Intel Test System
CPU	Pentium 630 Retail (3.0GHz 2MB L2 1.31Vcore) Socket-775
Mainboard	1.) Asus Asus P5ND2-SLI Deluxe (BIOS 0605 also tested on 0802)
Memory	Corsairmicro 5400UL (2x512MB DC CL3-2-2-6)
Graphics	2x Gigabyte Gigabyte GV-NX68T256DH (SLI)
Power Supply	PCPower&Cooling TurboCool 850 SSI
Cooling	Alphacool 12V Cora passive H20-system, mated with Alphacool CAPE KC-42 X2-Pro
Operating System	Windows XP SP3

The photo above is somewhat misleading since all tests in this review were conducted with Asus stock passive NB cooling. I included this shot as a teaser to an upcoming DDR2 Overclocking Round-Up on the C19 I plan for this December (an Overclockers Stocking Stuffer). Under passive cooling I was able to boot into 838MHz memory speed raising the NB voltage to 1.5V, unfortunately this wasn't stable through all benchmarks. After completing the tests I discovered it wasn't the Crucial 5300 holding me back, but in fact is was the NB temps.


Speeds achieved on our test bed (stock NB cooling) are summarized below:

CPU = 4.0Gz, FSB = 1066QDR/267FSB, Memory speed = 533MHz, CL = 3-2-2-8, VDIMM = 1.95V
CPU = 4.0GHz, FSB = 1066QDR/267FSB, Memory speed = 600MHz, CL = 3-2-2-8, VDIM = 2.0V
CPU = 4.0GHz, FSB = 1066QDR/267FSB, Memory speed = 667MHz, CL = 4-4-4-10, VDIMM = 1.95
CPU = 4.0GHz, FSB = 1066QDR/267FSB, Memory speed = 800MHz, CL = 5-6-6-19 (SPD), VDIMM = 2.1V

For each memory speed and latency I've included CPU-Z, WCPUCLK thumbnails below correlating to the data above.




While the P5ND2-SLI Dlx allows VDIMM adjustable to 2.30V anything above 2.1V seemed counter productive. That is until the NB was cooled with H20, ironically this is one chipset which truly benefits from this; however; I haven't seen many articles on it. I've eschewed benchmarks obtained with NB H20 cooling as feel it's not representative of what most PC-Enthusiasts (even those cooling with H20) are using. Initially I had to modified an Alphacool GPU water block for the NB running it in-line with the CAPE KC-42 X2; however, H20 temps climbed to the point there were detrimental effects on CPU temps! now that's hot. I eventually ran a dedicated system.

Regardless of the anomalies mentioned in the introduction, the C19 chipset (at least on the Asus and Gigabyte motherboards) performs spectacularly in many respects. It out-performs the 925XE which was our last Intel test bed. With the Asus P5ND2-SLI Dlx I've bested every benchmark set on the Asus P5AD2-E Premium including single 16x PCIe 3D scores, memory speeds (as high as 850MHz) and overall tighter latencies at higher speeds. SLI performance is excellent and all tests in this review were run in SLI mode using the Gigabyte 6800GT's.

Our first series of benchmarks covers bandwidth and latencies beginning with SiSoftware Sandra 2005.SR2.



Further bandwidth tests included Lavalys bandwidth and latency benchmarks included in Everest Home Edition. I've eschewed the Latency benchmark placing that in another graph since it doesn't conform well within the Excel formula. Also of note is Everest Ultimate Edition which features a convenient taskbar GUI displaying just about every temp imaginable, you can download a 30-day free trial here.



Our last bandwidth benchmark from ScienceMark.org tests algorithms from one vector to another, ScienceMark 2.0 membench was written to be as objective as possible. ScienceMark 2.0 was designed to be the antithesis to programs optimized with assembly coding to favor one platform over another. The last (perhaps best IT) journalist whom exposed such Benchmarketing issues ended up in the middle Bapcogate, his home Vandalized twice. Membench results.