Athlon 3800+, OCZ VX 4000+, and OC'ing

[Hajime]

Ran into a real corker of late.

I've got a pair of OCZ VX PC4000 512mb modules.

Just purchased a Athlon 64 X2 3800+

Currently running them on a DFI Ultra-D NF4 mobo.

Now, if I run the RAM at a divider, (I.e. 3/4) and push up the fsb, I can run the cpu stably at 2.6gz+.

However, right now I'm running the RAM at 240fsb/480mhz, 2-2-2-6 timings with 3.0v.

If I push it to even 245fsb, 2-2-2-8 timings, I can up the volts to the RAM as high as 3.3v, I still can't test reliably with it.

Initially, my gut told me this was a problem related to the processor - I've alread ran this ram stably for months @ 3.2v with 2-2-2-7 timings at 245fsb, with the Athlon 64 3000+ cpu prior to the X2. With the 3000+, I was cpu-limited - I just couldn't push it any further.

However, as I noted earlier, if I run a divider, I can run it at 260fsb easily, and didn't bother to push it further, but I did not hit the limit.

I updated the BIOS on the mobo, still no luck - even if RAM performance actually managed to go down .

 

[HamidFULL]

look here :
www.dfi-street.com

 

[customcoms]

Check their and use memory timings from the Overclocking database. However, I'm guessing you've been to the street already Here are my two guesses (I'm running basically the same ram-yours is rated to do DDR500 at 3.2v I believe-with your previous cpu and the same board)

1)The tref and trfc settings aren't right for this new proccessor.

2) Weaker memory controller in the X2. Did you have a winchester or venice 3000+?

3) The VX has just slowly died do to voltage creep and it can't maintain the same speed. Are you running any cooling on it specifically?

 

[myocardia]

Originally posted by: customcoms
Check their and use memory timings from the Overclocking database. However, I'm guessing you've been to the street already Here are my two guesses (I'm running basically the same ram-yours is rated to do DDR500 at 3.2v I believe-with your previous cpu and the same board)

1)The tref and trfc settings aren't right for this new proccessor.

2) Weaker memory controller in the X2. Did you have a winchester or venice 3000+?

3) The VX has just slowly died do to voltage creep and it can't maintain the same speed. Are you running any cooling on it specifically?

Very good points, though from what I know, tref and trfc settings shouldn't need to be changed because of processors, only because of RAM. With that said, are you running your tref at 0648 and your DRAM Strength at 8? Those are the "recommended" settings for your RAM and your (our) motherboard.

 

[Hajime]

Originally posted by: customcoms
Check their and use memory timings from the Overclocking database. However, I'm guessing you've been to the street already Here are my two guesses (I'm running basically the same ram-yours is rated to do DDR500 at 3.2v I believe-with your previous cpu and the same board)

1)The tref and trfc settings aren't right for this new proccessor.

2) Weaker memory controller in the X2. Did you have a winchester or venice 3000+?

Winchester

3) The VX has just slowly died do to voltage creep and it can't maintain the same speed. Are you running any cooling on it specifically?

I have a little bit of airflow directed over it - not a whole lot, mind you, on the order of 5cfm - but some. However, by the same token, I've never taken these sticks above 3.0v. I've never HAD to - it did 250fsb @ 3.0v with 2-2-2-8 timings just perfecttly untill recently.


The drive strength is at 8, I'll try changing the tref from 2560 to 0648, though.

 

[myocardia]

That RAM requires a fan, man. 3v is alot for RAM, and 3.5v isn't even safe for VX, WITH a fan. Also, I just thought of something. Running boards with the 4v jumper set to 4v has been killing/damaging quite a few boards. I'd set it back to it's stock setting (not the processor, the jumper), and only give that VX up to 3.2v, with a 7v modded 60mm fan blowing across it.

 

[anandtechrocks]

Originally posted by: myocardia
That RAM requires a fan, man. 3v is alot for RAM, and 3.5v isn't even safe for VX, WITH a fan. Also, I just thought of something. Running boards with the 4v jumper set to 4v has been killing/damaging quite a few boards. I'd set it back to it's stock setting (not the processor, the jumper), and only give that VX up to 3.2v, with a 7v modded 60mm fan blowing across it.

Eh... I've had my VX running at 255 MHz w/3.5V for over a year and it's still running strong. I have 2 80mm fan each pushing about 30CFM on them though.

OP if I run any dividers below 11:12 (183) my memory craps out on me.

 

[myocardia]

I've heard plenty of stories of VX being ruined with 3.5v, although neither of us know whether or not they were cooling it properly, do we? Are you sure that you actually need that much voltage, for only 255 Mhz? I'm willing to bet that it would do 255 @2-2-2-7 with alot less voltage.

And concerning the 4v jumper on the motherboard, I saw a link somewhere, a few months back, where either DFI or a DFI employee, or at least someone really reputable, was saying that it isn't really safe to use on an Ultra-D. It is the cheapest of their good Skt. 939 boards (which is exactly why I bought it). I haven't heard of it causing problems with the more expensive boards, but I have heard of it causing problems with the Ultra-D, just FYI.

 

[Hajime]

Ouch. I'm changing the 4v jumper back as soon as I finish installing drivers.

However, I'm running the RAM at 2.9v @ 480mhz, 2-2-2-6 timings...The heatspreaders aren't even hot to the touch.

There's like an absolute barrier @ 480mhz with this RAM. No matter what I tweak on it or change, I can't get it past 480mhz for the life of me.

 

[Munky]

I've ran my VX (when I had it) for 1.5 years at 3.3v, 250mhz, 2-2-2-7 timings, and no fan (only case fans). I used both a single core Venice and an Opteron 165, on a DFI Ultra-D mobo. The recommended bios options did not work that well for me, but the default ones did, and I found my own, better, settings through trial and error. Here's what I used:

CPC: enable
CAS: 2
Trcd: 2
Tras: 7
Trp: 2
Trc: auto (found out later it was 13)
Row refresh: auto (= 15)
Trrd: auto (= 2)
Twr: auto (= 3)
Twtr: auto (= 1)
Trwt: 3
Tref: auto (=648)
Twcl: auto (= 1)
Bank interleave: enable
Skew control: auto (= disable)
Skew value: 0
Drive strength: 8
Data drv strength: 2
Max async lat: 7
Response time: fastest
Read preamble: auto (= 6)
Idle sync limit: 256
Dyn. count: disable
R/W queue bypass: 16
Bypass max: 7
32B granularity: disable

Also, there are different bios versions for different memory types. Make sure you're using the right version for the VX mem.

 

[Hajime]

Well, tweaked some stuff.

RAM has about 8-10cfm blowing on it now, running at 2.9v

Heatspreaders aren't even hot to the touch, and I can prob jack it down as low as 2.8v.

However, I STILL can't get it past this 240fsb barrier, even with the suggestions in this thread - swapping to my old proc, I could kick it up to 250fsb easily.

Is it just a weak memory controller? I can run the proc easily at 2.6ghz, if I run the memory at a divider? Whereas I'm stuck at 2.4ghz if I'm using a 1:1 divider?

 

[Hajime]

It was voltage creep.

The RAM can't even test at 2-2-2-6 fsb200 w/3.0volts stably now.

RMA'ing it, thanks for the help.

 

[customcoms]

Yeap, voltage creep kills!. And that 4v jumper crap on the Ultra-D is a bunch of BS!!!! It was a problem with certain PSU's and certain memory sticks. The ultra-d is the EXACT SAME BOARD as the UT sli-d, sli-dr and straight lanparty SLI-DR. The only boards that are different are the venus and expert, which don't even have a 4v jumper (its just a straight bios option). Plus, we never know exactly what the cooling is (unless u get from a dfi rep, in which case the ram was properly cooled I'm 100% positive)!