Logo   Banner   TopRight
TopUnder
Transparent
OCZ DDR2 PC2-6400 Platinum EB
Transparent
Date: Sep 28, 2005
Section:Misc
Author: Robert Maloney
Transparent
Introduction

 

Last month we took a look at two recent sets of memory from Corsair and Kingston that were pushing DDR2 technology ahead, offering higher clock speeds and lower latencies to complement leading edge motherboard technologies.  Each set did very well in our benchmarks, offering up premium peformance with solid stability. But, besides death and taxes, the only other constant in this world is that the speed of most PC components increase every six months or so.  While Kingston was the first to reach the 750MHz barrier with their HyperX KHX6000D2K2/1G memory, we now have faster sets on the market, including today's entry from OCZ Technology, a 1GB set of 800MHz rated DDR2 called the OCZ PC2-6400 Platinum Enhanced Bandwidth Dual Channel.  It's quite a long name for two shiny, short sticks of RAM, but breaking it down to basics means we should be looking at high speed RAM with low latencies.  

OCZ DDR2 PC2-6400 Platinum Enhanced Bandwidth Specifications
http://www.ocztechnology.com/products/ddr2/
Each OCZ PC2-6400 EB Dual Channel Kit is hand tested as a matched pair across a wide variety of motherboards. In addition, OCZ DDR2-800 EB offers integrated platinum mirrored copper heatspreaders for efficient heat dissipation, a lifetime warranty, toll-free technical support and the exclusive EVP (Extended Voltage Protection) coverage.
Specifications:
·_800 MHz DDR2
·_CL 4-3-3-8 (CAS-TRCD-TRP-TRAS)
·_Available in 1GB (2x512MB) Dual Channel Optimized Kits
·_Unbuffered
·_Platinum Mirrored Copper Heatspreader
·_Lifetime warranty
·_2.1 Volts
·_240 Pin DIMM
Special Features:
·_OCZ Enhanced Bandwidth Technology - provides intelligent bandwidth management by offering the lowest possible page access times
·_2.2 EVP - OCZ's unique Extended Voltage Protection feature allows {C2-6400 modules to handle voltages up to 2.2V and still be covered by their Lifetime Warranty

Part Numbers:
·_OCZ29001024EBDCPE-K

    

    

As with most memory kits we've recevied, the two sticks are prominently shown in plastic clamshell type package that contains little else but the memory and fact sheet.  Luckily, the shell is not sealed in that "grab the knife and start cutting" method, so we were able to get the sticks out without any issue.  The pair of sticks are encased within platinum colored copper heat-spreaders with OCZ's logo machined right into the center on both sides.

Other than the logo, there's not much else to note except for the identification sticker placed on the left edge.  This sticker contains all of the necessary descriptive information, in an easy to read format and thus easy to identify parts later.  OCZ's nomenclature labels the memory as "PC2-6400", which should be read as DDR2 rated for speeds as high as 800 MHz.  Each stick consists of 512MB of memory that have been tested and paired for stable dual channel operation.  The CL timings listed here were 4-3-3-8.  Although we had seen lower CAS ratings with the Corsair memory we previously reviewed, these were timings for running at 675MHz, and not the 800MHz that the OCZ sticks can operate at. 

   

Using CPU-Z version 1.30, we can get an inside look at the programming of the modules.  This will tell us how the RAM is officially detected with the on board Serial Presence Detect (SPD) PROM.  The SPD timings are shown to be a little looser than what was officially stated.  According to OCZ's website and the sticker on the heatspreaders, these sticks were supposed to be running at 4-3-3-8, but in CPU-Z the SPD Timings Table listed the RAM as capable of running at 4-4-4-8 at 400MHz.  Even though the SPD timings on the RAM were close to the rated values, the Asus P5WD2 set the timings to 5-5-5-15 at 334.5 MHz, using the FSB:DRAM ratio of 3:5.  As the P5WD2 reported the same timings for all three kits that we will be testing, this is obviously an issue with the motherboard and not the OCZ memory.

Transparent
Test Setup and Compatibility

  

How we configured our test systems:  When configuring the test system for this review, we first entered the system BIOS for the Asus P5WD2 and changed all settings to their "Optimized" or "High-Performance Defaults".  The hard drive was then formatted, and Windows XP Professional with Service Pack 2 was installed. We then installed all of the necessary drivers, and removed Windows Messenger from the system. Auto-Updating and System Restore were also disabled, and we setup a 1536MB permanent page file on the same partition as the Windows installation. Lastly, we set Windows XP's Visual Effects to "best performance", installed all of our benchmarking software, defragged the hard drives and ran all of the tests. 

Throughout all of our standard benchmarking, we had the memory voltage set to 2.1v for the OCZ memory pairing, as this was the voltage recommended by the manufacturer. For the overclocking and lowest-latency tests, we raised the memory voltages to 2.2V.

Test System Specifications
Time to put the pedal to the metal
Memory Modules Tested:
OCZ Technolgies PC2-6400 EB Platinum Edition

Corsair XMS2 TWINX1024A-5400UL
Kingston HyperX KHX6000D2K2/1G

Common Hardware:
Asus P5WD2 Premium (Intel i955X)

Intel Pentium 4 550J Processor @ 3.4GHz

nVidia GeForce 7800GT

On-board audio & LAN
Seagate Barracuda 7,200rpm SATA Hard Drive

Software / System Drivers:
Windows XP Pro with Service Pack 2
DirectX 9.0c
Intel Chipset Software, v7.21.1003
nVidia ForceWare v77.77

Memtest86 Compatibility
http://www.memtest.org

Before we got down to testing, we first checked to see how compatible our the OCZ sticks were with a few of the recent motherboards we had in the labs.  We installed the memory into two different setups, using the Asus P5WD2 (i955X) and Abit AW8-MAX (i955X) with all BIOS detection methods left at 'By SPD'.  The first set of RAM we received were not accepted by the Asus P5WD2 when using 'SPD' timings, and the board would not boot unless we specifically set the timings and memory dividers.  OCZ assured us that this was an isolated problem, and the next set we received did not have any problems at all in either setup. 

To assure ourselves that we now had a working set, we then began checking for errors using Memtest86.  Memtest86 executes a series of read/write test patterns to check for errors. Using the Asus P5WD2, which will be the base for our benchmarks, we used a bootable CD-ROM to start the testing.   The OCZ memory kit passed through two loops of this procedure without any hiccups.  Satisifed with the results, we now turned to some popular benchmarks to compare the performance of the OCZ PC2-6400 with two sets we had reviewed previousy, Corsair's XMS2 TWINX1024A-5400UL 1GB kit and Kingston's HyperX KHX6000D2K2/1G. 

Transparent
Performance at SPD Settings

Performance Comparisons with SiSoft SANDRA 2004
Synthetic Raw Bandwidth

We began our testing with SiSoftware's SANDRA, the System ANalyzer, Diagnostic and Reporting Assistant, with the memory configured by their SPDs. SANDRA consists of a set of information and diagnostic utilities that can provide a host of useful information about your hardware and operating system.  We ran SANDRA's Memory Bandwidth test with OCZ, Corsair and Kingston memory modules installed on our Asus P5WD2 Premium.

(Note:  Although we had the Asus motherboard we used for testing configured to detect memory timings by SPD, it ran all memory kits with 5-5-5-12 timings, unless we manually specified otherwise.)

Corsair's TWINX set was the overall leader in bandwidth, placing first in both Integer and Floating Point metrics.  In the Integer test, the Corsair modules lead OCZ's set by merely 16MB/s, which in turn beat out Kingston by an additional 40 MB/s.  The same pattern persisted with the Floating Point test, with the order going Corsair, OCZ, and finally Kingston.  Although all sticks were rated equally within the BIOS, OCZ's sticks found their way to just better than middle of the pack.

Performance Comparison with PCMark05
Overall Memory Score

For our next round of benchmarks, we ran the Memory performance module built-into Futuremark's PCMark05. For those interested in more than just the graphs, we've got a quote from Futuremark that explains exactly what this test does and how it works...

"The Memory test suite is a collection of tests that isolate the performance of the memory subsystem. The memory subsystem consists of various devices on the PC. This includes the main memory, the CPU internal cache (known as the L1 cache) and the external cache (known as the L2 cache). As it is difficult to find applications that only stress the memory, we explicitly developed a set of tests geared for this purpose. The tests are written in C++ and assembly. They include: Reading data blocks from memory, Writing data blocks to memory performing copy operations on data blocks, random access to data items and latency testing."

The PCMark05 scores were right on target with the Sandra 2004 scores.  Once again, the Corsair TWIN2X1024A-5400UL kit posted a score that was just a bit higher than the OCZ PC2-6400 set.  Kingston's KHX6000D2K2/1G kit, however, found itself in last place by a hair.  The differences between the highest and lowest kits still came out to less than one percent, so there's not too much to be concerned about.

In-Game Performance Comparisons With Unreal Tournament 2004
Memory Bandwidth and Frame rates 

We also ran through a batch of time demos with the popular shooter Unreal Tournament 2004.  We ran these tests at the "Fastest" setting with a low resolution of 640 X 480, using 16-bit color and textures.  Running Unreal Tournament 2004 with a high-end graphics card at these minimal settings isolates processor and memory performance without being limited by the graphics subsystem. 

In Unreal Tournament 2004, we can see how the variances in bandwidth affect gameplay.  Corsair's memory kit, which led in both the SANDRA and PCMark05 tests, unsurprisingly gave us the highest frame rate as well, topping the charts at 140.38 frames per second.  We were about a half of frame behind when using OCZ's PC2-6400 kit, and Kingston's HyperX modules were almost two frames back from the leader.

 

Transparent
Performance at Lowest Latency

To get this next batch of results, we manually lowered the latencies of each memory kit as low was the BIOS would allow, bumped up the memory voltage by 0.1V for each set of RAM, and slowly raised their clock speeds until our test system was no longer stable. We found that the Corsair XMS2 and OCZ's PC2-6400 modules were both able to hit the tightest timings at 3-2-2-4 on the Asus P5WD2, with a maximum bus speed of 204MHz at a volatge of 2.2V.  As we reported before, the Kingston RAM would only be stable at 3-4-4-4 timings at these voltages; anything lower would prevent us from booting.  We were able to hit the same bus speed at these timings as well.  So, in the graphs below, we have all memory kits performing at the lowest latencies possible while overclocked slightly to a 204MHz FSB (340MHz Memory Clock).

Performance Comparison with SiSoft SANDRA 2004
Synthetic Raw Bandwidth

With the BIOS set at a 3:5 clock/memory divider, a 204MHz FSB equates to a 340MHz memory clock (680MHz DDR) for all modules.  Not surprisingly, the two sets that were able to run at the tightest timings, Corsair and OCZ, came out on top, although the margin of difference appeared to be roughly the same as before.  Corsair maintained its lead over the two other sets in both SANDRA benchmarks.

Performance Comparison with PCMark04
Overall Memory Score

We've included the original 'By SPD' results in these graphs to show how the memory kits performed better with lower latencies and higher speeds.  Each set gained about 80-90 points in this second running of PCMark05, so the order of the kits remains the same.  We noticed that the Kingston set actually gained the most of the three kits, pointing out that there was no big disadvantage in running at 3-4-4-4.

In-Game Performance Comparisons With Unreal Tournament 2004
Memory Bandwidth and Framerates

When we re-ran the Unreal Tournament 2004 benchmarks, every memory kit saw frame rates increase by about 2 frames per second.  There's still a tighter grouping between OCZ and Corsair's kits, as opposed to Kingston's set, which remained in third place. 

 

Transparent
Overclocked Results

For our last batch of tests, we tried running the OCZ kit with SPD timings, and raised the front side bus on the Asus P5WD2 Premium as high as possible.  As we've seen in the past, we appear to be somewhat limited with overclocking our Pentium 4 550 on this motherboard.  The ceiling with the current setup appeared to be a 234MHz FSB, resulting in our CPU running at 3.98GHz and the memory at 780MHz effective.  That gave us our second frame of reference.  We then backed the FSB off slightly, and changed the memory divider from a default setting of 3:5 to 1:2.  Using this ratio, we were able to push the memory up over 900MHz, stopping at 920MHz.  One last time we ran our benchmark suite to see how the performance changed.  As with the Low Latency testing, voltages were raised 0.1V, so that the RAM was running at 2.2V.  Please note, the results below reflect only the overclocked performance of OCZ's modules, the overclocked performance results for the Corsair and Kingston memory kits mentioned earlier are available here.

Performance Comparison with SiSoft SANDRA 2004
Synthetic Raw Bandiwdth

 

The scores increased expectedly in the overclocked testing as far as SANDRA was concerned.  The largest jump occurred with initial overclock to 780MHz where the divider remained at 3:5.  Overclocking to memory to 780MHz gave us a 500 point boost over the original SANDRA scores. However, moving upwards to 920MHz only gave us an additional 70 points.

Performance Comparison with PCMark04
Overall Memory Score

The overclocked PCMark05 results were nearly identical in relation to performance gains.  As one would expect, overclocking the system increased the overall performance, by nearly 600 across the board.  Once again, we noticed a much smaller increase in our second attempt at raising memory speeds.

In-Game Performance Comparisons With Unreal Tournament 2004
Memory Bandwidth and Framerates

In our final look using Unreal Tournament 2004 while overclocking the system we saw an increase of just over ten frames per second when we reached 780MHz, and then another five frames at 920MHz.  Combined, that equalled about a 16 frames per second improvement, or roughly a twelve percent increase in memory bandwidth. 

 

Transparent
Benchmark Summary and Conclusion

   

Benchmark Summary: OCZ's PC2-6400 modules placed a solid second in almost every benchmark, usually mere points behind the top performing Corsair XMS2.  Both the OCZ and Corsair kits were able to reach the tightest timings possible to us, running at  3-2-2-4 at a 204 MHz FSB, although the increases in performance compared to Kingston's kit at 3-4-4-4 were negligible.  We were able to overclock the OCZ memory up as high as 920MHz with our current setup, and we might have gotten even further with a different CPU and/or heatsink.  Even at these levels, we managed to increase memory bandwidth by approximately 12 percent.

OCZ's PC2-6400 EB Platinum Edition memory kit was able to compete directly with Corsair and Kingston at standard operating speeds and timings.  While Corsair's TWINX1024A-5400UL memory usually edged out OCZ, the performance deltas were slight, and real world results could equal a frame or two in gameplay at best.   Hitting 3-2-2-4 timings is always a plus, as these are currently the best timings that we've been able to achieve with DDR2.  What many will find most appealing is the higher speeds that OCZ provides. 

Unfortunately, a major downside of higher speeds along with low latency, comes in the form of a higher price tag for these types of modules.  We had some trouble locking down an actual street price for this set, but it's likely to be upwards of $265 for the kit.  At this price, it may be harder to justify not going with Corsair's TWINX1024A-5400UL, which is selling for a slightly higher premium but run a hair faster at stock speeds.  So is the slight gain worth the extra $20 or so?  We suppose how you answer that question will depend on the bulk of your wallet and more importantly wider availability of the product!  We've been told by OCZ that they'll be available within a week of this article's post date, at places like NewEgg.  So in conclusion, based on their very good performance and excellent overclockability, we're giving OCZ's PC2-6400 EB Platinum Edition a 9 on the HotHardware Heat Meter.

_Near the top of the charts in every benchmark
_Top rated memory when it comes to pure speed
_Platinum Mirrored Copper heatsinks
•  Relatively high price still compared to standard DDR
_First set we received would not allow test motherboard to boot at SPD settings

Looking for advice on hardware and nowhere to turn to?
Get into the HotHardware.com PC Hardware Forum!

 



Content Property of HotHardware.com