680i chipset burning up RAM? Nvidia responds!

[Quakindude]

Important message regarding memory and the EVGA 680i motherboard

NVIDIA has investigated end user reports of high performance DIMM failures on the NVIDIA nForce 680i SLI-based platforms. During this process we have been in close contact with DIMM manufacturers and the DRAM manufacturers they rely on to understand the failure scenario. By working with our community, we believe that the observed failure is a breakdown of the silicon in the DRAM caused by the prolonged application of 2.4V on the voltage rails of the DIMMs.

NVIDIA’s own internal testing has observed this failure on multiple motherboards using different chipsets (both NVIDIA and non-NVIDIA chipsets). This issue is not directly related to motherboards using the NVIDIA nForce 680i SLI MCP or other chipsets.

If you are using this type of memory and are experiencing this issue, NVIDIA recommends contacting your memory manufacturer or system manufacturer for additional information and warranty information.

You can see it for yourself here. Seems like it may be a quality control issue with the silicon used by memory manufacturers rather than a chipset issue. Of course, the Dominator PC2-8500 memory I'm running is at 2.2V. I don't like overvolting my components at all due to past experiences like this. Plus, it isn't covered under warranty when you do.

[gvblake22]

Wow, I knew 2.4v is overvolting, but I didn't think it was enough to fry the RAM! All the more reason to not be crazy with voltage; VOLTAGE KILLS!

[Lead Head]

There are reports of ram rated for 2.4v, that have been fried in the 680i. When people ran those same sticks at 2.4 in other mobos. Sounds more like nVidia has no idea whats wrong, so they are pinning the problem on someone else.

Thats how I look at it atleast.

[Capper]

Originally Posted by Lead Head

There are reports of ram rated for 2.4v, that have been fried in the 680i. When people ran those same sticks at 2.4 in other mobos. Sounds more like nVidia has no idea whats wrong, so they are pinning the problem on someone else.

Thats how I look at it at least.

Whoa, whoa, whoa......if you are going to put stuff like this out there.....cite your sources....do not say "reports" if you are referring to a forum thread at various overclocking communities. First of all, you have to consider the test enviroment (Do these guys know what they are doing)? Then you have to consider the history (Were these guys doing it right? Were they pushing things too far, then backed off when they had problems?). I've become kinda cynical with stuff like this, and I have a hard time trusting people when they say they never voided their warranty or blew something up themselves. I'm not calling anyone dishonest, but have seen people put information out there based on rumor and tainted sources.

On the flip side, I've had problems with the 680i boards I'm working with running any memory in excess of DDR2-1000, to the point where I won't use the chipset on the test bench anymore. eVGA people are telling users to underclock their memory and tighten the timings....which works, but won't make a lot of overclockers happy....this after the whole LinkBoost debacle.

The 680i board has really shown some strange characteristics to me......I'm currently testing a DDR2-1000 kit, that fails memtest at rated speed (5-5-5-15), but will run DDR2-984 @ 4-3-3-8. On an INTEL i975 board that same kit of memory will only hit DDR2-836 @ 4-3-3-8, but maxes out at DDR2-1032......all at the default voltage of 2.2V As to stability, I call memory stable after 25 full passes of Memtest86, 3 runs of Super Pi to 32M, and our full range of benchmarks

Originally Posted by gvblake22

Wow, I knew 2.4v is overvolting, but I didn't think it was enough to fry the RAM! All the more reason to not be crazy with voltage; VOLTAGE KILLS!

Lets also consider that the rated voltage for DDR2 is 1.8V....and 2.4V and beyond is an awful lot of voltage. Look back at the end of DDR memory......there were enthusiast kits out there with people pushing 3.3-3.6V through them, and killing them....

[Lead Head]

Originally Posted by Capper

Whoa, whoa, whoa......if you are going to put stuff like this out there.....cite your sources....do not say "reports" if you are referring to a forum thread at various overclocking communities. First of all, you have to consider the test enviroment (Do these guys know what they are doing)? Then you have to consider the history (Were these guys doing it right? Were they pushing things too far, then backed off when they had problems?). I've become kinda cynical with stuff like this, and I have a hard time trusting people when they say they never voided their warranty or blew something up themselves. I'm not calling anyone dishonest, but have seen people put information out there based on rumor and tainted sources.

On the flip side, I've had problems with the 680i boards I'm working with running any memory in excess of DDR2-1000, to the point where I won't use the chipset on the test bench anymore. eVGA people are telling users to underclock their memory and tighten the timings....which works, but won't make a lot of overclockers happy....this after the whole LinkBoost debacle.

The 680i board has really shown some strange characteristics to me......I'm currently testing a DDR2-1000 kit, that fails memtest at rated speed (5-5-5-15), but will run DDR2-984 @ 4-3-3-8. On an INTEL i975 board that same kit of memory will only hit DDR2-836 @ 4-3-3-8, but maxes out at DDR2-1032......all at the default voltage of 2.2V As to stability, I call memory stable after 25 full passes of Memtest86, 3 runs of Super Pi to 32M, and our full range of benchmarks

I would cite the sources, but I think they would conflict with the forum rules to? So I am not sure really what to do in a situation like that. But there are people who have said that even with ram with a 2.4v default voltage, the 680i has killed them. Even some not voltage bumped memory has been killed. Its not like every board kills ram though, its just that a few of the enthusiasts boards killed the ram, But even people running 2.1-2.3v have had their ram die. So who knows . Could have just been a bad batch of mobos

[gvblake22]

Originally Posted by Capper

Lets also consider that the rated voltage for DDR2 is 1.8V....and 2.4V and beyond is an awful lot of voltage.

Well it depends a lot on the particular memory set. I've seen different DDR2-800 CAS4 kits range in voltage from 1.8 to 2.1 volts. Upon further inspection, Corsair's top end Dominator DDR2-1000 kit is rated at 2.4V and OCZ's DDR2-1000 CAS 4 kit is rated at 2.35V.

But all in all, it is really going to depend on the specific kit and the PCB it is built on. OCZ and Corsair must have put a lot of work into their kits to rate it at 2.35v - 2.4v compared to the usual 1.8v - 2.1v offered on just about everything else.

[Capper]

Go to JEDEC and research the specs....

I'm not quoting yahoos on other forums or hearsay, I talk to these companies on a daily basis, and have been doing this awhile.

Theres a difference between a standard, and what an individual kit is "rated" to run.

Look at instances before DDR1 left us....remember those super enthusiast kits that were released right at the end? Those 3.3-3.4V kits? Most of those were on the shelf for a very short period of time because many crapped out.

Memory is like any other part, and from that point of view GPU cores are rated at a spec, and if you remember there was a big controversy awhile back about companies overclocking GPU cores and them dying on consumers.....if you see my correlation, then you see that memory companies, to make their kits more attractive than their competition, heavily screen their kits....they take the small percentage that will run heavily OCed and release those as an enthusiast part.....those parts are really no different than any other.....they just pass the testing phase......

Lets look back on this in about 6 months, and lets look at how many of these kits are still around and in heavy OCed use.

[Capper]

Also, LH....the reason we don't want people linking or quoting other forums is I honestly don't trust a lot of what is said.....No offense to anyone, and I'm not saying anyone is dishonest or not correct, but I look around to see what people are talking about, and I'm honestly shocked about some of what I see posted around the web ....

I have a real hard time with people who go to a forum, and take any information they see and regurgitate it somewhere else, as they know from experience. Its a bad habit.....

I think we do a really good job managing the information on HL, and make sure that what you read here is true, verified by us personally, or comes from sources we know or trust.

[onebxr]

Because if its not true there are about 8 people that will tell you that its not :) <-----this is what is good about this board, no one lets misinformation flow. I trust this site implicitly and I come here for all my computer needs and then some. I may go to other sites to get different opinions but I always come back here for fact checking, and also to spout off what I have learned. HAHA
-1

[Quakindude]

Something else to keep in mind. It seems to me like the NVidia 6xx chipset motherboards have garnered more attention and buyers than any chipset in the last 5 years that I can think of. I've not seen this level of enthusiasm for a new chipset in a long time. There's really no other chipset with the level of flexibility and enthusiasts oriented build specs that this chipset enjoys.

So why did I say all of that?

I said it all to illustrate why I think a good many people have the very wrong impression that the 680i chipset is eating a ton of memory. There's been a LOT of these boards sold through various manufacturer's. ASUS, MSI, Gigabyte, EVGA, ECS, etc... are all pimping the 600 series chipsets. The 600 series is still selling like mad because it's an awesome performer, highly adept at everything an enthusiast demands of their equipment and let's face it, with SLI capabilities, it punks out other chipsets and takes their electricity money.

Once you consider the high volume of sales these chipsets are enjoying, you should quickly realize that if there's been 500,000 of these boards sold, and there's a 3% failure rate, that means there's 15,000 enthusiasts out there bitching up a storm about it. That leaves 485,000 peeps that aren't getting involved in the discussions about problems because guess what? They aren't having any and don't need the forum resources around the 'Net to express their frustration.