i7 upgraded DDR3 but no faster ?

digitaldreams

digitaldreams

digitaldreams

Associate
Joined
5 Jul 2007
Posts
512
i7 upgraded DDR3 sticks but no faster ?

i7 965 CPU
Asus p6T WS Pro Motherboard
Corsair Dominator TR3X6G1600C8D Ram (8-8-8-24)
Basic Bios settings, no overclock.

Memtest 86+ gave 17431 MB/s, great !

Read a recent article on Toms Hardware where they tested PC3-16000 (2GHz) modules. Noted that the results for most at 1.6GHz were an even more impressive 27240....so baught the 2GHz OCZ blade set to get the most from 1.6GHz.

Memtest 86+ gave 17431 MB/s, what !

Timings had been manually set at 6-6-7-16 as in the article & CPUz confirmed this plus the correct DRAM frequency of around 800MHz (1600).

Any clues ?. I must be missing something here as the timings alone are much better (e.g. CAS 8 down to 6).

Regards
 
Last edited:
First thoughts are that memtest is a very strange tool to be using as a benchmark, what do other programs make of the change?
 
Not the point !

I'm missing something if for the first time I experience no improvement in memory bandwith by decreasing CAS !. I'm talking of many years of experience.

There must be a bios setting that's throttling/overriding my settings....the bios does seem to confusingly have many more settings than my old 775 socket board.

Regards
 
Well

Tried both sets again @1600, this time with Everest and Sandra benchmarks.

Both Everest and Sandra showed a 'slight' improvement (Everest giving different readings each time it was run, Sandra more stable)

So back to the latest Memtest 86+ & the Blade set. I can't accept that the bandwidth doesn't change with CAS, it has on all my previous rigs. So I pushed it wider, 9-9-9-24 (auto in bios), lo and behold Memtest showed a poorer 15722 MB/s. So it DOES change. So why does it seem to hit a limit of 17431 @ CAS 8 & 6 ??

Regards
 
Hey digitaldreams,

you got such an utterly amazing system there, holy **** :eek:

It seems such a pity that your enjoyment is being spoiled by a small technical hiccup! :(

It's fair enough though man I understand, I get a similar thing where I get niggled by something not acting as I want it to heh! :p

I too use the bandwidth readout in Memtestx86 as a quick gauge of tweaks, obviously the Memtest readout is *unbuffered* compared to SiSoft Sandra which is *Buffered*, its two different things. SiSoft Sandra can be configured to take off the Buffering and the results then tally with what you see in Memtestx86.

Your making some assumptions based on your previous experience with older technology when actually you have some killer state-off-the-art hardware that is still not widespread and holds a lot of mysteries . . . . your the dude at the cutting edge so sadly there isn't much support and its gonna be down to you to sort things out and then tell us lot how to do it in several months heh! :D

That thing you were saying about reducing the memory subtimings to improve bandwidth is not entirely true, as far as I know its the memory speed (MHz . . .now GHz) that offers a lot of the bandwidth, as well as the tRD setting on the old Northbridge which also increased bandwidth but relied on low memory timings. Now the Memory controller has been moved from the Northbridge to the CPU I ain't got a clue how its all works! :p

Good luck though, brave new world and all of that! :cool:
 
Are timings that impoertant on i7? i have tried cas 7 vs 8 on my rig and have noticed not much real world gains, hardcore benching doesnt interest me, im interested in best performance with 1.65v and mhz seems to win.
 
Thanks Big Wayne...

Might try overclocking the ram interface next, tho I demand uber stability at all times.

For those wondering why I bother with all this....I'm currently an IT engineer ;), used to design & build electronic/computer systems from the bare components up for the aerospace industry (miss that badly). Yup, if a processor was involved I had to program the end result in machine code too. Hence I'm a bit of a perfectionist and twiddler to boot. I also run some demanding ray tracing, taking all day to produce the end result, so an hour or two saved is worthwhile....

Regards
 
You must log in or register to reply here.
Back
Top Bottom