HWDB Explores Retail Versus Media Sample DDR3 Memory Kit

HWDB, a website run by household overclockers Christian Ney, Tapakah, and El Genieben, just published a remarkable article. In their latest memory review, they stumbled upon a memory kit with an incorrectly flashed SPD. The kit rated at DDR3-2400 had in fact the spec of DDR3-2933. As something smelled fishy, they went out and bought the same kit in retail.

As it turns out, the retail kit has different ICs on board than the media sample kit. The difference between Hynix MFR (media sample) and Hynix CFR (retail kit) isn't big, definitely not for regular gamers and other non-hardcore enthusiasts, but it may highlight a better overclocking range. While the manufacturer is not exactly lying or, even worse, producing false advertisement, there is difference between the product reviewed and the product you'll find in the retail store.

Providing media with high(er) quality hardware samples isn't uncommon practice. In very few cases the hardware is specifically binned for reviews and produces exceptional results as usually we see the samples that are amongst the better within the normal distribution. In this particular case the overclocking results may not reflect that of what you can expect with retail kits as the ICs used are different. We applaud HWDB for this investigative research. Too often we come across reviews from "trusted" sources which write about improbable (euphemism) overclocking results. We also strongly encourage hardware manufacturers to be more open and transparent about their products. In the end, customers don't like being deceived.

For more information, check out the full review over at HW-DB.com.


24

Belgium Massman says:

Good investigative research!

Slovenia tiborrr says:

I see yet another site not getting media samples in the future :D

TaPaKaH says:

If Avexir's policy is to send super handpicked kits even for low-end models, they are useless as a review partner anyway :)
The 2400C11 they've sent have dominated all MFR that I had previously, including some 2666C11 and 2933C12 rated sets, plus 2800C12 handpicked by Corsair.

United States sabishiihito says:

I suspect you would have caught this even if Avexir hadn't forgotten to cover its tracks with the SPD, no company is going to put MFR that good on a midrange speed bin!

websmile says:

Avexir is notorious for cheating on samples, but now they shot themselves in the foot, what I really like is that they obviously think everyone apart from themselves are idiots - the reason they gave for the 2933 XMP and the partnumber is a major insult to everyone who ever did a minimum of ddr3 memory oc... I really appreciate the work and time and money HWDB invested to make comparison with a retail kit, and the courage to do so

Germany der8auer says:

Good job for making this public! reliability >> samples

United States Splave says:

Gskill never sends better samples than retail (2000 c6 flare) and why even have 0000 serial numbers. Asus I know never sends gpus that actually say what speed voltage and temperature the card will run at. Galaxy doesn't send out cherry samples with special firmware that is necessary for high clocks. Gigabyte didn't used to send cherry cpus with their sample boards. Intel doesn't release a0 stepping 980x into the wild knowing they rape retail chips because hey to everyone that doesn't know it doesn't matter. I just find it funny that now you guys find a conscious.

They support me sure so anything I say doesn't count. But why don't you compare the 2 kits at 2400 and enjoy the benefits of double sided vs single sided. I'd take cfr over mfr any day.

Maybe hwdb would like to try my review sample hyperx 2666c11 4x4gb that do 2900 9 12 12 17 versus retail Hynix single sided same part number.

websmile says:

The problem is that you don´t get the point, Allen - these were sent for a review to show customers how this particular series and bin does, and if you think it to be correct to send faked kit to deceit customer about retail kit quality, I am simply sorry for you, support is obviously no good for integrity :) - if you compare these to evaluation samples that go to pro ocers, you obviously don´t understand what´s the problem in this particular case^^

United States Splave says:

The problem is you are focused on one company. Why not do an expose on the industry in total. Bro I know you, you can't be so naive to think this. I know you want to support your buddies' website and that's cool. Can you answer do you think other companies do this? ;)

TaPaKaH says:

The main problem here are actually not the ICs, it is the quality of components. It is possible that some 2400C11 retail kits might be MFR (since Avexir were clearly dumping CFR rejects on retail 2400C11 that we got, same might happen with MFR), but it is not possible that a retail 2400C11 kit, be it CFR or MFR, would easily run 1400 11-13-13 below 1.60V.

I never said that every sample out there is special binned, most of them are not better than retail. But when you get a 2400C11 kit that is flashed with 2933C12 SPD, clocks like hell and maker claims that this is absolutely normal, you'd have to be "very uninformed" to believe and praise this in a review.

websmile says:

I am much longer in business than you think, Allen^^ - I already kicked gskill for example at several forums years ago for doing things similar, sending out xpsc on kits or d samsung samples only to sell tpsc and samsung b retails ;) - there are indeed some companies who do or did this, but I see no reason to to keep quiet about this particular ase because of this. You know that I am no big fan of Kingston for example, but at reviews I did for four of their kits in 2009 and 2010 for example, I got normal kits, using same ics as retails and doing around the same as retails I know did, same with GEIL and Exceleram, just to give small examples. The fact that deceit and lying is part of business for some companies does not justify keeping quiet about this, and it is btw not first time I saw this with avexir, and not first time I criticized this. Obviously they do not want to learn or deceit is part of their business methods, if they think that´s OK they will have to live with criticism like other manufacturers too. Their major mistake this time was that they forgot to conceal traces, and being caught they lied insulting everyones intelligence. If they wanted to show what they can do in highend, they should have sent evaluation sample or only highest bin (they also sent 3100c12 kit for review iirc). This has btw nothing to do with pushing the website, as stated I criticized behaviour like this long ago and on various websites, even on some I didn´t like :D

Slovenia tiborrr says:

I think we all agree cherry picked samples will always be given away for promotional (OC) purposes in order to best promote the manufacturer. But I also hope we all agree that review samples should represent retail samples. Always. These aren't even the same modules.

Germany der8auer says:

Well Allen is still right tho. It's not like Avexir is the only "company" doing this. And you should not forget that usually it depends on the guy who sent the kit to you. Stuff like this is often more an ego thing for a marketing guy to pull great reviews/awards than a company policy. I heard several stories like this over the years and from almost all big companies. This case is just quite funny because they really fked it up lol. But nothing really surprising. Still good that hwdb made it public and didn't shut their mouths. But I can even understand the guys sending out special binned kits. For sure if you know you will send out a sample for test purpose you don't want it to look bad. But it's not the right way to gold plate a pile of crap.

Slovenia tiborrr says:

No, Allen is not right. We're not talking about cherry picked samples given away for the OC purposes. Everybody knows these are creme de la creme when it comes to high bin.

We're talking about the issue of buying a retail product, expecting it to contain certain renown components, which should yield certain overclocking potential as stated in the review and then getting something completely else.

The defense argument of Avexir not being the only company doing it does not justify such actions, it's actually quite hypocritical and quite disturbing for the customer.

Everybody caught with their pants down should be exposed, regardless of the reasons. Otherwise you soon won't be able to tell what the hell are you even buying.

Germany der8auer says:

I didn't want to justify what they did. Just saying that people should not be surprised that this stuff is happening everywhere and think all other vendors have a clean slate.

Slovenia tiborrr says:

Well for one I am glad we both agree on this! :)

United States Splave says:

Im not justifying it either lol just saying everyone does it already why pick on one company. Its like 4 people robbing a bank and only sending 1 to jail.

TaPaKaH says:

I've reviewed nearly 100 memory kits in the last couple of years and this is the first case of a severe overclockability/spec mismatch. I've seen other testers/reviewers end up in similar situations (on kits from Corsair, G.Skill and Kingston) but how they handle the case is their own business.

United States Splave says:

I suppose I could go pickup a set of 2933 gskill and from newegg and they will hit 4200mhz on ln2 then? or do we all know that special bin ones that "pros" get are "different." Does a rambler that semi knows about computers know when they google gskill 2933 that they arent getting the same quality ICs no. As a memory master such as yourself Sam you should know that low level kits are grab bag of ics. IE DDR3 1600 can be pretty much anything, and the same will go for ddr4-3200 kits (hynix turning to samsung it seems). Is it the fact that the spd was wrong the issue? because I also know that binning its common place to flash a spd and load XMP on a board rather than setting all timings manually. how do you know they dont bin out 2400c11 that far?

websmile says:

Splave said: I suppose I could go pickup a set of 2933 gskill and from newegg and they will hit 4200mhz on ln2 then? or do we all know that special bin ones that "pros" get are "different." Does a rambler that semi knows about computers know when they google gskill 2933 that they arent getting the same quality ICs no.

As a memory master such as yourself Sam you should know that low level kits are grab bag of ics. IE DDR3 1600 can be pretty much anything, and the same will go for ddr4-3200 kits (hynix turning to samsung it seems). Is it the fact that the spd was wrong the issue? because I also know that binning its common place to flash a spd and load XMP on a board rather than setting all timings manually. how do you know they dont bin out 2400c11 that far?


Did you even bother to read the review, brother? Then you might have seen retail 2400c11 failing even 2800c12-14-14 at 1,65v, what makes you think that avexir bins their kits for 2933 at same voltage and timings as a normal routine then? Do you think they sell retails that fail their binning pattern that hard and by chance send out sample kit with 2933spd that easily does more than this? It honours you that you support your sponsor, but it doesn´t make you look too good if you ignore the facts... I can only tell that I had several 2400c10 Core series retails with cfr for example, which is much better bin than the 2400c11 for hynix, and even these were far from doing 2933^^ - sometimes a cigar is just a cigar, no need to search for more..

United States sabishiihito says:

I still can't get over Avexir's response.

"AVEXIR tests all 2400 module on z77 instead of z97 and 2933 z97 equals 2400 z77. Our production department forgot to flash SPD."


United States Splave says:

websmile said: Did you even bother to read the review, brother? Then you might have seen retail 2400c11 failing even 2800c12-14-14 at 1,65v, what makes you think that avexir bins their kits for 2933 at same voltage and timings as a normal routine then? Do you think they sell retails that fail their binning pattern that hard and by chance send out sample kit with 2933spd that easily does more than this? It honours you that you support your sponsor, but it doesn´t make you look too good if you ignore the facts... I can only tell that I had several 2400c10 Core series retails with cfr for example, which is much better bin than the 2400c11 for hynix, and even these were far from doing 2933^^ - sometimes a cigar is just a cigar, no need to search for more..

I did read. I'm just saying since when is an overclock guaranteed? If they do spec then? Ics aren't guaranteed either. Spd being wrong could be many things, and maybe we will never know for sure..

United States nnimrod says:

Thanks for doing this :) Just because everyone is doing it doesn't make it ok. Also a little insulting that they apparently thought you'd either not notice or just play along with it.

Australia Jimba says:

Great article! This is bad practice full stop. Sure have cheery for OC events but not retail reviews. Bad move avexir. Same goes for gskill, corsair and others.

Please log in or register to comment.

Leave a Reply: (BBCODE allowed: [B], [QUOTE], [I], [URL], [IMG],...)