Tweak3D - Your Freakin' Tweakin' Source!

In operation

Our sample motherboard (Asus P5WD2 Premium, BIOS 0709) detected the memory without an issue and so we proceeded to perform the benchmarks and stability testing at various combinations of frequency, timings and voltages - it was a task that eventually required several cups of coffee and oh so many hours it isn't even funny. During tests involving a high DDR2 voltage of 2.3V and including dizzy operational frequencies (1000MHz effective or higher), the memory was never really overly warm hence there is no need for further cooling through the use of extra fans. The only instance where heat was the most likely the culprit was when testing speeds beyond 1066MHz (around 1080MHz and beyond). Here regardless what valid voltage or timings were used, the memory would always lock up after a few random minutes of running MemTest86 V1.65. Given that Corsair recently showed off a memory cooling solution at Computex, it's needless to say that pushing DDR2 speeds much further is going to require a little more than just the standard heatspreader. Back to life beyond Computex, this memory proved to offer considerable flexibility in the frequencies and timings it can operate. For example, at 667MHz (the current official Intel standard), it will quite happily do 3-2-2-8 at 2.2V and what's more the fun doesn't stop there. How about 5-5-5-15 at 1066MHz using 2.3V or 4-4-5-15 at 1000MHz equally using 2.3V. During testing it became obvious the memory preferred a higher RAS activate precharge timing when used at very high frequencies. Luckily for Crucial, this value is the least influential in final performance with CAS latency the most important, followed by a close second and third place held by RAS to CAS latency and RAS precharge respectively. Results like these make one wonder why Crucial only sells this memory at 5-5-5-15 or from a slightly different angle, why is it not stated to run at PC2-8500? The answer seems to be quite simple, to be on the safe side due to a) the various motherboards out there and their varying specifications and b) users who donít quite yet know how to fine tune memory to best performance and are perhaps a little hesitant to go changing voltage levels etc. The results obtained are all worthy in their own way and prove the flexibility and versatility of this offering. One may either select the tightest timings at a lower frequency or go for gold at the heights of the Himalayas receiving in reward a hefty memory bandwidth and fine latency. Our test results later on in this review showcase the kind of bandwidth and latency we were able to attain.

More than you paid for? Although Crucial sells the memory as PC2-8000, it happily runs at PC2-8500 speeds.

Test setup

Processor: Intel Pentium Extreme Edition 965 3.73GHz ES (1066MHz FSB, using 14x & 16x multiplier)
Operating System: Windows XP Professional SP2 (32 bit)
Memory: 2GB (2 x 1024MB) Crucial Ballistix PC2-8000 (default 5-5-5-15 at 2.2V)
Graphics card: Gigabyte 7900 GTX 512MB, 84.21 version driver
Motherboard: Asus P5WD2 Premium (i955X chipset), 709 BIOS. PCB revision 1.02
Power Supply: Akasa PowerPlus 650W ATX 2.01

Software #1: MemTest86 V1.65 (to test memory stability)
Software #2: Sisoft Sandra 2007 (to measure memory bandwidth)
Software #3: Lavalsys Everest Ultimate Edition 2006 version 2.80.534 (to measure memory latency)

Test methodology

Test 1: (stability):
We tested each of the memory configurations up to the highest stable speed and timings using MemTest86 v1.65. In order for any given memory configuration to classify as stable it had to complete 1 hour of running this software. Unless a given configuration was proved to be 100% stable, tests 1 & 2 below were not executed.

Test 2: (memory bandwidth)
We tested the highest stable speed and timings obtained within test #1 using Sisoftware Sandra 2007.

Test 3: (memory latency)
We tested the highest stable speed and timings obtained within test #1 using Lavalsys Everest Ultimate Edition 2006 version 2.80.534.

Definition of memory timings
All references to memory timings made within this article correspond to the following syntax: CAS latency - RAS precharge - RAS to CAS latency, RAS activate precharge.

Other information relating to testing:
All tests at a memory frequency of 1000MHz effective or over (2x500MHz DDR) were performed using a Northbridge voltage of 1.65V to minimize any negative influence the test motherboard may have on stability. All tests with a memory frequency below 1000MHz were performed using a default Northbridge voltage 1.50V. In either case, all tests featured HyperPath3 as enabled for best performance. The CPU speed was kept as close to consistent as possible to reduce impact within the memory & cache benchmark. Yields of RAM will vary from sample to sample as does motherboard stability at high frequencies. For this reason the results verified and subsequently displayed within this article should be used as an indication but not guarantee of the capability of this memory.

Next Page: Test Results

  • News
  • Forums
  • Tweaks
  • Articles
  • Reviews