8350 help cant get 4.8 stable?

[god_43]

i just got an 8350, 4.6 is stable at 1.44 vcore, but anything after that errors out (occt). i have a crosshair formula v mb, CPU Load Line Calibration - Ultra-High. i used the extreme oc profile so everything is pretty disabled/stock, i am only using vcore and multiplier. i dont get it?



thanks.

[tbone8ty]

do you have latest 1703 bios?

try bumping CPU-NB volts a little? whats stock 1.225v? try 1.25v

whats batch number of cpu?

[god_43]

do you have latest 1703 bios?

try bumping CPU-NB volts a little? whats stock 1.225v? try 1.25v

whats batch number of cpu?

stock cpu-nb was/is 1.187, i tried both 1.225 and 1.25; but that didn't work.


# is 1244PGT.

thanks.

[tbone8ty]

hrrrmmm

all cpu configurations disabled? APM off?

memory settings okay?

i dunno

yur on watercooling so temps shouldn't be an issue right?

i did here piledriver cores dont like prime95...doesnt occt use this method?

[god_43]

hrrrmmm

all cpu configurations disabled? APM off? - yes and yes

memory settings okay? ye....well they should be, but you are right maybe the oc profile changed something there. i'll check it out later.

i dunno

yur on watercooling so temps shouldn't be an issue right? temps are great, the core temps dont go above 38 (new pump and res).

i did here piledriver cores dont like prime95...doesn't occt use this method?

i know prime does not work well with bd/pd, but i thought occt would be different. occt is harder than prime...so idk. what else can i stress test with? everyone in other threads are using prime?

[Zeus]

Welcome to the wonderfull world of Vishera. I've had two FX 8350's yet, one 1236PGN and 1244PGT. Both won't run stable at 4.8GHz. 4.6-4.7GHz is where the fun ends.

[FlanK3r]

As The Stilt write some weeks ago, 8-cores FX has usually one or two core worse than others. How did you tweak LLC (not extreme!) and CPU current protections? What about CPU VDDA (AMD PLL)?

[god_43]

Welcome to the wonderfull world of Vishera. I've had two FX 8350's yet, one 1236PGN and 1244PGT. Both won't run stable at 4.8GHz. 4.6-4.7GHz is where the fun ends.

....that sucks...wtf i could do 5ghz on my 8120?

As The Stilt write some weeks ago, 8-cores FX has usually one or two core worse than others. How did you tweak LLC (not extreme!) and CPU current protections? What about CPU VDDA (AMD PLL)?

llc is at ultra-high for this board (that seems to keep the voltage stable), my cpu over-current protection is disabled. the vdda is left at stock, it was at 2.495, i tried to change it to 2.85; but that only bought my overclock 20s more. memory settings are all normal and on auto (they are stock). that really sucks if this thing cant go past 4.7.

[AMDforME]

God_43-

Where did you get the idea that all FX processors OC'd to 4.8 GHz. ????

Some do, some don't. 4.5-4.7 is the norm. 4.8+ GHz. may be possible with 1.5v or not but you'll need water cooling to run OCCT/P95 for any length of time at that frequency with that voltage. In normal PC use you typically can use just a highend HSF @ 1.5v, but not for stress testing for any period of time.

IME and that of many reviewers, the Vishera CPUs are slightly different when it comes to OC'ing than past CPUs. Too much Vcore or CPU-NB voltage can actually cause system stability or errors especially in P95/OCCT. Finding the exact ideal Vcore/CPU-NB voltages and the maximum CPU frequency is a long trial and error process. Most people are happy just to run stable at 4.7 GHz. but in most cases you can tweak a little more from the CPU if you get the above voltages optimised and also by tweaking the FSB setting.

Asus mobos/BIOS vary in capabilities/stability just like other mobos/CPUs so you get what you get.

[tbone8ty]

try:

cpu over-current protection to 130%
cpu/nb over-current protection to 110%

[FlanK3r]

and next things, whats your PSU? Do you can testing how your CPU scaling with voltage (from example 1.3V, 1.325V, 1.35V, 1.4V, 1.425V) and temperatures. No need for this PRIME95 (its very bad stress for FX and for Ivy B.), try for keeping time x264FHD benchmark with this voltages and finding clocks.

[AMDforME]

IME there is no issue with using P95 nor OCCT to stress test the FX/Vishera processors as long as you don't run excessive core temps. Many folks do this without issue. CPU stability depends on many factors. Only you can determine how much stability is right for your PC , based on your needs and priorities.

[FlanK3r]

but many with issues at stock settings too (FX-8150)...Or next example is i7-3770K OC, PRIME test OK, but some BSOD or frozed in idles sometimes ...Frozed in x264 video encoding. Something must be wrong.

http://www.mersenneforum.org/showthr...t=16779&page=4

[AMDforME]

Just because some folks have issues doesn't mean the software is defective. I suspect that the stability issues are due to other issues since the stress test software runs fine for most people. If it was a software issue then you'd expect it to apply to most people not just a small percentage. If the same people that are having issues with P95 are also having issues with OCCT, then it's almost certain that their hardware or a corrupt O/S is to blame, not the stress testing software, which justs runs numerical calculations and checks for a proper result.

[tbone8ty]

Hey god_43

I'm having the same problem with my new fx-8350 (1249pgt)

Anything past 4.7ghz is not "stable" in prime95 or occt or stress test in Amd overdrive.

Im on air cooling though.

Blaaaah driving me nuts. Ah we'll ill keep at it until this chip breaks in. I'm no where near a pro at this lol

[tbone8ty]

Wonder if the "clock mesh tech" has anything to do with it. Considering the powe consumption is a lot lower when OC'd compared to bulldozer

[Zeus]

I think you guys expectations of these chips are a little too high. I've had 3x FX 8350, non of them ran 4.8G stable at reasonable voltage. Pretty sure i could have gotten 4.8G by brutal force like 1.5V+ and LLC set extreme but that was not what i was looking for. 4.6-4.7G seems to be the sweet spot.

[AMDforME]

Wonder if the "clock mesh tech" has anything to do with it. Considering the powe consumption is a lot lower when OC'd compared to bulldozer

There is no "clock mesh tech" in Vishera CPUs. Vishera generally will OC a little better than BD and the higher frequency increases the total TDP so that the two are very similar in TDP.

I think you guys expectations of these chips are a little too high. I've had 3x FX 8350, non of them ran 4.8G stable at reasonable voltage. Pretty sure i could have gotten 4.8G by brutal force like 1.5V+ and LLC set extreme but that was not what i was looking for. 4.6-4.7G seems to be the sweet spot.

I agree that expectations are too high. I'm not sure where people got the idea that most of the FX-8350 CPUs OC'd to 4.8 GHz. ??? As noted 4.6-4.7 is the sweet spot and where power consumption is the lowest for the best gains. Temps are manageable also with the lower Vcore and CPU-NB that is required.

[tbone8ty]

didn't FX-8150 overclock better or about the same? or am i remembering wrong?

i guess i got my hopes up with u guys posting suicide runs and CB runs at 5.2ghz+

[tbone8ty]

Whats suitable cpu-nb freq? For 24/7 air? 2400mhz? Or can I go higher? what about HT link freq? Keep it on auto?

Auto cpu-nb volts puts it a 1.4v wth? Isnt that really high?

[The Stilt]

Usually 2600MHz is doable depending on DRAM frequency and installed ranks per MCT.
2400MHz (2 ranks per MCT) / 2400MHz NCLK is usually doable with NB VID + 10x VID (0.1250V).
2133MHz (2 ranks per MCT) / 2600MHz NCLK should also be ok with the same voltage.
Raising the VDDCNB higher than that usually does not help on air.

[vario]

i just got an 8350, 4.6 is stable at 1.44 vcore, but anything after that errors out (occt). i have a crosshair formula v mb, CPU Load Line Calibration - Ultra-High. i used the extreme oc profile so everything is pretty disabled/stock, i am only using vcore and multiplier. i dont get it?



thanks.

Your result is pretty much normal.Going into vishera i too had higher expectations, of most of them doing 4.8 (easy) with lower power consumption etc.
After my own failure i started reading more on the subject, it turns out almost no one does any kind of normal stability testing with their clocks, and the ones that do, more often than not are on water or high end water.
Plus 8320 consistently oc worse than 8350.My kinda results arent that rare either, thing is vishera will do some benches even most of them on clocks that are not fully stable ,it has biger margin for that.
1.4v for cpunb is very high, i do ~2600 at 1.287 - 1.32v depending on memory settings although i dont have memory/nb clocks nailed down stability wise.I could bench at 2800 @1.5v .
One more thing to add, up to 1.4v cpu vcore ,power consumption is OKish , after that it gets really high, after 1.5v it gets insane, in excess od 300W for cpu alone (400W total running only prime no 3d load)
And yea, OCCT loads cpu similarly like prime95 x64.Nothing wrong with it, works fine.
All in all, you got 200mhz more than me, seems ok than

[tbone8ty]

okay i set 2400mhz cpunb @ 1.25v(in bios) and left HTlink on auto for 2200mhz...stable at 4.6ghz @ 1.45v(in bios) for now


been playing with LLC for CPU and CPU-NB....setting auto for these overvolts them alot when checking in hwmonitor or asus suite probe II.

so i set them at high (cpu llc), regular(cpu/nb llc) weird how they dip down when under load

[Zeus]

Ht link should be 2600 default.

Funny how software reads out voltage dipping down, DMM tells a whole different story. I'm really afraid to go by what e.g. CPU-Z tells me for Vcore, if i set LLC to high like you do it reports voltage drooping under load whereas my DMM tells me it's actually overvolting 0.04-0.05V over the value i've set in bios!

For example 1.4875V set in bios droops down to 1.44V according to CPU-Z but my DMM tells me i'm running 1.53V under load! I tend to put most trust in my DMM to be honest.

[The Stilt]

I have never seen the ITE8721F monitoring misreading the voltages at idle or under load on any of my Crosshair V / -Z boards.
Also setting the CPU LLC to "Ultra High" and NB LCC to "Regular" yields almost exact output voltage (within 7mV) when measured from Asus specified Vdd and Vss points with Fluke DMM.

Yet everyone seems to be saying the monitoring readings undershoot under load and overshoot when measured with DMM