Ever heard of a motherboard model that eats ram?

[Brando]

I went through a never ending fiasco of rma's trying to figure why my system was having issues and thought I finally had it nailed down. In the beginning I had a set of g skill ddr3 1866, same as what I have now, that couldn't run full speed without failing prime. It took awhile to track it down to the ram but finally I rma'd the ram and thought everything was cool. Then more issues popped up and long story short I rma'd lots of stuff and bought a new ssd. Thinking everything was fresh in my system and that it should be trouble free but alas, my replacement ram from newegg is now failing prime95 blend at it's rated speed like the first set. When I got this new set I ran memtest all day while I was at work for 13 hours with no errors. Now maybe a month later it seems to have degraded to the state of the first set and may have caused me to replace other parts without reason. Just because it's not on the approved list for this motherboard doesn't mean it should be killing it right? I've never heard of such a thing. Any idea what's going on? I set the ram to default conservative settings and prime blend is passing now for 15 min when it was failing within a minute at 1866 cas 8 per the xmp profile so the cpu seems to be fine. WTF?! This is the ram. Any ideas?

http://www.newegg.com/Product/Product.aspx?Item=N82E16820231538

 

[moocow]

It wouldn't surprise me at all if the board is killing ram.

 

[Brando]

It wouldn't surprise me at all if the board is killing ram.

How does that work exactly? I already swapped the motherboard out for another of the same model which SHOULD rule out a bad component. Does that mean motherboards destroy ram that isn't on the recommended ram list for that model? I've never bothered with it before and never had one bad set. I just make sure it's running at spec. It says right on the box that it's for Z77.

 

[Brando]

Well this is just weird. I know I had seated the ram properly and decided to triple check it and re do it just to see what happens. It was seated fully but I tried taking it out and putting it back in 3 times. It f#$@king worked (I think). I'm running prime 95 blend with maximum ram with no errors for 7 minutes when it failed within seconds before reseating. Thed only thing I can think of is there must be some kind of anti rust coating or something that had to be rubbed off of the ram slots. It makes sense because when I tested this same kit in my pre-RMA motherboard it passed memtest all day. The only thing different is the motherboard. It's the same model but still had whatever they spray on it at the factory or maybe a bit of corrosion from the salty air on the boat ride over. LOL. This Actually happened with a mag lite I bought once. The damn thing wouldn't turn on no matter how many batteries I tried until I tried the contact points against each other for a minute. Mystery coating for the fail.

 

[Brando]

Crap. Finally got an error on one thread but it went a lot longer than before. I'll try rubbing the contacts more and see if it gets even more stable.

 

[freeagent]

My old P5K Deluxe consumed about 18gb of d9s

Tho I'm not 100% it was entirely the boards fault..

 

[Brando]

I don't know where else to go with this. I already swapped the motherboard so either it's a horrible design or there's something going on with the back plate of my megahalems touching the back of the board shorting stuff out. Prolimatech swears the insulating material on the plate keeps this from happening but it's the only other factor I can think of unless the cpu is bad and killing the ram somehow or I really did get 2 bad sets of memory in a row which seems almost impossible. This is lame. I might actually have to buy a whole new megahalems rev b to get a proper back plate that doesn't brush the back of the mobo.

 

[PolRoger]

That sounds like an odd and frustrating predicament to have to deal with. Did you try testing your combo at stock settings with the OEM cooler?

 

[Rotary]

This seems crazy, If i were in your you, I would just start over with everything.

 

[Brando]

That sounds like an odd and frustrating predicament to have to deal with. Did you try testing your combo at stock settings with the OEM cooler?

That's may be next. When I swap the ram again it'll either be with the stock cooler or a different back plate. Hopefully prolimatech will send me the plate that goes on the mega rev 2 but it's the weekend and they haven't answered yet, plus I don't know if it even fits the original or not. No way am I putting more ram in with that back plate. It may not even be the problem but I have to assume everything at this point.

 

[Brando]

This seems crazy, If i were in your you, I would just start over with everything.

I almost have. Everything but the case, bluray drive, storage hd, and cpu. I've got a new mobo, power supply, video card, ram, and ssd. Time to try a new cooler back plate with more ram I guess. I didn't even see any traces between the plate and the ram though. Just the lower right corner.

[/IMG]

 

[Metlcub]

Did you manually set the Dram voltage? I had a issue with non vender approved ram on a older set up that was setting the ram at 1.65v and not it's rated 1.5v. that could slowly kill ram

 

[Brando]

Did you manually set the Dram voltage? I had a issue with non vender approved ram on a older set up that was setting the ram at 1.65v and not it's rated 1.5v. that could slowly kill ram

XMP profile. Everything checked out in the bios for timings and voltage. 8-9-9-28 with 1.5v I think.

 

[larrymoencurly]

In the beginning I had a set of g skill ddr3 1866, same as what I have now, that couldn't run full speed without failing prime.

What's the full speed of that 1866 MHz RAM? I don't mean what G.Skill says the speed is but what the manufacturer of the chips says because G.Skill has shipped products with chips having 1066 MHz JEDEC profiles but with XMP profiles 30%-60% faster. Some XMP profiles can be wildly optimistic, and G.Skill has been known to overclock without testing very thoroughly (they don't have the equipment to do that). Then there's the matter of some module companies shipping products that have produced up to 2 error bits during testing, at least for products with XMP profiles slower than 2133 MHz.

Is it also possible that the backing plate for that Megahalems heatsink is creating capacitance that's fouling up the signals running through the copper traces next to it? I might try raising the heatsink a bit away from the motherboard by placing a plastic or fiber washer at each screw hole.

 

[Brando]

What's the full speed of that 1866 MHz RAM? I don't mean what G.Skill says the speed is but what the manufacturer of the chips says because G.Skill has shipped products with chips having 1066 MHz JEDEC profiles but with XMP profiles 30%-60% faster. Some XMP profiles can be wildly optimistic, and G.Skill has been known to overclock without testing very thoroughly (they don't have the equipment to do that). Then there's the matter of some module companies shipping products that have produced up to 2 error bits during testing, at least for products with XMP profiles slower than 2133 MHz.

Is it also possible that the backing plate for that Megahalems heatsink is creating capacitance that's fouling up the signals running through the copper traces next to it? I might try raising the heatsink a bit away from the motherboard by placing a plastic or fiber washer at each screw hole.

How do I find out what the chips are actually rated for as opposed to what they're selling them as? Also, why would they pass stability tests for an entire day and then less than a month later can't even go a few minutes? As far as the back plate it doesn't actually push against the motherboard but rather the raised metal plate behind the cpu. I would have to create some distance with a large square spacer like the ones they give you for 775 and AMD (I think). They warn against this though because it causes too much pressure to the cpu. I'm thinking about electric tape on that entire side, just a paper thin layer, for some extra protection. Or maybe something I can spray on. The weird thing is that the traces going towards the ram aren't even on that side of the board so idk wtf is going on.

 

[Brando]

Welp, found the memory on the gskill website and it shows it working with my mobo. At least I can rule incompatibility out. It says it's rated for 1600mhz at 1.5v SPD rating so they may just need to bin them better.

http://www.gskill.com/products.php?index=462

 

[Sabishiihito]

My old P5K Deluxe consumed about 18gb of d9s

Tho I'm not 100% it was entirely the boards fault..

Hah I think I've heard of that. My Maximus Formula ate its share of D9 sticks too.

 

[larrymoencurly]

How do I find out what the chips are actually rated for as opposed to what they're selling them as?

The only way is by reading the labels printed on the chips, meaning any heatsink has to be removed (and the warranty voided and chips possibly ripped from the circuit board), but APHnetworks.com found some G.Skills were made from Hynix chips with a speed rating of H9, or 1333 MHz, even modules that G.Skill rated for faster speeds, up to 2133 MHz. With some other modules, the chips had no factory markings but only G.Skill's, making it impossible to determine the real speed ratings.

The back plate doesn't have to short against the motherboard to hurt the signals because even the capacitance from the metal can do that. So I'd install a cardboard or plastic washer, at least 1mm thick, under each screw hole, without covering the whole back plate area.

The real puzzler is why the G.Skill memory didn't go bad until a month has passed and then stayed bad. Maybe the memory bus was loading down the signals too much, or the signals going to the memory had voltage spikes that exceeded the absolute maximum allowed 1.975V. I'd also measure the DIMM voltage with a digital multimeter, but an oscilloscope is the only way to check for voltage spikes.

Have you tried that G.Skill memory in a motherboard of a different design or brand? Or how about testing with known good memory (chips marked with actual manufacturer's part numbers, no overclocking) in your current motherboard?

Have you tried different DIMM slots and run other diagnostics, such as Gold Memory, which beat MemTest86 in RealWorldTech.com 's comparison of diagnostics? (only PhD's RST products did better) One person had memory that always passed MemTest86/86+, but Gold Mem found one bad bit. It had to run another 9-10 hours before reporting that error again. Also I test the memory at 50 Celcius case temperature (hard disk disabled, fast video card replaced with a low power card). To run even hotter, cover each DIMM with anti-static material, like pink foam wrap or pink bubble wrap. The chips are rated for operation at up to 95 Celcius.

 

[moocow]

Have you checked the memory voltage with a DMM?
Pretty much all of my Z77 boards have undervolted by a little bit and have caused instability.

 

[Brando]

I haven't tried other memory in my board but I can probably try this ram in somebody else's pc if I try. I'd be afraid of messing up somebody else's stuff. Haven't tried a dimm either but maybe I can borrow one. The easiest thing might be to just turn the viltage up on these a tiny bit but I don't want to hurt the cpu and I'm not 100% sure about going over 1.5v. I see recommended ram with 1.65v for this mobo and read 1.5v max in other places. By "under each screw hole" do you mean on the back of the motherboard between the plate and the mobo around the little peg that goes through the hole? The screws don't actually go through the motherboard because the pegs that that screws go in to stick through the motherboard a little.