To the 550Mhz fray and beyond with OCZ
OCZ Technology Group, has given to us some of the fastest memory on the Planet and with their Performance Series 4400 the tradition continues. Faster then any PC4500 on the market it screams on Canterwood platforms, with the right stepping P4, and appropriate BIOS. Currently most DDR makers are putting all their resources into the development of DDR-II, which promises to pick up the MHz gauntlet where DDR dropped it, at 550MHz.
I'd received this Beta4400 kit some time ago, but postponed the review believing they may release PC4500 instead. Although the memory performs at PC4500 speeds, OCZ decided to label it PC4400, leaving some head-room for the Overclocker. Sure we know it may run at PC4500, but it makes good engineering sense to leave a performance ceiling above the rated speed. There are a few companies "experimenting" with PC4500, and if you buy it, you’re most likely part of the experiment. There's a good chance PC4500 being sold is simply using the same Hynix IC's found beneath most high speed DDR for Canterwood/Springdale platforms these days. Those IC's, HY5DU56822CT-D5, (5ns) DRAM, are pictured below
The industry has, however; pushed these chips about as far as their going to go. With people pumping 3.30V into these, and getting 290FSB to 300FSB or 600MHz , can we really expect much more? As is been reiterated time and again, and since I originally came across this article at Lost Circuits
which details the behaviour of DRR IC's under high voltage. Written by Michael Shuette, who I consider to be one of the most knowledgeable experts in the field. The review excerpt basically states; current DDR utilize internal voltage regulation, which keeps core voltages at 1.8V. VDDQ voltage to the I/O Buffers are the only devices which see the increased VDIMM. While this increase in voltage may have some effect on speed, and/or stability at increased speed, the net effect is an increase in heat generated by the internal voltage regulators, which slows the memory, defeating the overall purpose. To obtain such bandwidth we’re going to need a CPU with a low multiplier and for our testing purposes we'll be using the Pentium-4 2.4C, running at 12x200MHz.
Test System:CPU - SL6WF MALAY L310a735-0101
Cooling - Danger Den RBX, Hydor L30, Black Ice Extreme, 2x12mm Sunons
Motherboard - Asus P4C800E-Deluxe (BIOS ver.1014)
Motherboard - Abit AI& (BIOS ver.16)
HDD - Maxtor Diamond Max 9 Plus (SATA150, 120GB)
VGA - Asus V9950 (Detonator 53.03)
PSU - TTGI USA TT-550TS (550W)
Coolermaster Musketeer LLC-U02
OS – Windows XP
I began running the memory on the Abit AI7, based on the 865-chipset, and featuring Abit's overclocking utility known as µGuru. µGuru is much more then a BIOS or Software based program. Through the use of a dedicated microchip, it allows the user to adjust voltages, and manipulate FSB speeds from within Windows.
It's quite a revolutionary utility; however, it has its deficiencies. First of all I purchased the board primarily for reviewing purposes, and essential to that issue was a VDIMM adjustable to 3.20V. The µGuru overclocking utility does in fact work, and one may raise the FSB, and change all essential voltages from within Windows as is claimed, however; there are problems. Utilities such as WCPUID, and Sandra SiSoftware benchmarking, fail to recognize the changes. CPU-Z will, however; SiSoftware is an essential tool along with WCPUID for measuring results. Without the benefit of such programs recognizing changes, the µGuru utility is no more effective then ClockGen. It's true while using ClockGen voltages aren't accessible through Windows XP, but if one needs to reboot to measure performance using µGuru perhaps changes aren't taking effect beyond the software level? The BIOS is always the safer bet. I digress as these are issues which Abit, H. Oda and SIS may need to work-out. For our purposes all changes were made within the BIOS;
The screenshot below emphasizes the AI7's Abit EQ facility. The display is quite extensive, although I've yet to determine the location of the thermal sensor, and/or diodes? I believe Abit may actually have a diode in, or a thermal sensor near the North Bridge. Since the former seems unlikely as it would be the biggest news since Janet Jackson removed her IHS. The system does have an interesting array of temp measuring devices, none-the-less. The only discrepancy was DDR-voltage which seemed to fluctuate. While Abit EQ indicates DDR voltage at 2.95V, it was set to 2.90V in the BIOS.
Onto the testing!