Gigabyte GA-EP45-Extreme Appreciation Society

Associate
Joined
8 Apr 2008
Posts
1,490
Location
Berks+Powys
MB-132-GI_400.jpg


OcUK is selling this really over the top board for a very competitive price. So join the club here !

I wanted to replace a P35 board, wanted 2 Gigabits, firewire, and good cooling support to run an operating system that is neither Windows nor Linux (:D). This one was out of the equation until I realised it was cheaper than the DS5 !

Please post any tip & tricks here !
 
Yes I think the OcUK picture I link to is some sort of prototype, my production board is the same as DJ's
I'm not sure myself I'll ever use the watercooling at all, but the heatpipe seems to be doing quite a good job as regulating the temperature. Also it prevents "hot spots" on the board, it must be significant for longevity as there is a lot less thermal stress.

On a side note, I also installed one of these :
BB-003-SY_400.jpg

I wanted mostly to use the temp sensors to monitor places where there isn't any. So far the memory runs at 36C, the PSU at 30, the case at 25C.

Edit: found a lame bug in this controller. I'm returning it.

It's a reasonably good buy, the quality externally is fine and I like VCDs. The soldering at the back is a bit less polished, but I can't complain. Plenty of cables (including a pair of spare sensors!)

Someone also lent me one of these:
MY-123-CS_400.jpg


This one is more doubtful. it's quite bulky and sits in front of the fan of the CPU cooler. Also, strangely enough, it does /not/ nake the memory any cooler. With the fan at 3/4 speed using the fan controller, the memory temo sensor still reports.... 36C !
I think that what happends is that it's it's bulk cancels the normal case airflow, and the fans just compensates for /that/. I don't think I'll keep it or buy one for myself...
 
Last edited:
Any of you guys can shed some light on the voltage "drop" that is measured, compared to the voltage that is set in the bios ? I get .02-.04 or so difference, it looks significant..

Could it be my PSU that is not beefy enough ? I got the PSU because it has a nice 120mm on it and is truely silent, however it's nowhere near the most powerful on the shelves.. I'm also wondering if that 500W is now under what I should use...
 
Did you have any problems updating the bios from a USB, BusError?

I unpacked the bios (correct one) onto the USB stick although QFlash just can't seem to detect it or the USB for some reason. The stick was FAT32 formatted and the file name is correct. Any help would be greatly appreciated.

I think you need the "USB Legacy" option turned on or something similar. Also in QFlash, you select "HDD" and not floppy. It's not very user friendly :D
 
hmm you copied it to the root of the drive, and kept the name exactly the same ? You checked the size too ?
I used a crapo usb stick myself, with just that file on:
Code:
[michel@yelp ~]% ll /Volumes/USB\ DISK            
total 2048
-rwxrwxrwx  1 michel  staff  1048576  4 Nov 21:15 ep45ex.f8
[michel@yelp ~]% md5 /Volumes/USB\ DISK/ep45ex.f8  
MD5 (/Volumes/USB DISK/ep45ex.f8) = e472e9fbc32b58590e75da6dfc2017f4
 
Welcome! I must have been lucky using that old, crummy 128MB stick that I had lying around :D

Do you notice anything different on the board behaviour ? Since I updated mine right away, I don't know how significantly different it is...
 
Wierd that your vcore would jump; mine is steady (save for the rather constant "drop" I mentioned before). Do you have a modern, trusted PSU ? 1.8 or 1.6V are /bad/ too (way too high), if the reading is true...(!)
You sure it's not 1.164 ? If so it could be explained by the motherboard using speedstep to underclock/undervolt the CPU when not in use.

I tried a bit more play yesterday; I tried to see how much I could push on stock voltage (1.22 here) and it goes very nicely a 3.43Ghz ! I'm quite impressed with that...
 
For that difference in voltage, I now know it's "normal" and that most mobo do it, so I won't change PSU just yet :D

Anyway, I've been trying downclocking configurations, and I'm rather unnimpressed so far. I downclocked the Q9650 to 2.2HGhz (that means speedstep runs it at 1.5 at idle) and reduced the voltage a notch, on the ram too...
But the temps stays the same (24C at idle, just a couple C over ambiant) and it's not considerably more silent (it's not loud to start with)

So I guess I stay at >3.0Ghz <4.0Ghz -- it works very, very well like that :D
 
pretty good results indeed ! Excellent even. How far do you push your memory ? I know mine runs fine at 960, but higher with just V1.9 becomes a problem.

In any case it's a cracker of a setup we've got :D ideal combo !
 
BTW I'd be curious i you (or anyone else) could have a look at their temperatures in the bios... Theres something a bit strange going on here:
+ The bios read a "cpu" temp of about ~50C
+ The board is bout 40C

However, coretemp reports cores (at idle) of <25C, and my various temp sensors report temperatures like:
+ Northbridge : 31
+ PSU : 28
+ Memory : 38
+ "case" : 22 (in the airflow)
Also, the air that comes out is cool...

So I wonder what these temperatures are coming from, I'm surprised to see such a difference...
 
Thats interesting and seems to match what I am seeing. In any case seems our result seems to match won't worry and will trust all the other sensors :D

The only 'nitpick' here is that since it seems to auto-regulate the fans from that sensor, maybe it's just overdoing it. Having core temperatures in the 25 is a bit 'extreme', I wouldn't mind trading 5C for 1/3 of the RPMS :D
 
darn, guess the one I wanted to buy but was out of stock ? :D

I think I'll wait until I reach the proverbial 250 posts and place a celebratory order :D

I only use the coax SPDIF myself, I got a 48 bits digital mixer with all my office audio; so I just need straight 48khz digital out really. I wish there was a coax /in/ but well, it's not stricly mandatory either.
 
On a side note I realized when playing in the bios that the Q9650 allows use of the C2E and C4E in the Extreme BIOS, and activating that seems to gained me a couple of degree C at idle (in Leopard)

Linux has a problem reading the ACPI properly and fails to use speedsteps 10 steps; it uses on 2 of them and beleive they are 2 and 3Ghz, failing to see the overclock completely...
 
BTW I solved my ACPI/cpufreq linux recognition issue by compiling my own kernel and activating the 'tolerant' acpi parse kernel option. Now the kernel recognizes the overclock properly.

You also need a new kernel to backport the support for the ITE IT8720F chipset to allow lm-sensor to work and recognize all the voltages, fan speeds and non-cpu temperatures.
 
I finally spent a bit of time pushing the q9650 on the -Extreme. It's now stable at 4.0Ghz (FSB 445), 1.30V (idle 1.26, full load 1.22) temps are 54-48 at full load. It's surprisingly easy to push that thing...
 
Back
Top Bottom