• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Seeking someone's successful experience re OCZ latencies

BonzaiDuck

BonzaiDuck

Lifer
Somebody should have some insights on this question.

I have the following configuration:

ASUS P4P800 mobo
P-4 Northwood 3.0C @ 3.75 and Vcore 1.55V -- FSB @ 1,000 Mhz
OCZ EL "Gold" DDR500/PC4000 @ 2.5, 4, 4, 7 (stock latency settings)

Both my intuition and experience inform me that I can get tighter latencies without too much trouble by lowering the external frequency and FSB -- thus running the memory below its specification.

But this configuration is so stable and rock-solid. And COOL: my load temperatures are peaking at about 91F/37C @ room temperature of 73F. I'm wondering if anyone with a similar configuration -- and relevant variations would include the ASUS P4C800 boards -- has had any success tightening these memory latencies.

What latency setting would you suggest I try?
 
Well Intel and really tight timings (especially at the overclock your currently at) are not going to be as easily obtainable. Right now I'm running OCZ 3200 rev. 2 with 2-2-2-5 but my processor is at stock settings.

Have you tried setting your timings at 3-3-3-8? That might relieve some of the stress. My experience is that Intel's really don't like the tight timings.

Try the 3-3-3-8 and see where that gets you. Your current overclocks with the stock latencies are nothing to scoff at.
 
Yo, TylerDurden

Lotta AMD aficionados here.

I'm looking at this -- your recommendation. You're suggesting I increase the Row Precharge delay from 7 to 8, increase the CAS latency from 2.5 to 3, and decrease the other two latencies by one each?

I'm still relatively novice at this, and I need to ponder the integer arithmetic. It would seem that reducing the RAS-to-CAS delay by one and bumping up the CAS latency by 0.5 is overlapped by the Row Precharge Delay, so I can't see what that would gain. Well -- wait a minute. What do you think the gain would be -- two clock cycles?

I'll think about it some more, maybe try it in the next day or two.

The stock settings -- under the existing regime -- seem totally stable. Ran MEMTEST86 several times allowing about 10 iterations of the several tests -- no error. Stress tests -- some 20 iterations -- under SiSoft Sandra '04 -- no error. Can't get the load temperatures to go over 91F or so at 73F room.

If you think that's nothing to sneeze at, I may try the other settings, but will probably be ready to leave it alone at the stock values if I run into trouble.

Some user-review post -- think it was NewEgg's page for the rev.2 OCZ gold DDR500's -- said he had been able to run the modules at 2, 2, 2, 5, but no other information, and I guess that he wasn't OC'ing the processor and (for that matter) the FSB and memory.
 
Well if your looking to overclock more loosening the timings might help in reaching a higher FSB. I'd say you've got your system pretty much as far as she will go on air cooling. Be happy and game away. 🙂

Basically you probably won't get tighter timings at the o/c you are currently at.
 
Originally posted by: BonzaiDuck
Somebody should have some insights on this question.

I have the following configuration:

ASUS P4P800 mobo
P-4 Northwood 3.0C @ 3.75 and Vcore 1.55V -- FSB @ 1,000 Mhz
OCZ EL "Gold" DDR500/PC4000 @ 2.5, 4, 4, 7 (stock latency settings)

Both my intuition and experience inform me that I can get tighter latencies without too much trouble by lowering the external frequency and FSB -- thus running the memory below its specification.

But this configuration is so stable and rock-solid. And COOL: my load temperatures are peaking at about 91F/37C @ room temperature of 73F. I'm wondering if anyone with a similar configuration -- and relevant variations would include the ASUS P4C800 boards -- has had any success tightening these memory latencies.

What latency setting would you suggest I try?
Click to expand...

2.5-4-4-7 are pretty decent timings for 250 fsb. You won't get much tighter timings, and you wouldn't see much of a difference anyways even if you could. Great OC btw...be happy with it. 🙂
 
Well, Charloscarlies --

I think you could be right. Everything I've read suggests you get more from the faster FSB and memory bus speeds than the few extra pieces of data by dropping a few clock cycles from memory access.

I HAVE been lucky with this. I have respect for the ASUS P4C800 boards, and friends who chose that board with the i875 chipset sort of poo-poo-ed my P4P800. But the i865P chipset and P4P800 turned out to be a winner -- for my purposes, anyway.

Now I have to start planning both the budget and the component choices for a P5AD2 build late next year. . . . . 🙂 [I could try a different board, but so far, that one gets rave notice. Well, we'll look a little harder.]
 
For anyone interested, I found the following post at the OCZ Forums. The only reservation I have about the poster's configuration is that his VDIMM voltage is outside the tolerance for the OCZ EL Gold rev.2 DDR500 modules. Also, intuitively and tentatively, it would seem that his ability to tighten the RAS-to-CAS and RAS Precharge values might hinge on his 2.9V VDIMM setting.

Even if I wanted, my ASUS mobo won't give me options above a 2.85V setting, which is also the upper tolerance limit for the OCZ modules. Still . . . maybe . . . wouldn't hurt to try it after recording a PCMark04 score and noting a PassMark 5 memory performance result, then checking again if indeed changing the pair of 4's to 3's will work.

**** LIFTED FROM OCZ's FORUM ****
I am hoping there are some ideas out there as to why I can't seem to get any more than 252 out of my FSB. My system is as follows:
3.2E C0
OCZ PC4000 gold rev 2 2x256 2.5-3-3-7
500W Ultra-x connect ps
Water cooling
Asus P4C800-ED Vcore and Vdimm mods done and functioning
Vcore 1.55V
Vdim 2.9V
Vagp 1.7V


My problem is that I can't seem to get any higher FSB even by increasing the Vdimm to 3.2 volts. In fact if the voltage to the memory is increased past 2.94v memtest locks up imediately after it starts. I must say I am happy at 4 gigs but I think that it should be possible to go a bit higher. I have also tried relaxing the memory timings, to no effect. temps are around 35c idle and 45c loaded.

Any suggestions would be welcome
Thanks

************************
 
To see if you're at the limit of your chip, run a divider and keep pushing it higher. Then you can decide about timings. 🙂
 
Now THAT idea -- I should file away. Hadn't thought of it, since I would prefer running the memory at 1:1.

Even so -- this latest development shows how we can humble ourselves by failing to review "old reviews". The OCZ DDR500 Rev. 2 modules were reviewed in Tom's Hardware Guide mid-January, 2004. That's when I started seriously considering them for this build.

And in that very short article was a statement that they had successfully tested the modules at their rated bus speed of DDR500 -- external clock setting of 250 Mhz -- and bumped the stock latencies down from 2.5, 4, 4, 7 to 2.5, 4, 4, 6.

I tried it. Punched "ESC" after 10 iterations of the seven MEMTEST86 v. 3.1a tests. (With no errors, of course.)

My PCMark04 score went from 5674 to 5699. I think I could break 5720 and another "Project" at the FutureMark database for a 3.2 processor over-clocked to 3.759 Ghz. That published project got a score of 5716, but with only an external clock setting of 235.

Anyway, that's only a little more than 200+ points away from Intel's own PCMark04 score for their 570J -- 3.8 Ghz LGA775 processor and a 925 chipset (most likely, almost certainly, positively absolutely an Intel motherboard -- what else would they use for the benchmarking?)
 
Say -- Beatle -- You went to VPI??

I went to George Mason -- I had profs from VPI . . .

. . . And Blacksburg -- not a bad place to be, either !!
 
Increasing your VDIMM to the max the motherboard will allow is the first step for faster frequencies and timings if you use good high end memory. Most high end memory chips will work with VDIMM's up into the 3.2 to 3.4 range. I have seen some people get 2-2-2-5 at 260 with vdimm at 3.4 on OCZ and G.Skill memory modules. I think OCZ is now selling a memory with a lifetime warranty up to 3.5 on the vdimm or close to it!

I looked at your system specs and I say great job and enjoy for me please! LOL (Lowly XP2400+ on KM266 chipset)
 
Gee, thanks michaelpatrick and the rest of youse guys -- it gives me that "birthday boy" feeling! 🙂

I'm running SiSoft Sandra04 Burn-in Wizard right now. The temperatures don't want to go over about 91F. I'll probably bump up the clock rate a few megahertz, but I can't get the voltage to go beyond 2.85V for VDIMM. Somehow, I think the P4C800-E board allows it -- but this P4P800 doesn't. Maybe I should check for BIOS upgrades, but I thought I did that during the summer, and this 865PE chipset and board have been in the market for -- well, since around June '03 I think, or shortly thereafter. I know it was available by October 03, because that's when I ordered this one.

I probably don't need to push it any more. But my P4C800-E friend across town -well -- he'll rib me about the BIOS voltage limit.

I just know he will. . . . He's "threatening" to go "water-cooling" if my benchmarks top his, but he seems reticent about downloading mainstream benchmarking software -- although mainly because he can only get dial-up access out there in the sticks.

Next time I get asked for dinner by he and his wife, I'm going to bring a CD filled with this year's versions of all the benchmark programs 🙂
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top