Memory timings question

[footballrunner800]

Can 2gb dd2 modules do cas4?

 

[n7]

Definitely.

http://www.newegg.com/Product/...8080+1052508081&name=4

 

[footballrunner800]

thanks. I have some g.skill 2x2gb modules that have timings at 5-5-5-15 but run at 1066. I can't run them at cas4 at 800mhz. I even tried pumping 2.3v ( with active cooling). I need cas4 because want a lower trD.

 

[n7]

Eh, that seems a bit unlucky.

Few things to know:
Your G.Skill is NOT Micron, which in short means it doesn't really like voltage.
It (& nearly every other 2x2 GB kit) is PSC (PowerChips), & you won't see any real improvement past 2.1v usually.

CAS 4 won't get very far with those ICs...i'd say you'll be hitting DDR2-900 if you're very lucky.
Realistically, you won't even likely get past DDR2-850 CAS 4.

I'm not sure i follow your goal of lower tRD with CAS 4...

You'll get the best performance/lowest tRD at higher speeds CAS 5, not lower speeds - & consequently ratios - CAS 4.

So i can't agree with your plan at all...not going to work out so well, unless you're looking for a performance downgrade.

 

[footballrunner800]

i was hoping of doing a 1:1 ratio with an e8400 and fsb at 450 with and a tRD of 7. I need cas4 for it to work. can you recommend me some 2x2 modules that can do cas4 at high speeds? I will RMA these tomorrow. Thanks for your replies n7

 

[footballrunner800]

I dont know which are micron chips but here are some

Mushkin

OCZ

GEIL

these are just some i found. Which one would you recommend?

 

[n7]

Gah tRD 7

Damn low multis makes such high tRDs

Anyway, there's is no guarantee any of those kits will do DDR2-900 CAS 4.

As i mentioned though, you don't want that anyway.

You should be trying for 450 FSB 5:6 (DDR2-1080 5-5-5), not 450 1:1, at least if you want maximum performance anyway.

And the good news is, your current G.Skill or those kits you linked to should do DDR2-1080 for you.

If you're really concerned about numbers, you should try benching 1:1 vs. 5:6.

You'll understand why i recommend what i am then.

Only way you're going to get a guarantee of CAS 4 @ DDR2-900 is if you go to 1 GB dimms using Micron D9GMH.
But then you're stuck with either only 2 GB or 4 dimms for 4 GB, which is a PITA.

Just stick with your existing RAM, up the ratio to 5:6, & go for your 450 MHz FSB.

 

[footballrunner800]

ill run fsb 445 and 5:6 with trd6. I will post some benchmarks when im done. Thanks for your help. Saved me an RMA hasle. :laugh:

 

[n7]

Please do post your results for us...love seeing OCing result

 

[footballrunner800]

Some results

image 1

image 2

 

[ShadowFlareX]

Originally posted by: n7
Few things to know:
Your G.Skill is NOT Micron, which in short means it doesn't really like voltage.
It (& nearly every other 2x2 GB kit) is PSC (PowerChips), & you won't see any real improvement past 2.1v usually.

Do you think Corsair Dominator 8500 2x2GB kit use Micron chips? I can't find any reviews on these.

Originally posted by: footballrunner800
thanks. I have some g.skill 2x2gb modules that have timings at 5-5-5-15 but run at 1066. I can't run them at cas4 at 800mhz. I even tried pumping 2.3v ( with active cooling). I need cas4 because want a lower trD.

I have 4x2GB Corsair Dominator 8500, would run at 800MHz 4-4-4-12 1.8V. Still testing but looks stable on short memtest session.

Here is a good review closest to the Corsair I have, just that it's 2x1GB instead of 2x2GB. Plus they did overclocking at 3-3-3-8, 4-4-4-12, and 5-5-5-15 at voltages 1.9 to 2.9.

 

[n7]

Originally posted by: ShadowFlareX

Do you think Corsair Dominator 8500 2x2GB kit use Micron chips? I can't find any reviews on these.

They are higher binned PowerChips.
Corsair employees have confirmed this on XS.

Originally posted by: ShadowFlareX
I have 4x2GB Corsair Dominator 8500, would run at 800MHz 4-4-4-12 1.8V. Still testing but looks stable on short memtest session.

That's good.

How far can you get with 4-4-4 though?

Most 2x2 GB kits can do DDR2-800 4-4-4, but don't get much farther, as i mentioned.

And when it comes to OCing, you really cannot compare 2x2 GB kits to 2x1 GB ones.
Totally different ICs.

 

[ShadowFlareX]

Getting random BSODs. One was after an hour of OCCT (passed stability test), then after hours of usage (browsing + gaming), then the last one was 2 minutes after arriving on the desktop from a cold boot. And weird BSOD messages too: "Multiple IRP Complete Requests"and "Resource Managemen Not Handled", never had these before.

Upping VDimm to 1.9V.

I haven't tried pushing the RAM further than 800MHz though, since there's little point to overclocking the RAM. Even if I try, probably I can't get far with 4x2GB.
All I've done to my PCs over the years are overclocking, trying something new, running it with lowest voltage as possible while maintaining at least stock speed. Going green a bit

Just looked around the net and you're right, Corsair 2GB sticks use PowerChips, at the same time I found a short review on Corsair 2x2GB PC2-9136C5DF, well, closest to my 8500. The guy wrote it as PC3-9136, but I believe he meant PC2. Overclocked quite well.

 

[brencat]

Originally posted by: n7
You should be trying for 450 FSB 5:6 (DDR2-1080 5-5-5), not 450 1:1, at least if you want maximum performance anyway.

If you're really concerned about numbers, you should try benching 1:1 vs. 5:6.

You'll understand why i recommend what i am then.

n7 is right. For the first 2 months I had my system, I initially ran a 4:5 ratio @ 4-4-4-8 timings (320 FSB: 400 RAM) using 2 x 1gb + 2 x 512 Ballistix. For the heck of it one day, I decided to test the 3:5 ratio @ 5-5-5-15 timings (DDR2-1060 on RAM) and now my system feels snappier and benches a bit faster in SuperPi too. Widening the "pipe" sufficiently on Intel chipsets can offset the negatives of a higher CAS.

 

[n7]

ShadowFlareX, using OCCT for stressing 8 GB is generally a joke.

You really need to run a stresstesting application that truely stresses all your RAM, not one that uses maybe a quarter of it.

I strongly recommend HCI Memtest for testing 8 GB stability.

You have a quad; run four instances set to as much is needed to utilize most of the RAM.
(I usually run 4 instances @ ~1350 MB each at least)

You'll likely error out before you get anywhere close to 100%.

The errors may be caused by needing more MCH/NB voltage, or vdimm.
Or regardless of voltages, you might not be at stable settings.

As far as i am concerned, if you want to actually test for stability with 8 GB, you basically have to use either LinPack or HCI Memtest.

P95 doesn't use remotely close to 8 GB, & even if you run the 4 instances of old versions like 24.14 where you can actually set higher RAM use, i still find it a poor choice for catching errors.
Memtest86+ does test all the RAM, but not simultaneously.
It does it in small chunks, which can mean it isn't great for large amounts of RAM either.

I've run Memtest86+ 6+ passes all night long w/o errors, only to boot into Windows & run HCI & get errors 15 minutes in.

Orthos, OCCT, etc...those also barely utilize any of your 8 GB, & are again basically useless for really testing 8 GB.

Get HCI, run four instances, & you'll find out just how stable [or not] your RAM & NB are.
http://hcidesign.com/memtest/

Or get LinPack, configure it, & it'll make P95 testing look like a joke.
http://www.xtremesystems.org/f...5729&highlight=linpack