E8400: Everything stable, but my voltage is too low?

Al_aric

Distinguished
Mar 9, 2009
6
0
18,510
I took some of your guy's advice and it worked wonders. Thanks to the two guys who responded to my 1st post. This is now my 2nd post, evar.

My system is:

-EP45-DS3L
-E8400 Wolfdale E0 stepping
-4gb OZC PC8500 ram (max 533 fsb @ 5-5-5-15 @ 2.2v)
-Scythe Infinity cooler, awesome, silent, so much better than Zalman

And my OC settings are:

-9.0 x 445 Mhz = 4005.9 Ghz
-core voltage 1.25v set from 1.285 in the BIOS
-MCH=1.26v
-MCH ref=.74
-MCH/DRAM=Auto
-ICH I/O=Normal
-Strapping is 1:1 or "2.0B" in Gigabyte-speak

At 100% workload (Prime 95 Blend Operation) I'm seeing 47-48c on both cores. The CPU is drawing 35.57 amps at 43.28 watts.

So I've got a 4 ghz OC at only 1.25v, and here I was prepared to go to 1.365v !!!

Here is my question: It seems like too few volts are drawn by the CPU. Things are stable. Prime95 hasn't crashing in 2 hours and it used to crash in 10 minutes when I was using "auto" to set my vCore and getting 1.344v at lower speeds (i.e. 3.8 ghz).

I don't understand why I'm running more stable with higher speed and less voltage (4 ghz @ 1.25v) than I did with lower speed and higher voltage (3.8 ghz @ 1.34v). Oh, and I'd generate 60c heat at those settings, whereas now I'm at 4ghz and only making 48c in a pretty warm case. Can anyone explain?

Last question... if I'm only drawing 41 watts out of a potential 65w... AND I'm only making 48c temps, AND I'm only drawing 1.25v... how high should I try to go? As tickled as I am to be running 4ghz on air right now, I suppose that's only a 33% overclock from stock. Aren't these Wolfdales supposed to get 50% Is 4.5 ghz too crazy?
 
I hope your initial numbers are typos and later ones are correct - 1.750 and 1.48 are both too high for an Intel 45nm CPU. If the later 1.25-1.36V numbers are correct, that should be fine. You could definitely try for higher if you are currently at 1.25V and running cool.
 

B-Unit

Distinguished
Oct 13, 2006
1,837
1
19,810
Yea, the E0 steppings are in a word, amazing. Im running mine at its 1.25VID @4Ghz. Sadly, its a 650i board, so Im lucky to be getting 445 out of it. :(

Anyhow, higher voltage leads to higher temps more than higher clocks do. Thats why you were getting 60C temps @ 1.34v.
 

Fruity

Distinguished
Feb 20, 2009
200
0
18,680
Worth checking your voltage then lol.

40% to 50% OC for your CPU would be pretty respectable on air air cooling. I've had the Gigabyte EP45-DS3l running 2000MHz FSB as well, so if your voltages are at 1.26V that you can almost certainly push for 500MHz FSB with 9x to get something around the 4.5GHz mark. Just a quick caution though - watch your NB temp!! There's no temp readout from it so it's worth crow-barring a small fan on to it.

If you decide not to go that high, try to get a stable 500MHz FSB and drop your CPU multiplier to 8x. That will still give you 4GHz but better performance.
 

Al_aric

Distinguished
Mar 9, 2009
6
0
18,510
Thanks everyone who replied! Tom's Hardware draws the smartest crowd. To be fair, I haven't posted in Anandtech yet.

I fixed the typo. I get about a .03 vDroop, so my CPU sees 1.25 when I tell the mobo to feed it 1.285

Which is the NB setting, Fruity? It's my first Gigabyte board, so I'm not used to what they call MCH and whatnot. I think I'll strap the old 1" fan from my Slot-1 Pentium II to the NB.

Are there any benchmarks you can point me to where someone measured the performance of a Core 2 Duo system at 4 ghz with a high FSB/low multiplier vs higher multiplier/lower FSB (or something roughly simliar)? I mean, I understand that it will increase bandwidth, but is it really that much of an improvement?

People still debate the real/vs theoretical losses from reduced L2 cache in a E7400 vs E8400 and that's bandwidth inside the chip so it's affect on latency is much greater than the CPU to RAM bandwidth. A 55 mhz increase of the FSB can't yield much more than a few percentage points, can it?

And Fruity, going by your theory, seeing how I have sensors in my proc & mobo but NOT the North Bridge, wouldn't it be better to keep the higher multiplier and lower FSB to keep from straining the NB?
 

Fruity

Distinguished
Feb 20, 2009
200
0
18,680
MCH = Memory Controller Hub = Northbridge = NB

From memory, I'm pretty sure the EP45-DS3l has a Vdroop control. Can't remember what Gigabyte call it though. Incidentally, the difference between BIOS voltages and windows based monitoring progs is caused by the way that the voltages are calculated, not so much Vdroop.

3DMark06 is pretty good for a benchmark programme. The basic version is a free download / trial. Good though.
From past experience, I've had differences of 10% or so in CPU benchmark and about the same in graphics benchmark using lower multipliers and higher FSB. As to whether this would be noticeable in the real world is a whole different matter. I suspect that if you do alot of processor intensive work then it may be worth looking at.

The board (from limited past experience with it) handles 2000MHz FSB with a bit of tuning, but there is no way to tell what the temps are like. Saying that, if you keep it as far away from its max voltage setting as you can you shouldn't be in any danger. The additional fan is there as a precaution.

hth