Good memory for overclocking

metamorphpwner

metamorphpwner

metamorphpwner

Associate
Joined
6 Sep 2008
Posts
2,111
Location
Near Hull
I've been looking at THIS OZC 2x2 dual channel kit mainly because of its price.
Is it any good for overclocking with?
i'll be getting an intel Q6600 with either a Biostar TPower I45 or asus P5q Pro(probably the biostar).
don't really know much about the RAM especially when it comes to overclocking.
 
OCZ 4GB (2x2GB) PC2-8500C5 1066MHz Dual Channel Platinum Series DDR2

£55.19 inc

Kennysevenfold said:
why bother overclocking the ram.?

you would only benefit really if you overclock that cpu to up and beyond 3.0GHz mate to be honest :)
Click to expand...
I don't agree,

An Intel® Core™2 Quad Q6600 can make full use of PC2-8500 even while the processor is running at stock!

Stock FSB = 266MHz
Stock System Bus = 1066MHz
Stock PC2-8500 Speed = 1066MHz

Once you start overclocking the FSB and making the CPU and System Bus faster it makes sense you would also benefit from having faster Memory and therefore more bandwidth.

PC2-8500 (DDR2-1066) is de-facto for Intel® Core™2 systems these days, especially if overclocked! (if your motherboard can indeed run the memory at uBer Mhz!)
 
Big.Wayne said:
OCZ 4GB (2x2GB) PC2-8500C5 1066MHz Dual Channel Platinum Series DDR2

£55.19 inc

I don't agree,

An Intel® Core™2 Quad Q6600 can make full use of PC2-8500 even while the processor is running at stock!

Stock FSB = 266MHz
Stock System Bus = 1066MHz
Stock PC2-8500 Speed = 1066MHz

Once you start overclocking the FSB and making the CPU and System Bus faster it makes sense you would also benefit from having faster Memory and therefore more bandwidth.

PC2-8500 (DDR2-1066) is de-facto for Intel® Core™2 systems these days, especially if overclocked! (if your motherboard can indeed run the memory at uBer Mhz!)
Click to expand...


How have you worked that out ? I have read plenty of overclocking guides and I have never seen anything like that?

The answer to the question is how much are you overclocking? 3.6GHz (9x400) or below then ANY 6400 RAM will be fine. Even slightly beyond that almost all 6400 RAM will be fine.
 
mame said:
How have you worked that out ? I have read plenty of overclocking guides and I have never seen anything like that
Click to expand...
You could read all the overclocking guides in the world and not learn as much as actually performing your own testing. I always appreciate other people taking time to write guides but you can't always assume they are right! :p

The main limitation of the Intel® Core™2 platform is system bandwidth, doesn't matter if you got a 4GHz + processor speed or a 500MHz-FSB if your memory speed is crawling along.

Synchronous (1:1) memory ratios are only really for people to use during their basic overclocking tests and also for people who have cheap/incapable sticks. PC2-6400 is really old hat now but still suitable for a system using a Non Overclocked 800MHz-System Bus chip such as the E5200 (or for basic 1:1 O/C testing)

Intel® Core™2 E5200
FSB = 200MHz
System Bus = 800MHz
PC2-6400 Real = 400MHz
PC2-6400 Effective = 800MHz

If you want a faster computer then get some fast memory and overclock it to match your overclocked CPU and overclocked FSB! I always find most tasks improve a lot when the effective Memory-Speed matches (as close as is possible) the effective System-Bus speed! :cool:
 
so, Big.Wayne the RAM you posted would be good for overclocking with. i was sticking to a fairly tight budget but now i can go a bit higher.

edit: THIS posted above would be good aswell then right, 1066MHz PC2-8500
 
Last edited:
Big.Wayne said:
I think that OCZ platinum is good stuff and is a steal at that price!
Click to expand...
Been considering changing from my current ocz reaper hpc pc6400 to that ocz platinum, ive been offered a decent price for my current ram from a mate, and it would cost little to upgrade. Worth it or stick with what i have?
 
I'm currently using OCZ Reaper HPC 9200 (1150mhz) and they are running fine OC'd to 1200mhz. OK i'm not pushing them too hard, but seings as they like a lot of juice by default 2.1v-2.3v , if you wanted i'm sure they'd go further.

Using 375 FSB on my Q6600 and a 333 Ram bridge @ 1200mhz , i'm getting an OC i'm happy with, the machine seems to run lighting fast and now I have the right RAM voltage (2.3v), it's finally stable.

So If you want to , as you indicate,and try to OC to the MAX, you need memory that can keep up surely ;)

Plus 5-5-5-15 CAS isn't bad for the higher speed rams, you'd be lucky to get 4-4-4-12!

CPU-Z.gif
 
1dmf said:
I'm currently using OCZ Reaper HPC 9200 (1150mhz) and they are running fine OC'd to 1200mhz. OK i'm not pushing them too hard, but seings as they like a lot of juice by default 2.1v-2.3v , if you wanted i'm sure they'd go further.

Using 375 FSB on my Q6600 and a 333 Ram bridge @ 1200mhz , i'm getting an OC i'm happy with, the machine seems to run lighting fast and now I have the right RAM voltage (2.3v), it's finally stable.

So If you want to , as you indicate,and try to OC to the MAX, you need memory that can keep up surely ;)

Plus 5-5-5-15 CAS isn't bad for the higher speed rams, you'd be lucky to get 4-4-4-12!
Click to expand...
Im currently running my reaper 6400 @844mhz, 2.02v, 333mhz fsb-nb strap, q6600 @3.8ghz, (9x422) the default timings for my ram are 4-4-4-15, but im currently running them at 5-5-5-18.
 
Very little point in overclocking memory* - mainly because once you start increasing the voltage to it a lot of sticks tend to die very quickly - even with as little as +0.1v over rated in some cases (D9GMH, etc.)

Performance gains are small - for most realworld useage you won't even notice the difference - in benchmarks only marginally - i.e. the difference in 3D Marks 06 between 533MHz @ CAS5 and 1152MHz @ CAS4 (yes CAS4) is under 5% and the difference in something like superPi in the region of 200ms.

My advice usually is to run the RAM 1:1 with the CPU and get the timings down as tight as possible.

The only exception here is probably high (4+gig) clocked quad (or more) core CPUs in which case you'd have to do your own measurements to see if RAM bandwidth is a bottleneck.

* When using a Core 2 based system, for older P3/4, athlon and some newer AMD systems then it does have bigger gains. I should note that this is mostly based on my own experiences overclocking E6x00 and Q6600 cpus upto 4gig max. My highest (stable) RAM overclock was a set of G.Skills 6400Hz @ 1333MHz @ CAS5.
 
Last edited:
My advice usually is to run the RAM 1:1 with the CPU and get the timings down as tight as possible.
Click to expand...

Really? I've tried that doing 333 / 333 / 666 , thought it ran way slower than current OC.

1:1 might be a thought if I could get 500-550 CPU FSB ;)

Im currently running my reaper 6400 @844mhz, 2.02v, 333mhz fsb-nb strap, q6600 @3.8ghz, (9x422)
Click to expand...

not bad, I might consider geting a better CPU OC now I have the memory stable. though you are under volting them by .08v according to their website?, yet it shows mine if run @ 800mhz to be 1.8v

The OCZ PC9200 have stock voltage of 2.1v-2.3v , and my previous Cellshock took 2.0v-2.2v which seems normal for the higher frequency memory.
 
Last edited:
Rroff said:
My advice usually is to run the RAM 1:1 with the CPU and get the timings down as tight as possible.
Click to expand...
That's not good advice Rroff! :p

1:1 synchronous memory is always gonna bottleneck a Intel® Core™2 system and really should be avoided if you want the best performance. a lot of people still don't realise this unforunately :(

More Bandwidth = Faster Computer (faster games, encoding etc etc)
 
1dmf said:
Really? I've tried that doing 333 / 333 / 666 , thought it ran way slower than current OC.

1:1 might be a thought if I could get 500-550 CPU FSB ;)



not bad, I might consider geting a better CPU OC now I have the memory stable. though you are under volting them by .08v according to their website?, yet it shows mine if run @ 800mhz to be 1.8v

The OCZ PC9200 have stock voltage of 2.1v-2.3v , and my previous Cellshock took 2.0v-2.2v which seems normal for the higher frequency memory.
Click to expand...
Im undervolting them due to my p5q deluxe board supposedly overvolting upto +08v on some bios revisions, wether this is true enough or not im not 100% sure, my version of the reaper is rated from 2.1-2.15 evp, but ive never had any problems with stability with my current settings, may tighten the timings back to their proper 4-4-4-15 as 44mhz over stock 800mhz isnt a very large memory oc.
 
Im undervolting them due to my p5q deluxe board supposedly overvolting upto +08v on some bios revisions
Click to expand...
Hmm I hope my P5Q-WS doesn't suffer from the same thing as i've put the BIOS to max 2.3v :D

I might drop it to 2.22 to be on the safe side and see if it stays stable, thanks for the potential heads up.

I'd kill for a set of DDR2 that ran at 1333MHz sheeeeeeet!
Click to expand...
well PC9600 DDR2 are 1200mhz stock, do you think they would OC to 1333mhz?
 
Last edited:
Big.Wayne said:
That's brilliant! :cool:

I'd kill for a set of DDR2 that ran at 1333MHz sheeeeeeet! :o
Click to expand...

Not so brilliant... the voltage killed them inside 18 months even with active cooling... it was good while it lasted... that was kinda what put me off RAM overclocking tho - it needed that much extra to actually put up a realword difference instead of something that could only be noticed on benchmarks.

I've always found running 1:1 with fairly tight timings (I'm talking 4-3-3-7 kinda ballpark) gave a more responsive "feel" to general windows usage tho its hard to state it as fact and the fps difference compared the higher bandwidth was negligible - infact with only a minor FSB increase to get it to say 850MHz 4-3-3-7 it would often run as or almost as fast as 950-1000 @ 4-4-4-15 with the lower FSB.

Only screenshot near that speed that I can find off hand: http://aten-hosted.com/images/spi3.jpg (only 1272 there but memory was 100% stable there too - CPU wasn't 100% as it took 1.65v using water cooling to eventually get it tweaked for 3.825 stable - but it could pass 2 hours on orthos ok before failing using the settings in SS).

NOTE: A lot of these numbers will be very different depending on the system you are playing with, especially the memory latency and northbridge/memory controller performance - i.e. nForce boards will often put up a lot higher memory through put than intel boards.
 
Last edited:
Rroff said:
I've always found running 1:1 with fairly tight timings (I'm talking 4-3-3-7 kinda ballpark) gave a more responsive "feel" to general windows usage tho its hard to state it as fact
Click to expand...
That "feel" you describe or "snappiness" is most likely the lower nanoseconds response time you get when the Northrbridge and your system is tweaked just so, I found the FSB-Northbridge strap setting help things feel more "nippy" when it was running on a lower strap (i.e 266-Strap felt faster than 333-Strap etc), I have since found out about this NB timing called tRD (or PL/Performance level) which plays a big part in making a system responsive and improving bandwidth. There is a good chance that if you were running 1:1 with manually set low timings the tRD setting could have been quite tight and gave a "nippy" feel although bandwidth would have been less.

I used to run all my Intel® Core™2 systems using 1:1 sync memory settings and it wasn't until I bought my 3rd LGA775 mobo (P5Q-Deluxe) and started hanging out in the P5Q thread over at XS that I started to realise I had been doing it wrong all that time.

Before while using 1:1 sync memory ratio I had noticed my benchmarks always trailed behind others results, whether it was gaming or encoding etc but once I started running the memory async on an upwards divider and taking control of the FSB-Strap and tRD setting things were noticably faster!

Basically there is no right or wrong as long as it works good for you but now all my systems are tweaked a bit better than in the past and they are all using async fast memory and the computers "feel" and "benchmark" faster! I don't use 1:1 memory ratio anymore except for initial overclocking testing when you need to remove the memory from the equation!
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom