Undervolting Penryn?

[EIN502]

It just hit me. You were saying that you can not get your index 7 to drop right? If so uncheck index 8 and then see. Just forget about index 8 for now in all forms. I think I had this same problem, my system would say that I was at the multiplier of index 7, but at the VID of index 8. To prevent this I unchecked index 8 under "Profiles". I hope this works.

[rtrdogs]

Well, I took all these screenshots along the way to show what I was doing. And then I started Orthos again to take some shots during testing.

No change. Unchecked IDA (index 8) and so far, it's holding the VID on index 7 that is currently 1.125 in the CPU Info tab. The Maximal still says the default 1.1375 and will not change at all (this is still in the CPU Info tab).

So right now, fingers crossed, without IDA on it's working for testing. I'll keep you updated!


Okay an update....looks like RMClock, even the current beta version will not work with IDA turned on..don't know if that's just for Penryn, though, or just an IDA thing.

Here's a snippet from the official discussion among the beta testers and developer in the RMClock forum:

"stisev
One thing that doesn't work is IDA mode. I selected IDA mode @ 11.0x and even left it at its default voltage. I used SP2004 ORTHOS (dual core edition) to stress my CPUs but IDA 11.0x mode didn't work. My comp remains on 10.0x mode."

From my personal experience and the other observations, it is hardly possible that the IDA will really show up the highest FID (i.e., 11x on your system). One can rather observe its action when running a single-threaded application by the CPU load level values reported by RMClock like 110% (which is 100% * 11x/10x).


"John Ratsey
However, monitoring is showing 10x as using 1.25V although I have set it to 1.05V."

This happens when, and only when you're telling RMClock to use IDA state as well. Looks like there's no way around it, other than not to use IDA.

"Also, it may have been answered before but, why if we enable IDA, does IDA's voltage stick to the FID below IDA even if it's set differently and when the CPU is not even using IDA?"

Please see my answer to John Ratsey just above. Your CPU is actually attempting to use IDA as soon as you include it into the P-states list and when the CPU demand is high. Since it is attempting to use this state, it sets the IDA VID immediately, even if the current FID remains one step below the IDA FID.

And the whole discussion is here:

http://forum.rightmark.org/topic.cgi?id=6:1423

So this is making more sense now because I knew I was following the directions to a T!

[rtrdogs]

So.... is this too low for the 12.0 index on a T9300? I tested all the way down to 0.9625 before I got an error.

It just seems wrong for the voltage to be that low but Orthos gave me an error right away at .9625 and I'm running an extended blend test for the next couple of hours on top of the 40 minutes for testing at 0.9750.

If you think this is okay, what should I lower the other multipliers down to?
Their default was all the same at 1.1375, with the lowest down to 0.9250.

Right now all of these are the default except the 12.0x index.

index 0 ---- SuperLFM ---- 8.0x ---- 0.9250
index 1 ---- Normal ---- 6.0x ---- 1.1375
index 2 ---- Normal ---- 7.0x ---- 1.1375
index 3 ---- Normal ---- 8.0x ---- 1.1375
index 4 ---- Normal ---- 9.0x ---- 1.1375
index 5 ---- Normal ---- 10.0x --- 1.1375
index 6 ---- Normal ---- 11.0x --- 1.1375
index 7 ---- Normal ---- 12.0x --- 0.9750
index 8 ---- IDA ---- 6.0x ---- 1.2250 (OFF/UNCHECKED)

My highest temp was a whopping 79.8 before I started and the highest temp during the 40 minute test at 0.9750 was 60 even. That 79.8 was with IDA on, if that makes any difference. That sure seems hot for a Penryn but I ran Orthos twice at those settings and it was pretty scorching both times and showing those temps. I can definitely tell there's less heat output now. It's very cool to the touch.

So here's what it looks like right now (Orthos is running)...screen attached.


Edited to add a second screen. Look how far this drop in voltage is from the default. Can this really be right?

[flipfire]

Index--Type-------FID-------VID--------Original VID
0------SuperLFM---8.0x------0.9500V----0.9500V
1------Normal------6.0x------0.9500V----0.9500V
2------Normal------7.0x------0.9500V----0.9875V
3------Normal------8.0x------0.9500V----1.0125V
4------Normal------9.0x------0.9500V----1.0500V
5------Normal------10.0x-----0.9500V----1.0750V
6------Normal------11.0x-----0.9625V----1.1125V
7------Normal------12.0x-----1.0125V----1.1375V
8------IDA---------13.0x-----1.0375V----1.2250V

edit: wait sorry did you say your 12x can go as low as .975 O.O.. if so that is a very low voltage and a very good undervolt if its stable at 0.9750. You have to adjust your multipliers accordingly so its not higher than the next (eg. above)

[rtrdogs]

Sorry flipfire...let me clarify. I haven't changed any of the voltages at all except for the 12.0x so the numbers I posted are all defaults ,my laptop shipped with except the 12.0x index.

IDA is already disabled because RMClock wouldn't change the "Current" value of the highest multiplier unless IDA was unchecked. That's why RMClock wasn't working for me when I started this thread and I dug up a thread on the beta forum of RMClock that I referenced above that said IDA has to be disabled because it's not working with current RMClock version (not even the current beta).

So...back to my question. The voltages I showed were all the defaults that I haven't changed except the 12.0x that is in testing.

The voltages you referenced are EIN502s numbers, right? I've been testing for an hour and a half at 0.9750 for the 12.0x index with no errors. Like I said, it seems like that's a low voltage but I remember seeing in other threads that we shouldn't copy someone else's numbers because it will be different for each computer even running the same processor. Is that correct?

If I'm able to test at that low (0.9750) with stability (so far) should I stay that low? Thanks for the additional help.

[flipfire]

yeah sorry i misunderstood that.. the post was abit long.. having a slow day here at work x.x

If its stable @ .9750 for at least an hour on ORTHOS then dayumm thats a good undervolt. Disable IDA for now because it seems to cause abit of issues.

You have to test the other multipliers stability by unticking the 12x multiplier in RMclocks profile page. If you untick 12x and hit apply.. your computer will max out on 11x multiplier.. Run orthos for 45mins.. this will take some time for a full stability test

you probably wont be able to change the other multipliers much. the lower the multiplier the more sensitive it is to BSOD's

[rtrdogs]

No prob flipfire. As you can see from the posts, I spent the day pretty much talking to myself so sorry I was a bit long-winded and didn't make any sense!

Thanks for the additional help on testing each multiplier. I don't have much wiggle room with that voltage coming in that low but I definitely want to make sure it's stable and be thorough.

[flipfire]

Sorry i really need my caffeine dose today haha let me try make this clearer-er. Yes you shouldnt copy other peoples voltages but EINC had a safe set of voltages to start off with.

To test the other multipliers:

1)Untick 12x multiplier on RMclock profile page and hit Apply. This will disable the 12x multiplier which is the full power of 2.5ghz.

2)Set your 11x a little lower than your 12x. Maybe 9.500v' ish. Run ORTHOS and go to RMclock's CPU info page. You will see that its maxing out at 11x multiplier and maybe hitting 2200mhz'ish. Let ORTHOS run for 45mins or so to test stability. If it bsod, you know the drill. if it doesnt try go lower voltage.

3)Then do the other multipliers one by one (only test 7-12x multiplier). 6x multiplier is your idle multiplier, you cannot use ORTHOS to test this. Same with SuperFLM. You just have to set it on a certain voltage and let it idle. Set your superFLM to 6x multiplier btw so it idles at a super low 600mhz.. 8x will idle at around 800mhz or so

4)After you are done testing the multipliers. Tick all the multipliers again in the main profile page and hit Apply. Now go to the "power on demand" sub profile page and re-tick all the boxes again, as per undervolting guide

[rtrdogs]

Makes perfect sense. I'm on the same page with ya and appreciate the help. I'll update with numbers tomorrow when it's all said and done. Thanks again! I was about to give up on undervolting when I started this thread and I'm so glad it's worked out. This sucker's so cool now!

[rtrdogs]

Just a few follow-up questions, please...

Here's where I am at the moment (screen attached but I haven't lowered 6.0x-8.0x multipliers down yet).

I've gotten 12.0x to a stable 0.9750 after 40 minutes trying out the voltage and then another 2 hours on top of that testing in Orthos for stability.

11.0x won't go lower than 0.9375 and tested for 45 minutes in Orthos with no problems. 10.0x is at 0.9250, the lowest and tested for 45 minutes in Orthos with no trouble.

Do I need to test the other lower multipliers (9.0x, 8.0x, 7.0x...) at 0.9250 with Orthos?

Also, SuperLFM is already set at the lowest voltage of 0.9250. 6.0x under the CPU Info tab is listed as 1.0 V for "Startup." Will lowering the 6.0x to 0.9250 mess anything up?

These numbers just seem so low that I'm terribly afraid I'm going to try and turn on the computer and nothing will happen!

3)Then do the other multipliers one by one (only test 7-12x multiplier). 6x multiplier is your idle multiplier, you cannot use ORTHOS to test this. Same with SuperFLM. You just have to set it on a certain voltage and let it idle. Set your superFLM to 6x multiplier btw so it idles at a super low 600mhz.. 8x will idle at around 800mhz or so

Flipfire, the voltage for SuperLFM is on its default of 0.9250 and the FID is 8.0x. Are you saying to change the FID to 6.0x instead of 8.0x for SuperLFM? Just want to make sure I'm reading that right.