Corsair's QC

[Faethon]

I decided to start a new thread, just as a "warning" to whoever may be interested. I have no intention to battle eventualy fanboys that may come to fight for their brand of heart, i am too old for this stuff. I 've been Corsair customer since the original XMS (1) series. The first part i copy-paste it from here (i don't want to hijack that thread):

http://hardforum.com/showthread.php?t=1813998

I ve' had trouble to run Corsair RAM stable at the "official" 1600Mhz with 2 different CPUs undervolted and i think at stock too on 2 different motherboards. Sticks say they 're 1600, but it's only by XMP profile and for reasons i don't know, mine just don't like running at 1600. They can pass IBT, but not Prime. (yeah, i just load the XMP in BIOS and touch nothing else, still, on 2 different manufacturer mobos and both with Thuban and Piledriver, it won't work).

Practically, i have the same issue with this guy with Corsair memories (same kit)

http://www.newegg.com/Product/Produ...a&cm_sp=&AID=10440897&PID=3891137&SID=rewrite


If i run it at 1333 T1, runs fine and passes Prime.

And this is how much Corsair "tests" for QC her sticks nowdays, like they claim to that american buyer. This is one of my "normal" sticks. Look at part number:

This is it's sibling stick of the kit (look at part number and no, it's not a motherboard "strangeness", happens with both Asrock and Gigabyte boards).

Window's memory diagnostic couldn't find error. IBT passes, Prime95 doesn't at 1600 and i 've had occasions of corrupted Windows when trying to run 1600. Gone ever since i went to 1333. Needless to say, i won't be buying Corsair in the future (although for compatibility reasons alone, i now have 5x8GB Corsair DDR3 kits, counting the quad kit as 2xdual (total of 40GB RAM). But in the future i will switch to Kingston or GSkill or someone else. It's a good thing i am not overclocker or gamer, so i am not too picky if i have to run 1333. I thought of trying to bump RAM voltage to 1.65v (since there is another Corsair KIT that runs at the same timings with 1.65 and maybe they mixed the sticks, but why run them hotter, i didn't bother. Funniest thing, my Gigabyte board has this exact kit in the compatibility list of the mobo...Oh Corsair, how low have you fallen...

------------------------------------------------------------------------------------------------------------------------

Since February 2012 that i bought this kit, i thought it was my Asrock's mobo's fault. I was also told the same here, by some forum member. It was the logical thing to think. But after buying now a Gigabyte board and having the same issue (with all 4 sticks of the kit), it can't be.

Yesterday, i decided to run Prime again, raising DRAM voltage after all...

- DRAM@ 1.55v = Prime95 failed in Blend after 6,5 hours (rounding error in 1 core).
- DRAM@ 1.60v = Prime95 PASS after 10 hours Blend.

^ You can verify the "supposed" DRAM voltage at 1.5v and the effective in Gigabytes Easytuner, manually set at 1.6v in BIOS, in order to "pass".

So, i bought a 4x4GB kit 1600C9 from Corsair, that would run though only at 1.6v. 2 motherboards on "auto" (setting 1.5v) and with XMP profile loaded, couldn't run it at 1600Mhz. For a kit that nowdays here is at sale at 144 euros (198$).


Of course, i do understand Corsair's tough decision... You have a kit that runs stable at 1.60v. Where do you sell it? At the kits of 1.5v or at the kits of 1.65v?

Because the same kit, same latency, same heatspreader and brand name (XMS3), is also sold at 1.65v version:

http://www.newegg.com/Product/Product.aspx?Item=N82E16820145324

So i guess, Corsair thought "well, let's put it at 1.5v kit, sells for higher price too)...


Interestingly, in the XMS3 1.65v kit, i also found some interesting reviews from american customers (i excluded those about DOA or errors in sticks), with... problems running at rated speeds or instability, that MAYBE can be caused by the same cause...

Now, i 've been loyal Corsair customer, since...ever... Here are the receipts of my past DDR3 purchases alone:
http://hardforum.com/showpost.php?p=1040706415&postcount=14

And i have another 2x4GB kit already ordered, arriving this week. I don't know if these things happen with other vendors too, but at this point, i don't care. If anything, i would expect that the QUAD kits, would be more "controlled" , given the fact that nowdays they ask 198$ for a kit, before boasting this crap:

I haven't tested the rest of my kits at 1600Mhz till this day, because i have them running on Athlons and 1333 is their "officially" specced speed and am happy enough with that (yes, i know they can run 1600Mhz RAM, but with more stress on the IMC and i prefer undervolting and it's easier to undervolt when the IMC runs the RAM at stock speeds). After the "adventures" i 've had with this 1st kit, i was bothered to mess with the rest. But i am afraid of what i may find in the other kits too, if i try to run them at 1600Mhz.


At any case, take it for what you like, maybe the tales of a few unfortunate customers, i just write it as a word of caution.

As for myself, i 've given a total of 215,45 euros (counting the last kit arriving this week) to Corsair for DDR3 alone, i am not buying anymore from them in the future and just yesterday i adviced a friend to buy Kingston instead.

I can accept a RAM failure (anything can break). I can accept something coming DOA even. At least it's obvious. But i can't accept selling "1.5v RAM", which effectively needs 1.6v and no motherboard can run it on "auto". Because it goes beyond the "normal" user to find that. Don't come and crap me about "guaranteed compatibility", when your sticks can't run on "auto" voltage.

Bye bye Corsair!


P.S.: At least, after more than 2 years after the purchase date, thanks to the fact that i bought a Gigabyte board, i figured out how to FINALLY make it run at 1600Mhz! It could have been worse! I could have stayed forever at 1333Mhz...

 

[Javvis]

You're over looking one variable ...not all Motherboards will set the CPU/Memory & other voltages properly in auto-set mode. You should set them manually & check with an application like AIDA64.

 

[Faethon]

You're over looking one variable ...not all Motherboards will set the CPU/Memory & other voltages properly in auto-set mode. You should set them manually & check with an application like AIDA64.

Javvis, i may not be an overclocker, but i 've been around computers since Amstrad CPC6128 and know how to assemble one. I even know basic stuff in Linux. That is to say, i am not [H]ard overclocker, but not your average Joe either.

This happened with:
- Asrock 970 Extreme3
- Gigabyte 970 UD3P

Both motherboards have digital PWM. Both motherboards have their own monitoring programs (Asrock AXTU and Gigabyte EasyTune), where they show you in detail their readings of voltages, as if my "generic" HWMonitor and Speccy weren't enough. I ALWAYS doublecheck via software the setting of BIOS. I even doublecheck through CPU-Z to see whether the XMP profile loaded in BIOS is correctly read and applied by the motherboard (comparing the SPD and Memory pages of CPU-Z).

Both motherboards, applied correct voltage (unless all software is wrong) and actually since they both have digital PWM, the voltage in both was rock stable. If anything, this shows that the kit, is more towards 1.65v than 1.5v. Because it's "off" the 1.5v mark by 0.1 volt, while off the "1.65v" mark by 0.05v.

And as i showed, it doesn't seem that i am the only dumb guy that can't make them work, while everyone else is happy camper. And if we are not the only cases, who knows how many other people exist out there, who run them unstable and never run Prime95 for 10 hours to see the instability and just have mysterious "my Windows crashed" episodes. Or you know, i have seen cases in fora with people saying "my FX can't pass Prime95, well, i can't see why, Prime95 sucks for FX, there, i found why, tis Prime95's fault, not my rig's!"

I assure you, that the instability was a subtle one. But i do much video encoding on this computer while also doing other things (i multitask heavily) and over the years, i 've seen that even the slightest instability in Prime95, will eventually "hit me" , exactly because i stress the system daily with heavy lifting. Other people may run this kit in 1.5v, never stress their system a lot and get away with it and never notice anything. But on my computers where i do video encoding, i always notice eventually. Or others, may simply blame some sporadic crashes to the motherboard. I did that too for 2 years. My Asrock's memory compatibility list had only 2x2GB kits listed. A guy here also told me "it's probably compatibility issue with your motherboard that doesn't like the RAM", it was a very probable explanation, i blamed the motherboard too and kept it at 1333Mhz and didn't think about it again... I said "well, you got a cheap board, wasn't listed as compatible, bad luck, this is Corsair after all, not cheapo Asrock"...

Sometimes you blame the wrong thing...

 

[larrymoencurly]

Corsair isn't alone because every retail brand of memory is made from either subprime (no-name) or overclocked chips, and PC-10666 modules, unless they're very old, are almost always made from subprime chips. IOW even their claimed JEDEC profiles are too fast. Overclocked or junk chips are also why I wouldn't undervolt the CPU because even a difference of 0.05 - 0.1V can make enough of a difference with memory that's marginal at best.

 

[Faethon]

Corsair isn't alone because every retail brand of memory is made from either subprime (no-name) or overclocked chips, and PC-10666 modules, unless they're very old, are almost always made from subprime chips. IOW even their claimed JEDEC profiles are too fast. Overclocked or junk chips are also why I wouldn't undervolt the CPU because even a difference of 0.05 - 0.1V can make enough of a difference with memory that's marginal at best.

I can believe you that Corsair isn't alone is this, but you know how customers are. Usually, when you get bamboozled by a company you had in high regard, you get pissed off and say "time to give another company a chance".

I am 100% certain that i have tried these sticks with default CPU VCore in my Asrock board. Not in the Gigabyte, but at this point, i don't think it's the board.

I 've been undervolting since socket939 era. Things have always gone like this:

- I 've never had a CPU fail in undervolt, while keeping the RAM at the "stock" settings the IMC is supposed to run at. FX CPU with 1600Mhz (and 2 sticks only) isn't something that should overstress the IMC. If i was trying to run 2133Mhz, maybe i could blame the undervolt.

- I 've never had an undervolt fail at a certain voltage (with stock IMC speeds for RAM) and aftwards pass the stress tests with the same vCore, by raising only dram voltage.

And i 've undervolted Athlon 64 3200+ (Venice), Athlon64 x2 3800+ (Toledo), Athlon x2 5050e (Brisbane), Athlon X4 605e (Propus), Athlon X4 640 (Propus), Athlon II x2 240 & 270 (Regor) , Phenom II x6 1090T (Thuban) and now the FX6300.

I 've NEVER, ever had trouble with undervolting and setting RAM to auto. Even on supercheap Asrock boards that had analog PWM and vCore voltage was dancing all around the place. I actually always leave VCore a notch above the minimum, just to be sure.

Coincidence?

 

[NickJames]

Corsair's QC has gone to shit in the last couple of years. Constantly failing USB sticks, SSDs, annoyingly loud PSU whine and dead RAM sticks. They used to be my #1 brand of choice but now I buy from others before I go to Corsair. It makes me sad

 

[Faethon]

This, just for the history, to shake off any doubts "larrymoencurly" may have.


Prime95 running at stock XMP Profile, AUTO CPU VCore 1.34v (a VERY generous one, since probably Gigabyte wants to help overclocking in a 8+2 board. Other boards give MUCH lower auto voltage to this CPU and frankly, the chip itself at stock, can run below 1.20v). Turbo Core is disabled in BIOS (as if it mattered anything in a multithreaded test like Prime, but just for history).

Blend FAILED at AUTO DRAM voltage 1.5v after 4+ hours, DESPITE VCore raised to 1.34v. And this is the SECOND motherboard that fails also at stock CPU voltage and 1.5v DRAM.

Like i said, i am not a [H]ard overclocker, but i know my undervolting... DRAM voltage was a PERFECT 1.5v, stable as rock.

This kit, should be sold as "Corsair 1600C9 XMS3 1.65v". End of story. Because even 1.60v may be marginal and if i were to leave it 24h instead of 10h, maybe it would fail.

It even makes sense!

- 1.5v fails after 4,5 hours.
- 1.55v fails after 6.5 hours.
- 1.60 PASS.

And the irony, just ran Memaxx.

1333 T1

1600 T2

^ Maybe it needs 1.65v! LOL!


I m back to run it at 1333T1 @ 1.5v. At least it stays cooler...

Last of the series "Prime95 sucks, with other tests my rig is fine, so there is something fishy with Prime95"! (everytime i 've read this somewhere i 've always thought "why should a CPU have problem running... mathematical calculations? Isn't what they were built for?"

 

[ashmelev75]

Why don't you return this questionable RAM and get something decent from Tier1 manufacturer like
http://www.crucial.com/store/partspecs.aspx?IMODULE=BLT4KIT4G3D1608DT1TX0
Yeah, it is a bit more expensive, but at least it works as supposed to.

 

[Faethon]

Why don't you return this questionable RAM and get something decent from Tier1 manufacturer like
http://www.crucial.com/store/partspecs.aspx?IMODULE=BLT4KIT4G3D1608DT1TX0
Yeah, it is a bit more expensive, but at least it works as supposed to.

Because

1) at this point i have 5X2GB Corsair kits (despite the difference names they 're the same sticks) and it makes sense to have RAM that is interchangeable between kits. You know, in 4 years from now, a stick may go bad. It's a good thing if you can change it immediately with a 100% compatible stick. I always keep "spare parts".

2) Because i am past the official 2 years warranty that EU forces retailers to cover you with warranty. As the law here is, i should send the modules to the retailer, who should then test them himself to see if they are defective. If he doesn't think they are defective, he must send them back to me, where i get to pay his postal fees plus a "penalty" of 30-40 euros for making him lose time with a non existing defect. The funny is, that he may very well "test" them and say they 're ok... In Windows memory diagnostic tool, there were no errors.. From what i 've googled, there is not even a direct "Corsair Italy" center, where i could try to circumvent the retailer (i don't even know if it's possible, by law, the RMA is done through the retailer). I 've already had a bad experience with a retailer in the past, where he sold me obviously wrong RAM (DDR 400, where the last SPD was 333 and couldn't run properly at 400 without overvolting, but the sticker on it was saying it was DDR400. The reply of the retailer was: "the sticker says it's DDR400, it's not burnt, it's DDR400. I paied it as DDR400 too, i was in good faith".

3) I can live with 1333T1... Actually, when i am in the mood of taking the screwdriver in hand, i will open one of my Athlon rigs, take out the Corsair kit from there and swap it with the kit from the FX rig. Because at the end, i deliberately run 1333Mhz in the AM3 rigs. So, i may as well put the "defective" kit in the Athlon rigs and get one of the other kits in the FX rig and run it at 1600. Of course...there is always the possibility that my other kits are also defective and can't run at 1600 at stock. LOLOL!

4) In the worst case, if i need it, i will buy more RAM, it will be less hassle than engaging with arm-wrestling with the retailer here.


But thanks for reminding me of Crucial! I 've heard/read the best opinions about Crucial SSDs, maybe i should rely on them in the future for RAM.

 

[ashmelev75]

Because
1) at this point i have 5X2GB Corsair kits (despite the difference names they 're the same sticks) and it makes sense to have RAM that is interchangeable between kits.

Are they made using exactly the same memory chips? Probably not, especially if you buy them from different production batches. Corsair don't make their own memory chips, they have to buy them from the actual manufacturers (Samsung, Elpida, Micron, Hynix).

Crucial, on other hand, is owned by Micron, so they always use Micron chips and they offer a lifetime warranty.

2) Because i am past the official 2 years warranty that EU forces retailers to cover you with warranty.

That sucks.

 

[German Muscle]

Ive had 2 failures when using cheaper corsair memory in builds. Ive since switched to Kingston HyperX.

 

[Faethon]

Are they made using exactly the same memory chips? Probably not, especially if you buy them from different production batches. Corsair don't make their own memory chips, they have to buy them from the actual manufacturers (Samsung, Elpida, Micron, Hynix).

I 've no idea really, i haven't tried mixing them. I have:

- 6x4GB sticks that are XMS3 1600C9 1.5v
- 4x4gb sticks that are Vengeance Low Profile 1600C9 1.5v

But i 've only tried run at 1600Mhz the kit in the FX rig.

Timings are the same. If they are not even compatible between them, well, i will be very displeased. But still, i have some interchangeability between XMS3 (3 kits of 2sticks) and Vengeance (2 kits of 2 sticks).

I didn't even know that Corsair uses many different chip manufacturers for the same series... Damn!

Now that you 've put the doubt in me, i think i will try to run mixed sticks between XMS3 and Vengeance and see what happens. I will be quite disappointed if they can't run together.

Crucial, on other hand, is owned by Micron, so they always use Micron chips and they offer a lifetime warranty.

I wish i had known that before i bought my first Corsair kit... I guess, it's a lesson to learn for the future... And that next time i buy RAM, i 'd better ask here for advice. But you know. I was always dead certain that "Corsair is a guarantee of quality".

I also run 3 Corsair PSUs right now in various rigs, i hope they don't disappoint me.

That sucks.

TECHNICALLY, i should be still covered by Corsair's lifetime warranty. Only i can't find any reference on my retailer's recept (but it's on the box) and like it or not, here the retailer is "bound" by national law for the first 2 years. Beyond that, when there is no local branch of the company ("Corsair Italy"), you are more or less at his mercy. And retailers here always hate RMAs even withing the 2 years. Imagine when you are outside the 2 years...

 

[Faethon]

Ok, gathered up all the sticks (except from the FX rig of course). The 5th kit i ordred in April 14 and they haven't shipped it yet (shops here take their time and Easter is an even better excuse to delay further), but it's Vengeance LP.

By nude eye, the 16GB kit, is ver 3.19

The other XMS3 kit and the Vengeance LP are both ver 8.16

That can't be good, can it?

Anyway, i don't have neither the time nor patience to make all the possible stick combinations to see if they work with each other running Prime95 for 10 hours each time.

I will make combinatinons and run IBT at very high for 10 passes at 1600. If it passes, at least i know they will be able to live together at least at 1333...

Ready for battle:

Ive had 2 failures when using cheaper corsair memory in builds. Ive since switched to Kingston HyperX.

At least when the sticks are bad and produce easily detectable errors, it's easier to ID them quickly too and easier to convince an unwilling retailer... Here, often retailers know about computers less than i do. However, they all have a clause in their RMA procedures that they are entitled to do their "own testing in the way they want" to determine if your complaint is right or not. And reading my local fora, they very often have a very odd way of testing that doesn't quite seem right... They can't hide when there is an obvious defect (like HDD can't run), but when there is something subtle, they go to great lenghts to avoid substitution. It's a mentality a bit like "if it powers up and works, then it's OK!". It's a good advice for example, if you have a defective part that is a bit "strange", to make sure it's really dead before you RMA it. My last hard disk that came DOA and started giving "clicks of death", and some bad sectors after the 1st full format, i made sure to give it a good bang, format it non stop for 4 times, to increase the bad sectors and then send it to them. By the time i was finished, it had 300 bad sectors, rattling like snake and they didn't say anything to me, they just sent me a brand new without trying to oppose. That's how you do it here to play safe.

 

[ashmelev75]

TECHNICALLY, i should be still covered by Corsair's lifetime warranty. Only i can't find any reference on my retailer's recept (but it's on the box) and like it or not, here the retailer is "bound" by national law for the first 2 years. Beyond that, when there is no local branch of the company ("Corsair Italy"), you are more or less at his mercy. And retailers here always hate RMAs even withing the 2 years. Imagine when you are outside the 2 years...

Since you've got nothing to lose, create a ticket on corsair.force.com
As long as you have a proof of purchase, you should be able to use their lifetime warranty. You will be responsible for shipping the defective sticks to them.

 

[Faethon]

Since you've got nothing to lose, create a ticket on corsair.force.com
As long as you have a proof of purchase, you should be able to use their lifetime warranty. You will be responsible for shipping the defective sticks to them.

Ship them where? To USA? LOL! Because there is no local branch here. It's not worth neither the headache nor the money nor the waiting...

Inside one kit there is a little note for "Australian consumers". The interesting part is:

"You are entitled to a replacement or refund for a major failure and for compensation for any other reasonably foreseeable loss or damage You are also entitled to have the good repaired or replaced if goods fail to be of acceptable quality and the failure does not amount to a major failure".

This reminds me of my local retailers trying to dodge bullets. I may have better chance and headache with my local retailer than trying to solve my problem across the ocean. Because, my kit isn't exactly "defective" or with RAM that has errors. It's what we call here "defect of conformity". It's works, only it needs 1.60v instead of 1.5v... That's not "major failure" nor is "damaged".

It makes more sense to take this kit out of the FX rig and drop it to an Athlon rig, save myself the hassle...


EDIT: Oh what the hell, just out of curiocity, i will run Prime on the Vengeance sticks at 1600 1.5, to see whether they will pass or not. I will run them overnight.

 

[Faethon]

ashmelev75 was right.

There are differences between the kits.

Oddly enough, all the XMS3 kits, despite being different versions, present the same SPD tables.

The Vengeance LP on the other hand, is different:

The good news, is that they all seem to be able to run mixed with the others (because the higher JEDEC tables are same, fortunately). Also, interestingly, the 2 other kits, passed IBT "Very High" 1600 1.5v. This time, running 1 of the "defective" sticks with a Vengeance stick, failed at 1.5v - 1600, a bit before the end of the IBT test (i think in the 9th run), BUT passed at 1333. So, there is certainly a voltage issue with the first kit. An other odd issue, is that the Vengeance kit, also seems to score higher score in Memmax at 1333T1 than in 1600T2... But i don't know how reliable MemMaxx is...

I am too tired to change sticks again to run Prime tonight...