Core Voltage? Memory Bandwidth?

Kazeko

Couple questions that arose today, after installing my TR120 Extreme.

I did some testing with overclocks, memory bandwidth and what have you, and I'm a bit lost.


First and foremost - With heat as a non-issue, I try to push the 1.5v marker, but can't get the system stable at any OC with anything more than 1.45v going through the core. Admittedly I didn't try feeding the FSB more voltage. Perhaps it was a FSB heat issue?

Any ideas what could be causing this? Voltage Regs? FSB needs more voltage or is too hot? I'm not sure where to start with this short of just throwing something out and hoping it works - plus, given the question below, I"m not sure how far I should push this.


Furthermore, should I avoid pushing more voltage through it than Intel's specifications indicate? I.e. The voltage range for a G0 Stepped Q6600 is .90v to 1.5v I believe? Should I avoid passing the top range, or is there any danger besides the potential heat issues?


Next item on the list - Memory Dividers, and what have you.

I'm currently using 4 gigs worth of Crucial Ballistix @ 1066mhz, 5-5-5-15-2T

Would it be better to drop the clock on the RAM down to a 1:1 ratio and speeding up the timings for higher over clocks (i.e. 400x8 or 450x8) or leave it at the OE settings, and simply jockey with a 4:3 divider or what have you?

What limited memory bandwidth testing tools I have would lead me to believe a higher frequency for the RAM is better, but after trending many of the OC'ers on these forums, I'm beginning to wonder if it really is better?

PangingJr

about the Memory Bandwidth?

motherboard's chipset platforms and all of the memory modules are not created equal. there are some good reasons why motherboard's BIOS option of memory divider, system memory multiplier, DRAM timings, come to exist.
but for me, there is only one reason for that that seems to make sense the most, it's for you to try to figure it out, try it and see what is the best setting and the worst setting on your hardwares, and for your needs.

ChaosMinionX

Is that Q6600 you have a B3 or G0 stepping? And 680i if its not the later edition, sucks for overclocking quads.

Kazeko

G0 stepped, VID is 1.275v. The NF68 (680i) is the later, A1 revision.




I suppose my bigger question is whether or not exceeding the 1.5v limit is a terrible thing, if thermals are good.

ChaosMinionX

Ahh ok then.... mind listing off a BIOS template?

Kazeko

Not sure what exactly you'd be looking for, just settings and what have you? I'll just take a picture if it's all the same.

ChaosMinionX

Well I have never owned a 680i mobo, however I can point you in the right direction if I have the template (BIOS settings as they appear in BIOS pertaining to overclocking.....overclock settings, voltage settings, any special mobo settings in their OCing section.)

-Chaos

PangingJr

for air cooled processor overclocked and at 1.5V i can only assume you can use this high core voltage from time to time only, likes for a day, or use it for benchmark tests and numbers. but it's not that safe to be using 1.5V under 24/7.

ChaosMinionX

General voltages on Core2 based chips for Intel....

Air Cooling (Highend) - Dont exceed 1.55v 24/7 usage, if temps allowing and feeling bold go up to 1.6v

Water - 1.50-1.65.

Phase/LN2/etc- go nuts.

You wont lose a chip at 1.55-1.6v if your temps are in order, being that you have a G0 with a pretty low VID you shouldnt have any problem in all honest doing 400x9 3.6ghz on like 1.45v

PangingJr

i used to try 1.575V on E6300 and Gigabyte P965-based board for like a day, and another day, and again and also use the 1.575V core voltage very often. i remember all the temps were quite decent, at least that was all the progarms that i've used to monitoring temps at the time tells me. then one day, all of sudden, i found the hardwares have gotten so many FSB holes.
i have no idea what was causing that problem, whether the motherboard's chipset, components or the system processor itself have started to show damage from what i had done to them. i don't know. lol.

Kazeko

I'll get the bios info up later tonight, I'm betting it's the Mainboard it's self though.


It has no problems up to about... 1.45v//3.5ghz or so. After that it's just fun house mirror - Tried 400x9 on 1.5v and it posted without issue, but locked up as soon as I tried to run Prime95. - I'm presuming it's the mainboard.


Maybe I'll go drop the cash on an X38 mainboard.

ChaosMinionX

I could tell you lots on any Asus platform Specifically the Asus Maximus Formula if you get it

Kazeko

Alright, here's the bios information. (Pertinent, anyway, unless you need device and IRQ information.)

PCIE #1-3 - 100mhz
SPP<->MCP Ref. Clock - 200mhz

HT Multipliers
nForce SPP <-> MCP - 5x (Can change SPP -> and SPP <- )
PCIE Spread Spectrum (Auto)
All other Spread Spectrums - Disabled

Memory
1066mhz (533)
tCL - 5
tRCD - 5
tRP - 5
tRAS - 15
CMD - 2T
tRRD - 4
tRC - 30
tWR - 6
tWTR - 10
tREF - 7.8uS

CPU SETTINGS
------------------
Limit CPUID Max Val - Disable
Speedstep - Disable
Thermal Control - Disable
C1E Halt - Disable
Execute Disable Bit - Enable
Virtualization Technology - Enable

Voltages
----------
CPU Core - 1.17v (Setting @ 1.225)
FSB - 1.4v (Auto)
Memory - 2.20v (2.20)
SPP - 1.4v (Auto)
MCP - 1.5v (Auto)
HT - nForce SPP <-> MCP - 1.2v (Auto)


There was nothing else I could dig up besides the various PnP settings, and such, no memory multiplier option to change as well.

ChaosMinionX

Turn off PCIE Spread Spectrum, and increase your core voltage.... safe to go up to 1.55v on air with a TRUE and good temps... you should be able to do 400x9 or 400x8 np at 1.4-1.45v.


EDIT: And turn off Virtualization technology.

Kazeko

Alright, stable at a set 1.575v (1.534 actual, in the bios) @ 400x9. Atleast semi-stable. I'd have to run it for a day or two to test it for real stability, which I won't do. Too concerned it might blow up. To be fair, it reached toward the top end of what I want to see the temps at, (Got up to 57 at some points, no core beyond 55C... I could use a better fan on this thing.)

The goofy thing is, CPU-Z won't report past 1.504v. I'm assuming it's the chip's limitation.

http://img225.imageshack.us/img225/5...creeniehx6.jpg

There's the screenie. I'm thinking this mainboard is still holding it back... and a better fan.


Edit: CPU-Z did report higher after a bit. Not sure what the hold up was, but it was stable through about 2 hours worth of prime95 on the 400x9 @ 1.538v overclock. I wish I could do more testing, but I really need a better fan on this TR120.

Kazeko

Another point of note, after playing around with this.

It could be attributed to an unstable overclock, but I get amazingly low memory latency per Everest's memory benchmark and the read/write is nearing 10GB/s with a lower overclock. (3.31ghz @ 1.30v)

Also, I'm seeing better 3dmark06 scores @ 3.31ghz than at 3.4ghz @ 1.40v (tested stable with P95 for about 8 hours.)

I'm assuming this is what panging said earlier about "What works best varies." It just strikes me as odd that it runs better at a lower over clock than a higher. 11:13 divider with the FSB @ 1457 and the ram @ 1067.