OC Report :: Buffalo FireStix 2x1GB PC2-9600 C5
 

I am sure many of you have seen my threads at other forums and reviews that usually get posted a week later, but I figure I might as well start posting my overclock reports here too, I like this place much better than some of those other forums ;).

Buffalo Technology has developed a fairly well known brand with their FireStix line of memory and really gotten their foot in the door with the enthusiast. They were known as one of the better values throughout enthusiast forums while DDR2 prices were still fairly high offering D9GMH/D9GKX at a better price than most other manufacturers. Now that DDR2 prices have dipped, the value the FireStix line offered isn't as big but they are still some of the better priced modules out there.


As always, I provide full disclosure with any info I post and this report is no different. I received these modules directly from Buffalo Technology as a review sample but I was assured that they were your average retail sample sent from the same place they would have been sent from had I ordered them myself. Here are the basic specifications of these modules:

These are fairly new modules but I was still able to find them at NCIX up here in Canada. At the time of writing this, they were listed for $375 CND. Of course, you can also purchase these modules directly from Buffalo at an extremely decent price of $289 USD. Hopefully NCIX just has them listed for a higher price until they come in stock. I hate how NCIX gouges certain items because they are the only supplier in Canada but what can you do? Either way, let's have a quick look at the package and the modules themselves.


Package:
Click for full size...

Buffalo appears to have made a slight revision to the blister pack that they ship their modules in. You'll notice in the images above that the outside rim of the entire package is heat-stamped and sealed. On top of this, the back also has a couple of grooves molded in that really gives the package a lot of rigidity.

I don't really think the typical blister pack needed these enhancements for regular sized modules but this package certainly is the most secure I have seen memory come in. The only downside is the fact that it is impossible to open without a weapon and the package gets chewed up pretty good in the process.


Modules:
Click for full size...

If you saw my last OC Report of the PC2-8000 FireStix, you will notice that there have been no changes to the FireStix heatspreaders since then. These DDR1200 sticks come with a standard aluminum heatspreader secured with your run of the mill thermal tape. Popping the heatspreaders off these modules is fairly easy but you still have to take great care in doing so. As expected, Micron D9GKX are the ICs used on these modules. The PCB appears to still be made by Dynamic with the part #2D286NF-BA. Unfortunately, I cannot find any more info on this PCB so it is tough to say whether it is still a 6-layer design of if they have made the jump to an 8-layer PCB like many other manufacturers have with high-end modules.


P5B-Dlx Setup:
Click for full size...

I will be utilizing my trusty P5B-Dlx setup for testing and if I get lucky, I will have my hands on an IP-35 Pro for long enough to do testing on P35 while I have this kit. As you can see, I always use active cooling on the memory regardless of voltage being run and the testing will only go up to 2.45v on the P5B-Dlx as I haven't volt modded the vDIMM on this board. Here is the complete list of hardware used:
My standard disclaimer before we see the numbers:
First, the voltage listed in each screenshot is measured with a digital multi-meter from the vDIMM output coil. My board tends to over-volt slightly at the various voltage settings.

Second, I am only interested in dual 32M SPi stability for my overclocking reports. I have found dual 32M stable with tight secondaries usually means Prime / Win Memtest stable within a few MHz at the same voltage. Don't confuse dual 32M stable with 24/7 stable but it is pretty close most of the time.

Third, and most important, these results are from one kit clocked on one motherboard. Your results are certainly going to be different because no single kit of memory is the same and every motherboard handles memory differently. Again, better or worse, results will vary from user to user and kit to kit.


P5B-Dlx Results:
3-3-3-x:
Intel P965x:
Click link for full size screenshot including Memset...
375MHz :: 2.02v | 392MHz :: 2.12v | 421MHz :: 2.26v | 437MHz :: 2.37v | 446MHz :: 2.47v
It appears that my slump of getting dog sticks at anything but 5-5-5 has been broken. These PC2-9600 FireStix were quite impressive at 3-3-3 timings. Sub-timings are reasonably tight throughout the frequency scale as well making this one impressive kit. I did try running 3-4-3 timings after 3-3-3 and gained next to nothing at the few voltages stops that I tried. This had me concerned that 4-4-4 was not going to run well.


4-4-4-x:
Intel P965x:
Click link for full size screenshot including Memset...
481MHz :: 2.02v | 502MHz :: 2.12v | 537MHz :: 2.26v | 555MHz :: 2.37v | 574MHz :: 2.47v
4-4-4 timings started out relatively impressive at low voltage. They also scaled with voltage pretty well but nothing too impressive. Keep in mind, my standards are fairly high and I am still looking for that elusive 4-4-4 600Mhz @ 2.45v or less kit of memory for benching. being dual 32M stable means that at 2.6v, possibly even 2.55v, this kit should be single 32M stable at 4-4-4 600MHz which is mighty nice to be truthfully honest. i just don't like running that high voltage, even for benching.


5-5-5-x:
Intel P965x:
Click link for full size screenshot including Memset...
586MHz :: 2.02v | 607MHz :: 2.12v | 645MHz :: 2.26v | 651MHz :: 2.37v
This is where the PC2-9600 FireStix really light it up. At a paultry 2.12v, this kit spit out dual 32M Spi at 600MHz with ease. Sub-timings at 600MHz are even quite impressive. Naturally the sub-timings have to loosen up past 600MHz but remained fairly tight. Topping out at 650MHz is not the limit of these sticks rather my motherboard. I have seen it before and will see it again. I really want to get these sticks on a P35 chipset board that handles memory as well because they would absolutely scream.


P5B-Dlx Bandwidth:
My OC Reports are quickly evolving into full fledged enthusiast reviews it seems. I have had some requests for bandwidth numbers to accompany the overclocking results in the past so I will be adding these results to my OC Reports starting with these PC2-9600 Buffalo FireStix. I wasn't sure exactly how to tackle this Bandwidth section so please make any suggestions you wish as to how I can arrange results better or what results should be added.


I tried to keep the CPU frequency under 3600MHz by adjusting the CPU multiplier and using different dividers to provide similar results that the majority of users on air cooling are going to see. All of the bandwidth numbers were obtained on the 1066 strap (boot @ 400FSB or less) to keep the results as comparable as possible.

As from the results, it is evident that either the 2:3 strap reduces Write bandwidth or the tighter timings are simply superior for Write bandwidth because the 3-3-3 results really stand out with higher Write bandwidth.


Conclusion:
That about finishes up what I had planned for this kit and this OC Report. I hope you find it helpful in your search for DDR2 memory. Again, keep in mind, these results are all from a single kit of memory and like everything else in overclocking...nothing is guaranteed so don't assume that every single kit of this memory will perform the same. If your lucky, your kit will be better than mine :thumb:

Great review as always Jody. Kit looks decent, but it is also a bit steep considering you can get a great D9 kit for $150-160 right now. What kits have you tested that hit the ddr2 holy grail of DDR1200 @ 4-4-4?

Oh and wheres the maximizer, it not work on P5B DX? Very nice work on the pics too btw!

FYI this is 6 layer.

I like your ram reviews, I'm glad you brought them over here. I wonder if the PCB is better on this kit compared to the PC2-6400 kit.

645MHz :: 2.26v

Nice overclock with 24/7 voltage right there!

Great review!!!!!

They do seem a bit pricy, although the 3-3-3 at high clock speeds might make that worthwhile.....

thanks. yeah, it is a bit steep considering what you can get Ballistix 8500s for but just look at the price of anything else binned for PC2-9600. sure any Ballistix 8500 is going to run these same speeds but there is always that chance it won't...ridiculously slim chance.

kits that have hit 600MHz 4-4-4...
Team 6400 C3s @ 2.60v+ for dual 32M
Super Talent 8000 C4s @ 2.60v for dual 32M

the maximizer doesn't seem to be playing nice with the P5B-Dlx anymore. it use to provide the same clocking ability as the stock vDIMM regulation on the board (but never any better). now, it doesn't...i actually get lower clocks at a set voltage with the Maxi in place. i think it is because i never removed the output coil from the board...that or my Maxi just doesn't work as it should.

Probably the output coil, I think u need to remove it for it to work properly.

Thanks for the info btw. :)

no problem...yeah, the weird thing is that it worked for the longest time with the output coil...who knows. if i remove the coil and the Maxi still doesn't work right, it's not hard to solder back in place is it? oh yeah, i have a couple dead boards on my shelf that i will practice on first. man, i should be shot for not thinking of that sooner.

@ the other guys...thanx, i am glad you like the reports. nothing beats spending a few days clocking memory IMO, i am too happy to present my results.

the 3-3-3 results were quite surprising but even more surprising was that 3-4-3 did absolutely nothing for higher clocks at a set voltage. my Ballistix give almost 40MHz more going to 3-4-3 from 3-3-3...i guess they just don't like tRCD of 3 :)

Cant be too hard to put a coil back, I mean you do vmods so that should be easy for you. :)

These #s just point out to me how much work I need to do on my ability to OC memory..... :)

Currently working on some Balistix 8500s..... Got them up to 1158 for what appear to be partially orthos stable (longest orthos so far has been 20 min as I'm still working on the max OC) but I'm pumping 2.4 into them so there is a lot more work to be done......

hmm...you should not need near that much voltage for Ballistix. of course, every kit is different but 2.4v seems waaaay too high for only 580MHz. i have had 4 Ballistix PC-8500 kits, 1 tracer / 3 non tracer, and they were all Orthos stable at 5-5-5 2.25v (DMM measured) or less.

this is from the last kit of 8500 Ballistix (non tracers) i bought about 3 weeks ago from NCIX:

600MHz :: 5-5-5-8 :: Dual 32M SPi :: 2.12v

for multi hour Orthos / Win Memtest stability, i have to bump up to 2.17v (2.15v in BIOS) and slightly adjust some secondary timings with that Ballistix kit. keep in mind, i have clocked over 20GBs of D9*** memory on this particular P5B-Dlx so i know it intimately and can clock memory on it blindfolded almost.

the other thing i wanted to mention was that 2.4v for extended periods of time tends to be a bad thing, especially for Ballistix. i really don't feel the Levin PCBs they use handle voltage all that well. just my opinion and not really based on anything factual aside from a couple experiences with PC2-5300 and PC2-6400 kits. i wish i had some experience with a 965-S to perhaps provide some insight for ya but i have never touched one...but desperately want to ;)