Sign in with
Sign up | Sign in
Your question
Solved

Am I gimping my 4770k gaming system with 1333mhz ram speed?

Last response: in CPUs
Share
September 14, 2013 1:40:31 PM

Too lazy to take out the mushkin blackline 8 gigs, and too lazy to buy new ram (I took this ram from my old 2600k system). I have to take out the my nh-d14 heatsink to change the ram. This muskin ram is rated to do 1600mhz @ 1.65volts, it is meant for overclocking. I do not want to exceed intels specs of 1.5 volts (personal choice). This ram can easily do near 2000mhz though. My question is, am I gimping my game performance by having the ram speed at 1333mhz? I really do not notice anything bad in games atm, in fact... the performance in everything is amazing.

SPECS:
4770k
asus maximus 6 hero
8 gigs mushkin blackline ddr31600 @ 1333mhz.
gtx 680 sli

a b 4 Gaming
a c 169 à CPUs
September 14, 2013 1:44:25 PM

Yes.

There's about a 5% to 10% gap between DDR3-1333 and DDR3-1600. There's a further 5% to 10% gap between DDR3-1600 and DDR3-2133.

This gap is amplified when using SLI or CrossfireX (all of that PCIe bandwidth has to terminate somewhere).
m
0
l
a b 4 Gaming
a b à CPUs
September 14, 2013 1:53:19 PM

Pinhedd said:
Yes.

There's about a 5% to 10% gap between DDR3-1333 and DDR3-1600. There's a further 5% to 10% gap between DDR3-1600 and DDR3-2133.

This gap is amplified when using SLI or CrossfireX (all of that PCIe bandwidth has to terminate somewhere).


Where did you get that? Its been shown numerous times that anything over 1600, and even 1333 is almost entirely irrelevant unless using integrated graphics. To imply he's losing 20% performance due to using 1333 memory is down right dishonest
m
0
l
a b à CPUs
September 14, 2013 2:01:35 PM

unksol said:
Pinhedd said:
Yes.

There's about a 5% to 10% gap between DDR3-1333 and DDR3-1600. There's a further 5% to 10% gap between DDR3-1600 and DDR3-2133.

This gap is amplified when using SLI or CrossfireX (all of that PCIe bandwidth has to terminate somewhere).


Where did you get that? Its been shown numerous times that anything over 1600, and even 1333 is almost entirely irrelevant unless using integrated graphics. To imply he's losing 20% performance due to using 1333 memory is down right dishonest


Might you have a link to the study on integrated graphics? I knew memory speed wasn't very important but I never considered what effect it might have on integrated solutions.
m
0
l
a b 4 Gaming
a b à CPUs
September 14, 2013 2:18:13 PM

What ever you DO NOT go to 1.65V on your memory, it's hard on the controller. If it is working fine do not worry about it, go with it. The memory bus is faster than the PCIe bus right? It should not affect gaming at all. In heavy video work it would.
m
0
l
September 14, 2013 2:20:34 PM

endeavour37a said:
What ever you DO NOT go to 1.65V on your memory, it's hard on the controller. If it is working fine do not worry about it, go with it. The memory bus is faster than the PCIe bus right? It should not affect gaming at all. In heavy video work it would.


Ya, I won't. It is the defaul/auto setting. THe maximus 6 hero set it to 1333mhz, and I am sure the volts is NOT at 1.65 (when in auto/default mode).
m
0
l
a b 4 Gaming
a c 169 à CPUs
September 14, 2013 2:35:24 PM

unksol said:
Pinhedd said:
Yes.

There's about a 5% to 10% gap between DDR3-1333 and DDR3-1600. There's a further 5% to 10% gap between DDR3-1600 and DDR3-2133.

This gap is amplified when using SLI or CrossfireX (all of that PCIe bandwidth has to terminate somewhere).


Where did you get that? Its been shown numerous times that anything over 1600, and even 1333 is almost entirely irrelevant unless using integrated graphics. To imply he's losing 20% performance due to using 1333 memory is down right dishonest


Entirely irrelevant for playing mediocre console ports perhaps, but some heavily tweaked PC games can show noticeable performance improvements up to around 15% that I mentioned. This is especially true where software physics are in use and is very evident in 3DMark tests.

http://www.tomshardware.com/reviews/memory-bandwidth-la...

endeavour37a said:
What ever you DO NOT go to 1.65V on your memory, it's hard on the controller. If it is working fine do not worry about it, go with it. The memory bus is faster than the PCIe bus right? It should not affect gaming at all. In heavy video work it would.


It's really not hard on the controller. I've never heard of anyone damaging a CPU from running 1.65 volt memory. I'm not sure about IB/HW as the relevant value is not published in the specification and the IB-E datasheets aren't available yet but SB-E allows for up to 1.8 volts DRAM supply before the safety margin is exceeded. The package will put out a bit more heat for certain, but its lifespan will not be significantly affected.

I've been running it for years on my PC with no issues at all.
m
0
l
a b 4 Gaming
a b à CPUs
September 14, 2013 3:52:18 PM


endeavour37a said:
What ever you DO NOT go to 1.65V on your memory, it's hard on the controller. If it is working fine do not worry about it, go with it. The memory bus is faster than the PCIe bus right? It should not affect gaming at all. In heavy video work it would.

You say:
"
It's really not hard on the controller. I've never heard of anyone damaging a CPU from running 1.65 volt memory. I'm not sure about IB/HW as the relevant value is not published in the specification and the IB-E datasheets aren't available yet but SB-E allows for up to 1.8 volts DRAM supply before the safety margin is exceeded. The package will put out a bit more heat for certain, but its lifespan will not be significantly affected.

I've been running it for years on my PC with no issues at all.
"

I agree with you on faster memory gives an overall performance increase to any system on everything you do. But the OP said everything was running great and just wondering if he needed to change memory, I say no if he is happy. If in the future he encounters problems then he can swap it out for faster stuff, perhaps 1866 CL8.

On voltage, again I don't argue with hard data if it is available or personal history you have presented. Yet I have read memory controllers on IB CPUs have been killed by over voltage, that leads me to assume lower voltages help keep it healthy. The new LP sticks are not about heat, I think they are about the controller, could be all wrong.

If I had a 4770 I would stick with 1.5V at the expense of speed to air on the safe side over the long term, just me. If you say it is fine to go 1.65 then it's fine with me also but I would not do it on a Haswell. Sure the OP can look into it and figure out what he wants to do, we have given him the pros and cons I guess...
m
0
l
a b 4 Gaming
a c 169 à CPUs
September 14, 2013 4:40:30 PM

endeavour37a said:

endeavour37a said:
What ever you DO NOT go to 1.65V on your memory, it's hard on the controller. If it is working fine do not worry about it, go with it. The memory bus is faster than the PCIe bus right? It should not affect gaming at all. In heavy video work it would.

You say:
"
It's really not hard on the controller. I've never heard of anyone damaging a CPU from running 1.65 volt memory. I'm not sure about IB/HW as the relevant value is not published in the specification and the IB-E datasheets aren't available yet but SB-E allows for up to 1.8 volts DRAM supply before the safety margin is exceeded. The package will put out a bit more heat for certain, but its lifespan will not be significantly affected.

I've been running it for years on my PC with no issues at all.
"

I agree with you on faster memory gives an overall performance increase to any system on everything you do. But the OP said everything was running great and just wondering if he needed to change memory, I say no if he is happy. If in the future he encounters problems then he can swap it out for faster stuff, perhaps 1866 CL8.

On voltage, again I don't argue with hard data if it is available or personal history you have presented. Yet I have read memory controllers on IB CPUs have been killed by over voltage, that leads me to assume lower voltages help keep it healthy. The new LP sticks are not about heat, I think they are about the controller, could be all wrong.

If I had a 4770 I would stick with 1.5V at the expense of speed to air on the safe side over the long term, just me. If you say it is fine to go 1.65 then it's fine with me also but I would not do it on a Haswell. Sure the OP can look into it and figure out what he wants to do, we have given him the pros and cons I guess...


There were some reports several years back about Sandybridge CPUs burning out when used with 1.65 volt memory. These were traced back to faulty manufacturing at Foxconn with some motherboards being recalled (not related to the sweeping Intel PCH recall). At least one Intel service representative claimed that any CPU failure encountered while running 1.65 volt memory with an XMP profile would be replaced under standard warranty as long as there were no burn marks on the package.

XMP profiles are tested and certified by Intel before the manufacturer is allowed to use the trademark. If a 1.65 volt profile posed a risk of damage, it wouldn't be certified. Intel even lists tested and certified memory in a datasheet to prevent vendors from abusing the trademark

http://www.intel.com/content/www/us/en/gaming/gaming-co...
m
0
l
a b 4 Gaming
a b à CPUs
September 14, 2013 6:07:38 PM

"
There were some reports several years back about Sandybridge CPUs burning out when used with 1.65 volt memory. These were traced back to faulty manufacturing at Foxconn with some motherboards being recalled (not related to the sweeping Intel PCH recall). At least one Intel service representative claimed that any CPU failure encountered while running 1.65 volt memory with an XMP profile would be replaced under standard warranty as long as there were no burn marks on the package.

XMP profiles are tested and certified by Intel before the manufacturer is allowed to use the trademark. If a 1.65 volt profile posed a risk of damage, it wouldn't be certified. Intel even lists tested and certified memory in a datasheet to prevent vendors from abusing the trademark
"

Point taken and thanks for expanding on this with good info, I humbly stand corrected. But many say 1.5v is preferred for 1866 and under, if going to 1.65v the CPU should also be OCed to balance out the voltage of the controller. Not being argumentative on this as spec sheets are hard facts, just talking about what I have read.
m
0
l

Best solution

a b 4 Gaming
a c 169 à CPUs
September 14, 2013 6:49:33 PM

endeavour37a said:
"
There were some reports several years back about Sandybridge CPUs burning out when used with 1.65 volt memory. These were traced back to faulty manufacturing at Foxconn with some motherboards being recalled (not related to the sweeping Intel PCH recall). At least one Intel service representative claimed that any CPU failure encountered while running 1.65 volt memory with an XMP profile would be replaced under standard warranty as long as there were no burn marks on the package.

XMP profiles are tested and certified by Intel before the manufacturer is allowed to use the trademark. If a 1.65 volt profile posed a risk of damage, it wouldn't be certified. Intel even lists tested and certified memory in a datasheet to prevent vendors from abusing the trademark
"

Point taken and thanks for expanding on this with good info, I humbly stand corrected. But many say 1.5v is preferred for 1866 and under, if going to 1.65v the CPU should also be OCed to balance out the voltage of the controller. Not being argumentative on this as spec sheets are hard facts, just talking about what I have read.


On that point you're correct. If the only difference between two profiles is the voltage, then 1.5 volts would obviously be preferred but they often cost a lot more. There are a number of modules from various vendors (most noticeably Kingston though) which which have multiple XMP profiles. Several of the XMP profiles will be below the marketed speed (eg, if a module is marketed at DDR3-1600, it may have JEDEC profiles for DDR3-800, DDR3-1066, and DDR3-1333, as well as XMP profiles for DDR3-1333 and DDR3-1600). The JEDEC profiles will of course be fully DDR3 standard compliant, but the XMP profiles may feature any combination of tighter timings, higher voltages, and higher frequencies.

For example, the Kingston HyperX DDR3-1600 4GiB SO-DIMMs in my laptop have the following profiles, all at 1.5 volts (it's a laptop).

DDR3-800E
DDR3-1066F
DDR3-1333H

There's also two XMP profiles which more or less correspond to the following JEDEC standards

DDR3-1333F
DDR3-1600H

Notice that there are two profiles for DDR3-1333 on there, H (9-9-9) and F (7-7-7). These all run at 1.5 volts, so it benefits us to always run the first XMP profile rather than the third JEDEC profile if the second XMP profile is not an option.

The same logic extends to many other DDR3-1600 kits, but on the desktop many vendors choose to tighten timings even further by increasing the supply voltage. For example, Mushkin Redline 4GiB DDR3-1600 modules can run at DDR3-1600H at 1.5 volts, or when XMP is enabled and running at 1.65 volts, at speeds which would correspond to DDR3-1600F which is the tightest timing bin that Hynix modules support at that frequency. Very high quality chips may be able to run DDR3-1600F at 1.5 volts, but these chips are typically hand picked for very expensive lineups such as Corsair Dominator Platinum, so be prepared to pay a hefty price for them.

With all that said, there are only a handful of scenarios where memory bandwidth becomes a necessary consideration. The biggest ones are compression, encoding, rendering, and simulation (physics, electrical, chemical, etc...). Multi-GPU systems also benefit from greater memory throughput, though not as much.
Share
a b 4 Gaming
a b à CPUs
September 15, 2013 6:17:26 AM

Thanks very much for the detailed information, must commend you on your knowledge about memory and I have learned from your post here today. Going to have to dig into this a bit deeper to fully understand what you are saying, but you have opened my eyes :) 
m
0
l
!