With the introduction of Intel's 865PE and 875P chipsets, and the 200MHz FSB Northwood processors, enthusiasts have consistently pushed their systems to the 250MHz FSB barrier. One of the main ingredients that has held us back, has been memory. The ability to overclock to speeds in excess of 250FSB typically requires you to run memory in 5:4 or 3:2 ratios, basically defeating the overclock performance you would gain.
Since that time many memory manufacturers have introduced Dual DDR kits that are able to exceed the current maximum standard of PC4000 (250MHz FSB or DDR 500). We have seen PC4400 (275MHz FSB or DDR 550) from the likes of OCZ and Corsair to name but two. Today I am reviewing one such kit, the (266MHz FSB or DDR 533) Dual DDR kit.
Why a PC4300 kit from Kingston you ask? Why are they not touting PC4400 as some of the other manufacturers out there? This is actually pretty simple according to Kingston; very few motherboards can sustain a stable 275MHz FSB (DDR 550), and just about as few memory modules can sustain a stable environment at that rate as well. When looking at DDR 533 (266MHz FSB), the amount of motherboards that can sustain a stable overclock increases substantially, as well as the number of memory modules. It certainly makes sense to target the larger audience.
Although you can run non Dual Channel memory in Dual Channel mode, you don't always get the performance of a matched pair of sticks. This is especially apparent when trying to exceed 250 MHz FSB. Dual Channel sticks are still a little more costly than buying two sticks of the same make and speed, but that is for a reason, they are matched to perform at the rated speed in Dual Channel mode, and then some. If you are still using a motherboard that does not support Dual Channel mode, than I would stick with the more cost effective non-kit memory sticks.
Specifications
• 2x512MB
• 533MHz (PC-4300) Unbuffered DDR Modules
• Aluminum Heat Spreader (Kingston Blue)
• Latency: 3-4-4-8
• VDDR: 2.7V
• Lifetime Warranty
Kingston's HyperX PC4300 comes in their usual clear packaging that places the 2 matched sticks next to each other with pretty much everything in sight. The packaging is designed to hang from a shop hook and still show off the impressive blue heatspreaders, which you can't fail to notice. You also get a leaflet giving you the pertinent info of the ram.
Closer inspection of the ram sticks shows a sticker in the right corner of one side which gives us such information as the default voltage and memory speed. The part number is given, KHX4300K2/1G, which tells us it is PC4300, and one half of a 1GB kit (hence 512MB). The maximum voltage is indicated as 2.7v, which is more or less in line with what we've seen with other PC4000+ modules. The timings aren't indicated on the module, but according to specifications, the timings are 3-4-4-8-1.
In the opposite corner is the red fisted logo for Kingston and of course being HyperX ram, we have the white HyperX logo in the center.
Test System
ABIT AI7 Motherboard, 1GB (2x512 in Dual DDR Mode) Kingston HyperX PC4300 Kit, Intel Pentium 4 2.4e (800MHz FSB), ATI Radeon 9600xt 128mb, Hitachi Deskstar 80GB 8MB buffer 7200 RPM SATA Drive, Windows XP SP1 and Pre-SP2 hotfixes, ATI Catalyst 4.4.
Comparison
I will be comparing the Kingston HyperX PC4300 Kit against my OCZ PC3700 Kit.
Overclocking
This is a tough one, the sticks come prepared to overclock, and the fact that I am testing this on a Springdale and not a Canterwood might come in to play as well. Initial attempts to clock up to even the rated speed was met with resistance, I had issues going over 260MHz FSB, knowing that this was not a limitation of the processor and/or motherboard (I had reached 274MHz on the original OCZ PC3700 at 3:2 ratio) I recruited Scott to assist me in researching HyperX and my particular board for possible issues.
After upgrading to BIOS V.17 and a few other tricks to no avail, Scott found out that apparently the AI7 clocks better if the memory is positioned in Slots 2 and 4. Well, it worked; I moved the memory to slots 2 and 4 and voila, the memory not only clocked to 266MHz FSB it went all the way to 269MHz FSB, 1:1 ratio. Granted this is only 3MHz FSB over the rated speed, but considering that this is the highest the motherboard has been at 1:1 ratio, it is good. Trying to move the ratio to 3:2 or increasing the VDDR resulted in not even 1MHz more overclock.
Being that I was able to get the OCZ PC3700s to 275MHz at a 3:2 ratio, I would not think it is the board that is my limitation here. That being said, it could just be a further mismatch of the ABIT AI7 and the Kingston HyperX, I have seen weirder things for sure.
Testing
Time for the testing phase, all tests are run 3 times and results are then averaged (unless otherwise noted). VL's testing suite includes the following:
SiSoft Sandra 2004
PiFast 4.3
TMPGEnc (AVI to MPEG2 encoding)
PCMark02
I popped out my 1GB of OCZ PC3700 sticks and popped these in, the first thing you notice is the weight of them , the HyperX is about 1/3 the weight of the OCZs. This is mostly because Kingston uses aluminum spreaders as opposed to OCZ's copper spreaders. Whether this actually equates to higher heat dissipation is still up in the air.
I booted my system at stock settings. I ran a few tests to make sure everything was kosher and proceeded to step up by 5MHz increments until I reached 260MHz FSB, at which time I incremented by 1MHz (already documenting the VCore requirements there was no rebooting needed to produce appropriate testing at each FSB setting).
I stopped at 266MHz FSB (the rated speed for the HyperX memory) and ran the full slew of tests. I then stepped the memory to its maximum (boot failure) and attempted to increase VDDR (I could not relax the timings as they were already at my motherboards maximum of 3-4-4-8), I attempted voltage adjustments all of the way to 2.9 VDDR, to no avail, 272MHz FSB was just not going to happen. I then backed off until I could get a full set of tests finished, and also run stable for at least 12 hours. The result was 269MHz FSB (DDR538). I determine it to be a stable 12 hour run when I have successfully run Folding @ Home (2 clients, one for each logical processor) as well as running a few games here and there to ensure that everything is properly communicating throughout the system.
NEXT