- Back to homepage
- Related articles
  Related articles

Corsair TwinX-1024 3200XL PRO: nForce3-250 v. Intel-875

Memory by KeithSuppe @ 2004-07-15

Liquid3D takes Corsair latest memory for a spin on an Athlon 64 and Intel Prescott system. Do tight timings still provide the best performance? Read on to find out

Intel Testing

Now onto our Intel system.

CPU -

TIM - Arctic Silver AS5

Cooling - Danger Den RBX, Z-chipset, Danger Den double heater core, 2x120mm/90CFM Sunon's (pull), Hydor L-40 , 1/2" tubing.

Motherboard - Asus P4C800E-Deluxe (Ver.1.2 BIOS 16)

RAM - Corsair Twin-X1024 3200XLPRO

VGA - Sapphire Radeon X800Pro (Catalyst 4.6)

HDD - Maxtor Diamond Max Plus 9 (SATA150 120GB)

PSU - OCZ Technology Power Stream OCZ420ADJ 420W

Software - OS Windows XP SP1, AsusProbe v2.20.07, Aida32, Podien CPUCool, CPU-Z, WCPUID, SiSoft Sandra, Aida32, Sciencemark 2.0, Futuremark 3DMark2001SE 330, 3DMark03, Aquamark 3, Gun Metal, [H]ardOCP UTK3 Benchmark. Note: I had some problems with CPU-Cool voltages. Although OCZ Technology Power Stream 420 has a 12V rail adjustable to 13.+V I can assure you it was not running at 13.19V as indicated. The temps, VCore, and frequency measurements are accurate, as I verified these in the in the BIOS, and through Asus Probe.

Testing 3200XLPRO on the P4C800E-dlx with its Dual Channel MCH (Memory Controller Hub) should (in theory) effectively double the memory bandwidth over that of an A64/nForce3-250 system. This is where we shall determine the flexibility of Corsair's 3200XLPRO, and where Corsair's Plug and Frag technology can be fully tested. Capable of 500MHz performance at 1:1 aspect ratio, the 250FSB setting necessary for such speeds will place a heavy demand on our Prescott 3.0E (15x 200FSB =3750MHz). While I could have simply installed a 2.4C for this part of the review, I wanted to dispel some false impressions concerning Socket-478 Prescott's purported "thermal issues". To date, there have been some irresponsible figures tossed about, misdiagnosing the source of "excessive heat" associated with Prescott.

It's been claimed Prescott's on-die temps are exceedingly high, when in fact on-die temps and wattages between Athlon64 and Prescott, are very close. In fact I've found Athlon64 to run a few degrees warmer then Prescott. Where problems do arise, and misconceptions originate, are the excessive operating temperatures found on Socket-478 motherboard power circuitry, while running Prescott.

Six months ago Abit released its AI7 motherboard. I purchased the board, in part due to its µGuru technology. µGuru offered an extensive array of temp, voltage, and overclocking options, all based on a Winbond monitor/controller chip made specifically for Abit. The AI7 was one of the first motherboards equipped with a device which specifically measured power circuitry supplying the CPU. Abit labled this value, "PWM".

In my article explaining Prescott's thermal and power management features such as VRD, and On Demand mode, were conflicting with mobo power circuitry. During my initial testing of 478-Prescott, AI7's µGuru utility indicated PWM temps climbing as high as 97C! Prior to my purchasing my Abit AI7 and 3.0E, [H]ardOCP indicated in their article Prescott Tweaks Dangerous they had literally experienced motherboard melt-downs. The problems arose from the failure on behalf of motherboard maker's to implement power circuit modifications to "Prescott ready" Socket-478 parts.

Manufacturers announcing Socket-478 mobos to be; Prescott ready had merely made ad hoc changes via a BIOS update. And it wasn't as simple as the difference in current requirements between Northwood, and the 90nm Prescott. Prescott's D-VID feature was another contributing factor to these excessive temps on power circuitry. The following quote from Intel's (sec. 2.4) VRD 10.0 (Voltage Regulator Down) white paper, should clarify the matter.

Madshrimps (c)

Without delving too deeply into the subject, those of us whom purchased Socket-478 Prescott had the rug pulled from beneath us the moment we dropped our new Prescott's into our Socket-478 motherboards. Yet, even with the aforementioned handicaps I've had some excellent overclocking experiences with my 3.0E. This was the first time I reached 4.0GHz. Running Corsair Twin-X1024 3200XLPRO in concert with Prescott, has resulted in excellent performance. As usual, I digress.

Our first set of benchmarks begin again with the synthetic, SiSoftware Sandra memory benchmark. For the remainder of tests to follow, the system was run at default speeds 15x 200FSB = 3000MHz, at default Vcore, and memory was placed in DIMMs 1,3 at 2.75V on SPD (Serial Presence Detect) at CL2-2-2-5 1CMD. PAT (Performance Acceleration Technology) set to Standard

Madshrimps (c)

SiSoftware Sandra Memory benchmark UnBuffered 200FSB (3010MHz) SPD CL2-2-2-5 1CMD.

Madshrimps (c)

Aida32 READ benchmark 200FSB (3010MHz) SPD CL2-2-2-5 1CMD.

Madshrimps (c)

Aida32 WRITE benchmark. 200FSB (3010MHz) SPD CL2-2-2-5 1CMD.

Madshrimps (c)

ScienceMark 2.0 Memory Benchmark. 200FSB (2996MHz) SPD CL2-2-2-5 1CMD.

Madshrimps (c)

ScienceMark 2.0 STREAM Benchmark. 200FSB (2996MHz) PAT/Standard, SPD @ CL2-2-2-5 1CMD.

Madshrimps (c)

Hexus PiFast Challange 200FSB (3010MHz) PAT/Standard, APD CL2-2-2-5 1CMD.

Madshrimps (c)

3DMark2001SE 200FSB (3010MHz) Pat/Standard, APD - CL2-2-2-5 1CMD.

Madshrimps (c)

3DMark03 200FSB (3010MHz) Pat/Standard, SPD - CL2-2-2-5 1CMD.

Madshrimps (c)

Aquamark 3, 200FSB (3010MHz) PAT/Standard, SPD CL2-2-2-5 1CMD.

Madshrimps (c)

[H]ardOCP UTK3 Low. 200FSB (3010MHz) SPD CL2-2-2-5 1CMD.

Madshrimps (c)

[H]ardOCP UTK3 CPU. 200FSB (3010MHz) SPD CL2-2-2-5 1CMD.

Madshrimps (c)

Onto the Intel Overclocking ->

All information and graphics contained in Madshrimps are sole property of the Madshrimps crew and may not be reproduced or copied in any manner without written permission from us.