Epox Socket 939 vcore fluctuation - a straw poll

[5t3v0]

I have an Epox 9NPAJ SLi with Opty 144 running at 2.5GHz on 1.55v. My issue is that the vcore fluctuates dramatically between cpu idle and load states, to the point that I cannot get a decent stable overclock without raising the voltage above sensible limits.

When set at any voltage, the vcore will be over-volting at between 0.015 and 0.025 when idle and then under-volting by a similar amount when under full load. The voltage varies depending on which tool is used to measure it, typically it shows higher in Epox's own Unified System Diagnostic Manager (USDM) than it does in Speedfan or cpuz, but the trend is always the same.

Using the charting feature on Speedfan I recorded this change in vcore from idle to a Prime95 small fft test. You can see why I have to over compensate with the vcore in order to get stability under load.

Now as far as I can tell, there would be 2 possible reasons for this, bad psu or bad mobo. My money's on the latter here. My psu, incidentally, is a brand new Enermax 485W.

Now I've already read about people with similar woes on Epox boards but what I'd like to know is how widespread the problem is.

What I'd like to hear from anyone is the following info (this is mine):

Epox Model: 9NPAJ SLi
CPU & Speed: Opteron 144 @ 2.5Ghz
Chosen vcore: 1.55v
Average vcore at idle: 1.56v
Average vcore at load: 1.54v
Minimum vcore: 1.529v
Maximum vcore: 1.578
Diagnostic tool used: Epox USDM
PSU rating: 485W

Any similar results and experiences may help should I decide to confront Epox support, but with an Opteron on the board, they're just going to give me the finger aren't they 🙁

 

[Wentelteefje]

I have the same issue with my MSI K8N Neo4 Platinum... It fluctuates as well, and sometimes even more... I never noticed any problems however... I'm running a Venice 3000+ @ 2.4GHz @ 1.44V... Sometimes it dips to 1.38V, but it stays stable (these voltages are taken with CoreCenter, so that could be causing it as well), so I suspect it's mostly related to a software bug, as 1.38V should normally let it crash... My PSU is definately up to the task... Yours should also largely suffice...

Software monitoring programs are unfortunately not known for their accuracy... Try reading the voltages in the BIOS (although that will only give you the idle state)...

EDIT: BTW have the same RAM... Nice timings and speed at that voltage! :thumbsup: Mine only goes up to 2.8V... 🙁

 

[5t3v0]

Originally posted by: Wentelteefje
I have the same issue with my MSI K8N Neo4 Platinum... It fluctuates as well, and sometimes even more... I never noticed any problems however... I'm running a Venice 3000+ @ 2.4GHz @ 1.44V... Sometimes it dips to 1.38V, but it stays stable (these voltages are taken with CoreCenter, so that could be causing it as well), so I suspect it's mostly related to a software bug, as 1.38V should normally let it crash... My PSU is definately up to the task... Yours should also largely suffice...

I guessed this wasn't going to an Epox-only problem. Isn't the default voltage for the socket 939 A64s 1.35-1.4v? If so then 1.38v should be well within tolerance. Would explain why it's still stable.

Originally posted by: Wentelteefje
Software monitoring programs are unfortunately not known for their accuracy... Try reading the voltages in the BIOS (although that will only give you the idle state)...

BIOS readings show the same, albeit to only 2 decimal places. I dont trust the actual values from any of these progs but the trends are the same on all of them. It can't be normal to fluctuate 0.025v when idle and then drop 0.03 - 0.05v when the cpu gets loaded. Is that bad enough to request an RMA?

Originally posted by: Wentelteefje
EDIT: BTW have the same RAM... Nice timings and speed at that voltage! :thumbsup: Mine only goes up to 2.8V... 🙁

yeah, i only dropped to cas 2 a couple of days ago after I noticed another punter in these forums had the same ram & timings in his sig. It has run 24 passes of memtest at these settings so I'm pretty pleased. It has added a whole 26 3dmarks to my 3dmark05 score (woohoo) which makes up for the loss I suffered when I dropped the cpu from 2.6GHz to 2.5GHz (done in order to lower the vcore). I have had this ram up to 237MHz on cas 2.5 & I'm sure it will do more. Its great stuff. So is it your board or ram that's limiting your vdimm to 2.8v?

 

[torpid]

You don't have Cool N Quiet turned on do you? That will alter your voltage and other parameters dynamically to "adjust for load".

 

[5t3v0]

Originally posted by: torpid
You don't have Cool N Quiet turned on do you? That will alter your voltage and other parameters dynamically to "adjust for load".

No, that's disabled.

 

[Allio]

Ok, here ya go.

Epox Model: 9NPA+SLI
CPU & Speed: A64 3200+ @ 2.55ghz
Chosen vcore: 1.525v
Diagnostic tool used: Speedfan
PSU rating: 485W (same as yours, incidentally, assuming you have the 485w Noisetaker)

As I used speedfan I didn't get the exact values like you did, but you can get an idea off the graph. Similar, but not quite as big a variation. I'm actually always at load (distributed computing) so I never even realised that the vcore rose up to near 1.54 at idle. Annoying, but it's not a dealbreaker; I love everything else about this board. Plus I don't think it's THAT bad compared other nf4s.

On a side note, I left my ipod shuffle plugged in when rebooting to check the vcore. Holy crap, heart attack. Thank god for that diagnostic LED 😀

Another side note, I'm incredibly jealous of your ram. Value ram shouldn't be able to do that 🙁

 

[Wentelteefje]

Originally posted by: 5t3v0

Originally posted by: Wentelteefje
I have the same issue with my MSI K8N Neo4 Platinum... It fluctuates as well, and sometimes even more... I never noticed any problems however... I'm running a Venice 3000+ @ 2.4GHz @ 1.44V... Sometimes it dips to 1.38V, but it stays stable (these voltages are taken with CoreCenter, so that could be causing it as well), so I suspect it's mostly related to a software bug, as 1.38V should normally let it crash... My PSU is definately up to the task... Yours should also largely suffice...

I guessed this wasn't going to an Epox-only problem. Isn't the default voltage for the socket 939 A64s 1.35-1.4v? If so then 1.38v should be well within tolerance. Would explain why it's still stable.

Originally posted by: Wentelteefje
Software monitoring programs are unfortunately not known for their accuracy... Try reading the voltages in the BIOS (although that will only give you the idle state)...

BIOS readings show the same, albeit to only 2 decimal places. I dont trust the actual values from any of these progs but the trends are the same on all of them. It can't be normal to fluctuate 0.025v when idle and then drop 0.03 - 0.05v when the cpu gets loaded. Is that bad enough to request an RMA?

Originally posted by: Wentelteefje
EDIT: BTW have the same RAM... Nice timings and speed at that voltage! :thumbsup: Mine only goes up to 2.8V... 🙁

yeah, i only dropped to cas 2 a couple of days ago after I noticed another punter in these forums had the same ram & timings in his sig. It has run 24 passes of memtest at these settings so I'm pretty pleased. It has added a whole 26 3dmarks to my 3dmark05 score (woohoo) which makes up for the loss I suffered when I dropped the cpu from 2.6GHz to 2.5GHz (done in order to lower the vcore). I have had this ram up to 237MHz on cas 2.5 & I'm sure it will do more. Its great stuff. So is it your board or ram that's limiting your vdimm to 2.8v?

Yes, my Venice "defaulted" at 1.36V, but I have it overclocked to 2.4GHz, and IMO, 1.38V shouldn't be/isn't stable anymore...

EDIT: Board, sadly... 🙂 My RAM went up to 240MHz at standard timings, but I couldn't lower 'em...

 

[5t3v0]

Originally posted by: Allio
Ok, here ya go.

Epox Model: 9NPA+SLI
CPU & Speed: A64 3200+ @ 2.55ghz
Chosen vcore: 1.525v
Diagnostic tool used: Speedfan
PSU rating: 485W (same as yours, incidentally, assuming you have the 485w Noisetaker)

As I used speedfan I didn't get the exact values like you did, but you can get an idea off the graph. Similar, but not quite as big a variation. I'm actually always at load (distributed computing) so I never even realised that the vcore rose up to near 1.54 at idle. Annoying, but it's not a dealbreaker; I love everything else about this board. Plus I don't think it's THAT bad compared other nf4s.

Thanks for the post. You're also using quite a high vcore there & you're dropping 0.03v under load like me. If I reduce my vcore to 1.525v with this overclock, I fail the Prime95 torture test within the first few tests. Go back up to 1.55v and its solid. But its the overvolting that worries me as my PC is idle more than its loaded. But like you, I do like the board for everything else - it's a no frills version of yours (basically minus the firewire) and GBP 25 cheaper. A 939 nforce4 with good overclocking features and SLi (if I ever need it) for < GBP 80. Can't beat that value (in the UK anyway).

PSU is the Noisetaker, same as yours. I love how quiet it is.

Originally posted by: Allio
On a side note, I left my ipod shuffle plugged in when rebooting to check the vcore. Holy crap, heart attack. Thank god for that diagnostic LED

Why, what happened?

Originally posted by: Allio
Another side note, I'm incredibly jealous of your ram. Value ram shouldn't be able to do that 🙁

This ram has surpised me too. It came from my previous Athlon XP rig whare I ran it at stock. I was considering replacing it until I discovered what it could do on this new rig. I bought after reading this review on 3dvelocity, although this is for the ram with the slower timings than mine (2.5-4-4-8 as opposed to 2.5-3-3-6). I dont think GEIL get enough recognition in these forums. I would recommend them to anyone.

 

[Allio]

Originally posted by: 5t3v0
Why, what happened?

It freezes during POST! I was about to have a heart attack before I checked the LED and noticed it was freezing on USB initialization.

I realise now I made a pretty bad memory choice for my new rig, and the Geil would have been a better deal. But what's done is done. At least the motherboard is good 😉

 

[5t3v0]

Well, I have found a workaround for my vcore overvolting problem in the shape of AMD's "Cool 'n' Quiet". Now I know that eveyone is going around saying that you shouldn't use it when overclocking & I just accepted this at face value. However, I installed the driver from the mobo CD and set the power option to "Minimal Power Management" and presto, it works a treat, even with the voltage manually set. So now when the cpu is idle, CnQ lowers the multiplier down to x5 (giving 1.4GHz) and the voltage to ~1.26v, then under load, it clocks back up to x9 and ~1.55v. Does anyone know if it can be made to go any lower than this?

 

[gfngfgf]

I've used a program called RMClock in the past as an alternative to CnQ. I don't recall, though, if RMClock could take the voltage any lower than CnQ already did (my instinct says no, but it's worth a look).

 

[5t3v0]

Originally posted by: gfngfgf
I've used a program called RMClock in the past as an alternative to CnQ. I don't recall, though, if RMClock could take the voltage any lower than CnQ already did (my instinct says no, but it's worth a look).

Looks like a neat program. It has a "P-state transition" option where it looks like it would be possible to set up a user-defined number of "P-states" where the multiplier & voltage can be set. I'll give it a whirl. Thanks for the tip.