Anandtech Reviews Mushkin RAM

EisleyRocks writes "Trying to find the right RAM to satisfy your overclocking needs? Anandtech has reviewed the latest offering from Mushkin. From the article: 'For gamers who seek fast timings and high bandwidth at stock memory ratios, the Mushkin XP2-5300 is a very good choice. The same can be said for overclockers looking for the highest DDR2 clock speeds that they can find. There are a few memories that can go higher in speed than Mushkin DDR2, but there are very few current DDR2 modules that can do 3-3-3 timings to DDR2-700 and above, or DDR2 memories that can handle higher voltage as well as this Mushkin. For now, the Mushkin XP2-5300 is a good choice among DDR2 1GB modules in 2GB kits.'"

Comments

Misreading the title...

[unkaggregate]

Who here quickly glanced at the title and saw "munchkin RAM"?

Slashvertisment

[Inverted Intellect]

I don't go to /. for advertisments. Yes, I know that high-end computer gear keeps improving, and that's all very exciting for rich computer enthusiasts and gamers. But I don't need to read about it on /. as it's not particularly exciting for me unless I just happen to be in the market for high-end gear (which I'm not at the moment), and any info on the subject is almost completely uninformative.

(Score -1, Redundant)

Re:Mushkin

[RupW]

So, do they actually manufacture RAM, or just retail it?

Judging by the article, they take Elpida's RAM chips and put them on their own custom "Brain Power" PCBs.

I don't see how much difference the PCB can make though: it's just an electrical connection to the chips, right? Sure, you can keep the circuits short and use really high purity copper but that's about it isn't it?

Re:Mushkin

[JonyEpsilon]

I don't see how much difference the PCB can make though: it's just an electrical connection to the chips, right? Sure, you can keep the circuits short and use really high purity copper but that's about it isn't it?

Not at all! It's all about preserving signal integrity between the components. Remember, at these speeds you shouldn't really think of the signals as plain-old electrical currents flowing down the tracks. They're really high-frequency radio waves propagating down waveguides. Think of where 400MHz lies in the radio spectrum - it's well above fm radio (in the uk, at least) and that propagates pretty well as a wave :-) The pcb's job is to guide these waves around, and this is trickier than you might think. You have to consider the effect of the dielectric circuit board (the fibre glass bit) and coupling between various tracks and layers in the board. Jony

Does it really matter?

[ami-in-hamburg]

I may be ignorant on this subject, but considering the speeds of non-overclocked hardware, is there really a human noticable difference when you overclock your hardware compared to if you don't?

I mean seriously, is the human eye capable of noticing a few more frames per second. I am almost certain that your keyboard, mouse and joystick cannot be overclocked. Therefore, if your input is limited by your input device speeds, does it make sense to increase your memory performance by 10% if it will burn out 50% sooner than if it were not overclocked?

No, this is not sarcasm, I really don't know the answers and am genuinely curious.

Re:Does it really matter?

[enigma48]

Well, Obsessive Math Freak's reply was a waste.

In most cases, you're right - overclocking isn't useful. The edge cases where it would help tend to *require* better accuracy/stability, which you risk when you overclock. Manufacturers have expensive equipment to benchmark their own products so they can sell them at the highest speed, I really hope people don't think burn-in testing comes close to what manufacturers do. About the only time you could really luck out is if a manufacturer decides to under-clock a bunch of chips because they can't tolerate an inventory shortage at that pricepoint/speed in the market.

The CPU, RAM, or GPU-speed limited programs I can think of would be video/audio rendering, gaming, heavy math/scientific work, etc. The only one people would "allow" errors in would be games.

Even if you overclocked everything 25% (a *very* nice overclock for the majority of equipment, but enthusiast/high-end stuff tends to overclock better) you normally won't see a 25% performance increase. Memory performance up by 20%? Great - in most cases, you'll see around a 5% general improvement. CPU performance is largely irrelevant these days, 25% won't get you much. GPU overclocking is the only thing that makes much sense - 25% overclocking will get you approximately a speed bump of the same size.

Big-spenders and media companies are the only ones who are thoroughly concerned with overclocking. Everyone else just gets by with their 5-second Word startup instead of 4.6s. Jocks are picky about their equipment, musicians even more so, and many computer users like tweaking what they've got.

Seems to be a pretty "basic" psychological condition, to tweak, experiment, etc. No one cares about the speed, they care about the bragging, the learning, and the process of doing it themselves.

Re:ECC?

[miller701]

I don't know about ECC (or even just parity) support in current desktop chipsets, but shouldn't overclockers who are worried about reliability (and all of them should be) start using ECC memory so that they are at least likely to notice when they are going too far?

As I remember it, ECC is signifcantly slower than non-parity. Parity RAM is pretty much non existant, and has been for quite some time.

From my perspective, most overclockers seem to be only worried about going from 80 fps to 83 fps in their First Person Shooters, I don't think they really care that much about stability.