Corsair Twin2x2048 - 8500C5D

[Nexus327]

HI Ramguy

 

We have talked earlier on this Ram timing and was coming back to show ya the benchmarks I got from it, also to maybe get some more information about the ram because I am questioning on why its not possible to overclock it more then 1167mhz without any issues, and to change timings on it to get better results.

 

there are the results of the over clock of my Quad G0:

(somebody was saying in a post about 3600mhz, i have done 3501mhz stable on a 2.4ghz Quad G0 :cool:

 

http://www.evga.com/community/ModsRigs/comment.asp?sysid=3790

 

 

Also I set the memory to 2.2v like you told me but I have seen posts that where saying it should be at 2.1v , what is the best one for over clocking ?

 

the best timing i got was 5-5-5-15 at 2.2v and at 905mhz

is this normal ? shouldnt it be best at 1066mhz ?

 

is it possible to change the timings to 4-4-4-12 T2 or even get T1 ?

 

Thanks for replying to all my numerous question :sunglasse

 

 

-DJ 5th

[DerekT]

T1 is pretty much impossible with Intel chipsets and Core 2. Nvidia chipsets sometimes can make T1, but not with any real higher overclocking. The new X38 should make T1 at stock settings.

 

This 8500C5 RAM is set for 2.2V if it is a Version 1x. High end overclocking is a YRMV (Your Results May Vary) and as long as the hardware makes spec, then it is all a bit of a crap shoot. It depends is the best answer you can really get. I can buy two sets and find that one set overclocks better than another. But as long as both sets run the stated settings cleanly, then such is the life of the enthusiast overclocker.

 

When I buy parts that do not do what I want but which do handle the stated specs error free, I sell them and buy another set until I find the set that does what I want.

 

That is why I am an enthusiast. The time, energy and money I put into the fun and art of overclocking is more than worth this.

[sparkinark]

In your BIOS, setting the DDR to 1066, 800, etc. only serves to change the CPU/x divisor. All other things being AUTO, some BIOS's will let you run a 1066 on 800 timings, maybe others will put you on EPP timing and let you OC the CPU to get max results (e.g. my "SLI" ready thingy on my Asus M2N32-SLI mobo).

 

Regardless, there is a linkage between the cpu speed (thus the fsb times the cpu multiplier), the DDR input/output (regardless of the speed or timings), and the various HTT busses, depending on which chipset you have. For example, maxing out your fsb with a low CPUx and an overall HTT of like 3x may give you 1183MHz (I got that) but I still got lower Everest scores than with other configs because the HTT3x put me at like 700 something when I max at 1000.

 

Likewise, you can max out HTT(s) and CPU and run at like 805Mhz on the DDR2 with std. 5-5-5-18-23 settings and kill. Depends on where the bottleneck is.

 

And for those who haven't figured this out (took me a while, I assure you), 4-4-4 at 800MHz is probably equivalent to 5-5-5 at 1066MHz. That's' because the "5" is really "5T", where the "T" varies as the DDR bus speed varies.

 

If you really want to know what's going on (and you may find you don't after getting this deep in it), go to Everest and pull put the Corsair timings in "nanoseconds", then re-compute for your DDR speed. I'll post what I got for mine below, but I think either it's not quite right or there is something going on that I dont' know about.

 

Also, and finally, ALL memory settings are important, not just the ones we quote. I have and EE so I have an inkling of what they mean, but I guess until I get the big Corsair "So ya wanna know more about RAM?" book I'll just keep being semi-ignorant and semi-blissful.

 

Remember, the main thing is how well your BIOSHOCK game plays -- to hell with what the numbers are! Let's put things back in perspective here!

 

Enjoy these -- plenty of great late-night sleepers:

 

 

=====================================================

 

 

 

Memory type: DDR2 SDRAM

Error correction: NONE

Voltage interface level: SSTL (1.8V)

 

Data width: 64 bits

Number of physical banks: 2 (double sided)

Number of logical banks: 4

Number of row addresses: 14 14

Number of col addresses: 10 10

 

 

Bank density: 512 512 MB

Primary RAM Width: 8 8 bits

Error checking RAM Width: 0 0 bits

 

Module density: 1024 MB

 

Refresh rate: 7812.500 us

 

Self-refresh: Yes

 

Burst lengths supported: 4 8

 

 

CAS 4.0:

Cycle time (tCYC) = 3.700 ns fMax = 270.270 Mhz

Access time (tAC) = 0.500 ns fMax = 2000.000 MHz

 

CAS 5.0:

Cycle time (tCYC) = 2.500 ns fMax = 400.000 Mhz

Access time (tAC) = 0.400 ns fMax = 2500.000 MHz

 

 

Not reported

Not reported

 

Min delay, Bk/Bk random cols: 0 clocks At 800MHz (/2.5) At 1066/EPP (/1.88)

 

Min row precharge time (tRP): 12.500 ns 5T 5T (7T)

Min row active delay (tRRD): 7.500 ns 4T 3T (4T)

Min RAS to CAS delay (tRCD): 12.500 ns 5T 5T (7T)

Min Actv to precharge (tRAS): 45.000 ns 18T 15T

 

Addr/Cmd setup time: 0.180 ns

Addr/Cmd hold time: 0.250 ns

Data setup time: 0.050 ns

Data hold time: 0.130 ns

 

 

Write recovery time (tWR) 15.000 ns 5T 6T

Int Wr to Rd Cmd Dly (tWTR) 7.500 ns 3T

Int Rd to Pchg Cmd Dly (tRTP) 7.500 ns 5T **change to 3T

 

Min Act/Auto-Rfrsh Time (tRC) 55.000 ns 23T 22T

Min Auto-Rfrsh/Act Cmd (tRFC) 105.000 ns 42T (not input anywhere)

 

Max Dev Cycle time (tCK max) 8.000 ns 4T

 

Max DQS-DQ Skew (tDQSQ max) 2.000 ns 2ns/2000ps

 

Rd data hold skew Fctr (tQHS) 3.000 ns

 

PLL relock time 0.000 ns

 

Memory analysis probe chars: 00

DIMM type: UDIMM (module width = 133.35mm)

FET switch external enable: No

Analysis probe installed: No

 

Includes weak driver: Yes Auto

 

Manufacturer's JEDEC ID code: 00000000 009e7f7f

Module manufacturing location: 1 (manufacture dependant)

Module part number: CM2X1024-8500C5D

Module revision code: 8224 (manufacture dependant)

Module manufacturing date: Wk 255 of 19255

Module serial number: ffffffff

SPD revision: 1.1

 

 

 

--- NVIDIA Enhanced Performance Profile SPD information ---

 

 

Profile 1 valid

Voltage Level: 2.2V 2.2V

Address Command Rate: 2T 2T

Address Drive Strength: 1.0x 1.0x

Chip Select Drive Strength: 1.0x 1.0x

Clock Drive Strength: 1.0x 1.0x

Data Drive Strength: 1.0x 1.0x

DQS Drive Strength: 1.0x 1.0x

 

Address Fine Delay: No delay

Address Setup Time: 1 MEMCLK

 

Chip Select Fine Delay: 1 MEMCLK

Chip Select Setup Time: 1 MEMCLK

 

 

 

Your's in MEMCLK,

 

sparkinark@yahoo.com

[eastcoasthandle]

I've seen 1.x versions do 1066MHz @ 4-4-4-12 at 2.30V/2.25xV with a northbridge of 1.60V, PL 7, etc. Not sure if they are rated at 2.30V or not. I've only seen 2.20V.

[sparkinark]

Run those timings at top speed (such that you don't make a bottleneck) and then run max EPP against like Everest. Would be interesting to see.

 

You understand that min. timings are given in ns and picoseconds, not # of "clocks", which is an arbitrary tern to further complicate the matter.

 

But, I've been out of commercial EE too long so I'm just remembering the pre-read command strobe, even more important to get now that the settle times are moving closer to pico every day (unless we once again open another door only to find 150 more.

 

It is cool that everyone is sharing timings here. It should be a sticky page, with verifications using somnething like a cpu-z "my computer kiiiickssss assss" and I'll show you, Take care and go for 2-2-2-3. Now THAT is a MAN' S core!