EggXpert

Setup:

AMD Athlon64 X2 5000+ Black Edition

Cooled by AMD/Dynatron Stock Opteron 148 (Socket 939, kitbashed) CPU Cooler. (cpu clamp mod) with addition of a little Artic-5 paste. 

Asus M2N32-SLI Deluxe Wireless Edition Mobo

2x EVGA GeForce 7900GS 256

80GB WD SATA

GeForce 163.75 Drivers

AMD Cool and Quiet installed.

CAGG's Methodology:

BIOS to Extreme Tweak/Overclock section, Set to AI Overclock 10% (max out).

Final Results: 3.08GHz Stable. 3DMark06 @ 7855

Swaydude's Methdology:

Pull up Tom's Hardware page. BIOS to manual overclock. Start playing with Multipliers and Voltage, get frustrated and mess with FSB's.

Final Results:

3.08 GHz: Stable. 3DMark06 @ 7885 (Manual tunes, not my AI tune above)

3.11 GHz: Stable. 3DMark06 @ 7953

3.15 GHz: Unstable. 3DMark06 Incomplete. 

3.21 GHz: Unstable. 3DMark06 Incomplete.

3.25 GHz: Unstable. 3DMark06 Incomplete.

3.31 GHz: Unstable, 3DMark06 Would not start.

3.35 GHz: Unstable, BSOD.

3.41 GHz: Unstable, BSOD.

3.45 GHz: Unstable, did not boot. (Hard Freeze)

Conclusions: After much questioning of Swaydude's sk1llz, he threw a bunch of stuff at me while scowling in derision during his OC phase.  I was busy with my Intel Build in the corner, and left him to it. PC is currently manually tunes to 3.11GHz (stable). Suggested getting an actual AM2 socket Zalman cooler to achieve the performance points on Tom's.

EDIT: To clarify for Wannapiece's satisfaction, NO NTune WAS USED HERE, all overclocks were performed with the Asus BIOS that shipped with the M2N32-SLI Deluxe Wireless Edition Mobo. 

And the point of this is?

Oh, I get it.. your trying to find the most effective way of destroying a chip using a windows based overclocking utility? 

Uhm.. I know you dislike me on a personal level and all, and I admit previous prickly posts and sarcasm directed at anyone and everyone... but... uhm... there was no "Windows Based Overclocking" there.

That was all done in BIOS. It's explicitly stated.

My methodology was to use "AI Overclock" in the Asus BIOS. Not windows based there.

Swaydude spent half an hour slaving over manual Voltage, FSB, and Multipliers according to the charts on Tom's.

The point of that was to post some real-world non ideal metrics on a setup lots of people have/will buy, and to see how it will act in a non-lab real-user environment (No Zalman here, plain jane ram, just a Mobo and SLI).

The differences were surprising. My BIOS-AI overclock took about 30 secs and got decent performance, but I was busy w/ other stuff. Sway's into grinding every. last. bit. of performance he can out of it, and so he spent the time to "do it right". End result with plain jane gear: 0.03 GHz (30 MHz people, that's a whole 486SX) and 100 3DMarks difference, and a good idea of what the gear will do in "real life", "non ideal" situations and rigs.

To conclude: I know you don't like me, and I have no problem with personal attacks. But If You Want To Question My Competency And Methodologies, "bring more ammo". this wasn't the way to do it. 

"New School" OC's were @ 42/43C idle. 51/52C load.

"Old School" OC's were @ 43-45C idle, 53C load. 

"Load" was simulated by playing Company of Heroes Opposing Fronts for about 15-20 mins. 

Old School = going in to the bios and playing around with fsb, multipler, and cpu voltage.

New School = software to overclock.

You can tweak a little more old school way. 

 
I just know that CAGG is lazy
 

Lazy enough to do the following:

Re-install WinXP x86.

Re-apply all drivers.

With AI Overclocks at 10%, NTune memory to 4-4-4-15 (860MHz)

3DMark06 = 8203.

That broke 8K, dude. With the CPU @ 3.08GHz. 

Hey Cagg,

I was just wondering, were the CPU, MoBo and 7900GS-2 flying around when you thought of making a system with Sway? XD

i could say that twin 7900GS is pretty impressive breaking the 8000point mark, Congratulations!

testify!!

That machine had too much c**p on it anyways.  Go back in to the bios.

FSB 203 mhz

X 15.5 multiper

Voltage = auto

Then we'll talk about it. 

Nah Swaydude is just a CAGG Labs neighbor, and knows when to drop by keep his sk1llz fresh.

Me, I just care about having things work. And I'm lazier. 

That was the point of the exercise. With non Zalman/Premium - class cooling solutions, you can't just "put the multiplier at 16 and call it a day". Thermal and voltage limitations exist, and Your Results May Vary. (AMD and Intel have been known to send "better" than retail CPU's to reviewers.)

Hence this Actual Real-World Review  

I would, but if I take my 9700 off my C2D rig, it might cry >.<

:p~  the idea to get the BlackSLI rig its' own Zalman has been tossed around, but it's no priority yet. 

I think the Swayster tried 3.11GHz first, then got all overconfident and cranked it like he was the Asian, CPU version of Mario Andretti and cranked it to the stops  at 3.45 GHz.

Then in the midst of my ridicule and derision (I was hurling insults and foam packing material his way while elbow-deep in my C2D Power Rig) he finally Got Analytical and cranked it down step by step. My final analysis was posted after all was said and done, but while going through it the settings tried were wildly divergent.

The only things that really flew were packing materials and insults. Compared to me E6850 rig, the AMD rig is 1/3 as expensive, but 1/2 the performance.

A nice power vs performance ratio, but not really a power rig.

I highly doubt it. The Proof Is Not In The Clock Speed.

It's all in the video cards. In Order to attain performance parity, I'd probably go w/ some decent 8800GT video cards instead of my "beater" 7900 GS's. Also some performance RAM might be nice for OC'ing. Just throwing a Zalman on that bizzle and getting the core speed to 3.45GHz stable isn't going to raise my 3Dmark score much.

Doing the parts as above will result in a system that's about 1/2 as expensive and 3/4 - 4/5 as powerful, however. A still excellent power/performance ratio.

But the $1100 and $1600 budget points *are* different for a reason. There are many moer $1100 systems out there than $1600, and very few $3000+ systems. (By comparison)

Nah, data was gather during the process and only compiled at the end, which sorta "made sense".

Hindsight however, is usually 20/20. 

I still think the old school way will get better performance.  It's all about the tweaking.

Just wait until we put a decent cooler on there.