours @4GHZ (gonna try 4.2ghz through the night) with all workers still going so I asked my mrs to stop the test so this has to rule out any other hardware instability like memory/power supply etc.

It's just that LLC... not like the old days of setting a vcore and it staying put at what you set in the bios... I will have to bit the bullet and get an AIO bit I've been wanting to upgrade to the 990 FXA UD7 rev 3.0. I have a spare UD7 1.0 but it dont support the 9590 and there's been no bios updates since 2012 so its highly unlikely gigabyte plan to add support for the 9590 eventhough I have read it's a high TDP board...

I suppose I could sell this redundant UD7 1.0 to someone who wants to stick an 8530/70 in it and then possibly use the funds to either put towards an AIO or possibly upgrade to the UD7 3.0 then save up for an AIO.

So many options...

Sorry for the long winded posts but hours have passed and brain ticks over.
 
Odd. I replied but it disappeared, I'll reply again later unless admin can figure out where it went.
 
Looks like you deleted it

Have you tried the 9590 in the UD7, there's really no difference aside from clocks and binning from any other Piledriver so I wouldn't be surprised if it works fine
 
Looks like you deleted it

Have you tried the 9590 in the UD7, there's really no difference aside from clocks and binning from any other Piledriver so I wouldn't be surprised if it works fine

Haha I must have pressed something whilst I went to edit out some typos and not realised. Thanks :)

I've not actually tried it mainly due to it not having a UEFI bios so I'm not certain my windows 10 would boot from it? Maybe I can be corrected.

I may try it on Sunday when I'm off work but I remember the UD5 when I bought it I had to flash the bios as it didn't support the 9590 out of the box so it thought it was an 8350 and auto clocked it at 4ghz. You would of though with the UD7 being their "flag ship" board that they would atleast support the 9590 like the rev 3.0 version does.. so yeah now your talking me into pulling the ud5 and swapping it for the 7, I'm still not expecting to hit the 4.5/7 or 5ghz mark but I'd atleast have a better board so to speak.

I just ran linpack xtreme 10 times over on 4.2ghz @ 1.404 vcore and it passed all 10 and core temp hit 54c (71c in core temp = 0 margin) so while I sort the kids out for the evening and shower, eat and communicate with the other half :( I shall stick prime on some small ffts and see what pops up if anything and get back in here.

Cheers Johan :)
 
The only issue I could forsee is if you installed Win10 in UEFI mode, you can tell in disk manager as UEFI always creates extra small partitions particularly one that's about 100 MB
 
The only issue I could forsee is if you installed Win10 in UEFI mode, you can tell in disk manager as UEFI always creates extra small partitions particularly one that's about 100 MB

Yeah my install is UEFI based (can also tell in the boot selection in the bios), I have a digital insider program key which is tied to my machine hardware i.d and I'm not entirely sure if and how I can change critical hardware without breaking the terms of my insider agreement etc so I'll have to do some reading on this. It's not really an issue if I have to install a non UEFI os as I use Onedrive to store all my apps etc so i suppose i can back track to windows 7 for testing it out as I have a proper product key for that OS.

Getting back on topic now so at 1.404 vcore at 4.2ghz I ran prime95 small fft for a couple of hours without issues and my temps increased to 57c as shown in core temp so still a good 14c off the margin but now I'm actually pretty much doubting that I can run 4.5ghz without breaking the margin for a 100% solid stable clock without upping the vcore another 0.025volts as that's all the increments that I can go up in.

I've basically done pretty much all i can do for case fan cooling. I butchered the side panel and added 3 intake 120mm fans, i butchered the back side panel and added a 120mm fan for back cooling the socket, I have 3 120mm front intake fans and 1 120mm rear exhaust fan. I suppose i could butcher the top and add a couple more 120mm exhaust fans but that's for another day to play with the dremmel hahaha
 
You could always switch to the blend test, that would give you some headroom I would think. Small FFT is a pressure cooker and not realistic at all IMO
 
Yeah I suppose you are right. I will be running blend tonight while we sleep to see how 4.2 ghz goes but I use small ffts for quick dirty results. What are your thoughts on Linpack Xtream?
 
I prefer P95 myself, I have used IBT AVX on FX before and it works fairly well but still prefer P95
 
Back