The Socket 939 Appreciation Club and Knowledgebase [Official] - Page 156 - Overclock.net

**thlnk3r** · 04-03-08

Quote:

Originally Posted by blitz6804

I am now SnM stable at my present Sig stats. See attached:

Click me!

Blitz, congrats

**txtmstrjoe** · 04-03-08

I've been running some tests on my S&M-stable gaming rig (Opteron 170 @ 2.8GHz), trying to determine what is the lowest possible VCore BIOS setting which also is S&M-stable.

I've had a few interesting findings.

First is that Poser was right when he said (a few weeks ago) that VCore directly affects the effectiveness of the IMC. Intuitively I suspected he was right, and when you really think deeply about things, it's easy to see why. As I recall, Poser also suggested that since the IMC is integrated, you cannot starve it of VCore; if you do, RAM will be affected. This explanation is right on the nose.

My gaming rig stays stable through multiple S&M runs at 2.8GHz, 1.2875V VCore (BIOS setting). The problem, though, is that sometimes, when idle, the rig will lock up. It's a phenomenon that happens intermittently. I'm pretty sure it's a VCore-related issue since I've changed no other setting in the BIOS.

Another interesting finding is that VCore's effects on CPU core temperatures is not nearly as influential as ambient temperatures. In other words, I had expected a .6V drop in VCore to really be reflected in CPU temps. My findings were that the VCore drop of .6V drop in VCore resulted in, at most, a 1 degree C drop in core temps. Ambient temperatures still had a much greater influence.

I hope someone can find this information interesting and useful. Better yet, it'll be interesting if someone can run similar experiments and corroborate my observations. It's one way to eliminate dodgy systems/components, and we can definitively conclude that this is predictable system behavior.

iandroo888 · 04-03-08

Quote:

Originally Posted by txtmstrjoe

I've been running some tests on my S&M-stable gaming rig (Opteron 170 @ 2.8GHz), trying to determine what is the lowest possible VCore BIOS setting which also is S&M-stable.

I've had a few interesting findings.

First is that Poser was right when he said (a few weeks ago) that VCore directly affects the effectiveness of the IMC. Intuitively I suspected he was right, and when you really think deeply about things, it's easy to see why. As I recall, Poser also suggested that since the IMC is integrated, you cannot starve it of VCore; if you do, RAM will be affected. This explanation is right on the nose.

My gaming rig stays stable through multiple S&M runs at 2.8GHz, 1.2875V VCore (BIOS setting). The problem, though, is that sometimes, when idle, the rig will lock up. It's a phenomenon that happens intermittently. I'm pretty sure it's a VCore-related issue since I've changed no other setting in the BIOS.

Another interesting finding is that VCore's effects on CPU core temperatures is not nearly as influential as ambient temperatures. In other words, I had expected a .6V drop in VCore to really be reflected in CPU temps. My findings were that the VCore drop of .6V drop in VCore resulted in, at most, a 1 degree C drop in core temps. Ambient temperatures still had a much greater influence.

I hope someone can find this information interesting and useful. Better yet, it'll be interesting if someone can run similar experiments and corroborate my observations. It's one way to eliminate dodgy systems/components, and we can definitively conclude that this is predictable system behavior.

interesting. my rig is stable but soemtimes it jsut locks up randomly. temps are fine and im not doin anything intensive. or anythign at all. its random. liek last night.. i was studying and talkin to a friend for help on AIM. and it locks up randomly. im like uhhhhh @_@ lol

blitz6804 · 04-03-08

The entire extent of my voltage games have been due to CnQ.

At idle with CnQ shut off (1.340 V), I get 30/35º C.
At idle with CnQ turned on (1.088 V), I get 25/27º C.

So a 0.252 V difference amounts to a difference of 5/8º C.

An average of 0.0387 V per degree Celsius, all things being equal.

I do not know what will happen under load (because CnQ shuts off) but if I have a chance, I will try it for you later.

Oh, and small tidbit of information:

With my DDRV = 2.78 V (2.70 V in the BIOS), I am 100% S&M stable
With my DDRV = 2.67 V (2.60 V in the BIOS), I am about OCNS2 stable. I get that cursed blue screen with "IRQL_NOT_LESS_OR_EQUAL." What does that mean for me? It means I am going to keep overvolting my RAM and they are going to darn well like it!

**txtmstrjoe** · 04-03-08

Quote:

Originally Posted by iandroo888

interesting. my rig is stable but soemtimes it jsut locks up randomly. temps are fine and im not doin anything intensive. or anythign at all. its random. liek last night.. i was studying and talkin to a friend for help on AIM. and it locks up randomly. im like uhhhhh @_@ lol

I'm not at all surprised. I bet you that the random lock ups is due to insufficient VCore for the overclock.

We have the same motherboard, so I'm fairly certain that it's a quirk of the board more than anything else.

Parabellum · 04-03-08

CPU looks to be dead

Unless there's something I overlooked when putting everything back, it's not showing any signs of life..

**thlnk3r** · 04-03-08

Quote:

Originally Posted by Parabellum

CPU looks to be dead

Unless there's something I overlooked when putting everything back, it's not showing any signs of life..

Parabellum, what are the symptoms? Does the machine power on? Do you get a POST screen? Any beep codes?

Good luck

blitz6804 · 04-03-08

I have a question for all the 939ers. What sort of memory bandwidths do you get? Theoretically, DDR400 (PC-3200) should be 3.2 GBps no? How does dual-channel factor into things?

**thlnk3r** · 04-03-08

Quote:

Originally Posted by blitz6804

How does dual-channel factor into things?

Blitz, with dual channel you have seperate channels accessing the memory control which increases throughput (bandwidth). I think others here and myself would need to run some Sandra benchmarks for comparison...

**txtmstrjoe** · 04-03-08

Quote:

Originally Posted by blitz6804

I have a question for all the 939ers. What sort of memory bandwidths do you get? Theoretically, DDR400 (PC-3200) should be 3.2 GBps no? How does dual-channel factor into things?

When I get home from work, I'll upload a chart of RAM bandwidth data. I can't remember what the results are, of course (which is why I keep charts for such things), but manipulating such things as the various timings parameters has a discernable effect on bandwidth and RAM access latencies.

FYI, the spreadsheet was made with OpenOffice's Calc program (I'm not sure if MS Excel can read it).

Edit: The experiments where the data came from are on hold as I need some new RAM to test with. My ancient Patriot PC3200LLK set finally kicked the bucket during these tests.

		Overclock.net - Overclocking.net > AMD > AMD - General
The Socket 939 Appreciation Club and Knowledgebase [Official]

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)