-
Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
z97 Is here
- Thread starter Witchdoctor
- Start date
OP
- Joined
- Apr 12, 2014
- Thread Starter
- #22
Sounds like we will have a literal 3400 setting moving forward 
Possibly higher on the boards aimed at pushing the envelope.
Hopefully this is an indication that DC's IMC will be stout .....
Possibly higher on the boards aimed at pushing the envelope.
Hopefully this is an indication that DC's IMC will be stout .....
OP
- Joined
- Apr 12, 2014
- Thread Starter
- #24
Hmmm...
My z87M OCF has advertised OC memory profiles in the BIOS ?
when set the stix run at said speeds with out pushing bclk up ?
I must be looking at this wrong as I thought you would be able to start pushing bclk at 3400 Mhz as a starting point ?
making 4k easier to obtain ?
This would really put the pressure on the IMC of the new chips
My z87M OCF has advertised OC memory profiles in the BIOS ?
when set the stix run at said speeds with out pushing bclk up ?
I must be looking at this wrong as I thought you would be able to start pushing bclk at 3400 Mhz as a starting point ?
making 4k easier to obtain ?
This would really put the pressure on the IMC of the new chips
OP
- Joined
- Apr 12, 2014
- Thread Starter
- #26
I see, will read up, My thought process was you would need to push the bclk to take speed higher than setting in BIOS
never was any good at overclocking memory, although I am contemplating grabbing a memory pot
never was any good at overclocking memory, although I am contemplating grabbing a memory pot
No no... you are right. I am agreeing.
Assuming: multi works, sticks can handle it, IMC can handle it, set it to one of the highest memory multi/speeds then use BCLK to push it up past that point.
There may be more performance to be had with higher BCLK with lower memory multi/speeds though...that is not something I know however.
For example 167xmemory multi to reach 3200Mhz or 100xhigher mem multi to reach 3200.
Sorry I was vague earlier, LOL!
Assuming: multi works, sticks can handle it, IMC can handle it, set it to one of the highest memory multi/speeds then use BCLK to push it up past that point.
There may be more performance to be had with higher BCLK with lower memory multi/speeds though...that is not something I know however.
For example 167xmemory multi to reach 3200Mhz or 100xhigher mem multi to reach 3200.
Sorry I was vague earlier, LOL!
OP
- Joined
- Apr 12, 2014
- Thread Starter
- #28
Yep,
Thought I was loosing it
My whole thought here is if the MB makers are advertising higher memory speeds all be it under the (OC) moniker, is this simply marketing and if not will this mean stronger IMC's on the Devil's Canyon chips.
If so, one would conclude world records are going to be broken
Have a friend who will be rocking the new OCF and DC in a few weeks
Should have some definitive answers to these questions
Thought I was loosing it
My whole thought here is if the MB makers are advertising higher memory speeds all be it under the (OC) moniker, is this simply marketing and if not will this mean stronger IMC's on the Devil's Canyon chips.
If so, one would conclude world records are going to be broken
Have a friend who will be rocking the new OCF and DC in a few weeks
Should have some definitive answers to these questions
IB, IB-E and Haswell memory controllers have maximum memory ratio of x29.33. IB-E and Haswell have 3 straps 100, 125, 166MHz what means that:
-> 100 x 29.33 = 2933 MHz ( DDR ) <- this is max official memory speed for every Z87/X79 mobo and probably Z97 too
-> 125 x 29.33 = 3666 MHz ( DDR )
-> 166 x 29.33 = 4868 MHz ( DDR )
There were IB boards with higher memory ratio but these ratios were not working.
So when you set profile for higher clock then board will overclock bclk to required value and if needed will raise voltages.
ASRock OCF are easiest boards to overclock memory. Most settings are optimized or already programmed to change for optimal performance or max clock. You can adjust all manually but auto settings are usually best anyway.
IB-E is also running great with memory. I was able to reach ~3700+ clock on my 4930K+RIVE. The same memory on MSI board couldn't make more than ~2750.
It doesn't change fact that anything above 3000 has no point as current even best IC has to run at really relaxed timings to be stable what means that something like 2600/2666 CL10 or CL11 will be still better.
I don't think we will see any DDR3 that will be running at CL10/11 3000+. I mean like in mass production not single samples.
-> 100 x 29.33 = 2933 MHz ( DDR ) <- this is max official memory speed for every Z87/X79 mobo and probably Z97 too
-> 125 x 29.33 = 3666 MHz ( DDR )
-> 166 x 29.33 = 4868 MHz ( DDR )
There were IB boards with higher memory ratio but these ratios were not working.
So when you set profile for higher clock then board will overclock bclk to required value and if needed will raise voltages.
ASRock OCF are easiest boards to overclock memory. Most settings are optimized or already programmed to change for optimal performance or max clock. You can adjust all manually but auto settings are usually best anyway.
IB-E is also running great with memory. I was able to reach ~3700+ clock on my 4930K+RIVE. The same memory on MSI board couldn't make more than ~2750.
It doesn't change fact that anything above 3000 has no point as current even best IC has to run at really relaxed timings to be stable what means that something like 2600/2666 CL10 or CL11 will be still better.
I don't think we will see any DDR3 that will be running at CL10/11 3000+. I mean like in mass production not single samples.
Last edited:
- Joined
- Jul 29, 2011
- Location
- Buenos Aires, Argentina
I want an OCF z97 + unlocked pentium...hoping it'll clock mem just as good as its bigger brothers.
OP
- Joined
- Apr 12, 2014
- Thread Starter
- #31
Yea, MSI has always seem to be memory weak, my x79 GD65 can't even do the advertised 2400 ,
3700 is rocking on a x79 .........
If I remeber right 4400 and change is the world record,
I am thinking that will be broken at CompuTex if the new chips are in fact biined for the show, something tells me they will be
3700 is rocking on a x79 .........
If I remeber right 4400 and change is the world record,
I am thinking that will be broken at CompuTex if the new chips are in fact biined for the show, something tells me they will be
Right now there is little issue with new memory kits. Even though clocks are still high then as you may noticed 2666 are already CL12 and are based on a worse chips. If anyone is going to break world records then it will be on early Hynix MFR so probably cherry picked kits that some already stock for a while. Good were kits from middle of the last year. All in last half year are worse for max clocks. The same is happening with Samsungs. Higher density but lower max clock.
I had MSI X79A-GD45 Plus and I could make it run stable at 2666 with max clock up to 2750. That was with Hynix and Samsung kits but I couldn't make it run with PSC above ~2333 even though these kits were running up to 2800 on IB.
Actually on IB-E when you set higher clock then performance isn't raising. For some reason max possible bandwidth will be about 80GB/s while in theory should be 90GB/s+.
I was able to reach about ~75GB/s with Samsung kits @2800 CL10 ( tested in AIDA64 ).
I think I still have the highest result with memory on air on hwbot. My CPUs had too weak memory controllers and cooling memory with anything better wasn't helping at all. I almost burned one of my kits running it on dice and I barely made 10MHz better clock ( was some moisture in slot ).
Generally high memory clock is good only for max clock validation and nothing else, still some fun to make it
I had MSI X79A-GD45 Plus and I could make it run stable at 2666 with max clock up to 2750. That was with Hynix and Samsung kits but I couldn't make it run with PSC above ~2333 even though these kits were running up to 2800 on IB.
Actually on IB-E when you set higher clock then performance isn't raising. For some reason max possible bandwidth will be about 80GB/s while in theory should be 90GB/s+.
I was able to reach about ~75GB/s with Samsung kits @2800 CL10 ( tested in AIDA64 ).
I think I still have the highest result with memory on air on hwbot. My CPUs had too weak memory controllers and cooling memory with anything better wasn't helping at all. I almost burned one of my kits running it on dice and I barely made 10MHz better clock ( was some moisture in slot ).
Generally high memory clock is good only for max clock validation and nothing else, still some fun to make it
OP
- Joined
- Apr 12, 2014
- Thread Starter
- #33
No doubt about it, its all about the fun in getting there ...
even if it just to get a screen shot
got lucky and scored some cas 9 2400 and CAS 10 2666 PCS stix before they disapeared
Nice results on that MSI board, I have been beating my head against the wall trying to get to 2133 CAS 8 for my gamer using Team Extreem 2400 Cas 9 stixs , do you know whos chips are on these ?
Best I have been able to get is 2004 CAS 8
really have to loosen them up to get any higher
Thought it was the board, guess it is the operator .... XD
even if it just to get a screen shot
got lucky and scored some cas 9 2400 and CAS 10 2666 PCS stix before they disapeared
Nice results on that MSI board, I have been beating my head against the wall trying to get to 2133 CAS 8 for my gamer using Team Extreem 2400 Cas 9 stixs , do you know whos chips are on these ?
Best I have been able to get is 2004 CAS 8
really have to loosen them up to get any higher
Thought it was the board, guess it is the operator .... XD
- Joined
- Jul 29, 2011
- Location
- Buenos Aires, Argentina
No doubt about it, its all about the fun in getting there ...
even if it just to get a screen shot
got lucky and scored some cas 9 2400 and CAS 10 2666 PCS stix before they disapeared
Nice results on that MSI board, I have been beating my head against the wall trying to get to 2133 CAS 8 for my gamer using Team Extreem 2400 Cas 9 stixs , do you know whos chips are on these ?
Best I have been able to get is 2004 CAS 8
really have to loosen them up to get any higher
Thought it was the board, guess it is the operator .... XD
Aren't those hynix or samsung based?
OP
- Joined
- Apr 12, 2014
- Thread Starter
- #35
Not sure, I will have to do some research, never really looked in to it as they are only for my gameing rig.
Team Xtreem, G.Skill TridentX and Corsair Dominator Plat. 2400 CL9 2x4GB or 4x4GB kits were only on Samsung but mixed IC. The same as 2600/2666 CL10. About the same IC series is in all of them. Some better some worse. Earlier kits were mainly on HYK0. Newer sometimes other IC.
Generally most HCH9 or HCK0 can't make 2400 CL9 ~1.65V. They are running on 2400 10-11-11 or 2600 10-12-12 but CL9 is much harder.
To make 2400 or higher clock on MSI board you have to use 125bclk+ strap. At 100MHz most kits won't even boot. I was using 166+bclk for 2666+ clock.
Generally most HCH9 or HCK0 can't make 2400 CL9 ~1.65V. They are running on 2400 10-11-11 or 2600 10-12-12 but CL9 is much harder.
To make 2400 or higher clock on MSI board you have to use 125bclk+ strap. At 100MHz most kits won't even boot. I was using 166+bclk for 2666+ clock.
OP
- Joined
- Apr 12, 2014
- Thread Starter
- #37
- Joined
- Oct 11, 2002
- Location
- The Empire State
Hmmmm...I might see an upgrade in the near future I'm liking those Giga boards!
I'm liking those Giga boards!Similar threads
