Wrong DDR Voltage Reading in BIOS With MA770T-UD3P Rev 1.0

[furb1]

Have MA770T-UD3P with latest official F10 BIOS dated 7/2010.  I set RAM voltage to 1.5 in BIOS, but the "PC HEALTH" BIOS screen shows a reading of 1.61.  When I raise RAM voltage to 1.6 in BIOS, "PC HEALTH" shows a reading of 1.71.  In summary, the RAM voltage reading in PC HEALTH is always 0.11V higher than the value set in the BIOS.

Which reading corresponds to the actual RAM voltage?  Is there probe points on the motherboard where I can check actual working RAM voltage with a meter?

TIA,
F

[bytheway_r]

You know, if you bothered to do a search here you'd have gotten your answer sooner.

http://forum.giga-byte.co.uk/index.php/topic,484.15.html

To answer your question - the reported value is the real one ( real voltage your RAM is running at is 1.61v when set to 1.5v ).

[furb1]

Thanks for the link.  I was running F4 BIOS and the lowest available RAM setting @ 1.6.  Switched to F10 BIOS, which gave me the 1.5V RAM option, but still could not lower the reading in PC HEALTH.  Going to reflash to F4 and see how high I can run the RAM at 1.6.  If I can still run this configuration with F10 BIOS set at 1.5Vdimm, then we've just confirmed that the reported value is PC HEALTH is correct.

Thanks again for your valuable input.

F

[bytheway_r]

Don't bother, really. I've done exactly that on the very same board ( even revision ). This whole problem confused me a little at the time, so it took a while to figure it out. You can read about my "adventures" in that linked thread.

It didn't matter to me much in the end as my RAM required 1.65v to run factory settings anyway. I feel for people who have this problem and use RAM specified for 1.35v ( considering that such RAM overclocks best till ~1.5v and 1.6v could actually make it less stable ).

[furb1]

OK...I'll trust your research and set BIOS voltage to 1.54 to achieve a real-world voltage of 1.65.  That's a shame, because I was able to stabilize the system with 1519MHz RAM @ 7-7-7-20-1T using 1.6 in BIOS (about 1.71V actual pumping thru the memory).  NB voltage is only 1.14V.  Probably no permanent damage to those sticks since I've only been running them like this for two weeks.

[bytheway_r]

Considering this I've decided to stay close to 1333 Mhz. Just so you know that there are potential risks to running your memory at higher frequency.

About voltage, what memory do you have? Myself I'm using OCZ Platinum 1333 Mhz CL 7 with 1.74v ( stock 1.65v ) and they're fine. Running as 1360 Mhz 6-6-6-18-24-1T at these volts. Keep in mind I have good airflow but if your RAM has heatspreaders then it should be fine at 1.71v. Generally, it's exceeding the +0.1v barrier that may be risky ( depending mostly on how warm the RAM gets ).

[furb1]

I have a fan blowing down on the Patriot 1600 RAM rated at 9 with 1.6V.  1.71V actual will net 1519MHz @ 7-7-7-20-27-1T.  RAM get very warm/hot under Prime blend and S&M 1.91 memory test.  No thermal concern under normal use.  I could increase the CPU multiplier from 12.5 to 13 to lower RAM speed to the 1400s with the same timing hoping to lower Vdimm to 1.65 or 1.67 range.

1519 is about as high as I can push the system.  NB voltage is only 1.14 vs. 1.1 stock.  I also think it may be better to drop the RAM speed to at least the 1400MHz range to baby the memory controller.  If the controller was designed for 1333MHz operation, then it should handle a 10% bump without missing a beat. 

[furb1]

Done with final adjustment.  Reported votage is per PC HEALTH in BIOS.  V NB is 1.14.  Managed to hit 1600MHz 8-8-8-22-32-1T with Patriot RAM @ 1.62V, and 1460MHz 7-7-6-20-27-1T @ 1.65V.  Will use 1460MHz configuration to minimize damage to memory controller.  RAM is warm during S&M and Prime.  1M Super Pi in 19.3 sec with CPU @ 3.56GHz (unlocked PII with 6M L3 cache).

[bytheway_r]

That's some pretty sweet RAM you got there. What's the stock voltage and timings like?

Don't worry about RAM heatsinks being warm. As long as they're not hot to the touch you're fine. Remember that DDR2 runs at much higher volts while using similar heatsinks.

One thing, as you keep writing about NB voltage, you do know that it doesn't really matter much anymore for RAM? I mean, it's a kind of last resort thing to tweak if you can't get your RAM stable. Nowadays it's the CPU-NB that's managing RAM so it's CPU-NB voltage that you may eventually want to increase.

Actually, it'd seem that, on Phenom II, CPU-NB overclock can boost performance about as much ( if not more ) as RAM overclocking. Some even claim that it should be overclocked along with the CPU to improve CPU stability. What I have yet to figure out is whether it's better to overclock it to the limit or keep it in relation to CPU frequency. Info out there is sometimes contradictory about this. Anyway, myself I'm keeping it at a comfortable value of 2,65 Ghz @ 1.25v ( stock is 1.20v ). From what I heard, it's reasonable to not exceed 1.35v but I've seen people go about as far as 1.45v.

[furb1]

The Kingston HyperX Blu was on sale yesterday for $12/2GB (1600), but I figured that it's going to take a lot of luck to run the Kingston at 6-6-6-1T with only 1.65V.  Only paid $13/2GB for these Patriot PSD36G1600KH modules (6GB kit).  Rated voltage is 1.6V @ 1600MHz @ 9-9-9-32-2T, and 1.5V JEDEC @ 1333 or lower speed.

I need to bump the NB voltage up to stabilized the CPU when I raised core speed from 2.6GHz to 3.56GHz.  CPU-NB VID is stock @ 1.2.  Only need to bump to 1.225 when I'm running RAM @ 1600MHz. 

Right now, my CPU NB frequency is 2192MHz (8x).  I plan to raise this to 2466MHz (9x) by bumping up CPU-NB VID later this week.  Suspect I will need more than 1.25 to stabilize this rig.  No plan to mess around with HT link @ 2192.  Did you make any other voltage adjustment to achieve 2650?  I'm running 1.65Vdimm, 1.14NB, 1.425CPU, and 1.2CPU-NB.  Other voltage at stock.

Are you saying that I should dial in 10x to max-out CPU NB frequency and jack up CPU-NB VID to stabilize the system?  We're probably looking at some scary voltage that 20% higher than the reference 1.2.  10% on top of 1.2 should be okay, but I still prefer to limit to 5% over if possible.

[bytheway_r]

CPU-NB overclock depends on your processor. I have a 720 BE ( C2 ) in another PC that can't stand any increase to CPU-NB voltage or frequency higher than 2350 Mhz. Generally, the newer the processor the higher clocks it should achieve. For example, Deneb ( quad or lower use these cores ) of the C3 revision can do about 2.8-2.9 Ghz in most cases with few going to 3.0+. At the same time, Thuban ( hexa core ) is more likely to reach 3.0 and some can do 3.2 Ghz.

Myself, I didn't have to increase any other voltage to get my CPU-NB stable at that speed. I wouldn't touch your HT link either as it's at the upper limit of what's known to not impact performance negatively. It should stay between 1800 and 2200 Mhz. If you want, you could try to lower it a little and see if your performance drops or increases.

Are you saying that I should dial in 10x to max-out CPU NB frequency and jack up CPU-NB VID to stabilize the system?

No, I'm just saying that you'll get better performance by overclocking it. I think that ~2.5 Ghz should be fine with 3.56 Ghz on the CPU. Some tests show that high CPU-NB clocks with inadequate CPU overclock can lead to performance drops, actually.

[furb1]

This CPU is an unlocked Phenom II X3 rev C2, with 6MB L3 cache.  The ceiling appears to be low 2.4GHz.  That's why I want to get the RAM and CPU stable, then work on the CPU-NB.  So far, system is not stable @ 2466 with 9x multi.  Lowering to 12.5 CPU multiplier to raise CPU-NB will raise RAM speed, which may be bad for the memory controller.

So far, I've raised CPU-NB from 1.2 to 1.3, CPU from 1.425 to 1.450, and Vdimm from 1.65 to 1.71.  Still no joy at 2466MHz.

[furb1]

After extensive tweaking, I was able to run at 2466MHz @ 1.3.  Other settings remain unchanged.  Will wait a few days to check for any anomaly.