W00t - burning-in 4tw - 3.1Ghz on Water (Opty 146)

Soldato
Joined
18 Oct 2002
Posts
6,669
After some encouragement from Bundles, I decided to push some more voltage through my Opty 146.

I managed to hit 3.1Ghz with 1.65v, 4M SuperPi stable. (I'm running it @ 1.705v now, just to be sure).

:D

I'd never been able to get the chip beyond 3.07Ghz before, and that was only just 3DMark stable, and I'm running 2GB RAM now too (sometimes the memory controller can crap out before the rest of the CPU).

Got a 27.5 second 1M SuperPi time, now testing for stability in games and BOINC. (It's only a gaming machine, so I don't need 24 hours Prime stable - my important stuff is on a stock-clocked shuttle).

I'd been running the chip @ 3.05Ghz, 1.65v fully loaded (BOINC) for months, so it looks like it's another case of slight improvements in clocks from burning-in.

Sorry, I know it's 'only' an Opty 146 and they are, like, SO last week, but I'm bored waiting for Conroe :)

Voltage Scaling with this chip (for those interested):

voltage%20scaling.jpg
 
Soldato
OP
Joined
18 Oct 2002
Posts
6,669
sr4470 said:
Any benchmark screenshots\pics of the rig?

Not done any benching for a while. Old scores are in sig. (3.07 Ghz), so are not gonna be too much different.

trojan698 said:
Nice. What method did you use for burning it in?

Explained in first post. I run a distributed computing task - BOINC. I personally run Einstein@Home, but take your pick from a bunch, including the infamous SETI

CS||nuTs said:
Meh waves Willy :D

http://valid.x86-secret.com/show_oc.php?id=68021

Nice one mate what are the steppings ?

CABYE 0540 FPMW on this 1.4v does 3ghz needed 1.425 for 3.1ghz but RAM fell over :(

Good work fella. Can't remember stepping. Will see if I can fnd it.

EDIT - Can't find stepping, and I'm not taking the WB off to find out ;)

PICT0001_small.jpg


Rig (posted a few times before, and I'm moved house, but it's essentially the same as that).
 
Last edited:
Soldato
OP
Joined
18 Oct 2002
Posts
6,669
CS||nuTs said:
Got a chance of some GSkill LA off a mate quite cheap Jimbo how far does it go before you need to start playing with dividers?

?

No idea - are you asking about the RAM or the CPU? I'm using G.Skill ZX and I always use dividers on AMD64, often quite 'nasty' ones like 133, 140 and 150. I don't care (too much) about memory. CPU speed > *
 
Soldato
OP
Joined
18 Oct 2002
Posts
6,669
CS||nuTs said:
Ahh only just noticed you have the ZX doh!.What divider have you set for the RAM?

I'm using a 150 divider and 2.5-3-3-8 1T timings.

I think I could get away with 2-2-2-6 1T and a 166 divider on the 1GB BH5 I've got sitting around, but the performance impact in SuperPi, FEAR and 3DMark'01 is practically zero (really - it's like 0.5%), and obviously the 2GB has other advantages.
 
Last edited:
Soldato
OP
Joined
18 Oct 2002
Posts
6,669
Mikey1280 said:
Jimbo what is that strange contraption beneath your rad?

It's an IKEA wine rack housing an aluminium box, inside which are two AQX 50Z pumps in series. They are fed by the large tube you can see to the right. It's a DIY bleed tube/res, feeding the first pump with a 3/4" tube (ID). Rest of the loop is 1/2" ID.
 
Soldato
OP
Joined
18 Oct 2002
Posts
6,669
tom_nieto said:
My RAM is going to let me down,

No it won't.

EDIT: CS||nuTs, my memory is PC3200, how much does system performance suffer with slower speeds? Do the dividers affect system performance adversely?

No. Use a divider. Just keep the memory at whatever it will do, but get the CPU as high as possible. If that means running the memory slower, even below 200 Mhz, do it. CPU speed is everything on the 939 platform.
 
Soldato
OP
Joined
18 Oct 2002
Posts
6,669
Bundles said:
On a side note, if it does 3.1 stable at 1.65v, but you're running 1.7v to be safe, what does it do stable at 1.7v? then when you give it a bit more to be safe, what does it do stable at that voltage?

It doesn't. Even with just a shade over 1.7v, it won't even do 3.13Ghz. You can see on the graph it's topped out.

The only reason I go (relatively) carefully with CPU voltage is that (as I said) I killed a really nice 3000 Venice and it took until this Opty 146 to get a decent clocking chip again. OK, so the death was most likely due to the bug in the CPU Special on the DFI Ultra D, but even so.

My OCZ is overvolted to 3.6v on the 3.3v line, and I've done the jumper trick to feed my RAM as much as I want (upto ~ 3.4v anyway) and my X1900XT is overvolted as far as it will go without a hard mod (which I'm reluctant to do at its current price!). I have hardmodded at least 3 other gfx cards though, so I'm not averse to the idea.

Don't tempt me to the spark side! I'm already there!

:)
 
Soldato
OP
Joined
18 Oct 2002
Posts
6,669
Mikey1280 said:
Ahhh right, when you put pumps in series should they be right next to each other or spread out throughout the loop? Does it make a difference?

Don't know. I have then right next to one another.

On, and in reply to your question in another thread, that's not a 120.3 ;)

That's a car radiator with 4 x 120mm fans on it. It has something like 3 times the surface area of a 120.3 (I forget exactly how big the core is), plus it cost ~ £15 from a local scrappy, although it was brand new. :D Granted, I needed to bodge some 1/2" barbs into it and fashion my own shroud, but you can't come close to the heat dissipation, hence water temps, with a PC rad or heatercore.
 
Back
Top Bottom