My New Build

[TQ]

I found that I had one convenient bay left and was able to move the drive down.
Let's go look at that 8800GS again, now that I have room to spare.

[ctal]

Here's an update:

I noticed today that my cpu was running at 3.0 gHz, instead of the 3.8 I had set it at, and the memory timings were at defaults. That means that at some point the system had trouble starting, and the bios settings for cpu speed and memory reverted to defaults (a nice feature of this board, which avoids the need for clearing cmos, or retrying multiple times, to get into the bios to change overly aggressive settings).

So I decided to go with 3.6 GHz instead of 3.8, corresponding to a 400MHz fsb (1600 MHz quad-pumped). I'm sure I could go considerably faster if I increased vcore from the 1.34 I have it at (and possibly the northbridge voltage also), but I'd rather not do that.

Core temperatures are now 37 degC idle, and only 47 degC running one instance of Prime95. Really cool (pun intended)!

This is based on the core temperature monitoring program Realtemp. SpeedFan reads 5 degC higher, which if you read the following writeup (and anyone interested in E8xxx temperatures should read it) reflects SpeedFan assuming a 100 degC Tjmax, while Realtemp assumes 95 degC for Tjmax. Some other programs assume 105 degC (based on Intel numbers for laptop processors, which are probably different than for desktops). As you'll read, Intel has not documented Tjmax for the E8xxx desktop cpu's, leading to much confusion. But since Realtemp's assumption is based on measurement of an E8400 with an infrared thermometer, it has the greatest credibility as far as I am concerned. I then set a -5 degC offset in SpeedFan, so the two programs give me the same results.

The idle temps may be artificially high, because as explained in the writeup a calibration procedure has to be done to make them accurate, which I have not done because I'm mainly interested in the load temps.

Real Temp - Documentation

Regards,
-- Al

[TQ]

Thanks for the update, Al.

[ctal]

Another update:

I decided to do the calibration described in the RealTemp writeup that I linked to in my previous post. The upshot is that in my 63 degF/17 degC basement (checked against 3 different household thermometers), my core temperatures were so low, and the required calibration correction so large, that the correction was outside of the range that could be accommodated by the RealTemp program. Therefore I've included calculated temperatures, as well as the temperatures indicated by RealTemp and SpeedFan.

Ambient = 63 degF/17 degC

Idle (at 3.6GHz [actually 2.4GHz at idle due to Speed-Step]; vcore = 1.35V):

Tcore (calculated): 26 degC
Tcore (per RealTemp with idle calibration correction of "-3"): 30 degC
Tcore (per SpeedFan with Tjmax offset of -8 degC): 34 degC

Load (at 3.6GHz; vcore = 1.35V):

Tcore (calculated): 41 degC
Tcore (per RealTemp with idle calibration correction of "-3"): 43 degC
Tcore (per SpeedFan with Tjmax offset of -8 degC): 43 degC

Regards,

-- Al

[TQ]

No further problems with the board resetting itself to defaults?

[ctal]

Originally Posted by TQ

No further problems with the board resetting itself to defaults?

Fine so far. I would think it should be ok now, because I'm running 0.2 gHz slower than the 3.8 gHz at which I was able to pass Prime95 and Memtest86+ for several hours each.

Regards,
-- Al

[TQ]

I might have missed it, but what timings are you using for your memory?

[ctal]

Originally Posted by TQ

I might have missed it, but what timings are you using for your memory?

It's in my profile: 4-4-4-12-2T, with some voltage increase (2.27V vs. the default of 1.9 or so -- the memory is spec'd for up to 2.4V, which is higher than the rating on the one you went with). It's DDR2-800 memory, running at DDR2-800 now that I've lowered the fsb to 400MHz (1600 MHz quad-pumped) to put the cpu at 3.6gHz (with the E8400's multiplier of 9).

When I was at 3.8gHz, the memory was running at DDR2-846, which necessitated the voltage increase. I could probably reduce the voltage somewhat now at the slower speed, but there doesn't seem to be much point in taking the time to do that (which would involve re-doing the Memtest test, etc.)

But none of this is particularly relevant to the memory you are getting, which is spec'd for a faster clock rate at slower timings. Overall I would expect the performances of the two memories to be similar.

Regards,
-- Al

[TQ]

The X48-DQ6 sits next to me in it's shiny, yellow-gold box awaiting the memory, CPU and the OS.
Unfortunately, I had assumed the 64-bit disk of Vista Business would be included in the retail package. It was only after we ordered it that I discovered I'd have to special order the disk from Microsoft and that was only after I had my serial number in hand.
Seeing as how my OS order is now on backorder it may be some weeks before I can build my new machine.

[ctal]

Originally Posted by TQ

No further problems with the board resetting itself to defaults?

I think I've now discovered why that happened, and it was not because the mobo, cpu, or memory couldn't handle the overclock. As I mentioned in PolonyC2's "New System Build" thread, the one slight negative about the case is that the power switch seems a little flimsy. When I pressed it today to turn the system on, it stuck in the pressed in position, which caused the system to be unable to start. When it automatically tried to restart, after I free'd the switch by pressing it again, it had changed the overclock settings to defaults. I recall that happening with the switch once before. I didn't notice what happened to the settings then, but that figures to have been when the previous reversion to defaults occurred.

So I've gone back up to 3.8GHz, and we'll see how it goes.

Regards,

-- Al

[TQ]

Thanks, Al. That's good to know.

[ctal]

Wow! I've now done several video editing renders ("builds") in various formats on the new system. Compared to my P4 2.8C/Asus P4C800 Deluxe system (which was overclocked to 3.25 gHz), the new system (E8400 overclocked to 3.8 gHz/Gigabyte GA-EX38DQ6) reduces the build times by a factor of about 3.5!

Following are the specific results, for a 30 minute program with a moderately complex timeline, using Sony Vegas 8 Pro, together with DebugMode Frameserver + CCE Basic for MPEG2 encoding:

DV Build: about 55 minutes (old system); 16 minutes (new system)

MPEG2 Build (two-pass): about 170 minutes (old system); 46 minutes (new system)

WMV Build (350 kbps): about 100 minutes (old system); 29 minutes (new system)

WMV Build (56 kbps): about 60 minutes (old system); 16 minutes (new system)

Pretty impressive improvement!

-- Al

[TQ]

Very nice delta, there...

[ctal]

I just finished another new and very similar build, this time for my wife. It replaces her 7 year old Dell, which is still going strong, but I've been so pleased with the one I built for myself that I couldn't resist providing her with something similar. Some of the components are a bit better than in the one that I built for myself, due to technology marching on -- it's now 3 months later!

Case: Again a Lian Li PC-A70A with the optional dual-blowhole top, on which I again mounted two Scythe 1200 rpm 120mm fans. There are 5 other stock Lian Li 120mm fans in the case, as well. This time I chose the black case, which matches its surroundings better than silver would have.

Motherboard: Gigabyte GA-X48-DQ6. The GA-EX38-DQ6 I used previously seems to be no longer available.

CPU: E8500. I set it to a conservative overclock of 3.8gHz, with a very slight increase in voltage over stock, tested the rig with Prime95 and Memtest86+, and left it at that. No reason to be more aggressive considering the non-challenging uses it will be put to (no games, and probably no video editing).

CPU HSF: OCZ Vendetta 2.

Memory: OCZ 2 x 2gB DDR2-1066 (OCZ2P10664GK).

Video Card: EVGA 8800GS once again.

Optical Drives: 2 x Asus DRW2014BLT (SATA), once again.

HDD (system): WD 300gB VelociRaptor (SATA).
HDD (backups, images, misc.): 1 x WD 640gB SATA.

Floppy Drive: Generic.

Power Supply: Corsair CMPSU-520HX.

OS: XP Home.

The rig has been rock solid through several days of software installation and intensive use, as well with lengthy runs of Prime95 and Memtest86+.

Regards,
-- Al