project stabilize sabertooth fx990 with amd fx9590

hi there,
i already posted in another thread that was about the same problem that i have (unstable system: freezing/BSOD)
my last post and your suggestions are on this site of that thread: http://www.overclockers.com/forums/...-990FX-R2-0-Amd-FX-9590-Freeze-Unstable/page5

i finally got my thermal paste and switched my cpu back to the fx9590.
now i'd like to try your suggestions and this is what i set in BIOS now:




now, i don't know if that's normal, but the displayed CPU/NB offset voltage still shows 1.125V... do i have to add my offset to the shown voltage to get the actual voltage to be used when saving those settings?



my hardware is as follows:
mb: sabertooth fx990 rev 2.0
ram: corsair 32gb vengeance pro 1866MHZ (CMY32GX3M4A1866C9)
cpu: amd fx9590

os: win 8.1 (but that doesn't really matter - also had those freezing/BSOD problems with win 10)

now, i don't know if that's normal, but the displayed CPU/NB offset voltage still shows 1.125V... do i have to add my offset to the shown voltage to get the actual voltage to be used when saving those settings?

Click to expand...

Yes so with the .15 offset you're cpu nb voltage is going to be 1.275. Have you tested the system after setting the Cpu Nb voltage to .15? With my Fx 8350 I needed 1.3875 Cpu Nb Voltage in order to run my ram at 1866 Mhz. You will need to test your system to see what works for it, having 32g of ram will also stress the Cpu Imc more then say 8g. You can also try giving the dram voltage a bump to 1.55. Additionally can you add a signature to your post so the hardware you have is in every post you make. To do so, click on settings in the top right of the screen and edit signature. Include your case, fans heatsink Psu etc. All these components are helpful for us to help you.

thanks for your answer. i'm running the settings from my first post right now and haven't had any freezes or BSODs since then (played for one hour and then surfed the web and streamed a movie with those settings).
but i also haven't really stressed the cpu and ram yet. i'll do so in the next few days.

Ya, this chip 9590 gonna be hot and temperamental. I will admit that cool and quiet is a friend to this chip and to any cooler as well. So that will be your major challenge is the heat spikes the hwmonitor can't see. So you may experience some lock ups and freezes. It's normal. Get up to about 5ghz, need 1.5125v. It's going to make heat rather quickly.

Your main goal is keeping stable while running fast and hard. First, you want to OC manual not DOCP.

While some may notice high ram speeds may take HT while on auto down to 2200mhz match by NB speed of 2200mhz. This is the boards way to help keep stability with FX chips. You have then the option to OC up to 2.6ghz and leave voltage on auto. Not saying your chip can't do well over 3ghz HT, I'm saying it'll just create more heat and performance gains are minimal for daily uses.

NB will find stability under 2500mhz. Really keeping it near 2200mhz stock will also help you find a good cpu OC.

The IMC will handle all your ram. It's nothing really. Seeing your not looking for 2400mhz as I am running, manually set your ram and ram voltage. Use suggested voltage in the SPD tab for the XMP profile. Some ram like a tad extra as you may notice there's always a little vdroop on the bios readouts. It's a normal Asus thing......

Really, to get 4.8ghz from 4.7 and no boost is a waste of time really. Your not going to gain a lot. I've been trying to leave Cool and Quiet on along with C6. This is NOT normal overclocking proceedures, but really does seem to help with stability and keeps the water loop from being heat soaked constantly. Have managed 4.9ghz constant with a load voltage of 1.44 - 1.456v cpu v-core. Much more than that, it just heat soaks the loop and will reach maximum temps.

Being the chip is already a high clocker, there's really not much more to get form it.... At least that's my take on it. Hot sob.

So here's an example of my current evening putz oc. Quick simple, games great Boosts to 5200mhz pretty stable. Up time is 0:23:59:26 minutes. Prime95 tested for One hour only. Most here recommend 2 hours.

ShrimpBrime said:

Your main goal is keeping stable while running fast and hard. First, you want to OC manual not DOCP.

Click to expand...

I don't get it.. Why everyone says that. D.O.C.P is just an easy way to load the X.M.P profile for memory. Other than that it allows manual control of everything.

Waza said:

I don't get it.. Why everyone says that. D.O.C.P is just an easy way to load the X.M.P profile for memory. Other than that it allows manual control of everything.

Click to expand...

Probably because half the time it makes a users PC unstable with all the other changes it makes.

Johan45 said:

Probably because half the time it makes a users PC unstable with all the other changes it makes.

Click to expand...

Then what other changes does it make? Ram voltage, speed and timings are the only things it should touch if there is existing memory multiplier for the speed specified by XMP profile.

Something else you can try, just to see if you can get it stable. Set the Dram to 1600 Mhz and then test it. If it's stable at 1600 then at least we know it's the ram causing the issue.

Johan45 said:

View attachment 161257

Click to expand...

Yes it will adjust other settings to achieve the specified memory speed. But it will maintain closest possible settings to stock as possible.
The thing is that 95% of memory runs in standard speeds 1066,1333,1600,1866,2133,2400 etc.. Those multipliers already exists for the memory so there is no need to adjust the BCLK.
If there is need to adjust BCLK for the specified memory speed, then it will touch the cpu multiplier(lower it to maintain closest to stock speed).
Only thing to look out for when setting non standard memory speed is the CPU/NB and HT multiplier. It won't adjust those if they are manually set so they can go well over stock.
I would not recommend using the preset speeds of DOCP. But for loading the XMP it just makes things simple.

That's all well and good but I have seen BIOS versions that don't properly register XMP profiles in the first place. Which would really throw this off. I actually posted some results in the Just Putzen thread about a year ago. It's just something I don't use or recommend personally.

Johan45 said:

That's all well and good but I have seen BIOS versions that don't properly register XMP profiles in the first place. Which would really throw this off. I actually posted some results in the Just Putzen thread about a year ago. It's just something I don't use or recommend personally.

Click to expand...

That is fair. It is actually Asus specific technology that allows the loading of XMP profiles and it may alter some of the XMP settings to suit AMD memory controllers. It is possible that there have been somekind of error or bug in the earlier Bios versions that has resulted incompatible settings to be loaded.

Johan45 said:

That's all well and good but I have seen BIOS versions that don't properly register XMP profiles in the first place. Which would really throw this off. I actually posted some results in the Just Putzen thread about a year ago. It's just something I don't use or recommend personally.

Click to expand...

Been a while "johan" but I remember your talking of problems with D.O.C.P. I also know that some but not all have problems with Kingston Ram. That is lately though with the Kingston ram and is likely due to bios updates. When you see a bios that says it is working to allow "more" memory stability...well they are messing where we cannot see.

A problem we have in forums is one person has something to work out for them and then suddenly it is seen as the "in thing" but that seldom is ever the overall situation. When we dispense this free help, we have to or should suggest the thing or things that overall work for the majority of users. Going now on something like 3.5 years with these Asus boards with D.O.C.P. and we have refrained from telling newbies to use that setting. It has worked out remarkably well that way because they then must learn something about their bios. Manipulating those bios items that are there for all to use and not some widget that may or may not perform as expected for the majority. We steer away from overclocking genies and the like because of the unknown variable results that come from what a bios programmer says the board should do. I fully prefer to have; to the best of my abilities, a full on idea of what will happen next by manually adjusting the bios.

I have multiple sticks of GSkill ares ram. I just purchased a 4x4gig kit of GSkill DDR3-2400 ram for use in CHV-Z and bought it in that sizing since it would also work in X-99 (not going there now) and this ram has thrown me a big curveball. It AUTO adjusted to DDR3-1866 in my new CHV-Z with tight timings. Never seen AUTO ram adjusting to a lesser speed "with" tight timings. But the newer CHV-Z is on a bios I have never seen in use before. Only experience I have had was with last/latest bios for CHV-Z. This board shipped with the third bios from being the latest.

I know D.O.C.P. seems simple for a certain task, but I will continue to take that 40 seconds or so it takes me to manually setup my ram timings to specification and then have a pretty good idea that some background bios manipulation will not happen without my full knowledge.

It is what D.O.C.P. can or may do without our full comprehension across the full spectrum of newbies that we don't suggest it as a way to setup their systems. However once they are not in the "how do I overclock my shett mode?", then they are more able to adjust their thinking when using bios genies and may or may not be slapped with an oddity. Just how it seems I have had to come to see helping these guys that so often come with n0 knowledge of bios setup or overclocking or cooling or well you get the idea "johan".

Luck guys.

RGone...

Waza said:

That is fair. It is actually Asus specific technology that allows the loading of XMP profiles and it may alter some of the XMP settings to suit AMD memory controllers. It is possible that there have been somekind of error or bug in the earlier Bios versions that has resulted incompatible settings to be loaded.

Click to expand...

It was an issue with high speed ram ( many members in the forum coming with 2400 2x4, 2x8 kits and having nothing but issues) that got me testing different BIOS versions on my CHVz. Typically I had stuck with older versions since I liked the results I was getting. Any way what I found was kind of surprising really, it even affected the 2133 ram I tried after. here's the link I finally dug it up. Idirectly it has something to do with DOCP in that the bios isn't always right. http://www.overclockers.com/forums/...Putzen-Round?p=7657778&viewfull=1#post7657778

I can't claim the experience that Johan and RGONE have with this but can state I used to use the DOCP to set my ram prior to the FX release. For whatever reason be it the ram or motherboard or bios combo I just can't find that stability I had got when I was overclocking my PII 1090T. A little easier perhaps for some but I can't claim for all and assume that based on what I have encountered.

Johan45 said:

It was an issue with high speed ram ( many members in the forum coming with 2400 2x4, 2x8 kits and having nothing but issues) that got me testing different BIOS versions on my CHVz. Typically I had stuck with older versions since I liked the results I was getting. Any way what I found was kind of surprising really, it even affected the 2133 ram I tried after. here's the link I finally dug it up. Idirectly it has something to do with DOCP in that the bios isn't always right. http://www.overclockers.com/forums/...Putzen-Round?p=7657778&viewfull=1#post7657778

Click to expand...

I remember reading trough your findings. The thing with the NB adjusting to ram speed is normal(CPU/NB must be greater or equal to RAM speed) and I too have noticed that SDP values for XMP profiles can be dodgy for some reason.
One thing to remember is that while XMP states the settings that the memory is quaranteed to achieve in feature supporting Intel systems, AMD memory controller is different and those settings even if correct might not work for AMD systems.
Maybe DOCP corrects some of these issues by altering some subtimings or DDR driving strenghts that might even be hidden from the user.

OT: Brings back memories messing with my old DFI infinity board where there were separate channel specific alterations to correct for different lenght PCB traces.

It wasn't the NB that was throwing me off so much as it was dropping the HT to 2200 IIRC. That little bit didn't make any sense to me as to why it would drop since theoretically with the faster ram/NB the IO should be faster as well NO? .

Johan45 said:

It wasn't the NB that was throwing me off so much as it was dropping the HT to 2200 IIRC. That little bit didn't make any sense to me as to why it would drop since theoretically with the faster ram/NB the IO should be faster as well NO? .

Click to expand...

Hyper transport is separate from the memory interface. CPU communicates with PCI-E and other hardware trough HT. Memory writes and reads go trough the memory controller.

This I know but if the cpu is working faster it should communicate with the system faster not slower. It just didn't make sense to me that the bus would drop but it was consistent. Any Way we're kind of muddying up Wullxz's thread here.