What would you buy? Comparing RAM

X

=XoX=

n00b
Joined
Nov 17, 2006
Messages
40
I'm looking at RAM right now and was wondeing what you'd buy. I'm considering between the two:

Corsair DDR3 8GB (2x 4GB) PC-10600 (1333MHz) XMS Heatspreader @ 9-9-9-24
(CMX8GX3M2A1333C9) AU$220

-or-

Corsair DDR3 6GB (3x 2GB) PC-12800 (1600 MHz XMS Heatspreader @ 7-8-7-20
(CMX6GX3M3C1600C7) AU$200

Basically the 2X4GB leaves my 6 slots for future upgrade, maxing at 24Gig, not that I can see a build in the next two years needing that much.

The 3X2GB would max at 12 Gig total, and is faster.

What would you buy? Anything else on the market I should look at?
 
=XoX= said:
I'm looking at RAM right now and was wondeing what you'd buy. I'm considering between the two:

Corsair DDR3 8GB (2x 4GB) PC-10600 (1333MHz) XMS Heatspreader @ 9-9-9-24
(CMX8GX3M2A1333C9) AU$220

-or-

Corsair DDR3 6GB (3x 2GB) PC-12800 (1600 MHz XMS Heatspreader @ 7-8-7-20
(CMX6GX3M3C1600C7) AU$200

Basically the 2X4GB leaves my 6 slots for future upgrade, maxing at 24Gig, not that I can see a build in the next two years needing that much.

The 3X2GB would max at 12 Gig total, and is faster.

What would you buy? Anything else on the market I should look at?
Click to expand...

There is no tangible benefit with higher memory speeds or tighter latency timings with current Intel i7 platforms. In fact, if you let the motherboard set the memory speeds, both DDR3-1333 and DDR3-1600 RAM will default to DDR3-1066 speed (Intel's official maximum for the Socket LGA 1366 CPUs). The presence of six DIMM slots implies that you have an X58 motherboard. (Though there is at least one LGA 1156 motherboard with six DIMM slots - but the LGA 1156 CPU's IMC is limited to a maximum of eight ranks total, limiting the maximum memory capacity of that platform to 16GB total, and that you cannot use six double-ranked DIMMs on that motherboard at the same time, and no more than two of the DIMMs may be double-ranked if all six DIMM slots were to be used.)

Also, triple-channel memory controller operation does not improve real-world performance by anywhere near as much as the cost of an additional DIMM would have justified.

Therefore, it all boils down to the DIMM voltage that the memory is rated for. Unfortunately, that particular Corsair CMX6GX3M3C1600C7 kit needs 1.65V to run at its advertised speed and timings, which might actually hinder the overclocking of your system's CPU due to the memory controller hitting Intel's recommended maximum. (In fact, Corsair only recently introduced JEDEC DDR3-1600 modules that need only 1.5V for rated-speed operation - but those modules are currently hard to find.) All recent Corsair DDR3-1333 modules use 1.5V.
 
Last edited:
Thanks E4g1e. Your comments are usefull.

BTW, the mobo is a x58 classified.
 
=XoX= said:
Thanks E4g1e. Your comments are usefull.

BTW, the mobo is a x58 classified.
Click to expand...

In that case, based on your first post:

Corsair DDR3 6GB (3x 2GB) PC-12800 (1600 MHz XMS Heatspreader @ 7-8-7-20
(CMX6GX3M3C1600C7) AU$200
 
Mike Clements said:
In that case, based on your first post:

Corsair DDR3 6GB (3x 2GB) PC-12800 (1600 MHz XMS Heatspreader @ 7-8-7-20
(CMX6GX3M3C1600C7) AU$200
Click to expand...

I cannot disagree more about this, but I cannot recommend overclocking the CPU much, if at all, with high-voltage RAM. I have already proved in my testing that such high DIMM voltages in a system that has the IMC on the CPU die already heats up the CPU temperatures more than lower-voltage RAM does - at the very same CPU and memory speed. And the only purpose of the XMP profiles is to run memory at speeds above Intel's official maximum while keeping the CPU speed at stock. But if the CPU is overclocked even modestly, all bets are off.

On the other hand, filling up all six DIMM slots with double-ranked memory will also hinder overclocking: The loading on the IMC will be increased under such a condition.

Third, using a dual-channel kit on a system with a triple-channel memory controller isn't optimal although the performance loss is much less than what the memory throughput loss implies (less than 5 percent in most applications).
 
Last edited:
"High-voltage RAM" is a relative term. 1.65v is in fact within what Intel rates for use with their Core i* IMCs. So, I don't think 1.65v qualifies as "high" voltage.
 
Mike Clements said:
"High-voltage RAM" is a relative term. 1.65v is in fact within what Intel rates for use with their Core i* IMCs. So, I don't think 1.65v qualifies as "high" voltage.
Click to expand...

Actually, I consider anything over 1.5V to be "high voltage". You just have different standards than I do.

Plus, as I have already mentioned 1.65V is at the upper limit of Intel's recommended voltage range for its Core i* IMCs. (And my claim is based on the actual DIMM voltage, not necessarily the VDIMM setting which is almost always higher than what the actual DIMM voltage reads due to Vdroop). And if a memory kit requires 1.65V just to even run at stock speeds (say, Intel's official CPU core speed and the maximum officially supported speed of DDR3-1066), you'd end up with virtually no headroom for overclocking.
 
Last edited:
E4g1e said:
Actually, I consider anything over 1.5V to be "high voltage". You just have different standards than I do.

Plus, as I have already mentioned 1.65V is at the upper limit of Intel's recommended voltage range for its Core i* IMCs. (And my claim is based on the actual DIMM voltage, not necessarily the VDIMM setting which is almost always higher than what the actual DIMM voltage reads due to Vdroop).
Click to expand...

It's not my standard, it's Intels standard. ;) And we all know how conservative they are. If it was dangerous, they absolutely would not approve it for use in XMP profiles.
 
see my sig...that "high voltage" ram runs like a scalded cat , with my 920 running at 4.2 ghz....runs at 2005 mhz 7,8,7,20 C1 - 21511 mbs read , 17421 mbs write,39.3 ns latency in Everest..
 
Mike Clements said:
It's not my standard, it's Intels standard. ;) And we all know how conservative they are. If it was dangerous, they absolutely would not approve it for use in XMP profiles.
Click to expand...

Fair enough. I must have had a bum CPU which runs quite a bit hotter than the norm for that CPU even at stock speeds (and I already have a fairly high-end air cooler).
 
Mike Clements said:
That is odd :confused:
Click to expand...

With my current setup, it is already hitting 70°C at full load at its stock 2.67 GHz speed.

I am suspecting that the heatspreader on my system's particular CPU isn't making full contact with the CPU cooler. And yes, the cooler did reduce temperatures over the stock Intel HSF, but not by enough to justify the price that I paid for it. After all, I expected a 15° to 20°C reduction - but I could not get more than a 10°C reduction compared to the stock cooler.

As such, I am considering a replacement CPU cooler.
 
E4g1e said:
With my current setup, it is already hitting 70°C at full load at its stock 2.67 GHz speed.

I am suspecting that the heatspreader on my system's particular CPU isn't making full contact with the CPU cooler. And yes, the cooler did reduce temperatures over the stock Intel HSF, but not by enough to justify the price that I paid for it. After all, I expected a 15° to 20°C reduction - but I could not get more than a 10°C reduction compared to the stock cooler.

As such, I am considering a replacement CPU cooler.
Click to expand...

Okay, so I tried it with another CPU, and it still runs much hotter than normal. Then, I looked at the thermal compund on the CPU, and discovered that the compound spead very unevenly, with some parts of the CPU heat spreader not getting enough compound. My idle CPU temps hit nearly 50°C even at stock speed and in a cold room. Then I tried my hand at lapping the HSF, only to discover that the HSF is now no longer able to make proper contact with the CPU even with the mount fully tightened.

Hence, as I suspected in the first place, I am thinking about getting a new HSF.
 
You must log in or register to reply here.
Back
Top