Review: TwinMOS TwiSTER PC4000 512MB Dual Channel Memory Kit

by Tarinder Sandhu on 14 January 2004, 00:00

Tags: TwinMOS

Quick Link: HEXUS.net/qavi

Add to My Vault: x

Setup and notes

Here's a quick rundown of the test system should you wish to compare benchmark results with your own.
  • Intel Pentium 4 3.2GHz ES 800FSB CPU
  • TwinMOS TwiSTER PC4000 512MB Dual Channel Memory Kit (2 x 256MB)
  • Corsair XMS3500C2 (2 x 256MB)
  • ABIT IC7 i875P Canterwood (A18 BIOS)

Other hardware

  • ATI Radeon 9800 Pro (380/340)
  • Toshiba 8x DVD
  • Dell P991 19" monitor
  • Thermaltake AX478 cooler with 25CFM fan

Software

  • Windows XP Professional SP1
  • DirectX9.0b
  • Intel 5.02.1015 chipset drivers
  • Pifast v41 to 10m places
  • ScienceMark 2.0
  • 3DMark 2001SE v330
  • UT2003 Retail (Build 2225)
  • Quake 3 v1.30 HQ
Notes

The test is to see if what kind of benefit the combination of high bandwidth and high latencies (DDR500, 2.5-4-4-8 - TwinMOS) offers over lower bandwidth and extrmely tight latencies (DDR400, 2-6-2-2- Corsair XMS3500C2). With a semi-unlocked Pentium 4 3.2GHz CPU capable of operating with 12x - 16x multipliers, a compromise had to be found that kept overall CPU speed relatively constant but allowed for the TwinMOS' higher bandwidth and FSB setting.

With that in mind, both sets of RAM were run with the following combinations:

3204.2MHz - 246.5MHz FSB (13x multiplier) - DDR493 - 2.5-4-4-8 timings - TwinMOS TwiSTER PC4000

3207.7MHz - 200.5MHz FSB (16x multiplier) - DDR401 - 2-6-2-2 Enhanced timings - Corsair XMS3500C2

We want to evaluate which is better, an 'overclocked' pseudo 2.6GHz Pentium 4 'C' running with near-DDR500 FSB and memory speeds, or a standard 3.2GHz CPU paired up with low latency RAM.



That's how CPU-Z sees it.



And the memory speeds, correctly identified as being in dual-channel mode and with the TwinMOS' native timings. The modules, with the motherboard set to 250MHz FSB and DDR voltage to 2.8v, were subjected to a 20-minute Memtest86 test that showed no possible RAM deficiencies at DDR500 speeds. It repeatedly without highlighting an error. Oveclocking-wise, this is about as high as the modules would go in an OS environment. Basic stability consisted of a single run of 3DMark 2001SE b330.



Given their ability to run at relaxed timings, we tried to reduce the latencies and boost performance. It's unfortunate that high-speed modules often can't run with extremely low timings. 2-6-2-2, at DDR400, refused to even boot. Instead, we tried to see just how much tighter they would go at the tested 246.5MHz, which is still a short of spec. Stability with 2.8v DDR was excellent with 2.5-3-3-7 timings.

All this talk of timings, voltage, FSB and memory speeds brings us around to our test setup.

3204.2MHz - 246.5MHz FSB (13x multiplier) - DDR493 - 2.5-4-4-8 timings - TwinMOS TwiSTER PC4000 - Setup #1

3204.2MHz - 246.5MHz FSB (13x multiplier) - DDR493 - 2.5-3-3-7 timings - TwinMOS TwiSTER PC4000 - Setup #2

3207.7MHz - 200.5MHz FSB (16x multiplier) - DDR401 - 2-6-2-2 Enhanced timings - Corsair XMS3500C2 - Setup #3

The first is the TwinMOS modules are close to spec. The second is the tightest timings we could apply at that speed, and the third is some comparison Corsair run with tight, performance-enhancing timings and with Game Accelerator settings of Street Racer and Enhanced. They were left to auto for both TwinMOS setups.