Memtest reports errors in brand new Corsair

[mcv]

I just received my new PC, but while it boots fine from CD and managed to install my 64 bit Ubuntu linux, when it tries to boot from harddisk, it fails somewhere around where the kernel should start. Reverting to failsafe BIOS settings or even more seriously underclocking doesn't solve the problem. Since my install CD contains Memtest86+, I decided to give that a try, and got hundreds of errors per pass.

So what do you do? Figure out which module is the offending one. So I take half of them out, test again, and get the same errors. I remove those modules and put the other half in, and I still get the same errors, though slightly more of them. I put them in different memory slots (2 and 4 instead of 1 and 3), and I still get errors. I test with just one module in slot 4, then with just one module in slot 1, and always I get very similar errors.

What the errors mean I've got no idea (I don't know much about memory or memtest86+), but I keep seeing variations of 0007feedf18 (or 0003feedf18) in the "failed address" column, and 7feedf18, 3feedf18 or 00000000 in the "good" column, and in the "bad" column always a value that's 00000001 higher than the "good" column.

I also get the impression that it's not the memory modules that are bad, but that it's something else. But what? I don't know enough about memory to tell what the other options could be. Motherboard? CPU?

Anyway, these are the relevant parts of my system:

GA-EP35-DS3
Intel E8400 (really hard to find a replacement if this is the culprit),
4x1GB Corsair Twin2x2048-6400 C4 DHX

Less relvant parts of the system:
Antec Solo case, Scythe Ninja cpu cooler, Peak Radeon HD3850 GPU with Accelero S1 cooler, 2 Samsung Spinpoint T HD501J, Seasonic S12II 380 PSU, Pioneer 215D optical drive.

 

[DSF]

Did you manually set the memory voltage? If I'm not mistaken, those sticks are supposed to be fed 2.1V, but at default the Gigabyte board is only going to give them 1.8-1.9V.

 

[mcv]

I had it on auto, but I just tried it with +0.3V, but unfortunately Memtest still gives me very similar errors. Are there any other voltages I could be playing with? Should I raise this one even more? I already see all sorts of flashing messages, and RAM +0.3V has a warning pink colour.

I'm a bit disappointed that a quality motherboard doesn't detect these settings properly.

(I'm currently watching House on TV, and suddenly I feel a bit like him: trying to diagnose a fatal but unidentifiable problem with my memory.)

 

[n7]

I don't know of any motherboard that can magically alter vdimm to match what the RAM is rated for.
That's not how it works.

When you buy anything but lower end value RAM, you need to be able to apply the correct voltage, timings, & ratios in the bios...or it's not likely to work as well as it could or as it should should.

If you are getting errors with only one dimm installed & set to to the appropriate vdimm, i'd be a little concerned.

With four dimms installed, you might need to bump up NB/MCH voltage to remain stable, since four dimms puts a lot of stress on the NB.

But if just one is erroring out, that's likely bad RAM.

Test them individually in the first slot with vdimm set correctly & perhaps one extra notch higher vNB (shouldn't be needed for one dimm, but shouldn't hurt either).

 

[mcv]

Originally posted by: n7
I don't know of any motherboard that can magically alter vdimm to match what the RAM is rated for.
That's not how it works.

When you buy anything but lower end value RAM, you need to be able to apply the correct voltage, timings, & ratios in the bios...or it's not likely to work as well as it could or as it should should.

I was under the impression that's what SPD was for. If not, then what is SPD actually?

If you are getting errors with only one dimm installed & set to to the appropriate vdimm, i'd be a little concerned.

That seems to be the case.

With four dimms installed, you might need to bump up NB/MCH voltage to remain stable, since four dimms puts a lot of stress on the NB.

But if just one is erroring out, that's likely bad RAM.[/quote]

The thing is, they're all bad, whether tested individually or all four together. And the less modules I use, the more errors I find, which perplexes me even more. With all four, I quite consistently get 185-187 errors each pass, with only two modules, it's a lot less consistent and often more than 200, and with a single module, I've seen over 400 errors once or twice (I've been doing memtests all day now).

I've got the feeling that it's not the memory modules that are faulty, but I've got no idea what could possibly explain this.

Test them individually in the first slot with vdimm set correctly & perhaps one extra notch higher vNB (shouldn't be needed for one dimm, but shouldn't hurt either).

I'll try it again with higher vNB.

 

[BlueAcolyte]

Well it's either your motherboard or RAM crapping out on you.

 

[taltamir]

its only DDR2 where all the "high end" modules are basically overclocked and are NOT stable at the standard requirement of 1.8v
SPD allows the memory to specify timings and speeds, but not voltage. So almost all have an SPD fallback for people who don't set things manually.
For example my ram is rated for 5-5-5-12 @ 2.0v @ 400(x2, aka DD800) mhz
it's spd calls for 5-5-5-18 @ 1.8v @ 400(x2, aka DDR800) mhz
However, there is EPP which allows more settings, INCLUDING voltage. But some motherboards would only read the EPP speed settings but not voltage, leaving you with the faster timing but lower voltage, where trouble could occur.

So if I just plug it into most motherboards it will work without any errors... at 1.8v 5-5-5-18. with some it will be at 1.8v 5-5-5-12 (which still can end up error free depending on component quality)
There is a chance that your mobo is messing things up, so check that they match, but there is also a chance you got a faulty unit, it happens. It is called dead on arrival. If you manually set it to match their requirement and it is not stable according to memtest then replace it (RMA or return to store and buy another one)

 

[mcv]

When I set it manually to 2.1v and I keep the timings quite conservative (5-5-5-18), I get hundreds of errors in a single memtest pass. And each module is the same. How could my mobo be messing it up? Is it because I bought memory that's not on the QVL? Some people say it's not a problem if you buy memory not on the mobo's QVL, others say that's asking for trouble. I believed the latter group, because the QVL was a bit limited to my taste, but I'll guess I'll return these and ask for replacements that are on the QVL. Some plain boring Kingston, I suppose.

 

[taltamir]

http://www.corsair.com/_datash...IN2X2048-6400C4DHX.pdf

This is your ram... it is rated for:
Tested at latency settings (4-4-4-12) at 2.1V
SPD programmed at JEDEC standard 5-5-5-18 values at 800MHz (that would be at 1.8v).

Don't loosen the timings too much. If you set it too slow but with the voltage too high it will also not work... that being said. Corsair makes quality memory and this is their high end DHX line... that should be rock solid even at 4-4-4-12 @ 1.8v, 2.2v, and anything in between (assuming your mobo and psu aren't utter crap). Seeing as you have a seasonic and a gigabyte I would say they aren't. So definitely defective. The only question, is it the ram or the mobo that is defective. MOST likely the ram.

But you could test it in 2.1v 4-4-4-12 if you really want to... just make sure you are setting the correct things to those values, if you set the wrong items then it can really mess things up. But overall it seems like your stuff is fubared and needs replacing.

Just remember that memtest should never give even a single error, no matter how long it has been running.

 

[nevbie]

Did you disable the "legacy USB support" or something like that in BIOS?

 

[mcv]

Originally posted by: nevbie
Did you disable the "legacy USB support" or something like that in BIOS?

Not that i know. I did enable USB keyboard and mouse support, because I noticed that otherwise my USB keyboard and mouse wouldn't always work.

 

[mcv]

Originally posted by: taltamir
http://www.corsair.com/_datash...IN2X2048-6400C4DHX.pdf

This is your ram... it is rated for:
Tested at latency settings (4-4-4-12) at 2.1V
SPD programmed at JEDEC standard 5-5-5-18 values at 800MHz (that would be at 1.8v).

Don't loosen the timings too much. If you set it too slow but with the voltage too high it will also not work... that being said. Corsair makes quality memory and this is their high end DHX line... that should be rock solid even at 4-4-4-12 @ 1.8v, 2.2v, and anything in between (assuming your mobo and psu aren't utter crap).

Well, they didn't work at 4-4-4-12 @ 1.8V and they didn't work at 5-5-5-18 # 2.1V. Nor at 5-5-5-18 @ 1.8V. I didn't test 4-4-4-12 $ 2.1V, but considering the consistency of the errors, I doubt it matters.

Seeing as you have a seasonic and a gigabyte I would say they aren't. So definitely defective. The only question, is it the ram or the mobo that is defective. MOST likely the ram.

I really hope so. It's a lot easier to replace.

But you could test it in 2.1v 4-4-4-12 if you really want to... just make sure you are setting the correct things to those values, if you set the wrong items then it can really mess things up.

That is indeed an issue. Those BIOS options can be extremely cryptic so I don't touch them until I think I have some idea of what they do.

But overall it seems like your stuff is fubared and needs replacing.

Just remember that memtest should never give even a single error, no matter how long it has been running.

That's what I thought. And considering the enormous number of errors, and the consistency of these errors, I was wondering if it couldn't be something else instead, like the mobo, or perhaps even the CPU cache (all data passes through there, so if that fucks up, everything fucks up, right?). If it's the memory, then it looks like it's completely fried.
At least to my layman's eye.

 

[myocardia]

There's nothing wrong with your RAM, there's something wrong with your motherboard, assuming you have the BIOS set up correctly, which it sounds like you may very well not.

 

[taltamir]

Originally posted by: nevbie
Did you disable the "legacy USB support" or something like that in BIOS?

that has NOTHING to do with memtest errors...

Some motherboards incorrectly load stuff to mem early on when usb keyboard and mouse are supported. So when memtest attempts to boot it overwrites that data and causes a hardlock of the system (instead of memtest loading).

In my old gigabyte mobo it happened with usb legacy support, in my new gigabyte mobo it no longer happens with USB legacy support (they now properly register that part of the ram as being used). However it happens if I have RAID mode enabled on the mobo.

If memtest actually loads instead of freezing/crashing the computer when it detects the cd (but before memtest loads) then it will have no errors due to the usb stuff

 

[taltamir]

Originally posted by: mcv

Originally posted by: taltamir
http://www.corsair.com/_datash...IN2X2048-6400C4DHX.pdf

This is your ram... it is rated for:
Tested at latency settings (4-4-4-12) at 2.1V
SPD programmed at JEDEC standard 5-5-5-18 values at 800MHz (that would be at 1.8v).

Don't loosen the timings too much. If you set it too slow but with the voltage too high it will also not work... that being said. Corsair makes quality memory and this is their high end DHX line... that should be rock solid even at 4-4-4-12 @ 1.8v, 2.2v, and anything in between (assuming your mobo and psu aren't utter crap).

Well, they didn't work at 4-4-4-12 @ 1.8V and they didn't work at 5-5-5-18 # 2.1V. Nor at 5-5-5-18 @ 1.8V. I didn't test 4-4-4-12 $ 2.1V, but considering the consistency of the errors, I doubt it matters.

Seeing as you have a seasonic and a gigabyte I would say they aren't. So definitely defective. The only question, is it the ram or the mobo that is defective. MOST likely the ram.

I really hope so. It's a lot easier to replace.

But you could test it in 2.1v 4-4-4-12 if you really want to... just make sure you are setting the correct things to those values, if you set the wrong items then it can really mess things up.

That is indeed an issue. Those BIOS options can be extremely cryptic so I don't touch them until I think I have some idea of what they do.

But overall it seems like your stuff is fubared and needs replacing.

Just remember that memtest should never give even a single error, no matter how long it has been running.

That's what I thought. And considering the enormous number of errors, and the consistency of these errors, I was wondering if it couldn't be something else instead, like the mobo, or perhaps even the CPU cache (all data passes through there, so if that fucks up, everything fucks up, right?). If it's the memory, then it looks like it's completely fried.
At least to my layman's eye.

Hah, sorry, those are DELAY TIMINGS... so the lower they are, the faster the ram... so 4-4-4-12 is FASTER then 5-5-5-18... so you have been testing them backwards. It is supposed to be 4-4-4-12 @ 2.1v, and 5-5-5-18 @1.8v
Not the other way around... that being said. It should be stable at any combination of those considering the quality of your mobo, ram, and PSU (they are all top notch companies).

Also it should have been rock solid at the defaults of 5-5-5-18 @1.8v as set by the SPD. a percent or two slower then the ram is rated for, but rock stable. Try setting your motherboard to defaults. And then use CPU-z to get a reading of what your settings are in default. If those are indeed 5-5-5-18 @ 1.8v as it should default to (And you had errors in), then you have definitely got faulty ram or mobo... If you replace the ram with similar part and it is also bad, try replacing mobo.

If you installed the OS with ram that errors, or have installed any windows updates, then you WILL have corrupt OS files and they will cause system stability issues, that will get worse over time. A clear format is generally recommended when you fix a ram or CPU error related issue.

 

[DSF]

Originally posted by: taltamirSeeing as you have a seasonic and a gigabyte I would say they aren't. So definitely defective. The only question, is it the ram or the mobo that is defective. MOST likely the ram.

Considering that every stick produces errors regardless of voltage or timings, I would be more inclined to blame the motherboard than the RAM.

mcv, do you have any friends or family you can borrow a working stick of DDR2 from? Size really doesn't matter, the only purpose would be to pop it in and run Memtest on it. If it passes error-free, then I would blame the RAM. If it doesn't, then I'd blame the motherboard. Just because Gigabyte is a good company doesn't mean that they aren't going to put out a bum board once in a while. It happens with any company.

 

[mcv]

Originally posted by: taltamir
Hah, sorry, those are DELAY TIMINGS... so the lower they are, the faster the ram... so 4-4-4-12 is FASTER then 5-5-5-18... so you have been testing them backwards. It is supposed to be 4-4-4-12 @ 2.1v, and 5-5-5-18 @1.8v
Not the other way around... that being said. It should be stable at any combination of those considering the quality of your mobo, ram, and PSU (they are all top notch companies).

Also it should have been rock solid at the defaults of 5-5-5-18 @1.8v as set by the SPD. a percent or two slower then the ram is rated for, but rock stable. Try setting your motherboard to defaults. And then use CPU-z to get a reading of what your settings are in default. If those are indeed 5-5-5-18 @ 1.8v as it should default to (And you had errors in), then you have definitely got faulty ram or mobo... If you replace the ram with similar part and it is also bad, try replacing mobo.

I did test them with 5-5-5-18 @ 1.8V. I think that was the second test I did. After 4-4-4-12 @ 1.8V failed, I decided to set them back to their less agressive initial values.

If you installed the OS with ram that errors, or have installed any windows updates, then you WILL have corrupt OS files and they will cause system stability issues, that will get worse over time. A clear format is generally recommended when you fix a ram or CPU error related issue.

I'm amazed that with these memory errors, installing the OS didn't crash halfway through, but reinstalling it won't be a problem.

 

[mcv]

Originally posted by: DSF

Originally posted by: taltamirSeeing as you have a seasonic and a gigabyte I would say they aren't. So definitely defective. The only question, is it the ram or the mobo that is defective. MOST likely the ram.

Considering that every stick produces errors regardless of voltage or timings, I would be more inclined to blame the motherboard than the RAM.

Considering the errors and the fact that each stick had the same errors, I was also worried that it might be something else, but a friend said that it's not impossible that it's simply a bad batch. Or maybe an incompatibility problem with my mobo, which means I'm still going to replace the memory, not the mobo. But what if it really is a problem with the mobo?

mcv, do you have any friends or family you can borrow a working stick of DDR2 from? Size really doesn't matter, the only purpose would be to pop it in and run Memtest on it. If it passes error-free, then I would blame the RAM. If it doesn't, then I'd blame the motherboard. Just because Gigabyte is a good company doesn't mean that they aren't going to put out a bum board once in a while. It happens with any company.

That's a good idea to check. I'm not sure there's any DDR2 memory I can borrow on short notice, but I was already planning on buying the cheapest memory stick on the QVL, just so I can do some work on it tomorrow, while I'm waiting for better memory to arrive.
If I can quickly get 1 GB of cheap memory for $20, that wouldn't be too big a loss.

 

[mcv]

Assuming my Corsair Twin2x2048-6400 C4 DHX was a really bad choice, I've been looking around at other, hopefully more suitable memory modules. What I'm wondering is, should I just avoid weird DHX memory, or should I avoid everything that's not on the QVL? What I mean is, the QVL mentions Kingston KHX6400D2LLK2/2GN. Is this the same as the KHX6400D2LLK2/2G that I can see on the Kingston site? Apparently (browsing around on the Kingston site), that extra N means SLI-ready. I don't need SLI, can I simply take the one without the extra N? Why is that one not on the QVL?

Another very attractive one could be KVR800D2N5K2/4G, because it's a 2x2GB kit: easier to expand, and according to some people it uses less power and might be more stable than 4x1GB, but I have no idea if that's true. Anyway, it's not on the QVL, but there is a KVR800D2N5/1G on there. Is that sufficiently similar? I've got no idea, and now that I got burned, I'm a bit scared to deviate too much from the QVL.

 

[taltamir]

the "SLI ready" is an nvidia gimmic name that means a lot more then it indicates. On a VERY specific combination of RAM and motherboard you will get automatic and SIGNIFICANT overclock to your ram, that is tested to be stable with that very specific ram + motherboard combo.

Your motherboard uses an intel northbridge so it is of no special use for you, and you will just be overpaying.

Honestly, you really DON'T need super high end DD2 ram, the difference in performance is negligeble. Last I checked you could get almost any quality DDR2 ram for 70$-80$ for a 2x2GB. 2x2 is a better idea then 4x1 in terms of stability and compatibility. But if all four sticks individually error then indeed the chances that they are all defective is low. It is hopefully an incompatibility. If not, then its a defective mobo. You should go by fries and get some DDR2-800 ram of any decent brand and check it, if it doesn't work you can easily return it to the store for full refund without dealing with RMA and shipping fees and waiting. You could even try several different brands. If it works, you can still return and buy online, or just keep it and be happy (is it really worth your time to save 10$ by returning it and then ordering online?).

Anyways, I like patriot, it tends to be very stable and compatible. American made too.

Now you have an E8400 and a GA-E35-DS3...
The E8400 should OC from 3 to 4GHz, and that mobo is an OC board that gigabyte actually claims is good for those speeds... You just need to get:
1. Faster ram, as in DDR2 1066 ram...
2. A good cooler

http://www.newegg.com/Product/...e=DDR2+1066+(PC2+8500)

But I am not too sure about this. I am not really into the OC scene myself. But the difference between 4-4-4-12 and 5-5-5-18 DDR-800 ram isn't all that significant, not compared to getting some ram like that.

 

[mcv]

Originally posted by: taltamir
the "SLI ready" is an nvidia gimmic name that means a lot more then it indicates. On a VERY specific combination of RAM and motherboard you will get automatic and SIGNIFICANT overclock to your ram, that is tested to be stable with that very specific ram + motherboard combo.

Your motherboard uses an intel northbridge so it is of no special use for you, and you will just be overpaying.

But how different is it from the non-SLI-ready version? If KHX6400D2LLK2/2GN is on the QVL, shouldn't KHX6400D2LLK2/2G be on there too?

Honestly, you really DON'T need super high end DD2 ram, the difference in performance is negligeble. Last I checked you could get almost any quality DDR2 ram for 70$-80$ for a 2x2GB. 2x2 is a better idea then 4x1 in terms of stability and compatibility.

That's what I'm currently looking for. A shop in my neighbourhood that has good 2x2GB in stock so I don't have to wait another day.

Now you have an E8400 and a GA-E35-DS3...
The E8400 should OC from 3 to 4GHz, and that mobo is an OC board that gigabyte actually claims is good for those speeds... You just need to get:
1. Faster ram, as in DDR2 1066 ram...
2. A good cooler

Faster ram? Are you sure? As I understand, I already need to overclock my FSB in order to make full use of 800MHz ram, and I should actually be looking at 667MHz ram. Memory bus can run faster than the FSB, but from what I've understood so far, I get the impression that most of the time you still have to wait for the FSB, so there's little noticable gain.

But I am not too sure about this. I am not really into the OC scene myself. But the difference between 4-4-4-12 and 5-5-5-18 DDR-800 ram isn't all that significant, not compared to getting some ram like that.

I was under the impression that it was the other way around, because a fast memory bus doesn't help you much with a slow FSB. Ofcourse I can overclock my FSB, but even if I do that, I don't think I want to overclock it beyond 400 MHz (in sync with 800Mhz memory), because I want passive cooling.

Perhaps I should simply be looking at 667Mhz memory even. On the other hand, it could be that I've completely misunderstood how memory speeds work.

 

[Syzygies]

I had these exact sticks on a Gigabyte GA-EP35-DS3P mobo build this month, and this exact problem; I cringed as I started to read this thread.

I sorted it out, but only after ordering 4-4-4-15 OCZ Reaper HPC Edition 2 x 2GB (OCZ2RPR800C44GK) and setting up an RMA for the Corsair sticks. I went through with the RMA, and I love the OCZ sticks, but I did learn how to make the Corsair sticks work.

You have to completely give up the idea that the world is looking out for you, to use this Gigabyte board family with these sticks. Then the mobo is a dream to configure and overclock. The key is to understand and disable absolutely every friggin' "helpful" performance setting in the BIOS, and drive stick-shift.

Corsair explained to me that the SPD settings are only meant to get the machine into the BIOS on the default voltage of 1.8V, so you can pick correct settings with 2.1V. Their tech support is great. As lame as this answer sounds, some memory sticks require that you first boot with one stick, or worse, a cheap standard stick, just to prepare the BIOS for two sticks of the memory you bought. This is better. Errors until you have your BIOS settings exactly right don't mean you have bad memory; both sticks showing bad in the same place is the giveaway.

Are you on the latest BIOS?

My bet is that your "Performance Enhance" has been left at its default of "Turbo", and that's your problem. The shaky English makes it sound like only a loser would settle for "Standard", when a better translation would read "No, don't botch the Static tRead Value (tRD) by making a bad guess, let me take care of this!"

Set "Performance Enhance" to "Standard", and set the voltage and timings as Corsair asks. You'll be fine. When you get over this bad experience, you'll be overclocking in days. Just remember to disable any and all "help" Gigabyte offers.

To learn about tRD, read Kris Boughton's recent writings on AnandTech. A good place to start is "Overclocking Intel's New 45nm QX9650- The Rules Have Changed", starting with the section "The Origins of Static Read Control Delay (tRD)".

After reading the section on Voffset and Vdroop, you'll also want to disable "Load Line Calibration". As I wrote in my first post, much of what one finds on the web reminds me of gifted rock stars who can't read sheet music. In the fog of not really understanding the underlying electronics at a basic level, like having read "The Art of Electronics", this last recommendation is not without controversy.

 

[mcv]

Originally posted by: Syzygies
Errors until you have your BIOS settings exactly right don't mean you have bad memory; both sticks showing bad in the same place is the giveaway.

Maybe so, but it does mean that it's awfully hard to get them working.

Are you on the latest BIOS?

I hope so. I don't know how often BIOS updates appear (or even how to actually update my BIOS, for that matter), but it's a pretty new motherboard, and the latest version of its kind.

My bet is that your "Performance Enhance" has been left at its default of "Turbo", and that's your problem. The shaky English makes it sound like only a loser would settle for "Standard", when a better translation would read "No, don't botch the Static tRead Value (tRD) by making a bad guess, let me take care of this!"

Set "Performance Enhance" to "Standard", and set the voltage and timings as Corsair asks.

I did this, but it doesn't work. I still get tons of errors.

You'll be fine. When you get over this bad experience, you'll be overclocking in days. Just remember to disable any and all "help" Gigabyte offers.

Are there more "helpful" functions I need to turn off? And what do these helpful functions actually do anyway?

After reading the section on Voffset and Vdroop, you'll also want to disable "Load Line Calibration".

What is that? I can't find any option that sounds anywhere close to that.

I'm glad (in a way) that I'm not the first to experience this problem. But even if it is possible to get this memory working, I think it's just too hard. I think it was a mistake to buy Corsaid DHX memory, and a mistake to buy anything that requires more than 1.8V.
The idea of a 20% lower CAS latency appealed to me, but apparently you have to make sacrifices to get that, and it's not like I'm actually going to notice it.

But I've done more testing now. I used a completely different brand of memory (dirt cheap 1 GB 667MHz Nanya), and I still got the same errors. I'm not sure how much this says exactly, since this particular version of Nanya is not on the QVL (it's surprisingly short), and Nanya itself has a completely useless website (so I can't check compatibility there either), but I think this points to a problem with something other than the memory. Perhaps the mobo?

Anyway, I did an RMA request, and tomorrow I'm going to send it back to the shop. I'm afraid I won't be able to use my PC this week. I also asked if they could replace the memory with something more standard and reliable, preferably KVR800D2N5K2/4G, which, while not on the QVL, looks pretty good to me.

 

[LOUISSSSS]

have u tried resetting the CMOS for a few hours? i had the same problem... memtest86 errors all over the place, single stick, 2 sticks, 3 sticks, 4 sticks. reset the cmos and ended up only having problems with 1 x 1gb stick of Crucial Ballistix ram.

after testing each indiv stick in each dimm socket on the mb for 1 full pass of memtest86, i'm now up to 2 sticks (of the 3 good ones) for 1 full pass of memtest 86, next will be 3 sticks for 1 full pass.

after finishing that i hope to go into windows and do some OCCT RAM testing... i failed that lasttime too!

 

[anyweather]

I have the EP35-DS3 as well (EP35-DS3R specifically) and I encountered the same kind of problem. No matter what settings/timings/vdimm I use, memtest always fails test#3 and keeps reporting errors.

What I found out later was the bios version that is messing up with memtest. I was on F1 bios. So I flashed it to the F2 bios and all errors were gone. I ran memtest for 8hrs and there was no more errors.

So try the F2 bios, if you haven't done so already and see if it makes a difference. It did for me. I'm running a E3110 with 2x1gb G.Skill 8500CL5 on EP35-DS3R.