Crucial Ballistix PC4000 (DDR500) Review - Test Setup and Overclocking

Article Index
Crucial Ballistix PC4000 (DDR500) Review
Test Setup and Overclocking
Game Testing with the PC4000
Conclusion and Final Thoughts

Test Setup & Info:

Below is the specs for our standard test machine.  For each review we use a fresh install of Windows with all of the latest critical updates applied.  We also use the latest WHQL certified Forceware drivers for the nForce 4 and Graphics card.

Installed & Running On DFI LP UT NF4 Ultra-D
Installed & Running On DFI LP UT NF4 Ultra-D

For this review we tried to benchmark the demonstrate the ability of this RAM when running at stock speeds and also when overclocking.  We used a couple of synthetic benchmarks, Everest Home and SiSoft Sandra 2005 and verified the timings and speed of the memory using CPUz.  I'd like to thank the good people over at DFI-Street for some info and help setting up some good overclocking as well.

Overclocking Results:

I discovered some interesting things regarding overclocking and memory bandwidth on the nForce 4 platform when using an Athlon 64 processor.  In times past, memory bandwidth was totally dependant on Front Side Bus (FSB) speeds, and was irrelevant to actual processor speed.  With the integration of the memory controller right onto the Athlon 64 processor itself, there is not actual FSB on an Athlon 64 system at all.  Instead you have your Hyper Transport Bus and total memory bandwidth is quite dependant on the HTT Bus as well as the total speed of the memory controller, built into the processor itself.

Essentially there is still a "FSB" when it comes to clocking your memory, but it doesn't matter if it is synchronous with your CPU or not.  To find the maximum speed of the memory, I lowered the CPU multiplier and raised the HTT/FSB until things got shaky.  For final testing I used Memtest x86+ to determine stability.  If the memory yielded no errors after 2 passes of Memtest it was determined to be stable.

Once we determined maximum speed of the memory, we started raising the CPU multiplier to an acceptable level to get the most performance out of our memory controller.  For example, if you run DDR at 200HTT/FSB and the CPU multiplier at 9x your CPU speed is 1800MHz, and your memory bandwidth results in  X MB/sec.  If you change the HTT/FSB to 300 and drop the CPU multiplier to 6x, you CPU speed is 1800MHz, and your memory bandwidth is still the same X MB/sec.  Even though the memory is running much faster, the memory controller is limiting the bandwidth.  To gain bandwidth, both memory and CPU (thus memory controller) must be overclocked for maximum memory performance.  For this reason it is hard to get accurate "real world" memory bandwidth results, but we did include some Half-Life 2 and Doom 3 benchmarks regardless.  These will have to be taken with a grain of salt as CPU speed is increased as well as memory bandwidth.  However, this will give you an idea of what is available with Crucial's Ballistix PC4000 DDR.

I was able to overclock the PC4000 from Crucial up to PC4800 (DDR600) with 3-4-4-10 timings, but not with 100% stability according to Memtest x86.  The maximum I could get the Crucial Ballistix to run at stable was 580MHz DDR (PC4720)) at 3-4-4-8 timings.  This had to be done at 2.8v, and raising the voltage didn't increase stability in Memtest at 300MHz.  This is a pretty impressive overclock on this DDR, as it can also be ran at 400MHz DDR at 2-2-2-5 timings with as little as 2.6v.  This memory performs excellent accross a broad spectrum.

Just for fun, I'm dropping in some benchmarks and screenshots of the Ballistix at DDR600.  It ran in Windows just fine, but Memtest reported errors.  You'll notice that this is at a 1:1 with the CPU at a 8x multiplier.

CPUz Fastest (not 100% stable) OC

CPUz Fastest Memory (not 100% stable) OC

Everest

In the tests following we run the system at DDR400 (PC3200) 2-2-2-5 with a 9x multiplier = 1800MHz, and DDR500 (PC4000) 2.5-3-3-8 with a 8x multiplier = 2000MHz, and finally at our maximum stable memory capability at DDR580 (PC4720) 3-4-4-10 with a 8x multiplier = 2320MHz.  Loosening timings or increasing voltage beyond 2.9v didn't help us gain a single MHz in stability.  290MHz is wide open on these modules.

Synthetic Tests:

For these synthetic tests we used Everest Home Edition, formerly known as AIDA32, and the ever popular SiSoft Sandra 2005.  These programs give you a good idea of how overclocking affects memory bandwidth, but whether you'll see it in real world applications or not is a bit up in the air.

As you can see in the Everest Benchmark below, memory performance is directly linked to the total speed of the processor and not just the FSB of the memory itself.  The top part of the graph shows performance at PC3200 running at 1.8GHz.  The next section shows performance at PC4000 running at 2.0GHz.  The biggest jump comes with the biggest increase in processor speed at PC4720 running at 2.32GHz.  Regardless, the numbers speek for themselves with the PC4000 running at it's stock speeds and hitting over 5,800MB/sec read transfer.

Everest Home Benchmarks

SiSoft Sandra shows similar results.  It scores the memory slightly lower that Everest but the trend is the same.  The largest jump in performance comes when the CPU speed is increased.  Again, this is due to the integrated memory controller on the chip itself.  Compare your scores using similar CPU/HTT speeds to determine how the Crucial Ballistix stacks up.  I think you'll find it does very well.

SiSoft 2005 Benchmarks

Now that we've seen how it shapes up in the theoretical world, let's take a look and see how this memory can increase performance in some of todays latest games.