Should I RMA my RAM?

[fishnchips]

I just put together my first build last night - kind of a big milestone in my life. Anyway, I'm working on installing Ubuntu 9.04, and when I run memtest86, it shows a single error. After switching the modules around, I've isolated it to a single stick. One error on one stick in either test 6 (moving inversions) or test 7 (random number sequence). How much is that going to affect performance, and is it worth the hassle/time to RMA it? I'm using 3x2GB OCZ Gold 1600 Mhz, Asrock x58 Extreme motherboard, i7 920 CPU.

 

[badge]

Have you set up the RAM correctly in BIOS including speed, timing and voltage? Perhaps increase the voltage a notch and see if the errors are still present. If the errors persist when running memtest RMA the RAM.

 

[Crashman]

I just put together my first build last night - kind of a big milestone in my life. Anyway, I'm working on installing Ubuntu 9.04, and when I run memtest86, it shows a single error. After switching the modules around, I've isolated it to a single stick. One error on one stick in either test 6 (moving inversions) or test 7 (random number sequence). How much is that going to affect performance, and is it worth the hassle/time to RMA it? I'm using 3x2GB OCZ Gold 1600 Mhz, Asrock x58 Extreme motherboard, i7 920 CPU.

Yes, you should probably RMA it. RAM should never have to be "set up in BIOS" to be stable, changing the settings in BIOS should only be required to reach a higher "approved overclock" performance level.

There's an SPD chip on the RAM that's supposed to be programed with the slower timings and/or speeds needed to make it run stable at default voltage.

Anyway, you're lucky. I just got two sticks of bad OCZ Gold, one with only a single bad cell so far as Memtest could show, the other with three errors at close addresses (bad block?) Anyway, I tried entering BIOS with mine, leaving all the settings "auto detect" and increasing only the voltage to the rated setting, and the errors still occured. But I've seen that before, and I generally try to stay away from OCZ Gold.

 

[badge]

I have had a lot of problems and RMA experiences with OCZ over the years more than nay other company. The first thing OCZ suport will do is tell you increase the voltage to the RAM and NB to get the RAM stable using memtest.

 

[Crashman]

I have had a lot of problems and RMA experiences with OCZ over the years more than nay other company. The first thing OCZ suport will do is tell you increase the voltage to the RAM and NB to get the RAM stable using memtest.

Yes, but when they do that they're telling you to overclock. SPD is what it is, if the memory can't run at default voltage even using the slowest standard speed and timings, it's defective. The RAM I just got was rated at 1.60V, it wouldn't run even at SPD speed and timings using 1.65V DIMM and 1.30V memory controller voltage.

 

[badge]

No, I have had OCZ support explicetly direct me to increase the voltage to the RAM and the NB just to stabalize with no OC using memtest. Then, I explained the RAM showed errors no matter what increases in voltage I made. Do you believe it...OCZ allowed me to RMA the RAM and the replacement kit they sent has a bad DIMM! I RMA'd that kit and the second replacement kit performed without error. Just recently, I RMA'd some G. Skill PC2 8500 for failing memtest. The replacement kit would not stabalize until I increased the RAM voltage to 2.0v, 1.8v was BIOS default. This kit shows errors at 1.8v (JEDEC standard is 1.8v) and the errors cease at 2.0v. A clean run. Double checked.

The RAM I just got was rated at 1.60V, it wouldn't run even at SPD speed and timings using 1.65V DIMM and 1.30V memory controller voltage.

OCZ seems to release high transfer rate stuff that more than anything has the potential to fail in a lot of set ups. Your setup had nothing to do with failure, it was OCZ's fault. I hope the replacement OCZ sent worked out.

 

[fishnchips]

Thanks for the help! I'll go ahead and RMA it. Unfortunately, I think I have to RMA the whole set, not just one stick.

 

[Crashman]

No, I have had OCZ support explicetly direct me to increase the voltage to the RAM and the NB just to stabalize with no OC using memtest. Then, I explained the RAM showed errors no matter what increases in voltage I made. Do you believe it...OCZ allowed me to RMA the RAM and the replacement kit they sent has a bad DIMM! I RMA'd that kit and the second replacement kit performed without error. Just recently, I RMA'd some G. Skill PC2 8500 for failing memtest. The replacement kit would not stabalize until I increased the RAM voltage to 2.0v, 1.8v was BIOS default. This kit shows errors at 1.8v (JEDEC standard is 1.8v) and the errors cease at 2.0v. A clean run. Double checked.

The RAM I just got was rated at 1.60V, it wouldn't run even at SPD speed and timings using 1.65V DIMM and 1.30V memory controller voltage.

OCZ seems to release high transfer rate stuff that more than anything has the potential to fail in a lot of set ups. Your setup had nothing to do with failure, it was OCZ's fault. I hope the replacement OCZ sent worked out.

I hope so too, for OCZ's sake. It's going into a review!

 

[badge]

Thanks for the help! I'll go ahead and RMA it. Unfortunately, I think I have to RMA the whole set, not just one stick.

Yeah, the last time I RMA'd defective RAM to OCZ I had to go onto the OCZ website and make a RMA request. I was asked a lot of questions and asked to do some troubleshooting like I explained (increase voltage to RAM and NB). You will have to RMA the entire kit and the kit will be replaced. Last time one of the DIMMS in the kit OCZ sent out would not even boot the computer, total fail. The other DIMM worked. Second RMA replacement kit both DIMMs worked.

 

[OvrClkr]

yea I have had my shares of RMA's with OCZ as well....

sup badge?? nice to see ya again =)

 

[apache_lives]

Yes, you should probably RMA it. RAM should never have to be "set up in BIOS" to be stable, changing the settings in BIOS should only be required to reach a higher "approved overclock" performance level.

There's an SPD chip on the RAM that's supposed to be programed with the slower timings and/or speeds needed to make it run stable at default voltage.

Anyway, you're lucky. I just got two sticks of bad OCZ Gold, one with only a single bad cell so far as Memtest could show, the other with three errors at close addresses (bad block?) Anyway, I tried entering BIOS with mine, leaving all the settings "auto detect" and increasing only the voltage to the rated setting, and the errors still occured. But I've seen that before, and I generally try to stay away from OCZ Gold.

Not so - in the P4 days with SiS and VIA chipsets and FSB400/533 cpus the VIA/SIS chipsets would try to run DDR400 at 400mhz creating instabilities, same as some AMD K7/Athlon XP configs - things had to be configured manually to 1:1 or 2:3 etc to work properly, i have even seen an Albatron P4M800/775 not post with Celeron D's untill older ram (ddr-333) was fitted THEN configured to 1:1/2:3 before it was able to even post with DDR400 etc.

Latest cases i have seen are my own ASUS P5B Deluxe/Wifi-App with Corsair XMS6400 C5's - stock settings would not work, and ASUS P5GC/MX with FSB800 cpus and DDR2-800 - setting ram down to 533mhz and 1.9v solves it.

This is why i always recommend setting everything manually in the bios so you *know* what the system is running at, and for testing etc the 1:1 (or worse, depending on rig architecture, if QPI/HT then obviously 1:1 doesnt apply, and you want to set it to just about the lowest setting possible etc) ratio cant be beat for ram at stock settings, and a .1v increase in voltage for memory (unless more is required depending on ram manufacturer) etc.

 

[badge]

yea I have had my shares of RMA's with OCZ as well....

sup badge?? nice to see ya again =)

OCZ is a good company and has come a long way in recent years which is great to see. But, there have been some issues over the years with chipset compatibility specifically with their higher end DIMMs. Their Gold edition stuff had issues with some the early Nvidia 5xx SLI chipsets. But the stuff worked fine with what Intel was offering at the time 945/955. Around the time of C2D release. I still have a couple of 'Gold' PC25400 kits running from the C2D release era. God forbid I install it in an Nvidia chipset and turn my system into a gross failure pile of junk.

 

[badge]

Might mention OCZ owns PCP & C for a while now. I had to RMA my PCP & C 610 Silencer a couple of months ago. OCZ has their own an inhouse 'factory' to service warranty on their units. So they essentially do inhouse repairs under warranty where RMA is concerned. I have to install my returned RMA'd unit (same unit was returned to me after 'repair') and see how it works. No time for me lately.

 

[Crashman]

Not so - in the P4 days ...

You can disagree all you like, but you're just blowing smoke to say that any particular RAM shouldn't have to be stable using nothing more than SPD values.

If you have RAM rated at DDR3-1333 CAS 7 1.65V, that's an overclocked setting. When you install the RAM, it doesn't have to be stable DDR3-1333 CAS7 1.50V. It does have to be stable at some lower "automatic detect" setting, such as DDR3-1066 CAS 9 1.50V, else it's defective.

ALL overclocked RAM is based on some slower-speed RAM. If it can't run stably at the slower speed and default voltage, it IS defective.

If you have DDR3-2133 CAS 9 1.65V, it's "ok" for the company to program the SPD chip to tell it to run at half that speed, so long as its stable at half that speed and default voltage. Companies should TELL you the details, but at least most of us understand that when it says 1.65V, you're going to have to set it manually to reach the higher speed stably.

Here's an example of acceptable RAM:

DDR3-1600 1.65V that runs at DDR3-1066 1.50V by default, but it stable at both settings

Here's an example of defective RAM:

DDR3-xxxx that won't run stable at ANY speed using default voltage.

And the reason why it's defective is that it doesn't meet the MINIMUM INDUSTRY STANDARDS.

Feel free to blather on, you'd be right after all to say that using an overclocked setting requires BIOS configuration. But using defaults doesn't require BIOS configuration, because industry standards exist to protect us.

Industry standards are minimums for quality, if RAM doesn't meet those it doesn't matter how far you think you can overclock it.

 

[apache_lives]

Do you think every system's bios follows those SPD values strictly?

Im not saying his ram is NOT faulty, im saying some setup's need a manual configuration to get stability - as i said, my last rig - ASUS P5B-Deluxe Wifi/app + Corsair XMS6400 C5's (4x1gb - two kits, each seperatly would do it) would not run stable at all using defaults, yet the next motherboard i used (Gigabyte EP35 DS3R) worked flawlessly with it.

So tell me what the issue was there hey?

 

[badge]

ASUS P5B-Deluxe Wifi/app + Corsair XMS6400 C5's (4x1gb - two kits, each seperatly would do it) would not run stable at all using defaults

My ASUS P5B Deluxe 965 is up and running for yeas now. I have a Q6600 at 3GHz, Vista 64 with 6GB G. Skill PC2 6400. I got nowhere fast OCing the system until I set the RAM speed to 533MHz.(as you mentioned), increased RAM voltage and added additional voltage to the NB, etc. You are right also about some BIOS installations not picking up on the JEDEC SPD standards. Like I mentioned, watch out for any new release OCZ high performance RAM. I think Crashman is experiencing the OCZ 'curse' firsthand.

 

[Crashman]

Do you think every system's bios follows those SPD values strictly?

Im not saying his ram is NOT faulty, im saying some setup's need a manual configuration to get stability - as i said, my last rig - ASUS P5B-Deluxe Wifi/app + Corsair XMS6400 C5's (4x1gb - two kits, each seperatly would do it) would not run stable at all using defaults, yet the next motherboard i used (Gigabyte EP35 DS3R) worked flawlessly with it.

So tell me what the issue was there hey?

A motherboard engineer told me his company was cheating on default voltage because of complaints about a certain brand of RAM being incompatible. That motherboard manufacturer was forcing 1.58V at the 1.50V setting, and reporting 1.50V in BIOS, simply to make people believe the problematic RAM could be operated at 1.50V. He said the problem was caused by the RAM manufacturer using inappropriate SPD values, and the problem for his company was that people were blaming the motherboard rather than the RAM.

And we all know who that RAM manufacturer was.

 

[apache_lives]

Oh almost forgot - that ASUS P5B Deluxe Wifi/App board using the default "auto" values in the bios, would run the ram at 4-4-4-12 (@ 800mhz) whereas they were supposed to be running at 5-5-5-18 (or 12?) - had to set everything manually to get prime to pass beyond even 5 minutes - thats why i had a go at correct usage and application of SPD values etc.

When i replaced that motherboard to the Gigabyte EP35-DS3R it worked even at 1.8v and with the correct SPD values using auto settings.

The same ram is now in two older rigs (P4/800mhz FSB) running two gigabyte motherboards (2gb each) and i have to set them to run at 400mhz (1:1) to get any stability - 533mhz (proper native memory for the 915 series chipsets) gets BSOD's and 667 forget about it.

Never buying corsair memory again, and thinking twice about ASUS motherboards.

 

[ewood]

You can disagree all you like, but you're just blowing smoke to say that any particular RAM shouldn't have to be stable using nothing more than SPD values.

If you have RAM rated at DDR3-1333 CAS 7 1.65V, that's an overclocked setting. When you install the RAM, it doesn't have to be stable DDR3-1333 CAS7 1.50V. It does have to be stable at some lower "automatic detect" setting, such as DDR3-1066 CAS 9 1.50V, else it's defective.

ALL overclocked RAM is based on some slower-speed RAM. If it can't run stably at the slower speed and default voltage, it IS defective.

If you have DDR3-2133 CAS 9 1.65V, it's "ok" for the company to program the SPD chip to tell it to run at half that speed, so long as its stable at half that speed and default voltage. Companies should TELL you the details, but at least most of us understand that when it says 1.65V, you're going to have to set it manually to reach the higher speed stably.

Here's an example of acceptable RAM:

DDR3-1600 1.65V that runs at DDR3-1066 1.50V by default, but it stable at both settings

Here's an example of defective RAM:

DDR3-xxxx that won't run stable at ANY speed using default voltage.

And the reason why it's defective is that it doesn't meet the MINIMUM INDUSTRY STANDARDS.

Feel free to blather on, you'd be right after all to say that using an overclocked setting requires BIOS configuration. But using defaults doesn't require BIOS configuration, because industry standards exist to protect us.

Industry standards are minimums for quality, if RAM doesn't meet those it doesn't matter how far you think you can overclock it.

either way it wouldnt hurt to check the settings in the bios. i got ram a week ago that was rated at 1600mhz at 1.7v but for some reason the spd voltage values were 1.5v and 1600mhz... poses a problem because the ram wasnt stable with 1.5 v. a quick bump to 1.65 solved the problem. anyways its always worth making sure, but it sounds deffective to me.

 

[Crashman]

either way it wouldnt hurt to check the settings in the bios. i got ram a week ago that was rated at 1600mhz at 1.7v but for some reason the spd voltage values were 1.5v and 1600mhz... poses a problem because the ram wasnt stable with 1.5 v. a quick bump to 1.65 solved the problem. anyways its always worth making sure, but it sounds deffective to me.

You would think these companies could have figured out the whole SPD thing eh? I mean, EPP was invented to take care of automatic configuration for overclocked settings, SPD is supposed to take care of the unoptimized defaults.

 

[ewood]

You would think these companies could have figured out the whole SPD thing eh? I mean, EPP was invented to take care of automatic configuration for overclocked settings, SPD is supposed to take care of the unoptimized defaults.

in a perfect world... until then ill keep tweaking

 

[apache_lives]

i cant see a systems knowing better then a well seasoned tweaker

 

[Crashman]

i cant see a systems knowing better then a well seasoned tweaker

The system doesn't have to "know" anything, it's supposed to be able to read SPD. If the memory manufacturer doesn't know how to program SPD, that manufacturer is going to produce a slew of defective products.

Improper SPD values are a defect.

 

[apache_lives]

SPD is also bad when fitting older systems with newer ram - good examples are fitting old P4 (fsb400/533) rigs (with SIS/VIA chipsets) with DDR400 - the SPD says it can run at the full 400, and the system attempts to run it at 400 causing instabilities (note this does not apply to Intel chipsets thanks to limited FSB:RAM ratios etc like 1:1 and 3:4? only etc)

yet another scenario where SPD fails

 

[Crashman]

SPD is also bad when fitting older systems with newer ram - good examples are fitting old P4 (fsb400/533) rigs (with SIS/VIA chipsets) with DDR400 - the SPD says it can run at the full 400, and the system attempts to run it at 400 causing instabilities (note this does not apply to Intel chipsets thanks to limited FSB:RAM ratios etc like 1:1 and 3:4? only etc)

yet another scenario where SPD fails

No it's not. You get DDR-400, it comes with mutiple SPD values: 400, 333, and 266 MHz. It comes with timings for each of those values. If the memory isn't stable at stock voltage (2.60V) using 400 MHz SPD values, it's bad RAM even if the only defect is defective programing. If the board tries to run it at 2.50V and 400 MHz, and it crashes, it's bad motherboard BIOS. If the board needs to use a lower SPD value and doesn't, it's again bad BIOS programing.

No matter what the excuse for the system not working, SPD wins. In those odd circumstances, it's the programers who fail, not the program. That is to say, when SPD doesn't work its because either the motherboard or memory is defective.