Live Memory Test: Overclock 'Em Till They Crash

[pschmid]

In our latest test, we gradually push memory modules to their limits, and test their boundaries. This time, we do it live and on two systems simultaneously. We also pit special test modules against ones we purchased to find out the truth. Are the manufacturers cheating? And who's really got the fastest modules?

 

[TechnologyCoordinator]

I feel only one test system should be used. What if the motherboard or OMC manufacturing quality issues affect RAM performance. I'm not an expert though...

However, the results should be accurate enough to see if these vendors are giving THG the creme of the crop or the real stuff that we'll end up buying.

 

[bluntside]

I'm curious about my mobo, its a Dfi Lan PAry Nf4 ut sli-dr eXpert. The Dfi mobos i know hav e great features and lots of oc'ing potential not to mention its very stable. But what so great about this particular mobo?? :twisted:

 

[sojrner]

I am much more curious as to why the top two performers in the last memory tests are not represented here? The article says they went w/ what was widely available, but I know that corsair and patriot are at least as distributed as the five used in the test... just seems odd to me, as does the use of different procs. I know they are similar, but why not just use the same one? they stress that everything else is identical.

techCoord makes a good point too: why not just use the same system and eliminate all differences?

These are all quibbles on my part though, and the test is still a good reference for memory hunters. Just if you are going to point out how hard you are working to get total parity for the two test setups, then actually GET total parity... dont fudge it in some spots and hope no-one notices.

 

[dorion]

This is cool. I've never been able to access a computer for the last stress tests. Hope that some memory manafacturers get burned for the results here. :twisted:

 

[AlaskaFox]

I feel only one test system should be used. What if the motherboard or OMC manufacturing quality issues affect RAM performance. I'm not an expert though...

However, the results should be accurate enough to see if these vendors are giving THG the creme of the crop or the real stuff that we'll end up buying.

I agree. while you have dropped the multiplier down, I have noticed one of the system limitations in my overclocking experience to be the system bus. slight variations in the traces between the memory and the northbridge, or imperfections in the northbridge of the system itself, imho, could skew these results. I plan on keeping an eye on this and seeing if one system consistently crashes before the other.

as far as the cream of the crop goes, I still remember the junk power supply on the AMD vs. Intel stress tests...

 

[TechnologyCoordinator]

TechnologyCoordinator and Bluntside, they're using i975X.

Ahhh, yes, I did not notice it was DDR2 on an Intel system. Regardless, even though the memory controller is on the motherboard and is not an OMC, there could still be manufacturing quality differences.

I guess another good way to double check the parts being used is to swap the RAM from the systems and re-perform the test to see if similiar results are achieved.

 

[firefoxz]

To the editors:...

Keep up the good work

*edited

 

[sojrner]

mmmm... cherries... 8)

 

[delanooch]

Wow GEIL busted, SO I guess non retail samples will be slightly bloated over real world retail product?

Nice work THG. Can't wait for the Mushkin test, I have a lot of their stuff!

 

[Danijel_P]

I’m very curious as to why “you” (THG reps) chose a Gigabyte motherboard for conducting a stress test?
Especially since, “you” are well aware of the fact that they are notorious for their O/C shams when it comes to motherboards. I don’t want to sound like a cynic, but god knows what other
tricks they pull with their other products just to get an advantage to the competitors.
I can’t see how it is even remotely a surprise to you that the motherboards BIOS is rigged to perform better.
Need I remind you of the i865 and i875 lineups that came with the obvious FSB manipulation? (guess I just did…)

I am one of those “stupid” enough to have bought a GA-8KNXP Ultra. (i875 chipset with U320 SCSI onboard)
I wanted and needed to have the SCSI controller and why not strike two birds with one stone I thought. Big mistake but it will not happen again.

You might wonder how just “1”tiny Mhz boost on the FSB can have a negative impact on a system!?
Let me put it this way. The system gets an even bigger boost when starting up a program!
Clever engineering and good BIOS programming you say?
Sure, that “might” be true, but let me ask you this then. Is it stated on the retail box that the component has a default O/C that can’t be turned off and that you buy this product at your own risk? Maybe in the manual?? How about a simple piece of paper??? And most important of all! Did I ask for it!?
No, no, no and a definite no!

I could go on with more details but I’ll take a breather here. Surely there are more of you out there that disapprove of companies that lack the marketing skills required to stay in business.
That would rather cheat and lie than to make sure that their customers are satisfied.
And yes I have been in contact with Gigabyte and their official response will be posted if asked for.

Maybe its time for a petition here on THG to ask the readers what they think about getting tricked?
Then again if you guys need to stay neutral I’ll be more than happy to start it off right here and see who’s with me.

O/C is a privilege that should be left open as an option exclusive only to the consumer and not set to default by the manufacturer!
Not unless “clearly” stated that the product is boosted, optimized, O/C:d by default or whatever they want to call it.
Hiding so called “optimizations” that fall out of spec with other components, without taking any responsibility for hardware that gets fried along the way, Is nothing less than sabotage.

Happy Easter everyone.

//Danijel P

 

[Aragorn]

@Dani

The article clearly stated that they chose a gigabyte board because they were the only manufacturer that could proivide them with a bios that did what they need it to do. The other manufacturers either could not or chose not to.

BTW I own a gigabyte board GA-K8NF9 and I am not that happy with it. It does not have all the options in the bios that their site led me to expect. However the base clock was exactly what it should have been 200 MHz. Some of the missing options were made available with a bios update but not quite what I was expecting. I am not a fan of Gigabyte (I will go DFI or ASUS for my next board) but the reasons their board was selected is pretty clear in the article.

 

[delanooch]

comprehension of what he did or didn't read, I think he just saw gigabyte and freaked out.

In his defense , GEIL tried to dupe the THG crew me thinks.

20 lashings with a wet noodle.

P.S. Buy Mushkin!!

Unless next week they dupe us again!

 

[sojrner]

Personally I am partial to corsair or crucial. They have not failed me ever. Mushkin has failed my friend once, failed memory in his server.

question: was the noodle for the mem makers or the idiot above? :twisted:

 

[delanooch]

Personally I am partial to corsair or crucial. They have not failed me ever. Mushkin has failed my friend once, failed memory in his server.

question: was the noodle for the mem makers or the idiot above? :twisted:

LOL it was for GEIL!

 

[sojrner]

now i'm hungry for spaghetti... mmm.. pasta.

 

[Danijel_P]

Aragorn:

I admit that I did not read the whole article and Im shocked to hear that there was only one manufacturer who could or was willing to provide a mobo.

My intetion is not in anyway to flame the THG crew. If anything they have my sincerest an outmost respect.
What I was trying to point out was the fact that we consumers get cheated.
Your motherboard is as good an example as mine is.

Strange Stranger:
Do I even need to comment that? :roll:

 

[Danijel_P]

What Im trying to highlight are my concerns regarding the policy that many companies have adapted when it comes to overclocking PC components per default.
I was however specific about motherborads because I regard that specific component to be the core of a computer.

My goal is to open up a discussion.
Some of you apparently have the wrong idea thats this is a brand issue or even just-limited-to-the-test-issue when its in fact an overall quality of motherboards issue.
I hope it's all a litle clearer now, even for you strangeStranger.
Try reading my post again and see if you might have missread something.

If you still dont understand then please feel free to ask. A discussion is supposed to bring out facts, questions and hopefully som answers too.
If you simply couldnt care less, then thats your oppinion, and voice that oppinion it is your right.
Just try to do it in a not so obviously anal and teenager way, ok!?

Best regards,

Danijel

 

[Feanor1]

Good reviews are hard to find especially when building a new box. My pick for memory Corsair is not being tested, but some I have used some that are in the past. Thanks to THG we get to see true results. Keep testing! I would love to see THG build a complete system having tested all pieces, parts and software.

 

[a1ien]

I admit that I did not read the whole article and Im shocked to hear that there was only one manufacturer who could or was willing to provide a mobo.

If you had read the whole article, you would also have realized that the main limiting factor in their choice was which company was able to provide a bug-free BIOS in the given timeframe, and not that "only one manufacturer who could or was willing to provide a mobo," and I quote:

"Only Gigabyte was able to program a bug-free BIOS in the given timeframe. In our test, where the machines need to cope with a large range of clock speeds, extreme demands are made on the registry entries stored in the BIOS. For this reason, Gigabyte made a special BIOS just for this test, something not every company could achieve."

Some of you just can't judge when it's a good review or not.
You guys have got no idea have many reviews I've read in the past few days only to find myself trashing it in the bin after a quarter of way through because it was not up to my standard.

Good reviews are hard to find nowdays.
People should be thankful for this one.

I agree wholeheartedly.

 

[gudodayn]

I guess it aint a mystery that these memory manufacturers actually hand pick and test the modules before they are sent for reviews. This is why a random retail bought sample will most likely OC lower than the reviewed sample sent.

Believe it or not, before being sent for a review, hand picked modules are tested on various big brand/ top selling boards. The company I work for have boards from Asus, DFI, Abit, Gigabyte, MSI, Epox and Albatron. When a hardware review site requests our modules, they will be tested on most of these boards before being sent out.

There are instances when we didnt follow that procedure and in that particular review........we got annihilated!!!!

The bottom line is when buying memory, reviews can only be used as guide, for better performing modules - it simply means that their modules have better potential and quality. It is never never never never a guarantee that you can always get the same OC as the hardware review and vice versa, there's a chance you just might pick up a module that will perform better than the review samples....which I believe many exists!!!!

 

[mfg8876]

Could someone plz explain to me, how is it (at the time of the current posting) that the retail kingston memory has too, like the Geil retail one, got stuck at a 421Mhz memory bus setting?

Memory Stress Test: Currently Testing Kingston DDR2-900

Memory Bandwidth
Live Memory Bandwidth
Sample Memory Retail Memory
Copy 3722 MB/s 3385 MB/s
Scale 3811 MB/s 3474 MB/s
Add 4044 MB/s 3671 MB/s
Triad 4066 MB/s 3681 MB/s


Live System Clocks
Sample Memory Retail Memory
CPU 3360 MHz 3036 MHz
FSB 280 MHz 253 MHz
Memory 466 MHz (DDR2-932) 421 MHz (DDR2-842)

Edit: After visiting the Live Performance page at 12:48 GMT the kingston retail modules are on par with the sample ones.

 

[ara]

You will most certainly be interested:

GEIL responds to cherry picking sample accusations

i read some of the comments though... some of them have points, they are only guaranteeing the running speed. However, it makes the product look better that they overclock more... what do you guys think?

 

[gudodayn]

Trust me.....I work in the same technology park as all the other memory manufacturers, they all specifically pick them for reviews.
Thats why they dont guarantee OC speeds and they'll tell you that it will vary from batch to batch.....I think I saw a RMA letter from our RMA department to customer that was pretty much what Geil VP said on that response.
If they dont pick them, they'll loose face and others will perform better and whoever was in charge of that department can kiss their ass goodbye!!!!
This is how the question should have been put forward to the VP:
If a ultra series was purchased from a retail store and OCed but did not come close to a review sites max OC speed, will you refund the guy/gal his/her money + money for transportation to and from the place where he/she bought it from and $ cost of possible machine down time cause they weren't able to do work on it?

He should have just been honest and said " Yes, we do pick them! We do this because we want to let the viewer know what our modules are capable of but with that said, not every module is going to perform the same. A search on google will certainly tell you that there are a lot of retail sold Geil modules that are reaching OC speeds that review sites got no where near! That type of result speaks for itself"...............and it would have been a happy ending!!!

Instead, he's going around in circles avoiding the question " DO YOU CHERRY PICK THE MODULES FOR REVIEW SITES?".........I think the fact that he did not give a straight answer means a resounding "YES"!!!

Ah crap, I hope I'm not gonna receive "GET A NEW KEYBOARD" comments!!!

 

[benjamin]

Wow.

It's a test that needed to be done, to just see how big the differences can be... but this really isn't going to win you any friends. Pretty brave move, guys.

Of course, the very fact that these modules are being made shows that at least *some* customers are getting the real deal. I believe cherry-picking isn't necessarily a bad thing per se, either; the problem is manufacturers failing to be upfront with tolerance numbers which run true to life.

But the Gigabyte motherboard over-voltage cheat? That deserves all the blunt words you gave & more. Nice one.