Is 5:6 RAM Ration OK for Overclocking ?

[telefanatic]

Hey guys im getting a really poor 3dmark06 around 11k in Vista , im just wondering my ram is set at a 5:6 ration in bios can that hurt my performance ? The ram is a OCZ Platium PC2-8000 at 1020mhz 5 5 5 15 , also my E8400 is OC'd to 3.8ghz on 1.25V and my 3870 is OC'd to 856mhz.
Link to the CPUZ Picture http://s273.photobucket.com/al...=view&current=CPUZ.jpg

My Rig: Intel E8400 3.8Ghz , DFI LanParty X38 T2R mobo , Sapphire ATI 3870 , OCZ Platinum PC2-8000 1020mhz ram 5 5 5 15, PC Power and Cooling 750 Crossfire Ed PSU , BFG PhysX PPU, Thermaltake Armor case w/250mm fan, Thermalright Ultra 120 Extreme, Corsair memory cooler, Creative X-FI ExtremeGamer Fatal1ty, Acer 24" monitor, Logitech 7.1

 

[n7]

Your score is very fine.
5:6 doesn't hurt performance; it helps.

If you want high 3DMock scores, you'll want a heavily OCed quad (not dual), a better video card or two, & XP not Vista.

Fortunately, 3DMock scores mean absolutely nothing, so go play some games & have fun instead of worrying about a stupid rating that is meaningless

 

[lopri]

OCZ Platinum PC2-8000? Are those 2GB sticks or 1GB sticks?

 

[telefanatic]

2GB sticks

 

[superstition]

As far as I've read, there is never a good reason to go away from a 1:1 ram ratio. Running RAM asynchronously just generates heat, requires more voltage, and so forth. Make sure to test with real world benchmarks, not synthetics, if you think this isn't true.

 

[n7]

Originally posted by: superstition
As far as I've read, there is never a good reason to go away from a 1:1 ram ratio. Running RAM asynchronously just generates heat, requires more voltage, and so forth. Make sure to test with real world benchmarks, not synthetics, if you think this isn't true.

Oh please.

I'm well aware that there won't be a huge improvement with running higher speeds of RAM vs. 1:1 with DDR2, but why run a lower speed when it's capable of higher.

I tested 4:5 vs. 1:1 a long time ago back when C2Ds came out on my own system, as have many others, & while the benefits are very small, in general, high speeds will beat out lower speeds. (excepting when using 1T on nForce chipsets - had to add that just for lopri & those few nVidia chipset users )

The only way it would geneate "more heat" is if you were running higher voltages, which isn't always necessary to run higher speeds.

Don't rely on reading; benchmark for yourself.

Again, i know very well how minimal the gain from running over 1:1 is, but generalizing & saying that running over 1:1 isn't wise is just not accurate.

I know people hate this example, but i'm going to bring it up again because it shows how bandwidth can matter.

Intel + DDR3.

Please show me systems running that @ 1:1.
Yes, everyone wants to run their DDR3 @ DDR3-667 speeds, right?
No, they're running 1:2 usually.

Intel has promoted higher than 1:1 ratios for years & years now; there are reasons for it.

 

[Mondoman]

Originally posted by: n7
...
5:6 doesn't hurt performance; it helps.
...

Actually, running the RAM one "notch" higher than the FSB (e.g. DDR2-667 RAM speed with 1066MHz FSB) *can* reduce performance a bit, presumably due to synchronization issues between the memory bus and FSB. See for example these tests: http://www.madshrimps.be/?acti...tpage=1962&articID=472

 

[lopri]

Originally posted by: Mondoman

Originally posted by: n7
...
5:6 doesn't hurt performance; it helps.
...

Actually, running the RAM one "notch" higher than the FSB (e.g. DDR2-667 RAM speed with 1066MHz FSB) *can* reduce performance a bit, presumably due to synchronization issues between the memory bus and FSB. See for example these tests: http://www.madshrimps.be/?acti...tpage=1962&articID=472

That was true until 975X. (925X, 955X, etc.) Starting P965, Intel changed strategy and by the time we have P35, their chipsets are very much tuned for higher bandwidth instead of tight timings. Do I like that change, NO. Plus, Intel chipsets are very inefficient at handling dividers. (unlike AMD's integrated memory controller which has nearly zero performance overhead when using dividers) I'm a fan of tight timings and I hope we see the return of it with Nehalem, but with current Intel chipsets higher bandwidth will give better performance than tight timings.

 

[bunnyfubbles]

Originally posted by: lopri

Originally posted by: Mondoman

Originally posted by: n7
...
5:6 doesn't hurt performance; it helps.
...

Actually, running the RAM one "notch" higher than the FSB (e.g. DDR2-667 RAM speed with 1066MHz FSB) *can* reduce performance a bit, presumably due to synchronization issues between the memory bus and FSB. See for example these tests: http://www.madshrimps.be/?acti...tpage=1962&articID=472

That was true until 975X. (925X, 955X, etc.) Starting P965, Intel changed strategy and by the time we have P35, their chipsets are very much tuned for higher bandwidth instead of tight timings. Do I like that change, NO. Plus, Intel chipsets are very inefficient at handling dividers. (unlike AMD's integrated memory controller which has nearly zero performance overhead when using dividers) I'm a fan of tight timings and I hope we see the return of it with Nehalem, but with current Intel chipsets higher bandwidth will give better performance than tight timings.

that and you can actually run true memory dividers where the memory is running slower than the "FSB" and still not hurting performance which meant your overclocks were no longer held back by budget ram - you'd just try to keep the ram at its rated speeds or lower and then just tweak the timings to be as tight as stability would allow all while running the CPU at the highest speed it could go. People buying super expensive ram might only enjoy a few % points better in benchmarks and the only true 'value' of doing so would be for bragging rights and little else

 

[Zenoth]

Use ratios to keep you ram at desired speeds if you don't want to over-clock them too much or at all.

My E8400 runs at 3.6Ghz at the moment (400x9) and with my ram at 3:4 ratio it runs at exactly its rated default speed of 1066Mhz, but at 1:1 it runs at an over-clocked speed which needs a complete re-configuration in the BIOS (voltage, timings) to ensure stability, which in my book it just too long to do, and, after all, my memory is already over-clocked 6400 modules, basically, so I don't feel the "need" to push them above 1066Mhz.

 

[VirtualLarry]

Originally posted by: n7
Intel has promoted higher than 1:1 ratios for years & years now; there are reasons for it.

I have yet to see a real reason for it.

 

[VirtualLarry]

Originally posted by: Zenoth
Use ratios to keep you ram at desired speeds if you don't want to over-clock them too much or at all.

My E8400 runs at 3.6Ghz at the moment (400x9) and with my ram at 3:4 ratio it runs at exactly its rated default speed of 1066Mhz, but at 1:1 it runs at an over-clocked speed which needs a complete re-configuration in the BIOS (voltage, timings) to ensure stability, which in my book it just too long to do, and, after all, my memory is already over-clocked 6400 modules, basically, so I don't feel the "need" to push them above 1066Mhz.

That doesn't make sense. 1:1 is the lowest memory multiplier. It's not faster than 3:4, quite the opposite.

 

[Mondoman]

Yep, zenoth's RAM should be at DDR2-800 for a 1:1 ratio on a 400MHz FSB.

 

[superstition]

I'm well aware that there won't be a huge improvement with running higher speeds of RAM vs. 1:1 with DDR2, but why run a lower speed when it's capable of higher.

Because the "speed" depends on the ratio, not just the MHz. Again, if you can show some real world benchmarks that support your point, do so.

I'm running my RAM at 750 instead of 800 to be in 1:1, and I doubt the difference between 750 and 800 is going to be made up for by going asynch. One article argued that real world benchmarks didn't show a significant improvement between 667 DDR-2 and 800 DDR-2. I found that interesting. However, I did see one a while back that showed higher frame rates with overclocked ram, although there was a drop when going out of 1:1 for some time. One article I read recently said MCH voltage may need to be increased if running asynchronously as well, which is another downside.

 

[Zenoth]

You guys are right indeed... sorry about the mistake. I don't know where I had my mind exactly.

One thing is for sure though, is that I have to use a 3:4 ratio at 400FSB, or else I get some incredible memory performance loss.