New DDR Highs: Shikatronics, OCZ, and the Fastest Memory Yet
by Wesley Fink on June 29, 2004 12:05 AM EST- Posted in
- Memory
Final Words
OCZ and Shikatronics have produced the fastest memory that we have yet tested at AnandTech. The latest Hynix B chips used in both memories are incredibly fast, and both memories are performing very close to the DDR600 threshold. Across the board, the performance of the two new DDR550 memories is equivalent, and both outperform the earlier DDR550 memory, which we reviewed at Anandtech.We have come to expect OCZ to perform at or near the top of our memory review tests, but the fact that ShikaXRAM is also performing at the top of our review charts says worlds about the quality of the new Shikatronics memory. While ShikaXRAM is a new name to many in the US market, based on the performance of the PC4400, Shikatronics is a company producing very competitive Enthusiast memory products. We cannot draw a conclusion with just one product, but we are certainly impressed with what we found in quality and performance with the ShikaXRAM PC4400.
If you want the fastest DDR memory that you can buy, then either ShikaXRAM PC4400 or OCZ PC4400 EL Gold will fit the bill. The top DDR performance is a moving target, and both these new memories based on the latest Hynix B chips raised the bar a bit higher. For those of you who will fret that the timings for DDR580 performance are 3-4-4-7, we would point out that DDR2 requires even slower 4-4-4 timings. The additional latencies of DDR2 at the present time also cause DDR2 at 533 speed to perform about the same as fast DDR at DDR433. The 6400 Standard (buffered) Sandra 2004 memory bandwidth that we achieved at DDR580 is the highest result we have seen so far on our Intel test bed, as is the 3750 unbuffered memory bandwidth.
We have looked at the question of which is best - slower memory at fast timings or fast memory at slower timings - in many past reviews. We leave that argument to those reviews. 1:1 memory performance is always a bit faster than asynchronous operation due to latency differences. Also, we know of no fast CAS 2 or CAS 2.5 memory that can even approach DDR580 timings. In the end, these results are the highest that we have achieved with any memory we have tested. This, of course, comes with a price, since there are very few motherboards available that can run at DDR580, and only a very few CPU combinations that can achieve these kinds of overclocks. However, if your system can do it and top speed is the goal of your overclocking, you can't do better than the new ShikaXRAM and OCZ PC4400 memory.
Hynix memory chips have carried a very high tariff in the US, and this might give a pricing advantage to Shikatronics, which is located in Canada. However, we could not find ShikaXRAM PC4400 available for sale in the US market. This means that it is much easier to find OCZ PC4400EL Gold for sale at vendors in the US. However, Shikatronics does sell directly on their website, and you can request pricing information on their website. Whichever memory you buy, we do not think you will be disappointed in the high-speed performance.
Facebook
Twitter
RSS
11 Comments
View All Comments
jediknight - Wednesday, July 7, 2004 - link
I wish you would have put in some "generic" or inexpensive PC3200 RAM as a base to compare the performance of these chips to..Pumpkinierre - Thursday, July 1, 2004 - link
The smaller multiplier allows you to finetune better Trog that's all. Remember the mutiplier is locked. Even though the 2.4 gets all the limelight, some 2.6s get over 300MHz as well (and I did see a post on the net once where 2.4 would only get 230- a bit like an AMD ahmmm scratch that!). Its just the luck of the draw. Mine is a phillipines one and I think the Costa Ricans average better. I overclocked the 2.6 using 5:4 mem on my abit IC7-G which is good to over 300 MHz FSB using async. But at 1:1 it appears that the ASUS P4C800s are the better boards as the high end mem. manifacturers only seem to guarantee their modules on that particular board. My 2.6 gets to 246MHz (3.2Gig) on default voltage, 250 takes an extra .1V and 260 an extra .125V past that again. I use an Aero4 copper heatsink with the turbine removed and an external tube to a fitered Sunon 50cfm on the case (without a fan directly on the heatsink to avoid the dead spot). Temperature is 35-45C (core sensor idle- load) with ambient of ~22C (far worse in summer). Naturally liquid N2 like the cold finger on one of the mass spectrometer sources I worked on would be better but lets be serious!I'd be happy with the 246 as I dont like overvolting (especially with the current leakage sudden death rumors) but I want the 1:1 low latency RAM to match that and that would be my final memory upgrade for the system (probably 2Gig). I thought a year after buiding my system it would be plentiful given the fast development curve from 266 to DDR400 with low latency. But I am still waiting.
MoToR - Thursday, July 1, 2004 - link
Wesley, please correct me or the text below:-quote-<i>
The 6400 <b>unbuffered</b> Sandra 2004 memory bandwidth that we achieved at DDR580 is the highest result we have seen so far on our Intel test bed, as is the 3750 unbuffered memory bandwidth.
</i>-quote-
6400 is buffered isn't it?
Regards,
MoToR.
TrogdorJW - Thursday, July 1, 2004 - link
That's what you get for going with the higher 13X multiplier of the 2.6C instead of a 2.4C, Pumpkin. ;) Of course, a better cooling setup might potentially help (not sure what you have). Anyway, considering 260 = 520 = ~PC4200, there's really not a whole lot more room to overclock anyway. As the article points out, only Asus and DFI boards are able to reach these high clockspeeds. Given the variation in quality among RAM and motherboard parts, a "lesser" Asus or DFI might not clock quite as high. CAS 2 at 250 MHz would be nice, though. Wonder when (if?) we'll actually see it? Maybe when we start getting 90 nm memory parts?Pumpkinierre - Tuesday, June 29, 2004 - link
Even though its nice to see your 2.4c approach the 300MHz FSB at 1:1, not all of us are so lucky. My 2.6c maxes out at 260MHz so I'm still looking for that DDR500 CAS2 holy grail. Along, I suspect with many others, I'll buy when I see good evidence of this in a memory module.pjwalker - Tuesday, June 29, 2004 - link
CompuSmart Brandon has dealt with Shikatronics for over 5 years and have always received top quality products and excellent customer service. We have purchased from several other memory suppliers over the 10+ years in business and Shikatronics has had the least DOA and quickest replacement time of any of them.Way to go Shika keep Kicking Ass!
Wesley Fink - Tuesday, June 29, 2004 - link
We seem to get this question in every new review with graphs that have a lot of bars. We can't provide a different color for the tested item right now no matter how much we agree with all of you. It is a limitation of our current graphing engine, which is used for all reviews.There is good news, however, in that a new Graphics Engine is now testing that WILL provide a highlight color for the tested item. It is expected to go live next month. Your patience will be rewarded :-).
Fr0zeN2 - Tuesday, June 29, 2004 - link
nonono whatever you do don't reduce the number of graphs -- it's much better seeing all memory on one chart than poking around the archives. I agree on the highlight thing though.unas - Tuesday, June 29, 2004 - link
It sounds great, but with the GIANT lots of memory which are included in every test, it becomes hard to find the graph which belongs to the actually tested RAM. Either you reduce the number of graphs to only four or five for each diagram, or you emphasize the tested memory's graph by making it light blue, red, green or whatever colour. It's even becoming hard in an Athlon 64 Mobo roundup. Please do something about that.Unas
unas - Tuesday, June 29, 2004 - link