OCZ PC3200 Gold won't work as a set

My luck with memory is not too good these days (except with my old Lv2s). My relatively new Mushkin Blues started producing repeatable errors, bad enough that I can't get into Windows. The set was previously stable up to 257MHz 2-2-2-8 at 3.5v (actively cooled). Before I say anything about the OCZs, my DFI rig is running stable right now (as I write this) with Mushkin Lv2s (BH5) at 260MHz 2-2-2-6 3.4v.

I just got a set of 2x512 OCZ PC3200 Gold (BH) to replace the Blues and tested them this morning. I tested them individually at 2-2-2-6, 3 clean passes of Memtest test 5 just to get an idea of how they stack up.

Stick 1_______________Stick 2
227________2.8v______224
233________2.9v______231
239________3.0v______241
244________3.1v______247
248________3.2v______251
251________3.3v______255
255________3.4v______258
256________3.5V______260

Looks promising, so I try them together (DIMM 1&2, that's what the Lv2s liked). Immediate problems as I could barely get to Memtest until I raised the vdimm to 3.0v (200MHz). 3v didn't make it to 210MHz, and 3.1v only 225MHz. This was only a couple of passes of test 5. So I went back to 200MHz and 3.0v (2-2-2-6) to make sure they would run clean as a set - errors after about 5 passes. I changed every variable I could while remaining at 200MHz and 3.0v - dimm slots, CPC, Bank Interleave, 2.5-3-3-8 and everything in between. Test 5 always produced errors sooner or later.

So I put the set in my AI7 (my other rig that likes BH5) to test them. Again, the sticks were OK individually. But errors popped up even faster on this board when run together, regardless of vdimm (up to 3.2v) and timings (kept speed at 200MHz).

I'm not asking OCZ to guarantee that the pair will overclock to a certain speed - any of us could get a set that just doesn't clock well over rated speeds. But since I can't use them together on 2 boards at any speed, voltage or timings, something isn't right. Especially considering their high performance individually. I hope one of the OCZ reps will chime in and let me know where to go from here.

First off they are not bh5, rather BH die, and yes there is a big difference.

AI7 isn't going to be the MB ti test them on. Mine refused to run 2-2-2-5 at 200. Only 2-2-2-6. Not sure why either.

By far the best MB to test ran on is the DFI Nf4 offerings.

DIMM 1&2 on any P4 MB is not DC, you need to use 1/3 or 2/4. Im pretty sure its the same on most of the NF4 MBs as well.

If the 2 sticks wont run rated timmings at rated speeds, OCZ will take it back.

The "DIMM 1&2" comment was in reference to the DFI, not a P4 board. I mentioned it because the board can behave differently depending on which slots are populated, even though it's single channel. I know the deal with "BH/CH die" as opposed to BH5/CH5, but referred to them as BH5 'cause AT stated that's what they were (but they've been wrong before). Speaking of the difference between actual BH5 and "die copy" ICs, I know there's a big difference in tRFC tolerance. My BH5s will run down to at least 12, while the "CH" UTT needs 15/16. What about "BH" die?

It just seems very strange that the sticks perform very well individually, but don't work together SC or DC.

Ufi Ut?

Hey

One thing you have to remember with TCCD is that they don't run very fast with 2-2-2 timings in terms of FSB. You should loosen up the timings to 2.5-3-3-7 and crank the FSB

Tony made a post with some other settings that make this combo fly. Give them a show and let us know how things are going for you
http://www.bleedinedge.com/forum/showthread.php?t=4443

SteveOCZ said:

Hey

One thing you have to remember with TCCD is that they don't run very fast with 2-2-2 timings in terms of FSB. You should loosen up the timings to 2.5-3-3-7 and crank the FSB

Tony made a post with some other settings that make this combo fly. Give them a show and let us know how things are going for you
http://www.bleedinedge.com/forum/showthread.php?t=4443

Click to expand...

TCCD? He is talking about 3200 Gold BH not plat rev 2 unless I missed something heh. Kris I think you will be able to get em running well together, just keep tweaking, took me a damn week to get my Gold BH to work the way I wanted em. Here are my settings which gets me a llitle above 260Mhz:

2-5-2-7-16-3-2-1-2-3072-auto-enable
auto-0-lev 8-lev 1-7-5-256-dis-16-7-dis

Dram response = fast.

These are not TCCx, but BH ICs. Thanks, Eldonko. I run my Lv2s at 1T-2-2-6-2-8-12-2(or 0)-2-1-1(or 2)-3120(or 1560)-1-BI En-Auto-0-Lv1-7-5-16-En-16-7-Dis. Tried these settings, and looser. After having to repair Windows because of this set, they're not going back in unless OCZ asks for specific Memtest results.

I found my 3200 Golds hated loose settings, every setting I loosened it made it worse, tightening things up gave me stability.

Ryder at BE is helping me right now. Running tests, so we'll see. I have to say he was quick to help. I've always read good things about their support, and now I'm getting it.

LOL u got it guys! telll that OCZ rep about his own memory! hahaha

so have you tried the RAM in different DIMM slots to see if you still get errors?

KrisMCool,
You aren't the only person who are experiencing these problems, I have a set of blues that will individually do 255 seperately but refuse to work together. What's even stranger is one of the sticks one combined with my lvl 1 will run flawlessly but the other stick will not, even though it runs flawlessly by itself. Could be something up with the memory controller combined with a non JEDEC reference PCB. Also the Winbond ICs that appeared to have issues where the ones with the larger wider staples while the ICs from the working Blue stick had the thiner staples like my lvl 1.

KrisMCool said:

I changed every variable I could while remaining at 200MHz and 3.0v - dimm slots, CPC, Bank Interleave, 2.5-3-3-8 and everything in between. Test 5 always produced errors sooner or later.

Click to expand...

Yes, different slots. Fldrice, I've read numerous posts about odd UTT problems like yours. I used to feel lucky that mine (Mushkin Blues) were so easy to work with...until they puked. After a night of more testing on the Golds, a couple of things are certain: as a set, they are voltage intolerant (anything over 3.0v causes thousands of errors that get worse when voltage is increased); CPC had to come off at 220MHz and above. It appears I may have found the settings necessary to get them stable as a set at 200MHz, even in Windows. But it took a lot of toying with settings - definitely not plug-n-play like other similar sets. I made it to 238MHz at 3.0v fairly stable in Memtest, but not Windows. On to more testing.

/does the walk of shame

I had a long day that day and must have mixed up this post with one on another forum..or emai...or msn...it gets confusing sometimes

NP, Steve. The more times I get to see your avatar, the better. Besides, I'm sure you're dealing with quite a few customer issues at once. Must be a lot to keep track of.

I've moved the Golds to my AI7 (ain't looking good so far either - rated specs yes, OC no). They just won't cooperate in my DFI rig (and can't compete with my 265MHz-capable Lv2s). The set is VERY picky about finer timing settings, and can't run 1T over 220MHz. The set is not RMA worthy, I guess, though it took over an hour of toying with timings just to get 200MHz stable. I'm very disappointed with the set as I can't get anywhere near the speed others are reaching. I don't expect OCZ to guarantee the set will OC to a certain speed, obviously. If they don't fly in the AI7, I guess I'll try to sell them to someone not looking to OC. What a waste of time and money.

Sorry, this isn't meant as a bag on OCZ at all. I'll still buy and promote their products. Anyone can get a set (from any company) that doesn't live up to expectations, but meets specs. I just wish these were even mediocre instead of shameful, OC-wise, especially since I bought them to replace a fast set of Blues that died. I probably shouldn't have spent the money, and I'm just sore that I did now. Expensive hobby, but I already knew that.

It might be the mem controller on your proc too. If you have a friend that would loan you a proc or maybe has a system that you could test your memory in, you could deduce what's holding you back.

It's not the CPU or mem controller. That's why I keep mentioning that my DFI rig works with Mushkin Lv2s up to 265MHz and tight timings. Unless this CPU just doesn't get along with this particular set. I don't think that's the issue, though. I'm burning the set on my AI7 now, and it has similar speed limits on that rig too. I'm at 234MHz, 3.2v, 2-2-2-6, A-A-A-D-D, NBstrap667, getting just a couple errors in test 5 per pass. At those settings, I had one error in test 6 in an overnight run of the full Memtest suite. Loops of test 5 bring out more errors, after the first 4 passes. I had to increase voltage very gradually to get to 3.2v, otherwise the set would spit out thousands of errors; at least 5 passes of test 5 at a given voltage and speed before increasing. The set doesn't like me adjusting GAT settings on the AI7; that's why Im using NBstrap667 to enable PAT. T-A-A-D-D at PSB800 1:1 won't fly. We'll see how the set does after burning like this for a while.

Sorry for the rant last night. I was in a foul mood.