Dual Channel Memory Shootout

MDT48 writes "Ever wondered if that expensive low latency memory was worth the cash? These guys have rounded up almost every memory module out there and hammered them. Must have taken them ages, and takes almost as long to read, well worth the effort though."

Comments

Results

[Anonymous Coward]

For those that don't have time to read 33 pages, or even 1 page:

The Winners

• Value Editor's Choice - TwinMOS PC4000 CL2.5
• Value Recommended - Buffalo PC3200 CL3
• High-End Editor's Choice - Corsair TwinX1024-3200XL Pro
• High-End Recommended - OCZ PC3200 Platinum Dual Channel

Re:Results

[Anonymous Coward]

Thank you. I'm on page 4. Of 33 -- why can't they give me ALL at once easily?

It's dead Jim.

Re:Results

[Anonymous Coward]

Thanks.

Only 9 comments posted, and they're /.'d already :|

Re:Results

[randyest]

They did.

Re:Results

[Cyno01]

Good to know, i just bought a 1GB stick of what the review says is "some of the fastest PC3200 memory on the planet". Sweet. I'm gonna go wait for UPS...

Re:Results

[randyest]

Super! But note that the article is old, or at least a bit behind the times (it takes time to compile, plot, and publish all that data!) Newer, faster RAM is already available and can allow even tighter RAM timing (2-2-2-2 anyone?) and more overclocking (my DDR500 runs stably at 325MHz FSB.)

For example, you can get PC4000 (DDR500) @ $259/GB, PC4200 (DDR533) @ $283/GB, PC4400 (DDR550) @ $314/GB, PC2-4300 (DDR2-533, if you happen to have a DDR2 mobo, the others above will all waor in any DDR mobo) are all available cheaper/MB than the high-performers reviewed.

What'd you pay for that "fastest PC3200 memory on the planet" and let me know how fast you can push it (and be stable) once you get it?

Re:Results

[Kulaid982]

It's a shame that the parent posted anonymously, because I believe this is the first time that I've ever seen a first post where the poster has actually read the article!


Note: My hoping that this will be modded "Insightful" will probably only result in this post being modded "Karma Whore"

Re:Results

[Anonymous Coward]

I'm a subscriber, but didn't want to karma whore.

Re:Results

[Cyno01]

I didn't buy the ram based on this article, just a strange coincidence, and you are correct. But Corsair's XL 2-2-2 memory isn't avalible in 1GB sticks yet. This is what i got. I dont mind paying a little more for something thats fast without overclocking(=fast and stable).

Re:Results

[Cyno01]

Oh, and my MoBo only supports DDR400.

Re:Results

[timts]

wow, thanks for the summary, funny enough,corsair and OCZ stuff are not all that expensive, not the platinum version though.

Re:Results

[randyest]

Good choice. Probably the best DDR400 RAM out there (for low latency.)

But, unlike some PC100 mobos that won't work with PC133, in the DDR SDRAM realm, it's always backward compatible (except for new DDR2, which is an all new interface protocol.) If your mobo/CPU works with DDR400, it will run the same (or slightly faster, assuming you're not at 2-2-2-2 with DDR400) with DDR500/533/550+

Your good Corsair will probably do the tightest timing your mobo can support, but it's about $100 more than DDR500 that definitely will.

Just FYI.

Re:Results

[homer_ca]

That's because the highest speed officially supported by motherboard chipsets is DDR400. Anything higher is overclocked and unsupported. Of course that doesn't mean that it can't run an overclocked memory bus reliably, and the DDR RAM itself can be rated to run at higher speeds. High end video cards have run faster memory bus speeds for years. If you mean your BIOS menu has not overclocking options for CPU or memory bus, that's a limit of the motherboard, but you still might be able to overclock through software like with Speedfan.

Re:Results

[Cyno01]

I'd rather not overclock, i'm paying a bit more for high speed and stability. I'm also going for the quietest air cooling avalible, fanless vga cooler, hard drive coolers, a huge cpu heatsink and only 2 case fans. I could have gotten an XP2600 and overclocked it to 3200 speeds, but i paid more for the 3200 because i know it'll be stable at 3200 speeds. If i ever need the power the overclockability is there, but for my dorm room i want it as fast and quiet as possible without sacrificing stability.

TrustedReviews?

[angst7]

Never heard of them. But they sound reliable!

Re:TrustedReviews?

[Nasarius]

they sound reliable!

Maybe, but their server sure isn't.

Re:TrustedReviews?

[Anonymous Coward]

Never heard of them.

Judging by the speed of their webserver, it will most likely stay that way.

- Seth

Re:TrustedReviews?

[Anonymous Coward]

Maybe, but their server sure isn't in Japan.

Re:TrustedReviews?

[PhrostyMcByte]

Hey, if it's anything like Trusted Computing, it has to be good!

Re:TrustedReviews?

[randyest]

I don't get it. Why Japan?

Re:TrustedReviews?

[tarunthegreat2]

It's one of the older memes that are floating around out there (shows the slashdotter belongs to a different era) :-). This might help!

Re:TrustedReviews?

[Ari_Haviv]

In other news...forthcoming Microsoft Windows Longhorn to be renamed Microsoft Windows Trusted Edition.

Now Nobody will RTFA

[IcEMaN252]

Most people don't even read the short articles!

Re:Now Nobody will RTFA

[eddy]

There's articles?!

Re:Now Nobody will RTFA

[roman_mir]

Some people read?!

Re:Now Nobody will RTFA

[Anonymous Coward]

There's articles?!
Yep, the Articles of Configuration.

painful to read waiting for pages to load

[Anonymous Coward]

i often run fbsd and the amount of processes i have running always fit into memory. Is that not the real goal behind running a fast server?

I know faster memory would help, but does it help that much when you follow the advice above?

Also, does anyone run a machine with > 500 processes using 20MB footprints?

l8r

Re:painful to read waiting for pages to load

[shaitand]

"I know faster memory would help, but does it help that much when you follow the advice above?"

It matters MORE not less. When you have swapping or vm going on your bottlenecked by the harddrive. When everything is in ram your bottlenecked by the Ram.

That's why if you increase the memory in a linux or bsd system you tend to see a much more dramatic performance increase than a processor upgrade usually yields (unless it's a several generation jump, like from a p1 200 to a p4).

The processor is much faster than the memory, since you've eliminated the hdd as a bottleneck (and everything else, the memory has a direct channel to the cpu), every increase in memory speed = actual performance increase.

Re:painful to read waiting for pages to load

[JollyFinn]

Half true. Amount of memory is important, relative memory speed, negligle after certain point... Well there is little problem. There is two kind of memory performance one is latency and one is bandwith. Lower timings don't give much in either way. A64 is faster because it gives 50% reduction in latency not some singledigit% fast modules gives over slower. The memory performance is improving PROCESSING performance, and processor matters more if you have REASONABLE memory speed=dual channel Remember Ahmdalls law, Time= Timenotaffected+Timeaffected/percentageofimproveme nt. Now if you hit cache most of the time. On the other hand some applications are limited by memory bandwith so that only way to improve them is adding improving memory, while processor means nothing, on the other hand MOST applications benefit from processing power.But remember to consider about % of improvement. 300Mhz in processor at 3Ghz is 10% improvement and means about nothing in responsiveness. Adding other processor, while getting slower model probably means that while a badly coded application locks your one processor the other processor can continue working for your tasks. Wan't fastest memory system. Go for dual opteron. ~200$ a peace, get ~200$ MB that has 2 channels per processor available [Some dual opterons boards use only memory controller on ONE of the processors see the layout to determine if they use both processors memory controllers]. You get more benefit from dual processing and get more than twice the memory bandwith than the getting 2 channel P4 system with fastest available memory. Id say amount of memory and dual processing means more than the few % of improvement the SPEED of memory gives.

Re:painful to read waiting for pages to load

[the chao goes mu]

The parent is a good example for those who claim punctuation and formatting are not important. (Also, if you are not going to use underscores in variables made up of multiple words, please use capitalization. It took me longer than it should to figure out what timenotaffected is...)

RAM Speed Differences

[HitByASquirrel]

Anyone know if 4 512MB PC3200 DDR 400 chips is faster/slower than 2 1GB PC3200 DDR 400 chips?

Re:RAM Speed Differences

[lessthanjakejohn]

Slower

Re:RAM Speed Differences

[Chuck Bucket]

Same

Re:RAM Speed Differences

[Anonymous Coward]

Faster

Re:RAM Speed Differences

[randyest]

Why do people make wrong claims in such a matter-of-fact manner? Is it ignorance or malice?

Whatever, more modules in general leads to slower access time. More sticks (modules) of RAM adds to the capacitance driven by the memory controller, and therefore increases the propagation delay on those traces. The more connected modules the longer total trace length driven by each IO buffer of the memeory controller. The difference is probably negligible for all but the edgiest overclocker, but there's always a speed advantage for fewer modules.

I did a quick google and found this example to back me up, but it's not really needed as you clearly have no fucking idea what you're talking about and could not cite one single reputable argument to back up your insane claim. One of your friends spewed that nonsense and you're parroting it because it sounded reasonable. Of course, that's only because you have no fucking idea what you're talking about.

Re:RAM Speed Differences

[Mindtoy]

Don't drive angry!

Re:RAM Speed Differences

[Anonymous Coward]

fnord=ehowa?

Re:RAM Speed Differences

[Detritus]

It may not be slower, and may well be faster, if you have a multi-channel memory controller. YMMV. BMSMA.

Re:RAM Speed Differences

[Lord_Slepnir]

I fully agree with you. It's times like this I wish I had a ( -1: talking out of his ass) mod option.

Re:RAM Speed Differences

[Anonymous Coward]

Could it be that parent misunderstood grandparent's extremely brief comment? Could grandparent be referring to the fact that with multiple, independently addressed modules, the memory controller can overlap accesses, leading to vastly improved memory bandwidth (and, given prefetch, latency) in the common case of accessing sequential memory locations? Could it be that grandparent is a system architect, and parent is a hardware engineer?

Nahh. Clearly grandparent is insane, and spewing friends' nonsense. And don't even ask me where the sexual intercourse comes in...

Re:RAM Speed Differences

[Anonymous Coward]

4 DIMMs would be faster than 2 if you could access them in parallel, which would require 4 separate memory busses, or if you overlap accesses on the same bus. However, DDR is divided into multiple independent banks, so any overlap that can be done between multiple DIMMS can be done within one DIMM. Ergo, no -- there is no speed gain from multiple DIMMs in a standard PC.

Re:RAM Speed Differences

[Anonymous Coward]

But aren't you forgetting that with larger memory modules, you have more beebees and thus the Flux capacitor in the memory controller has an increased magnetic field in conjunction with an increased hierachical structure of "its" cosmos and thereby the timing is increased? My reference here ;-)

Re:RAM Speed Differences

[randyest]

It may not be slower, and may well be faster, if you have a multi-channel memory controller. YMMV. BMSMA.

No, multiple modules will never be faster -- a mulit-channel controller will be the same with one or two modules; slower with more than two (which is possible.)

More modules is never faster from a memory access timing standpoint which is, again, what this discussion is (or should be) all about. You can, of course, increase overall system memory bandwidth by adding one each memory controller and SDRAM module, but that's irrelevant here (even in that case, you'd want faster RAM so that the CAS, RAS, and other latencies can be minimized and bandwidth increased a bit more.)

Re:RAM Speed Differences

[HitByASquirrel]

This would be in a Dual 2.5Ghz G5.

If im reading the block diagram correctly, it has only one 400Mhz memory bus, yet the chips must be installed in pairs (symmetrically with the top and bottom groups of 4 slots), something I still do not understand fully. This is described here.

So I assume I should go with whatever is cheaper and works with my expansion plan. In that case, the four 512MB PC3200 chips.

Thanks for the input everyone, I appreciate the insight.

Hmmm... I wonder... would the two dissipate the same amount of heat?

Re:RAM Speed Differences

[randyest]

Maybe, but that'd be missing the point entirely, since this article is about RAM access timing -- comparing different RAM brands and types in the same system.

Everyone who is paying attention (or thinks for a second) knows that adding memory controllers to the CPU can increase memory bandwidth (it's 2 pipes instead of 1, after all), and this increase is more than the gain found in overclocking your FSB (front-side bus, or memory controller) from 200MHz to 210MHz or going from 3-3-7-11 access timing to 2-2-2-2.

This article is all about how fast you can run your FSB and how tight you can set the timings for various brands of RAM. Not how much more memory bandwidth your system can get by adding memory controllers/channels. That's a CPU upgrade, not a RAM swap. Wrong topic.

Re:RAM Speed Differences

[Concerned Onlooker]

Why do people make wrong claims in such a matter-of-fact manner? Is it ignorance or malice?

It's learned behavior from watching too many political ads.

Re:RAM Speed Differences

[randyest]

ehowa==funny

ehowa!=fnord

But now I have to kill you :)

Re:RAM Speed Differences

[randyest]

I'm cool (and home) thanks. But somehow I just got +5 for telling a guy that he has "no fucking idea what he's talking about." Twice.

It was true, of course, but it's still kinda funny.

Re:RAM Speed Differences

[OneSeven]

seriously, put the flamethrower down, chill.
Next time, read the question before blasting away in a mad frenzy. The poor guy was just asking an inoccent QUESTION, not making an ill-informed statement as you seem to think. Go back, read it again..... think for 5/8th of a second......
don't ever do it again.

Re:RAM Speed Differences

[Animixer]

I agree with you for the most part....however, what if you had 64 dimms (60ns) that are 8-way interleaved like many of the machines I use? Certainly imagining a trace through all of them would have incredible capacitance....but I think the interleaving avoids this by using 1 module per bank of 8 and sort of 'stripes' the data across the dimms.

I need to read up on this further from an engineering point of view. :)

Re:RAM Speed Differences

[Anonymous Coward]

But aren't you forgetting that with larger memory modules, you have more beebees and thus the Flux capacitor in the memory controller has an increased magnetic field in conjunction with an increased hierachical structure of "its" cosmos and thereby the timing is increased?

That was true up until recently, but todays triple phase inverted wobulators cancel out ALL the capacitance via pre-linked kernel module dependancies being globally optimized in the 5th dimension.

I don't know what those dorks were thinking when they were researching along these lines in the 4th dimension. But I guess, "benefit of hindsight", right?

I am going to wait until ATI release the 7th dimension wobulators though... then I will r0x0r y0u @ll!

Re:RAM Speed Differences

[ThaReetLad]

slower, because of increased latency. Also memory timings usually have to be decreased in order to maintain signal integrity over the longer distance and through the first bank of RAM to reach the second, and that assumes that the 4 sticks are on a dual channel system. If you only have a single channel then many boards will really struggle to maintain good signal integrity to 4 banks of RAM at fast timings, or even at DDR400.

Re:RAM Speed Differences

[_Eric]

Amen!

Mod this one to the sky, as even the posting mixes up bandwidth and latency.

News for wannabe nerd...

Re:RAM Speed Differences

[dj245]

The more connected modules the longer total trace length driven by each IO buffer of the memeory controller. The difference is probably negligible for all but the edgiest overclocker, but there's always a speed advantage for fewer modules.

While you are correct that the parent poster is indeed full of shit, on many motherboards dual-channel memory is not enabled until 2 sticks of memory are inserted. If there is only one stick, there is only single-channel memory. The Nforce 1 chipset operates like this, and many other new chipsets with dual channel memory as well. With these boards, you really should have two sticks.

Re:RAM Speed Differences

[EulerX07]

You're aware you're setting a dangerous precedent by requiring slashdot posters to know what they're talking about, right? If this concept catches on, this place will be a ghost town, with only four posts per topic, all rated +5, informative.

Re:RAM Speed Differences

[Anonymous Coward]

While you are correct that the parent poster is indeed full of shit, on many motherboards dual-channel memory is not enabled until 2 sticks of memory are inserted. If there is only one stick, there is only single-channel memory. The Nforce 1 chipset operates like this, and many other new chipsets with dual channel memory as well. With these boards, you really should have two sticks.

The nForce1 chipset is also notoriously fickle about what makes/types of RAM it will accept when you use two memory modules. If you can't use the *exact* make and model listed on the manufacturer's support page, then you'll probably have to reduce your memory timings if you want a stable system.

Took me a few months to figure out that:

A) Neither of my two memory modules were defective

B) Using both of them at the same time always led to instability, but using either alone worked perfectly

C) MemTest86 is not perfect (it failed to show that having 2 in at the same time didn't work) and the Prime95 is a better tester of timing issues.

D) Backing off my CL refresh value in the BIOS (from 2.0 to 2.5) worked wonders for stability

(Symptoms were frequent reboots / BSODs in Windows on my A7N8X motherboard, without any overclocking.)

Re:RAM Speed Differences

[hamanu]

No, multiple modules will never be faster -- a mulit-channel controller will be the same with one or two modules; slower with more than two (which is possible.)


you seem to be confusing the concept of "multi" with the concept of "dual". My computer has quad memory channels. In my computer of course 4 will be faster than 2, not in latency, but in data transfer rate ("bandwidth").You need BOTH, I guess some riced-out overclockers might have missed the fact that MHz isn't everything.

Re:RAM Speed Differences

[randyest]

No, I realize there are mobos with more than 2 controllers (though not very common), but you're right that I should have said "Dual" in my first sentence.

However, if you scroll back up the thread and see the post that started all this, you'll note that the claim was "by separating out the memory into an array of smaller modules, you allow faster memory access." That is, for the same system (i.e., same number of memory controllers), this guy claims more modules will provide faster memory access.

This, of course, is fasle, as has been established. Then someone else chimes in with the claim that memory access time "may not be slower, and may well be faster, if you have a multi-channel memory controller." Which is also false, as this poster (along with the original article itself) confuses latency (memory access timing) with bandwidth (total memory throughput.)

To put this to rest, hopefully:

More memory controllers or channels paired with one module each is the fastest for both latency and bandwidth. As you add memory controllers (assuming only one well-matched module each) you get more bandwidth with the same latency (with a few exceptions -- some mobos get slower with more modules regardless). As you add more modules to each controller, you get more latency, but still roughly the same bandwidth (just slightly less.)

memory access time : memory bandwidth :: ping : network bandwidth

Re:RAM Speed Differences

[mollymoo]

Whatever, more modules in general leads to slower access time. More sticks (modules) of RAM adds to the capacitance driven by the memory controller, and therefore increases the propagation delay on those traces. The more connected modules the longer total trace length driven by each IO buffer of the memeory controller. The difference is probably negligible for all but the edgiest overclocker, but there's always a speed advantage for fewer modules.

Surely it's potentially slower, not necessarily slower. More specifically, you won't be able to push the timings as far with several sticks as you could with just one. If it works with several sticks at at a particular set of timings then surely it will be equally as fast as with one stick with those same timings. Please correct me if I'm wrong, but that seems logical to me.

Re:RAM Speed Differences

[randyest]

Surely [more RAM modules is] potentially slower, not necessarily slower.

No, it is necessarily slower. Maybe so small a speed difference that the available mobo RAM timing options aren't fine-grained enough to even tell the difference, but it will always take longer to propagate a signal from a driver to two loads (2 RAM modules), even if the modules are right on top of one another with no difference in trace length -- the second module adds to the input capacitance of the load seen by the driver, and all CMOS technology devices have delays that increase with load. If you add a buffer to split the signal, you indeed lower the load seen by the driver (one buffer), but then you add the propagation delay of that inserted buffer. Result: slower for a different reason.

More specifically, you won't be able to push the timings as far with several sticks as you could with just one.

Er, that's what I said. If you meant the reverse -- no.

If it works with several sticks at at a particular set of timings then surely it will be equally as fast as with one stick with those same timings. Please correct me if I'm wrong, but that seems logical to me.

Again, that's what I said. Right. I'm confused now -- you start off diagreeing, then repeat what you (seem to) disagree with as correct. AM I misunderstanding you?

Re:RAM Speed Differences

[Anonymous Coward]

you read it again, asshat

and learn how /. works before spewing nonsense out of your gaping maw

Re:RAM Speed Differences

[fitten]

This is a good explanation for unregistered/unbuffered DDR. Doesn't necessarily hold for registered/buffered DDR though.

Re:RAM Speed Differences

[randyest]

Thanks. And you are indeed correct sir. But the context of your comment puzzles me, seeing as how this article, the reviews and benchmarks, and every thread I've seen here refer to unregistered, unbuffered DDR SDRAM.

I guess we should also point out that it doesn't apply to EDO, FastPAGE, VRAM, VirtualChannel RAM, SRAM, FLASH, or any others either. Just to be thorough, you know :)

Re:RAM Speed Differences

[mollymoo]

Surely [more RAM modules is] potentially slower, not necessarily slower.

No, it is necessarily slower. Maybe so small a speed difference that the available mobo RAM timing options aren't fine-grained enough to even tell the difference [...]

I get that it will take the signals longer to propagate, but if you still get your signal within one clock cycle then surely it just doesn't matter from a performance point of view. You could see it with an oscilloscope, but not a benchmark. You seemed to be suggesting that more sticks automatically led to slower PC performance, rather than a reduced overhead for overclocking. Perhaps you were talking about a different "slower". But then...

If it works with several sticks at at a particular set of timings then surely it will be equally as fast as with one stick with those same timings. Please correct me if I'm wrong, but that seems logical to me.

Again, that's what I said. Right. I'm confused now -- you start off diagreeing, then repeat what you (seem to) disagree with as correct. AM I misunderstanding you?

I think we must be misunderstanding each other, because we agree on the above, which is for the 90% of computer users who run at stock timings (which will likely be fine no matter how many sticks you use) the important point.

Memory review cache?

[Bill_Royle]

I suppose asking for a cached copy of this article would be a bit too ironic, right?

Re:Memory review cache?

[swordboy]

Well, check out the article text once again:

Ever wondered if that expensive low latency memory was worth the cash?

Shouldn't that be cache ?

Re:Memory review cache?

[Anonymous Coward]

I suppose asking for a cached copy of this article would be a bit too ironic, right?

No. You flunk irony school.

Mid Line Recommendation

[Dominatus]

GEIL Not only do they have GEIL placing well in the middle of the lineup both in price and performance, but I have noted that it offers the best of both worlds. It doesn't cost nearly as much as the high high end stuff, and while it doesn't perform *quite* as well, it's still better than the low end stuff. GEIL presents a good opporutinity to chose between Value and High End

Re:Mid Line Recommendation

[darkain]

i second that. i personally run geil in my main machine (this one that im using right now).

i was a bit worried about heat tho when i first got them, because they are 216 (433)mhz chips w/o heat spreaders, but what they did with their golden dragon series was nice, altho the red eyes do clash with the blue themed computer to create an odd "purple haze" effect in the middle of the system.

Re:Mid Line Recommendation

[Dominatus]

Value RAM doesn't mean best value, it's a nice word for "cheap RAM".

Re:Mid Line Recommendation

[Dominatus]

Oh, in reference to the articles use of the word Value...sorry Yeah, they had different picks for perf vs price, but I stick with GEIL, all of the stuff I've bought last, and run well.

Re:Mid Line Recommendation

[Merovign]

My GEIL Gold Dragons work pretty well, and they are measurably faster than the cheap stuff. I'm not quite into high-end yet.

I do know the Gold Dragons make hash of the Kingstons in my machines (Kingstons work fine, just not as fast). And the price was right.

Re:Mid Line Recommendation

[Anonymous Coward]

Any memory brand who's name in Dutch and German means horney should be good.

Re:Mid Line Recommendation

[Anonymous Coward]

Grrrrr. The parent asked a simple question. "What does Best value mean?"

Answer: They split the memory at an arbitrary price point per MB (20 pence/MB IIRC)and the best performer in that group is Best Value.

33 pages

[pvt_medic]

Well since it is a 33 page read, how about the link to the end where the conclusions are.
Conclusion
The Winner

and for those too lazy to click
Conclusion
So there we have it, 18 different types of memory benchmarked to within an inch of their life and to prove what? Well, one thing we've proved is that while even value memory may offer sustained levels of high speed operation, when you want to make it to the scary end of the spectrum above 250MHz you generally need to pay the price premium associated with "enthusiast" modules. Cheaper stuff may get you close, but at the very high-end we're afraid you really do get what you pay for.

If you're happy to settle for fast rather than fastest, it seems that reputations and price tags count for little in this game. Good chips on a poorly designed PCB and poor chips on a great PCB will both leave you wanting. And even if you have the best and fastest memory on the planet you still need a motherboard that can do it justice.

Evaluating performance at more reasonable frequencies is slightly easier, but also slightly fuzzier. Some of the benchmark results varied so widely that it was hard to understand what was really going on, and with various tests favouring different attributes and the surprisingly similar stock performance from many of the modules on test, it was almost an exercise in identifying the "Top Dogs", the "Turkeys" and "The Rest".

We hope you enjoyed this roundup and gleaned at least some information from the effort that went into it. As we hope you've seen, or will see when you glance through the benchmark results, memory is a very complex subject and pinning it down in performance terms isn't as simple as you might imagine. We won't let that stop us trying though.

Finally we'll pick out some winners. Remember that for our purposes, overclocking is more a pleasant bonus than an essential feature, and though we have made allowance for overclocking performance it hasn't been given the same weight that it might have been given were we writing this roundup purely for enthusiasts.

Finally, benchmarks do a great job of highlighting strengths and weaknesses in products but it's important to keep things in perspective. Most of us would not notice any tangible performance differences between even the fastest and slowest modules on test here when run at the processor's default speed unless we were specifically looking for them. That's not to say that faster isn't better, it's just not as critical as some would have us believe.

The Winners
Value Editor's Choice - TwinMOS PC4000 CL2.5

Concerns over availability make us a little nervous about handing the value crown to TwinMOS, but we can't argue with the numbers and this is quite simply the fastest memory for the money that we had the pleasure of testing. We only hope the poor availability isn't due to production constraints and look forward to seeing more retailers pick up on what appears to be a very nice product.

Value Recommended - Buffalo PC3200 CL3

We may have had little experience with Buffalo before this roundup but both of the Buffalo modules we were sent acquitted themselves well in testing. For the most part, performance levels in the value group contenders were quite evenly matched but Buffalo did enough to scrape onto our leader board. A very impressive all-round showing earns Buffalo a well-deserved Recommended award.

High-End Editor's Choice - Corsair TwinX1024-3200XL Pro

Perhaps the toughest choice of all, we eventually opted for Corsair's 3200XL Pro for its shear arrogance. While far from cheap, it consistently set the standards for stock performance while flashing a knowing LED at the competition as it did so. For a change we see a product with a bite every bit as bad as its bark.

Hig

Re:33 pages

[Anonymous Coward]

What's so special about posting or not posting as AC? You can always mod up or down depending on weather you feel a copy of /.'ed article is useful. What do you care about author's motivation or amount of effort/intelligence if you enjoy the result? Your post, on the other hand, is offtopic. So is my, so I am not taking chances.

Slashdot editors, on the other hand, should always provide a cached copy of the article as a matter of social responsibility.

Re:33 pages

[Anonymous Coward]

If I had mod points.....

He posted the whole of the conclusions which added much info. Since the site is /.ed it was a + informative post.

You, judging by the tone of your posts here, are up way past your bedtime, nighty nite!

/.ed

[Quixote]

Their server could sure use some of that memory now...

Re:/.ed

[rd4tech]

And the winner of our favorite /. game, called "where goes the server", for today, is...

nice

[Chuck Bucket]

I've got some old OCZ (PC150!), but I've always liked them, their chips and their service. Cool to see them win the 'high-end' though not a surprise, I think they sell the best RAM out there today.

Anytime you see a 'smaller guy' win something, it's always a nice thing.

P

Re:nice

[lessthanjakejohn]

I've heard some bad things about OCZ being hit or miss. Can anyone fill me in on the deal about OCZ?

Re:nice

[randyest]

So, are you going to tell us from whence cometh this "understanding" or, by your silence, will you implicitly admit that you made that up, or heard it from a friend-of-a-friend, and actually have no fucking clue?

Your move! ;)

Re:nice

[randyest]

Thanks for replying. Not much there (silly allegation that OCA re-brands memory chips, which all SDRAM OEMs do,) but it is something, which I said it wouldn't be.

I take back my smart-ass vitriol :)

Re:nice

[irokitt]

They still owe me 512MB of RAM. I sent in 1 GB of defective RAM, they only sent back 512 MB. No joy getting that memory back from them, so they're on my shit list. That said, the RAM they did send back was remarked Samsung, very good chips (the chips I sent in were generic, godknows who made them). But I'll stick to Kingston next time.

Re:nice

[Tassach]

But I'll stick to Kingston next time.

Smart man. The questions you you should be asking when buying memory isn't "how fast can I overclock it", but rather "how stable is it" and and "how good is the vendor's quality control and customer service". Kingston wins hands down on these counts.

As for the tests in the article, they are flawed because they only tested one sample for each vendor. A single sample is not statistically significant -- you have no way of knowing if you just got lucky and got a particuarly "hot" unit, or got unlucky and got stuck with a dog.

Overclocking is overrated and is pretty much unnecessary anymore. Overclocking a P-75 to 100MHz, or a Celeron 300A to 500MHz made sense back in the days of Socket7 and Slot1 -- you could often O/C CPUS of that generation to 150% or more over their rated speed, which made a massive difference in observable system performance. Given my experience with the CPUs which are currently available, you're lucky to find one which will at run more at more than 110% rated speed -- and they are so fast to start with that you really won't notice a significant difference between a system running at 2.0GHz and one at 2.1GHz. Nowadays it actually makes sense (sometimes) to underclock the CPU to reduce heat and noise.

Re:nice

[randyest]

Not sure why that would matter. Neither OCZ nor winbond actually make DRAM chips. In either case they're made by Samsung, Elpida, UMC, or one of the other (relatively few) companies that own the insanely expensive ($billions) chip fabs. OCZ, winbond, Micron, Crucial, Mushkin, Kingston, etc. merely buy the chips and make SDRAM modules out of them.

It's not uncommon for one to buy spare stock from another (it usually benefits both -- the co. with the extra chips unloads excess stock and gets cash, the co. that buys them gets to fill outstanding orders and make important customers happy.)

It's also not uncommon for the chip fabs to change markings on the parts after production, especiallly parts that are the same for many customers (contrast with ASICs, where this is unheard of.) We (I work for a semiconductor company that fabs chips, including DRAM) may make a million chips with custom marking for "foo company," then if Foo Co. cancels or reduces an order, we'll charge them a fee to cover the re-marking (usually) and recoup some money by selling them to another company.

No harm no foul. If they changed the marked specs rather than the name, that might be interesting. This isn't.

Re:nice

[randyest]

So tell me, why do one company's ram perform better then anothers? or is it in the quality of the rest of the components as well?

Off the top of my head, in order of impact on performance (and cost):

• Speed rating of the chips (there aren't many manufacturers, but each has several different speeds available, and faster costs more, of course.) Often chips are made and tested until they fail, then "speed-binned," or grouped by max performance, and sold accordingly.
• Printed circuit board (PCB) material and layer count. More expensive epoxy / resin materials (like FR4) allow tighter and more consistent trace etching, allowing tighter design rules and more efficient and matched routing. Number of layers and ground/VDD planes affect potential quality as well, and cost of course.
• PCB design and trace routing. There's more to an SDRAM stick than just slapping on some chips. It's circuit design and, as such, there are better and worse designs. Moreover chip layout, trace width and length, placement and quality of passive components, etc. all affect performance. This is linked to PCB material above, since better material allows tighter design.
• Quality control and testing. The best design in the world can be screwed up by process and you'll never know until you test. More expensive RAM OEMs test every part over a wide range of voltage and temperature using expensive machines. More testing ensures better parts (or at least a better understanding of how good each part it, and thus how it should be marked and sold) but costs lots of money in tester time.

Hope that helps!

Re:nice

[DAldredge]

Please let me know which of my posts made you mark me as a foe.

Thanks,
DAldredge

Re:nice

[randyest]

This one.

I've been ignoring the silly sig, but that's just a huge stretch and one abortion reference too many. Your agenda is (1) too obvious and (2) getting on my nerves.

No offense intended, but you asked :)

Re:nice

[DAldredge]

IOW, you can't respond to what I say so you just block it. It's cowardly, but then again it's your right to be a coward.

I just find it funny/sad that that our elected officials think that games are more of a threat to kids than kids, with out having to notify their parents or a judge, having their babies killed.

Games = BAD
Abortion = GOOD

Now if you will excuse me, I have to get back to the D&D maps I am working on and turn up the metal I am listing to. Hell, later tonight I think I will have a drink. (Opps, there goes what you thought about me right out the window) ;->

Re:nice

[DAldredge]

Micron makes their own RAM. Here is ONE of their fabs, go to micron.com for a complet list.

Manassas, Virginia

Micron's state-of-the-art semiconductor manufacturing facility in Manassas, Virginia, is located approximately 35 miles from Washington D.C. The Virginia plant, Fab 6, is Micron's flagship 300mm manufacturing facility and is representative of the future of Micron semiconductor manufacturing. Currently, Fab 6 is going through qualification and aggressively ramping production.

/.'d already??

[Temfate]

Well I'm glad to see Corsair made it, according to the cheater post. But the site is not responding... Hmm...

Re:/.'d already??

[Anonymous Coward]

They're swapping out all the flavors to see how each handle a real world slashdoting....

Ooh, shiny thing!

[Animats]

Their criteria: "high-shine chrome-like heat spreaders".

This is for the crowd that puts neon lights in their PC. Not the people who buy ECC memory for their desktops.

Re:Ooh, shiny thing!

[MattyCobb]

They are cold cathode not neon... and I like them you insensitive clod!

Re:Ooh, shiny thing!

[Orick]

I'd tell you which stick was the best, but I don't remember. Must have picked the wrong one for my new brain mod....

--
Kirby Reviews

Re:Ooh, shiny thing!

[NickDngr]

They are cold cathode not neon

The lighting device commonly referred to as "neon" is indeed a type of cold cathode lamp. Not all cold cathode lamps are neon, as diefferent colors use different gasses, but neon and cold cathode are not mutually excludive.

Re:Ooh, shiny thing!

[jo42]

Yes, but which one has the sign of ultimate performance Type R sticker on it?

worth the cash?

[Anonymous Coward]

Looking at the benchmarks, in most cases I would say..... no

Single-page version

[Quixote]

Sinle page version of the article here

(just in case someone has the urge to actually RTFA...)

GB Micro?

[rngadam]

Just wondering if anyone can explain what GB Micro memory is? I just ordered some:

"GB Micro 512MB PC-3200 DDR400 SDRAM Dual Channel Kit"

which at 126$CAN was 23$CAN less expensive then Kingston and 43$CAN less expensive then the OCZ... Anyone know what the heck it is and how well it ranks?

Re:GB Micro?

[stratjakt]

Know what, in reality, it all works about the same. All this piddly-dee shit about timings and a handful of MHZ overclock is all about benchmarking and little else.

Re:GB Micro?

[ergo98]

It's generic memory. In the grand scheme of things the difference between the high and low end will be marginal in real world benchmarks (i.e. games, office benchmarks) -- it's only in extreme memory stress tests that any difference at all is revealed.

Though note that you dual-channel kit will likely come with two 256MB modules. You probably knew this, but I mention it just in case.

Cheers.

Re:GB Micro?

[DAldredge]

You have a 3 digit UID and you just ordered ram with out know how well it was made?

Sad, just sad. But, you are from Canada so perhaps that explains things ;->

Re:GB Micro?

[Anonymous Coward]

As someone who works at a computer store in Canada I can tell you that GB Micro is a distributer. Your memory is generic, it could be crap ram (LEI chips) it could be good ram (infinion). You won't know at least untill you see it.

But as others have mentioned, the differences are pretty marginal and in the generic ram segment reliability and compatibilty are much bigger concerns.

Re:GB Micro?

[Anonymous Coward]

Just wondering if anyone can explain what GB Micro memory is? I just ordered some:

Do you want to buy a bridge?

how convenient

[rd4tech]

they put everything into 33 darn pages so it can be mirrored easily...

Re:how convenient

[randyest]

They also put it into one page so that, er, well. I dunno why.

Re:how convenient

[Anonymous Coward]

bash$ wget --no-parent -recursive www.sitename.com/path/to/page1

what's so difficult about doing something like that?

Doom III

[iMaple]

Must have taken them ages, and takes almost as long to read, well worth the effort though.
Yeah .. Tell me what to buy for Doom III

Re:Doom III

[DigiShaman]

Though Doom III will be out soon, unfortunatly you will have to wait a bit longer for Intel's new quantium processor to run the damn thing smoothly.

Maybe this "killer app" (pun intended) will help give the PC industry a shot in the arm it needs. Then again...I suppose spyware is already doing that. Ohhhh the irony.

Re:Doom III

[Anonymous Coward]

Yeah, WTF! No game benchmarks in the review? Why else would you worry about memory latency that much?

Google saves

[rd4tech]

Here's a cached link on google for the first page here

Re:Google saves

[Anonymous Coward]

whoever moded the parent offtopic, have you at least clicked on the link?
The site is /.ed, the parent is posting a GOOGLE cache link.
How can it be offtopic?

Re:Google saves

[Frogbert]

whoever moded the parent offtopic, have you at least clicked on the link?
The site is /.ed, the parent is posting a GOOGLE cache link.
How can it be offtopic?

Did you even click the link?
"ViewSonic VP2290b - High-Resolution TFT"

The only way you could be more offtopic is if it contained gay cowboys eating pudding!

Re:Google saves

[rd4tech]

wtf.. I just copied the first page from google, sorry, I'm in no way selling 5k TVs. Appologies to everyone

Re:Google saves

[Anonymous Coward]

See? That's what you get for karma whoring

Re:Google saves

[rd4tech]

Naah, just the usual effects after 8 hours of coding :)

Great tests, but lacking the new goodness

[randyest]

PC4000 (DDR500) RAM and faster) is here, and it's a lot faster than anything they reviewed. Even if your CPU/Mobo don't support more than DDR200 or DDR333, you can get lower latency (i.e., 2-2-2-2 BIOS timing options will be usable at lower frequencies) and be ready for upgrades using faster RAM.

I just picked up two 512MB sticks of this amazing stuff to go with my AMD64 3200+ from, of all places, Circuit City (because it is only $129 each, cheaper even than newegg, which is usually a great deal for RAM and has amazing reseller ratings.) Though I just noticed there's a new $40 rebate on newegg's $299 price (for 1GB), making the net $259, or about the same as I paid at CC. Then again, I hate rebates.)

No shipping and insta-pick up at Circuit Shitty (though newegg is usally free for fedex saver), even with MA 5% sales tax it's a steal.

Sorry, I digress -- I'm full of gin, tonic, and Italian food. Back OT -- they got their best RAM up to 275MHz FSB (DDR550). The Kingston I got works stable up to 325MHZ (DDR650!!) and it's not much more (a few cents/MB) than the reviewed RAM.

Just thought I'd share all I recently discovered before someone runs out and buys the best of this review. I think it took so long to write all that review text and compile/plot the data that they missed the new goodies!

Re:Great tests, but lacking the new goodness

[momerath2003]

full of gin, tonic

You sure you don't mean jynnan tonnyx or tzjin-anthony-ks?

Douglas Adams, Restaruant at the End of the Universe:

It is a curious fact, and one to which no one knows quite how much importance to attach, that something like 85% of all known worlds in the Galaxy, be they primitive or highly advanced, have invented a drink called jynnan tonnyx, or gee-N'N-T'N-ix, or jinond-o-nicks, or any one of a thousand or more variations on the same phonetic theme. The drinks themselves are not the same, and vary between the Sivolvian "chinanto/mnigs" which is ordinary water server at slightly above room temperature, and the Gagrakackan "tzjin-anthony-ks" which kills cows at a hundred paces; and in fact the one common factor between all of them, beyond the fact that the names sound the same, is that they were all invented and named before the worlds concerned made contact with any other worlds./blockquote

Re:Great tests, but lacking the new goodness

[blair1q]

>Sorry, I digress -- I'm full of gin, tonic, and Italian food. Back OT -- they got their best RAM up to 275MHz FSB (DDR550). The Kingston I got works stable up to 325MHZ (DDR650!!) and it's not much more (a few cents/MB) than the reviewed RAM.

You got lucky. Sometimes low-latency memory doesn't even work at lower than rated speed with more than one module installed if you don't buy matched, pretested pairs.

Re:Great tests, but lacking the new goodness

[randyest]

You got lucky.

Maybe, but DDR500-qualified RAM is guaranteed to achieve better poorest timing (worst-case, of the worst sample of all batches) than DDR400, which was the fastest reviewed in the article. My point is that, given even DDR550 is available, they seem to have completely missed the "sweet spot." For the same price as their best DDR400, you can have DDR500 that will either (1) run at 250FSB if you have or are overclocking to that or (2) give better timing (closer to 2-2-2-2) in DDR200/333/400 systems.

Sometimes low-latency memory doesn't even work at lower than rated speed with more than one module installed if you don't buy matched, pretested pairs.

Yep, that's why you make sure the worst case performace specs meet your needs and are the best value for the buck. You can't predict how your part will fall in the min-max distribution of the batch in question, and returning RAM because you can't OC it to the level you want is dickish, so you have to pick something that is guaranteed to do what you want.

Re:Great tests, but lacking the new goodness

[kesuki]

The ddr 500 you linked to has a cas latency of 3-4-4-8 not 2-2-2-2 so in effect it can pump 25% mor data at 1/2 the speed in cycles consumed. so in the time a ddr 400 can have completed 800,000 operations on it's memory, the pc 500 will have only performed 400,000 operations, but the volume of data it could have performed operations on is still going to be 25% greater, so the 400 mhs could have performed 800,000 operatrions on say (just for arguments sake) 128 MB of memory, in the time the 500 could have performed 400,000 operations on 160 MB of ram...
better thruput, but inferior turnaround time on operations. this is why 2-2-2-2 ram is WORTH the money. most people aren't block erasing 10 GB of ram ever say ten seconds. Google might be...somme high end photoshop filters could possibly push that much data through the ram... but in terms of overal performance, for gaming and system responsiveness ONLY certified 2-2-2-2 latency ram is worth paying extra for.
you could be paying tripple the cost of pc 4800 and STILL be getting a better bargain, because these modules are Tested and gaurenteed to run at 2-2-2-2 latency while it's a total crap shoot trying to run a 3-4-4-8 timing module below that, and you CANT return it because it won't 'overclock' to 2-2-2-2 timings when they only gaurenteed it to run at 3-4-4-8.

By the way, of course low latency ram needs to be matched, unless your motherboard supports independant timining for each individual ram slot, all ram sticks must be gaurnenteed to run at a 2-2-2-2 timing to be run at a 2-2-2-2 timing. It's overclocking to try to run a 3-4-4-8 module at 2-2-2-2.
cas latancy is the true test of the speed of ram, and has been for a Long long time. there haven't been many programs that truly stress thruput capabilities of ram since we started using pc-100 and better ram modules.

Re:Great tests, but lacking the new goodness

[randyest]

The ddr 500 you linked to has a cas latency of 3-4-4-8 not 2-2-2-2 so in effect it can pump 25% mor data at 1/2 the speed in cycles consumed.

Do you know what the units of those numbers are (3-4-4-8 and 2-2-2-2)? Judging rom the rest of your post it seems you think they are time units, such as 3ns/4ns/4ns/8ns or something. That's not the case, though. The units are clock cycles. And a clock cycle at 500MHz is a shorter time interval than a clock cycle at 400MHz.

You seem to miss that critical issue entirely, and base your conclusions on an unwarranted comparison of 400MHz clock cycles and 500MHz clock cycles (sorta like apples and oranges in this case.)

Re:Great tests, but lacking the new goodness

[blair1q]

[3,8] clock cycles at 500 MHz = [6,16] ns
2 clock cycles at 400 MHz = 5 ns

[6,16]ns/5ns = [1.2,3.2]

The 500-MHz 3-4-4-8 takes between 20% and 220% longer than the 400-MHz 2-2-2-2 for comparable accesses.

So if your point was that the 500-MHz 3-4-4-8 couldn't be slower than the 400-MHz 2-2-2-2, then it was wrong.

Comment removed

[account_deleted]

Comment removed based on user account deletion

Comment removed

[account_deleted]

Comment removed based on user account deletion

A RAM Primer

[diagnosis]

And if you would just like a short introduction to what the heck RAM speed means, check out this excellent Arstechnica.com article:

FSB and memory speeds

---------------------
Freedom or Evil: Freevil.net
G. W. Bush says, "You decide!"

It's all about the CAS timings

[Thaidog]

The better the cas timing the better the ram... Also more sticks in the banks cases more latency... if you can do one stick 512 vs 2 256 it will give less sytem latency... I wonder what dual channel ram does to system latency... if there is more because 2 sticks have to be addressed...hmmm?

Re:It's all about the CAS timings

[Dominatus]

The fact that the bandwidth is increased on a system that can utilize it with dual channel obviously completely negates any latency issues.

Re:It's all about the CAS timings

[Thaidog]

Don't post shit like this if you don't have anything to prove it with. If you're going to say something like that, you need to have an example.

Re:It's all about the CAS timings

[Dominatus]

I need an example for something so blatantly obvious? Think about it. Dual Channel effectively doubles bandwidth if the system can handle it. Now, we all know that it doesn't achieve those results 100%, but it still increases bandwidth by a decent margin. Take for example a Pentium 4 with an 800 mhz (200x4) bus. It has the capacity for 6.4 gigs of bandwidth, but PC3200 RAM single channel only offers at most 3.2 gigs of bandwidth. Switching to dual channel allows a theoretical 6.4 gig of bandwidth to now be available. Of course, results don't achieve their theoretical limits, but that's true for everything. Latency issues are almost non existent, and certainly don't override even 1 gig of extra bandwidth.

Re:It's all about the CAS timings

[Anonymous Coward]

Dude, dual channel cann *not* improve latency. Dual channel basically means that data is split and stored on the two modules simultaneously. Obviosly this store operation takes time for each of the module. And it is *not* anywhere shorter compared to single channel for the same module. It's like you send two packages via FedEx. Single channel means that you send the second package after the first one is received. Dual channel means that you send the two packages simultaneously. In both cases it takes equal number of days to receive a package.

Re:It's all about the CAS timings

[Dominatus]

Ummm...wow Where..exactly in my post do I claim dual channel improves latency? *looks back* Oh right, no where. I said it impoves BANDWIDTH, which it does, which leads to higher performance.

Re:It's all about the CAS timings

[PenGun]

No ... bandwidth and latency are two different things. Ask any cable modem user that tried to run a game server. Bandwidth was never the problem, but cable modems have inherently high latency due to the cable, it's filters and the network topology.

PenGun
Do What Now ???

Re:It's all about the CAS timings

[Dominatus]

Once...again If you increase speed by a factor of X due to bandwidth and decrease speed by a factor of Y due to latency, if X>Y then the overall benefit is positive and the latency is negated.

Re:It's all about the CAS timings

[PenGun]

Banwidth intelligently managed can only reduce latency by prefetching. Not that useful in a game for instance.

Bandwith is how much data you can move / time. A T1 is 192 K/s for instance.

Latency is how long you have to wait for a response from a device, memory in this case. This is why the 939 pin Athy64s have less latency than a 940 pin Opteron. Registered memory which the Opteron requires has the same bandwith but as it is error correcting has more latency.

PenGun
Do What Now ???

Individual chip variation

[timeOday]

Anybody who's looked into overclocking a CPU knows that two chips are NOT necessarily the same just because they have the same model number. In fact no two are quite the same, even if they came from the very same wafer.

So now that they tested all these brands and models, how consistent are the results? If I go to newegg and get the one that worked best for them, will the one I get work equally well? On a different board? With a different chipset?

Re:Individual chip variation

[Artega VH]

RTFA! They tested everyone piece of RAM in every possible configuration...

DUH...

Re:Individual chip variation

[bersl2]

You're missing the point.

Every chip is different, so every RAM module will perform differently from any other RAM module for all configurations.

Re:Individual chip variation

[harmanjd]

Yep - pretty much all you can do is buy like a batch of 100 and benchmark them your self. They choose the 2-4 fastest ones depending on how many sticks you need. The rest? Well you might be able to return them under warranty.

Re:Individual chip variation

[iamacat]

RTFAC (article copy posted by a nice slashdotter). They pay little attention to overclocking. That would make brand comparison way too expensive - you would need like a hundred chips bought over the course of a few month, from different stores. With that out of the way, chips are tested to perform to stated specs. If yourth doesn't, you can probably exchange it as defective.

The main value of the benchmark than, is to check if the manufacturer is honest. I am sure front side bus has some concept of retries, variable wait states or error correction, so one could just claim higher speed and generate lots of the above.

Re:Individual chip variation

[timeOday]

That would make brand comparison way too expensive - you would need like a hundred chips bought over the course of a few month, from different stores.

I agree, it would be difficult and expensive to perform a useful test. That doesn't justify a useless test.

chips are tested to perform to stated specs. If yourth doesn't, you can probably exchange it as defective.

Yup. I bought a stick of Mushkin that didn't cut the mustard. I exchanged it for another stick (same brand, same model) that did. Testing a single stick doesn't prove anything either way.

So again, what is the point of this test?

Mushkin?

[SpootFinallyRegister]

Where's the mushkin at?

Does this seem like a pretty major omission to anyone else?

Re:Mushkin?

[Dominatus]

They prefer the term "Little People" now

Re:Mushkin?

[Falkentyne]

Yeah it does

They have some of the tightest timings I've seen on memory (See the black level II). Expensive though.

Re:Mushkin?

[randyest]

They asked vendors for (presumably free) samples from for the tests. My guess is Mushkin declined to waste money on the free samples -- they don't need to. People who buy Mushkin know why.

Re:Mushkin?

[agrippa_cash]

I'm surprised that you say that I've had limited but 100% bad luck with Mushkin. Maybe Fry's stores their Mushkin next to the microwave.

Re:Mushkin?

[Anonymous Coward]

Fry's -- There's the root of your problem.

If you're going to buy high-end memory then buy it from someone reputable.

Re:Mushkin?

[Ari_Haviv]

they are too busy representing the lolipop guild

Comment removed

[account_deleted]

Comment removed based on user account deletion

never saw the point of the super duper RAM

[alen]

I'd rather buy the generic stuff and save the money for a better cpu or video card. better price performance ratio i think.

Re:never saw the point of the super duper RAM

[shaitand]

Any current video card is going to perform about equally well for almost everything (including the highest end games on the market to date, the hardware is WAY ahead of the software in that area).

If your running linux or bsd or another OS which doesn't swap unless it has to, then there is a VERY big impact in memory. With a system like windows where you have swapping and vm going on your bottlenecked at the harddrive and memory performance isn't going to impact you so much.

In a linux system your cpu load is almost never all that high, the system is very efficient. Unless your doing a VERY substantial upgrade of processor it won't have alot of impact. In contrast putting in enough memory that you don't need to swap has a HUGE impact in performance, and you can definately feel+see it. Since you've eliminated the drive as a bottleneck then the memory is the bottleneck at that point, so every bit of ram performance increase results in a significant performance increase.

So in short, it's not really a windows versus linux versus bsd thing, it's a swap versus not swap thing. And windows swaps, period, no matter how much memory you have it swaps.

If you have an OS that doesn't insist on disk based memory schemes, and enough ram that you don't need those schemes, then the performance matters. I'd say (admittedly out of my arse, educated guess) that a boost in memory performance is about 50x more significant if everything is running out of ram than if some disk based scheme is happening.

Re:never saw the point of the super duper RAM

[Anonymous Coward]

You do know that the modules they tested are NOT the stuff you would actually use and in a mode you would use in your suggested Linux workstation. I would much more likely buy a normal brand of ECC memory then one of these elitest brands with 5% speed boost.
Also, Linux DOES swap even with more then enough memory, because if it can put things on disk instead of in memory it has more space for diskcache (note, the system for deciding this is rather intricate)

Re:never saw the point of the super duper RAM

[Anonymous Coward]

"Any current video card is going to perform about equally well for almost everything. (including the highest end games on the market to date, the hardware is WAY ahead of the software in that area)."

Wow, am I missing something or is that comment completely BS?

The software (games) constantly taxes video cards that are even only a few months old.

Play Far Cry on a brand new, recently released value level card ($100ish) and then on an enthusiast ($300ish) card and tell me you can't see the difference in framerate and or special effect/visual quality.

Remember the gf4mx series of cards. Those were way behind their time. (but cheap for those who didn't need the pixel shaders and what not).

Re:never saw the point of the super duper RAM

[Anonymous Coward]

why??

dont you want to pay 2X the price for your ram to gain a 0.5% percent increase in speed?

what's wrong with you man???

buying the high end uber-ram is stupid. it's a tiny increase in speed that would be better served by putting the cash into a better motherboard or better processor.

Re:never saw the point of the super duper RAM

[psyclone]

I/O performance:

The disk is still the slowest part of your system (well, after the network and external devices). Fast RAM is nice, but only after your data has been loaded from disk into memory! So the biggest boost in performance visible to the user would be a faster disk -- try a 10k rpm SATA drive for your OS/games/apps, and larger/cheaper 7200 rpm drives for data storage.

Does it really matter?

[Anonymous Coward]

I mean quite frankly, unless you're primary purpose is overclocking, does it really matter performance wise? I can live with a 1-2% hit by going with CAS level 3 rather than 2 and saving myself a couple of hundred dollars in the process.

http://www.gamepc.com/labs/view_content.asp?id=d dr 400myths&page=1

No Mushkin LL Dual Channel tested?

[Anonymous Coward]

I've got Mushkin Blackline low latency 2-2-2 dual channel DDR and I'm quite happy with it, and in my research, it is faster than the Corsair they claimed to be the top performer.

TwinMOS

[stimpleton]

Always nice to have a personal verification on the products you purchase.
This about TwinMos from overclockers NZ:

"Up on closer inspection, we found out that the TwinMOS's PCB is made by Tripod Technology. I've been to the Tripod plant in Taiwan and was pleasantly surprised by their workers self-discipline and how organised the plant was."

Re:TwinMOS

[Anonymous Coward]

Arbeit macht frei! Choose TwinMOS, or we will whip this self-disciplined and well organized slave!

Who gives a shit, as long as it's cheap. RAM is RAM. It works or it doesn't. Run memtest86. Let the drooling Type R people who put neon lights inside their destroyed cases worry about squeezing 0.5 - 2% extra memory performance from their hundreds(!) of extra dollars spent on chromed heatsinks and trademarks with lots of Zs and Xs in them.

Comment removed

[account_deleted]

Comment removed based on user account deletion

Comment removed

[account_deleted]

Comment removed based on user account deletion

Comment removed

[account_deleted]

Comment removed based on user account deletion

Comment removed

[account_deleted]

Comment removed based on user account deletion

They left out the best one

[melted]

Corsair Twin QuadraX 2048-5000XXXL Pro Ultra Mega Super Platinum.

I have one of these and it smokes everything else in their review.

Comment removed

[account_deleted]

Comment removed based on user account deletion

Nnnghaa!

[Anonymous Coward]

...In Japan.

Mirror

[McDutchie]

...is here (until the real site is back). Have fun.

If we're talking about slashdot?

[Anonymous Coward]

And don't even ask me where the sexual intercourse comes in...

I'm guessing in the ass with a greased Yoda doll. The things you learn about reading at -1....

Not to mention,

[Anonymous Coward]

Most chipsets nowadays run 400Mhz Ram at 333Mhz if you have 3 or 4 sticks instead of 2. Read your mobo manual. If it's recent, it may touch on this briefly.

I heard SIS had a chipset that doesn't drop the speed, but still there is the point you made. And the fact that SIS just generally sucks ass.

I was going to mod you up....

[Anonymous Coward]

but looking through this thread you account for a ton of the posts. I can only assume you're on a sugar high and felt that encouraging you could only do more harm.

Remember there are people who love you. So for the love of God switch to sweet-n-lo or splenda where ever possible. But if you don't and I see a story about how some kid had an aneurysm and died while reading slashdot, I'll send flowers.

Re:I was going to mod you up....

[randyest]

Keep the karma points; I don't need them. This just happens to be a subject I researched in detail recently and, instead of finding a great, in-depth, up-to-date review here on /. that would make me say "why didn't I read this before buying," I get this old article, with insanely misinformed posts that make me wonder why I bother with tech articles here.

But you're kinda right. I brought home chocolate mousse cake from Appetito and ate it a little while ago. It rocked, but was heavily sugary. And my wife fell asleep before I noticed, so no nookie tonight. I'll be OK -- thanks for caring though :)

Re:I was going to mod you up....

[ckaminski]

Dunno about your woman, but all the ones I've known have this built in "on" button... not even sleep will keep them down once sufficiently stimulated.

and now

[Anubis350]

These guys have rounded up almost every memory module out there and hammered them

and now they're being hammered....

Comment removed

[account_deleted]

Comment removed based on user account deletion

Mirror # 2

[Anonymous Coward]

http://www.deepblueproductions.net/mirror/trusted% 20reviews.htm
some pucs removed to conserver on bandwidth

Mirror # 2

[Anonymous Coward]

http://www.deepblueproductions.net/mirror/trusted% 20reviews.htm

Mirror # 2 now works

[Anonymous Coward]

http://www.deepblueproductions.net/mirror/

use this link

What's the difference between dual and not dual?

[Anonymous Coward]

Sorry if it sounds like a dumb question.
Is dual memory just 2 identical memory modules? The reason I ask because reading it seems to be that way, but look at retailers' listing, they all seems to list them separately. If they are the same, why don't they just say memory. Then I can just buy 2 of them. I did scan the first page of the article, but some how I didn't see the answer to this (I may have missed it). Thanks.

More info, please

[dargaud]

I have never been able to find a good explaination on PC memories. What are the timings, latencies, what are the numbers on their labels, etc... PC133 ? 4000 ? I put PCs together and usually stay with the recommanded motherboard manufacturer's recommendations but I'm curious for more.

The same but different.

[mollymoo]

Would somebody care to explain to me how different sticks of RAM at the same frequency with the same timings can perform differently?

I had presumed that the frequency and timings absolutely dictated the bandwidth and latency etc. For example, with CL2.5 RAM you have to wait 2.5 cycles after strobing the CAS line to get your data, so doesn't that operation take exactly the same ammount of time on every stick of CL2.5 RAM? Isn't the same true of every operation and therefore the overall performance? Why not? Do the modules themselves cache and pre-fetch data in the same way CPUs do?

Note I'm only talking about sticks at the exact same frequency and timings - it's easy to understand how faster timings and higher frequencies improve performance.

Another victim of /. moderation!

[Walkiry]

> Next time, read the question before blasting away in a mad frenzy.

I see where your reply comes from, but you're barking at the wrong tree. Here's why:

randyest is NOT replying to the OP question, he's doing so to a -1 modded answer to it that says more modules are faster because some bullshit he pulled out of his ass.

Of course, there are several things that have failed in this communication channel:
1. You didn't check who randy was replying to (by clicking the "parent" link at the bottom of his post).
2. The -1 modded post didn't change the subject line (left it with Re: whatever).
3. Neither did randy.
4. /. breaks thread integrity when showing anything with a threshold. Everything under the threshold is lumped under "X more replies".
5. randy should have quoted who he was answering to. Quoting is good.

Anyway, chill a bit guys.

wrong

[Anonymous Coward]

The only way you could be more offtopic is if it contained gay cowboys eating pudding!

Cowboyneal jokes are always on topic.

Two Questions for all

[bananacalvin]

Hey,

I'm not an avid RAM purchaser or anything. I don't care if my throughput rates are increased by 0.05%, or even 5%. My questino is, what is this , you guys always talking about timing? like, x-x-x-x??

Additionally, have you guys ever seen any significant difference with ECC RAM? The new workstation I am getting has this type of RAM. It's an HP xw4100. Let me know.

What about reliability?

[leandrod]

Performance is all very good, but I'm a professional and what I care most about is reliability.

Re:What about reliability?

[Anonymous Coward]

Performance is all very good, but I'm a professional and what I care most about is reliability.

And unless you're running ECC, you'll never get reliability.

No Mushkin in the test

[GuyFawkes]

and since mushkin has always been my personal benchmark of excellence I have to say that they products selected for review were based on fanboy junk that is heavily advertised and hyped rather than products that professionals (such as the slashdot readership hopefully) would select, eg based PURELY on fitness for use, value for money, performance in the real world etc

my 2p worth anyway.

Looking for more?

[Valkyre]

As usual, THG comes with a bit more practical angle:

Which DIMMS actually work with which motherboards:
http://www.tomshardware.com/motherb oard/20040602/i ndex.html

Also, effects of memory timing on performance:
http://www.tomshardware.com/motherbo ard/20040119/i ndex.html

Re:Looking for more?

[Valkyre]

And for the people too lazy to cut and paste:
Motherboard Compatibility
Memory Timings

Don't flame me.

Amateurs

[andrewm]

I didn't RTFA completely, but from my scanning of it (the OCZ part at least), the reviewers seem like amateurs to me.

FWIW I just got some "cheap", not-the-top-of-the-line, OCZ PC4200 3-4-4-8 "Performance" memory. Pretty close to the bottom of the OCZ line actually. It has shiny copper (elemental color) heat spreaders btw. My overclock goal was 266 MHz, the processor (P4 2.4C) isn't co-operating (won't overclock without dangerously high Vcore) though so I'm stuck around 262 MHz.

OCZ memories does NOT accept lower timings. Even OCZ has warnings about this all over their website.

OCZ memories benefit from higer freqencies though. , which more than makes up for high latencies. Any reviewer should know this... or at least have looked at the corporate website. It is almost as though the memory bus signals are synthesized, and not the result of direct memory chip outputs.

I think OCZ has the highest frequency capable memory available.

Free Memory?

[najay]

What a great way to score a bunch of free memory!
(as the article states: "Well, that's where we hope we can help. TrustedReviews sent out an invitation to memory manufacturers to submit 2x 512MB modules of their budget and/or enthusiast level memory for us to compare, and those who bothered to do that have had their memory put through an exhaustive collection of benchmarks and torture tests to try to sort the wheat from the chaff."

Re:What's the difference between dual and not dual

[Anonymous Coward]

dual channel is a bit like RAID 0 for memory, but like RAID 0 its best if you use 2 identical/similar modules, hence they are sold in "matched pairs". It splits chunks of data between them in little pieces and reads/writes simultaneously, thereby increasing bandwidth. sadly, it doesn't double it as you would expect.

Normally [single channel] as the memory is filled, it fills one module then the other, and the northbridge doesn't have to be so fussy.

Comment removed

[account_deleted]

Comment removed based on user account deletion

Comment removed

[account_deleted]

Comment removed based on user account deletion