PDA

View Full Version : DDR2-800/PC2-6400: A Brick Wall


Das Capitolin
09-22-2006, 01:21 AM
I have been struggling all week to fing some DDR2-800 RAM that works right out of the box. It seems as though DDR2-800 is a barrier point for RAM, because it is not compatible with everything. In fact, compatibility is a major issue for this RAM.

At this moment, DDR2-667 has very little issues with almost every motherboard on the market that supports it. Conversely, motherboards that claim to support DDR2-800 RAM have long lists of problems and incompatabilities.

Has anyone else noticed this yet? I have been stranded all week trying to put together a dozen Core 2 Duo systems, and so far OCZ and G.Skill have both failed.

markkleb
09-22-2006, 01:33 AM
I like Geil, but check this out..
http://www.anandtech.com/mb/showdoc.aspx?i=2797&p=18

They say the Value ram works very closely to the HI Perf stuff..

Kougar
09-22-2006, 02:49 AM
Corsair's RAM failed?? I would very much like to know the details, because their PC2-6400 memory uses some of the lowest voltages out there, and it OCs out of spec pretty well with little to no volts.

I will say I am just now realizing that Intel 975X boards do not natively support DDR2-800 RAM. In most cases the boards will default to DDR2-667 RAM, and OCing will mostly fix the issue...

I have noticed quite a few OCZ and G.Skill users having problems, but to me it seemed almost all the problems were the result of the different kind of RAM chips those companies use that require 2.2v and higher voltage to run, at the settings Corsair memory can handle at 1.9v, even 1.8v.

Das Capitolin
09-22-2006, 04:36 AM
Kougar, I had a mental typo. It was OCZ and G.Skill that have failed so far, and I have ordered Corsair as a replacement. It seems like the 965/975 boards I have been building (for customers- so overclocking is not an option), have not liked the majority of DDR2-800 modules that I try. DDR2-667 seldom has this issue.

I wonder who, if not both, are rushing out their products?

werty316
09-22-2006, 04:42 AM
As far as I know doesn't the C2D support DDR2-667? This could be why although DDR2-800 could come in handy with OC'ing.

Kougar
09-22-2006, 08:47 AM
The C2D doesn't have a memory controller, so it is the chipset that needs to be compliant. Intel 975X is only rated for DDR2-667, 965P is rated for DDR2-800... obviously you can plug almost any rated module into most boards, but only OCing will reach the stock memory speeds.

Das, rather glad it was a typo! I've grown very fond of Corsair and would hate to find out they are having the same issues.

The reason DDR2-800 is likely the problem is because the voltages on 667 and slower memory are within the realm of Corsair's 1.8-1.9v range. Many of the "cheaper" memory chips need the voltage ramped way above official DDR2 specs when pushed to 800mhz...

The official JEDEC specs for DDR2 RAM start at 1.8v, hence why many modules that need 2.1v and 2.2v out of the box won't work! I googled up a random Intel/JEDEC PDF on 667/800mhz RAM, and the MAX rated voltage for the spec was supposed to only be 1.9v. if I read it right... ;)

Miles
09-24-2006, 04:08 AM
The Mushkin Redline DDR2-8000 has worked perfectly for me in every Conrow board I've used it in, SOme have had issues trying to run this RAM at <2 volts, from everything I've red and experimented with 2 volts is the best place to start and move upward as necessary from there.

Mushkin's XPS 5300 (DDR667) is also very good RAM and will OC to DDR 1000 without too much difficulty.

Actually most of the Conroe boards specifically state that DDR 667 is the memory of choice...I have used my Redline 8000 in a total of 3 Conroe boards and not the sligtest hint of a problem even at 1.9 volts. There are many BIOS settings on the 975 boards too that can effect your memory and just how well it wil function at a given voltage.

I could help more if I knew what boards you were using and which Conroe chip. So far my biggest issue with the Conroe is cooling, but much of that is due to the fact this is the first processor I've ever owned that idles at > 30 degrees Celsius.

Das Capitolin
09-24-2006, 06:31 AM
I think you guys are missing part of my point: I have seen a large portion of the DDR2-800 RAM that is out there fail because of incompatability. Sure, getting PC2-8000 would be an answer, but I highly doubt that it will be practical for production purpose. I build several systems at a time for clients, sometimes one or two, or sometimes a dozen or more (like this week). Cost is a major factor, and overclocking is not an option. DDR2-800 should work out of the package, to the same extent that DDR2-667 does.

Kougar
09-24-2006, 11:49 AM
Well, it would be nice if that was the case, but nothing ever is anymore I think.

Just be sure to buy RAM that stays in line with the official PC2-6400 JEDEC specs of 1.8v or 1.9v. Motherboard makers are building their boards to this specification, not 2.1v, 2.2v, or 2.3v. It's not unexpected that some motherboard manufacturers, especially on the mid to low end price range will not build in the full tolerances for outside of spec RAM.

Das Capitolin
09-24-2006, 03:42 PM
Thanks Kougar, I am noticing that there are more and more manufacturers making RAM outside of the 1.8-1.9 spec in the DDR2-800 speed more then anything else; which is probably why I am seeing so many problems with this particular speed of DDR2.

GIBSON
09-24-2006, 05:42 PM
I thought most recent motherboards offered enough juice to give these babies a ride
Any idea how much voltage a asus P5B can deliver to the ram. I did a 'lil search yesterday or so but didn't come up with anything fast.

werty316
09-24-2006, 05:44 PM
Not if thay are not within JEDEC specs ;)

The Asus P5B can deliver a max 2.45V (http://enthusiast.hardocp.com/images/articles/1155576258xE5eSgnY3g_3_10_l.gif) to the memory and 1.7V (http://enthusiast.hardocp.com/images/articles/1155576258xE5eSgnY3g_3_11_l.gif) to the CPU through the BIOS.

Kougar
09-24-2006, 07:10 PM
I thought most recent motherboards offered enough juice to give these babies a ride
Any idea how much voltage a asus P5B can deliver to the ram. I did a 'lil search yesterday or so but didn't come up with anything fast.

Define enough juice? If the RAM needs 2.1v or more often the machine won't even post with a single module, so even though the motherboard does support voltages above (Or far above) 2.1v, the user can't get into the BIOS to set the board to a high enough memory voltage. If they can't change the voltages to POST, they can't get into the BIOS. It's a chicken and the egg problem. ;)

GIBSON
09-24-2006, 07:59 PM
Define enough juice? If the RAM needs 2.1v or more often the machine won't even post with a single module, so even though the motherboard does support voltages above (Or far above) 2.1v, the user can't get into the BIOS to set the board to a high enough memory voltage. If they can't change the voltages to POST, they can't get into the BIOS. It's a chicken and the egg problem. ;)
I C, guess you have to lend someone else's DDR2 for a sec then. :lol:

Das Capitolin
09-24-2006, 11:56 PM
I C, guess you have to lend someone else's DDR2 for a sec then. :lol:

But thanks to SPD, when you remove your temporary RAM modules (which were used to increase voltage) and insert your out-of-spec modules, the BIOS will detect the SPD of the module and likely wipe out those previous settings.

Kougar
09-25-2006, 02:29 PM
But thanks to SPD, when you remove your temporary RAM modules (which were used to increase voltage) and insert your out-of-spec modules, the BIOS will detect the SPD of the module and likely wipe out those previous settings.

Exactly! That trick sometimes works, and sometimes it doesn't. I can't even begin to put a figure on the numbers of RMA'd and returned RAM and motherboards due to this problem. Heck, every single major tech forum is still having these threads created each day!

GIBSON
09-25-2006, 06:15 PM
That sux. How come memory manufacturers didn't have that in mind when manufacturing those sticks. Hell, why didn't they program the SPD so that it gives enough juice.

Kougar
09-25-2006, 09:46 PM
I'm sure the memory chips in the OCZ/Geil/G.Skill and other similarly high voltage RAM is cheaper for those companies to buy... which makes sense since those manufacturers often had/have some of the cheapest RAM, especially DDR2-800... and the price was exactly why almost everyone bought them, and later regretted it. :roll:

I'm sure they did program the modules with the right voltage settings, but when those settings are way out of spec not everyone is going to have equiped their motherboards to be prepared for it, that costs the motherboard makers money if they have to upgrade their own hardware to power RAM that is .04v above JEDEC specs just for some select companies that refuse to meet standards.

GIBSON
09-25-2006, 10:26 PM
Yes, but in case of a board like an asus P5B which i saw werty post giving voltage up to 2.45 or something like that, it should be possible to let it work with SPD. However I thought to understand from the previous posts that it does not, how come?!

Das Capitolin
09-25-2006, 11:45 PM
Gibson, I think that you might have misunderstood. Werty can boot his system at its default settings with his RAM, and then he sets the voltage higher. The problems we are having with RAM won't even allow us to boot up with it. We have to install different RAM. When you remove and reinstall the RAM, sometimes the settings are reset for the RAM in the system BIOS. Make sense?

werty316
09-26-2006, 02:10 AM
With high performance parts these days you have to cross your finger sometimes hoping that your ram will work. Take a look of the DFI LanParty for instance as it is picky when it comes to memory sticks.

GIBSON
09-26-2006, 05:20 PM
Gibson, I think that you might have misunderstood. Werty can boot his system at its default settings with his RAM, and then he sets the voltage higher. The problems we are having with RAM won't even allow us to boot up with it. We have to install different RAM. When you remove and reinstall the RAM, sometimes the settings are reset for the RAM in the system BIOS. Make sense?
Yes, I know. But I also read that the reason why these settings are reset when other ram is put in the motherboard is because of the SPD reporting those values. So why isn't the SPD programmed like it should be then, after all there should be no problem if it was, the board is able to supply enough voltage, and as far as I know all boards run SPD settings when they come out of the box...

werty316
09-27-2006, 03:58 AM
Having a BIOS set to "SPD" settings doesn't include voltage, just timings only.

Running a BIOS to "SPD" just means that it will run memory to its default/SPD setting if it is programmed with its "SPD" setting. If a memory stick isn't programmed with a "SPD" setting the BIOS will just set it to whatever.

GIBSON
09-27-2006, 01:00 PM
well, couldn't they make the spd run it at very lax timings, or would that not be sufficient to make it boot?

werty316
09-28-2006, 02:44 AM
It could boot but main issue would have to be compatibility.

Kougar
09-28-2006, 04:15 AM
well, couldn't they make the spd run it at very lax timings, or would that not be sufficient to make it boot?

The RAM would have to "know" when to use lax timings, and when it can use normal timings. RAM isn't that far along yet, the best I see is that they simply program lax timings, and everyone ends up with slow RAM whether they need it or not, which is a bad thing and something they'll obviously never do. Even then, simply loosening the timings will not make a unbootable system bootable the majority of the time.

GIBSON
09-29-2006, 05:37 PM
Well, after thinking about all this. How exactly do the manufacturers test the modules then. As far as I know most manufacturers test all their dual channel kits in real systems.

Kougar
09-29-2006, 11:39 PM
Yes, they do randomly test samples of their RAM in real systems, part of their testing process... But no company can afford to test their RAM in every motherboard currently made. The cost of buying a few thousand motherboards and paying a building full of people to switch modules and test each board out is just not possible. Automation would be impractical because most boards are not the basic ATX form factor and most never have the RAM slots in the same positions or places.

They take a sample of boards that use the various chipsets on the market and verify their RAM works.

werty316
09-30-2006, 02:20 AM
They just tets memory so no faults are found and ensure the memory chips aren't faulty. Memtest (http://www.memtest86.com/) is your friend ;) As Kougar said compatility is up to the motherboard maker and not the memory maker.

GIBSON
10-02-2006, 05:29 PM
IMHO, they both have a crucial role in it. If they decide to produce something no-one can ever get to work, they are fooling around with themselves instead of the customers.
And kougar, yes, I know they don't test them in all motherboards. I just mean, for instance corsair tests all their dual channel kits in real motherboards. If they can't make them get enough juice, how can they test them?

Kougar
10-02-2006, 08:04 PM
Gibson, that's a good question. I would assume that this is why not many companies claim their RAM works on specific motherboards like Crucial does, and why motherboard makers themselves usually have very selective lists of compatible RAM that they will guarantee works with their motherboard. ;) Almost every board I've look at recently has a little insignia saying guaranteed to work or guaranteed to work best with Corsair RAM, for instance.

GIBSON
10-02-2006, 09:22 PM
Only thing I can think of is that OCZ simply doesn't test them in real boards and that is why they're cheaper as well. Either that or they voltmodded their boards.

EDIT: when i'm thinking bout it, wouldn't it be possible to make them boot by using that ocz thing you plug in a free ram slot and gives your sticks a voltage bump. Shouldn't that do the trick?

Kougar
10-03-2006, 01:00 AM
Probably, but those were nearly completey bug free themselves. I wouldn't want one of those in my system, I only use 100% "stable" computers. If I can't leave it alone for two weeks and expect to find it running like normal then it's not "stable". ;)

OCZ probably uses normal boards, it'd be kind of pointless to mod a board just to make their RAM work, then sell the RAM anyway. That makes extra costs for them for no reason.

Look at it this way, they probably expect their customers to know if their RAM will work the buyer's motherboard. If it doesn't, then it's the customer's fault for not doing his research. Just like it'd be my fault for buying a AGP graphics card for my current system. Whether or not you agree with that philosophy is another thing altogether, but it doesn't mean their RAM won't work if you use it in the right mainboard, you just have to know or research it first, or otherwise be prepared incase it doesn't.