What's with Nforce4 SLI and 2 x 1GB RAM?

[Sorin]

At the end of my rope here and very frustrated. My pair of 1GB memory modules refuse to run at DDR400 when used together. They are both OCZ4001024ELPE.

In both an Abit AN8 SLI and an Asus A8N-SLI Deluxe (in my system, so all other hardware besides the motherboard is identical), when either module is used alone, it runs at 400MHz, DDR400. When used together, they run at DDR200 (yeah 200, as in 100MHz clock frequency). When forced in the BIOS of both motherboards to be DDR400, the system refuses to boot and just sits there, powered on but not booting. So with these motherboards, I am forced to run my nice DDR 400 2 x 1GB modules in dual channel, DDR200 mode, otherwise system won't boot.

Bad RAM right? Well, no. When tested in an Abit NF7-S v2.0 they run flawlessly. They are immediately detected as DDR400 and they run at DDR400, dual channel.

Before I got the ASUS board, I thought the Abit was just bad. The AN8 SLI (Nforce4 SLI) would not run them together at DDR400, only seperately. When together, they were forced to DDR200. So I tried the Abit NF7-S v2.0 (note Socket A and Nforce2) and they ran totally perfect at DDR400, 2 x 1GB. So I concluded bad motherboard.

So I got the Asus and guess what? Lo and behold, exact same problem as the Abit AN8.

So two different Nforce4 SLI motherboards, completely identical problem with RAM. They won't run 2 x 1GB DDR400 RAM at DDR400 speeds. They run each module seperately at DDR400, but when used together, both boards refuse to run them faster than DDR200, otherwise the system won't boot.

So I really don't know what to do at this point. Is there some stupid setting I'm missing somewhere? I'm not as familiar with Nforce4 and the whole AMD 64 stuff as I was with Socket A and Nforce2. I'm not buying into the thought that I just HAPPEN to get 2 shoddy motherboards from two totally different manufacturers.

I originallly posted in the Abit subforum here and got no replies. I posted on Abit's (the site) forums and got 1 useless reply that basically told me to go away and post at OCZ. I also posted at OCZ's forums, which is where I got some real help and came to the idea of trying the RAM in a different board (which was the NF7-S v2.0). Naturally concluded bad board and got a new one, but now I'm back at square 1.

Help :(

[cruiza120]

That is really wierd..I'm thinking possibly a motherboard driver problem? Umm...ram compatibility error (though that is less likely as you tried with two motherboards)...bad ram? even though you tested it there could be some problem that the test didn't find. I would try getting a replacement pair of the ram, and see how that works. Get the latest drivers for the mobo, too. Good luck with it.

[Sorin]

Already have latest chipset drivers and BIOS. I don't know how the RAM would be bad, because they run perfectly in DDR400 and dual channel on a Socket A Nforce2 board. I don't even have to touch the BIOS. It's just that these two Nforce4 SLI boards don't like it for some reason. Does Nforce4 not like two modules being different revision numbers? (I bought these two modules 9 months apart. One is rev 1, the other is rev 2.2. Identical model numbers though).

I really don't see too many other explanations. Other than a fantasically insane coincidence of having two different boards from two different manufacturers having an identical bug.

1) Nforce4 doesn't like different revisions of identical model number RAM?
2) Neither board is fully compatible with OCZ? (BS though, I've seen many sigs with 2 x 1GB and OCZ in these boards)
3) Extreme coincidence with having the same bug on different models of boards from different manufacturers?

me: -----> :bang head

[Cjwinnit]

which graphics cards did you use for each board?
What power supply did you use?

[Sorin]

Hardware for both was identical, as per signature. Both of these boards were in my system.

eVGA 256MB 7800 GT PCI-e

PSU is a 550w Fortron Sparkle 24-pin.

[Cjwinnit]

hat was in the NF7-S? ( I assume AGP...)

[Sorin]

In the Nforce2, Abit NF7-S v2.0 it was a PNY 6800 GT 256MB AGP, the PSU was a Fortron Sparkle 500w, and the CPU was a mobile Athlon XP 2500+ Barton at 9.5 x 200 (1900MHz).

[soulfly1448]

Have you tried flashing the SPD? You could dump the SPD from one module using a prog called Taiphoon and flash it to the other. That would result in both mods having the same SPD and thus the same info.

Now, I can't see that that would really be the issue but, as long as the specs (speeds/timings) are identical, this wouldn't cause an issue.

[billb]

using a prog called TaiphoonFantatic (and fantastically damgerous for nOObs!) program. ALL the SPD timings!

[Sorin]

Fantatic (and fantastically damgerous for nOObs!)

Which is exactly what I am when it comes to that program! Someone wanna baby step, small words, and slow speech me through it? I really don't want to blow up my memory.

[apesoccer]

I've had a similar problem...

And different...

I have ocz 2x1024 Gold 3200 stuff 2 3 2 8 stock
and ocz 2x1gb Plat 3200 stuff 2 3 2 5 stock.

I have an epox, jetway, dfi and asus nf4 board. I generally refuse to use the jetway anymore... But when either pair of ram are paired with the 3800+ x2 it fails in my dfi. When they're paired with the 3800+ x2 in the epox or asus they're fine. When in the dfi with my 3500 or 3000 they're fine. I also run similar sparkle psu's...both 550's, one is sli, the other is not. Anyway...I don't think we're having exactly the same problem, but since it's similar and we have such similar systems i thought i'd post it.

[Sorin]

Yeah, it's bizzare. But really, I think my best shot at this point is that Thaiphoon program. Even though the two modules have identical model numbers (OCZ4001024ELPE), they are different revisions. One is 1.0 and the other is 2.2. The 1.0 has timings of 2-3-2-5 and the 2.2 has timings of 2.5-3-2-5. So the only real difference is the CAS. Maybe that's the problem, maybe the nforce4 is being bratty and picky about not running them together correctly because of the different CAS (i.e. different revision)? Although if that was the problem, I'd think it'd refuse to do dual channel, rather than nerfing the clock speed by 50%. Anyways, I want to give it a shot.

n00b WTB walkthrough of Thaiphoon so I don't blow up my memory ;)

[masterwoot]

Yeah, it's bizzare. But really, I think my best shot at this point is that Thaiphoon program. Even though the two modules have identical model numbers (OCZ4001024ELPE), they are different revisions. One is 1.0 and the other is 2.2. The 1.0 has timings of 2-3-2-5 and the 2.2 has timings of 2.5-3-2-5. So the only real difference is the CAS. Maybe that's the problem, maybe the nforce4 is being bratty and picky about not running them together correctly because of the different CAS (i.e. different revision)? Although if that was the problem, I'd think it'd refuse to do dual channel, rather than nerfing the clock speed by 50%. Anyways, I want to give it a shot.

n00b WTB walkthrough of Thaiphoon so I don't blow up my memory ;)


Have you tried manually setting the RAM timings to 2.5-3-2-5?
Also, I presume you do have the RAM stix in the same colored dimm slots, correct?

Since the two modules are not the same revision, you'll be lucky to get dual channel out of them. Manufacturer's even want you to have pairs that are only different by 1 serial # (that is, they come off the same batch).

My guess is that you'll need to make sure you get most all your BIOS options for your RAM out of "auto" mode and force them to cooperate--definitely your general timings.

Good luck.

[Sorin]

Have you tried manually setting the RAM timings to 2.5-3-2-5?
Also, I presume you do have the RAM stix in the same colored dimm slots, correct?

Since the two modules are not the same revision, you'll be lucky to get dual channel out of them. Manufacturer's even want you to have pairs that are only different by 1 serial # (that is, they come off the same batch).

My guess is that you'll need to make sure you get most all your BIOS options for your RAM out of "auto" mode and force them to cooperate--definitely your general timings.

Good luck.

They do run in dual channel, just not at 400MHz (200MHz clock). They run at 200MHz (100MHz clock), but still dual channel. I manually set the CAS down to 2.0 from 2.5, and the command rate from 2T to 1T, and all is well. It's only when I force the BIOS to manually run them at 400MHz (DDR400, 200MHz clock) that the system wont boot and I have to clear CMOS. And yup, they are in the correct slots, that was the first thing I checked (slots A1 and B1 for my particular board. A1 and B1 are one color, and A2 and B2 are another color, so the colors match too).

[masterwoot]

I manually set the CAS down to 2.0 from 2.5, and the command rate from 2T to 1T, and all is well.

Have you tried setting it to 2.5? I've only been able to run mismatched ram in dual channel by going to the lowest timings...

[Sorin]

Mildly disappointing to go from 2.0 (that's the default CAS of the 1.0 revision) to 2.5 (default revision of the 2.2), but if it gets me going to 400MHz, they I'll be happy with that. I'll try it later on when I get home.

[Sorin]

Using Thaiphoon and CPUz, I wrote down as many of the timings of the slower, later revision module as I could, after which I went into the BIOS and set as many of the timings as were there to what I wrote down. I then rebooted (still at DDR200) to make sure things were ok, and all was well. So I restarted again and set the clock rate to DDR400 and rebooted.

It seems to be working for the moment. I ran Sandra a couple of times and sure enough, it seems to be DDR400 dual channel. Bandwidth now is around 5.5GB, when before at the dual channel DDR200 it was 2.9GB (2.2GB in DDR200 single channel, for reference).

I've rebooted a few times just to make sure it would, and it looks okay for now.

</fingerscrossed>

[masterwoot]

Good news!

I'd burn a copy of memtest (http://www.memtest86.com/) to a CD, boot off that CD, and then run test #8 for a few loops. That'll show you if your memory is doing ok at those timings/speeds.

[Sorin]

Already got memtest, pimp little tool ;) I'll try that.

[billb]

Even though the two modules have identical model numbers (OCZ4001024ELPE), they are different revisions. One is 1.0 and the other is 2.2. The're are probably(?) different ICs on the modules. It will be problematic to OC with much success, as the max OC for each module will be different (even if you Thiophon the SPD timings). You'd have more success if you bought a matched set or could find a module to match one you have.

Really, when you buy memory, you're just buying ICs. The manufacturer just builds modules from the cheapest and most available (at the time) ICs that will yield modules with the specs he wants. If it's to add inventory to a part No. it may or may not be the same ICs as used in previous production runs...hence the Rev nos.

[Sorin]

The're are probably(?) different ICs on the modules. It will be problematic to OC with much success, as the max OC for each module will be different (even if you Thiophon the SPD timings). You'd have more success if you bought a matched set or could find a module to match one you have.

Really, when you buy memory, you're just buying ICs. The manufacturer just builds modules from the cheapest and most available (at the time) ICs that will yield modules with the specs he wants. If it's to add inventory to a part No. it may or may not be the same ICs as used in previous production runs...hence the Rev nos.

Won't be overclocking these. Whatever CPU I upgrade to this fall (165/170, or maybe 4600/4800 now that they are recently dirt cheap - comparatively), I was going to upgrade the memory to an actual dual channel kit (2 x 1GB DDR500). But for the rest of the summer, everything is currently stock. I'll move this cpu/ram to another box, where it'll still be stock speeds/timings.

[mrgreenjeans]

Let's back up. When you said you 'tested' the RAm I would assume you meant MemTest or another such program that actually tested the RAM. What did you use to 'test' the RAM? Just cause they boot and run doesn't mean they're OK. My money says Memtest is going to fail one stick if not both.

[Sorin]

Different kinds of tests. In this case, not test as in Memtest, but test as in "plug it in and see if they run at proper speeds without backhanding them into submission". I was going to do memtest last night, but something isn't right with either the CD or booting from the CD drive. There would be a very brief, split second flash of a blue dos-like screen (memtest, obviously) that said "loading", but then it just went straight into windows. Since I'm out of CDs, I tried using my USB flash drive, but apparently a flash drive doesn't count as a "USB Disk" as far as my motherboard is concerned (invalid system disk error). Gonna need to get some blank CDs and try again.

[masterwoot]

There is the windows diagonstic test for memory (http://oca.microsoft.com/en/windiag.asp)...

[Sorin]

After much conversation with an official OCZ person and trying different things, I did as he suggested and used Thaiphoon to write the EEPROM information from the rev. 2.2 module to the rev. 1.0 module. I then reset the memory config option to auto in the BIOS and now it automatically detects and runs them at DDR400.

I also had to burn a new memtest cd; I guess the other one went bad. I ran memtest through one full loop of all 8 tests and then looped test #8 as masterwoot suggested five times and there weren't any errors. I might fire it up tomorrow morning before leaving for work and have it loop the entire series of tests all day while I'm at work just to make sure (though of course I know memtest isn't a 100% sure thing, it's still pretty good).

At least now I don't have to brute force it in the BIOS by tweaking all the specific timings.

[mrgreenjeans]

Cool. I learned something here. Didn't know about transferring the EEPROM between sticks. Thanks for the update.

[Sorin]

As a final note, I asked the guy about the differences between the two modules, and he had this to say:

"There is not much difference between the 2, just a different batch of the IC's and we rated them at a higher voltage spec.

Sometime after you purchased your first stick, we discovered some boards on the market would not boot if you cleared the CMOS and the SPD was programmed at CAS 2 on a 1GB stick of ram. We switched to CAS 2.5 on those for that reason."

So, no big deal for me. Not a respectable enough difference between 2.0 and 2.5 on the CAS to worry about. Now 1T and 2T on the command rate, THAT was a monumental difference.

[masterwoot]

Nice work, Sorin!
Glad you've got them cruising. :)

[Sorin]

Thanks. It's pretty cool, the whole flashing-the-RAM thing.

You know, completely on a whim last night I figured I'd throw an overclock at it and see what happens. I had been running everything at stock speeds from the time I got it, up until now (got it sometime earlier this year). I left the multiplier at 11 and bumped the bus to 219 (so now the memory is at 219MHz too) just to get to around 2.4GHz (it's at 2409MHz). When I left for work this morning, it had been priming for 13 hours, so by the time I get home it will be almost 24, if it's still running. I guess my next project will be having it run memtest all day while I'm at work.

So anyways, it's all good now. I was going to wait until I dropped in a dual core and got some faster RAM before overclocking, but eh, what the hell?

[SuperDave1685]

One is 1.0 and the other is 2.2. The 1.0 has timings of 2-3-2-5 and the 2.2 has timings of 2.5-3-2-5. So the only real difference is the CAS. Maybe that's the problem

yup... that'd be the problem... I had kind of the same deal with my Corsair (3-4-4-8) and my OCZ (2.5-3-3-7). The nForce4 chipset is picky when it comes to trying to run two different CAS latency RAM modules. If you'd be willing, I can take you step by step to flash the SPD of the newer, faster, stick to the 2.5 CAS latency one so that you don't have to worry about setting the BIOS everytime you clear the CMOS...


holy dead thread lol

[RollingThunder]

Thanks. It's pretty cool, the whole flashing-the-RAM thing.

You know, completely on a whim last night I figured I'd throw an overclock at it and see what happens. I had been running everything at stock speeds from the time I got it, up until now (got it sometime earlier this year). I left the multiplier at 11 and bumped the bus to 219 (so now the memory is at 219MHz too) just to get to around 2.4GHz (it's at 2409MHz). When I left for work this morning, it had been priming for 13 hours, so by the time I get home it will be almost 24, if it's still running. I guess my next project will be having it run memtest all day while I'm at work.

So anyways, it's all good now. I was going to wait until I dropped in a dual core and got some faster RAM before overclocking, but eh, what the hell?

Sorin,

I learned a lot form your misfortune too, more than I have the skills to deal with. Nice job and I am happy for your success. I would have smashed it long before that.

SuperDave,

I saw your same flashing the RAM sticks in another thread. Congrats to you too! Nice job. RT

[Sorin]

Thanks. I just refuse to give up on a piece of hardware unless there is absolutely no other choice! It's been running solid since that fix. As an update, I'm running at 2.9Ghz (11 x 264) with the memory on the 333MHz divider (it runs at 207MHz) since the memory is not stable at any settings above 225MHz. But yeah, problem solved, because I hate giving up, especially on new hardware. I :bang head and :mad: until it works, then I'm :D