Six Low-Voltage Dual-Channel 8 GB Memory Kits, Overclocked

[Crashman]

We've abided by Intel's 1.55 V recommendation for two architectures and two die shrinks, yet most performance memory manufacturers ignore it. Recent problems with one of our builds raised the question, how far can we push RAM without killing CPUs?

Six Low-Voltage Dual-Channel 8 GB Memory Kits, Overclocked : Read more

 

[excella1221]

Nice article.
A bit surprised that Corsair Vengeance didn't make an appearance though.

Been using G.Skill RAMs since I can remember, and they've never failed me!

 

[DarkSable]

Very interesting read, thanks Tom's! Guess we've been bothering you enough about, erm, "half-height" overclocker's ram enough for you to want to do an article on it. Shame that the mystery modules chose not to participate... perhaps I'll send you my 8 gigs. Want to swap? Ship me some ram to use in the mean-time and I'll ship you my sticks.

 

[rmpumper]

In other words, RAM speed is irrelevant if you are not using an APU.

 

[Haserath]

No power consumption numbers? :/

 

[cobra5000]

Good testing but at the start of the article, much emphasis was placed on the degradation of the cpu due to high memory voltage and then it just turns into a memory speed test. What about the cpu degradation?

 

[envy14tpe]

Good article. Still left wondering why anyone would choose low voltage RAM when costs do not outweigh the benefits.

 

[slomo4sho]

It is only worthwhile to pickup low voltage ram if you can obtain it at the same price as 1.5v ram. Then again, the Ballistix is around the same price point as other similar timing ram.

 

[allan_hm]

Pointless analysis for too little performance AND price difference...
Would be a bit less irrelevant if more brands were tested.
To be honest I was more interested on that "dirty little secret" details than the test itself.
Also, an article named "who is who in the RAM marked" would be awesome, just like that one about Power Supplies...

 

[Crashman]

[citation][nom]excella1221[/nom]Nice article.A bit surprised that Corsair Vengeance didn't make an appearance though.Been using G.Skill RAMs since I can remember, and they've never failed me![/citation]Here's from the article:

We invited every major manufacturer (including a crowd-favorite ODM) to this round-up, and a few (including that ODM) chose not to participate.

Corsair said they had some new modules coming out and didn't want to focus on older models.[citation][nom]cobra5000[/nom]Good testing but at the start of the article, much emphasis was placed on the degradation of the cpu due to high memory voltage and then it just turns into a memory speed test. What about the cpu degradation?[/citation]What about it? It's never been seen at 1.50V.

 

[Crashman]

[citation][nom]allan_hm[/nom]Pointless analysis for too little performance AND price difference...Would be a bit less irrelevant if more brands were tested.To be honest I was more interested on that "dirty little secret" details than the test itself.Also, an article named "who is who in the RAM marked" would be awesome, just like that one about Power Supplies...[/citation]Article intent was to find modules within Intel's 1.55V limit that pulled top numbers in frequency and/or latency. Some were found, article is a success.

 

[Onus]

Comparing data within each chart was not as interesting to me as comparing data between charts, which once again showed that although differences exist, they are tiny.
Put a lot more thought into your platform, CPU and graphics card; for RAM, get the cheapest DDR3-1600 CAS9 and call it a day.

 

[JackNaylorPE]

Anybody read Intel's web pages in this regard ?

1.5 volts is for the JDEC profiles .... but most RAM speeds are advertised and labeled based upon their XMP profiles......most XMP profiles for i5 and i7 RAM is 1.65. In fact, over 2/3 of the RAM on Intel's XMP compatible list are over 1.50 volts.

I have seen many posts saying "Well that's a MoBo manufacturer thing". No....Intel sets the XMP profile. I agree, Intel's JDEC specifications stipulate 1.5 volts .... for JDEC profiles ..... but XMP "allows compatible DDR3 memory to perform beyond standard specifications" and is "predefined and tested" by Intel

http://www.intel.com/content/www/us/en/gaming/gaming-computers/intel-extreme-memory-profile-xmp.html

Intel® Extreme Memory Profile (Intel® XMP) allows you to overclock compatible DDR3 memory to perform beyond standard specifications. It’s designed to enhance the gaming features built into Intel® technology–based PCs. If you like to overclock and squeeze as much performance from your PC as possible, then memory based on Intel XMP gives you that extra edge you need to dominate—without breaking a sweat.

Predefined and tested Intel XMP profiles can be loaded via BIOS or a specific tuning application through a computer’s operating system. Often the easiest way to load Intel XMP profiles is using a tuning utility, which may be available depending on the particular board manufacturer. To learn whether a tuning utility is available on your system, you should contact the board manufacturer.

Again, most listed compatible i5 / i7 RAM is 1.65 .... at least according to Intel's compatibility lists

http://www.intel.com/content/www/us/en/gaming/gaming-computers/core-i5-processor-memory-datasheet.html
http://www.intel.com/content/www/us/en/gaming/gaming-computers/core-i7-memory-suppliers-datasheet.html

Intel's approved i7 Compatibilty List (April 2012) includes:

34 1.65 volt modules
07 1.60 volt modules
19 1.50 volt modules
01 1.25 volt modules

 

[jee_are]

Maybe I'm being anal but, Westmere (Nehalem) was the die shrink to 32 nm (tick). Sandy Bridge was the following change in microarchitecture (tock).

 

[DjEaZy]

... i got me the crucial balistix 3~4 months ago... a 16Gb of them... and they OC good too... i have them running @ 1926Mhz 8-8-8-26

 

[Evan Richardson]

[citation][nom]Haserath[/nom]No power consumption numbers?[/citation]

Perhaps because memory doesn't use that much power to begin with? Be a cold day in hell when you have to buy a 1kW PSU just because you add another stick of ram.

 

[happyballz]

Where are the cost for each kit? Did I miss it? Flipped through pages and could not find anything on the cost.

 

[gondor]

[citation][nom]JackNaylorPE[/nom]Anybody read Intel's web pages in this regard ?1.5 volts is for the JDEC profiles ...[/citation]

It's "JEDEC", the first 'e' is not silent/missing.

http://www.jedec.org

 

[m32]

[citation][nom]Haserath[/nom]No power consumption numbers?[/citation]


3-6 watts isn't going to kill anyone.

 

[pazygozo]

Exactly the article i wanted to read, i was asking myself what clocks could the low voltage modules achieve at high voltage.

God, i wanted to buy the ballistix tactical in february, i bought the ballistix sport instead because i found 32GB for 158$ but they are taller and 1.5v rated, glad to know it doesn't make much of a difference in games past 1600Mhz.

Great work!

 

[Crashman]

[citation][nom]happyballz[/nom]Where are the cost for each kit? Did I miss it? Flipped through pages and could not find anything on the cost.[/citation]
On the day they were tested, Adata was $72, Crucial $70, Geil $75, G.Skill $68, Kingston $70. That changes pretty quickly though.

 

[slicedtoad]

I could be wrong but it's been my understanding that the only time you see any practical difference between kits above 1600mhz (with intel 2nd and 3rd gen i5s) is if your cpu is running at particularly high oc. The 4GHz in this article is a bit disappointing since I thought the idea was to remove all possible bottlenecks.

 

[daglesj]

Wow is it 2004/the DDR era again?
.
No.

Just buy on price and what looks coolest in your motherboard. Stop wasting your life doing RAM reviews.

 

[Aegean BM]

I really like how you take the time to do a gazillion tests that I'd never do. I also like that you don't assume much from the vendors' self-reported numbers. Again you do the tests.

Where I'm not aligned is with your conclusions. For example, averaging stock number with overclocked number to pick a winner? That's neither fish nor fowl. Your answer is guaranteed to not totally satisfy both types of consumers. Average can be a weird beast. The average American has 1.98 legs and 2.1 kids.

It's hard to read comments like "A performance spread of 3% difference isn't huge, but it is noteworthy." and "We understand that most users won't care about price differences this small, but that's not an excuse to suppress this information." 3% and $2 are not worth mentioning. NO ONE cares about it to that level. Don't get off in the micro weeds.

OK, sorry, we know some people do care. That's fine. But why wouldn't you write for "most users"? A good rule of thumb is most users only care about noticeable differences. And to be absolutely clear about "noticeable", it means to notice the performance or quality without a number attached. A person without a stopwatch can't tell you when their machine is 3% faster. And who brags about buying a $70 part for $2 cheaper? It seems like too much energy was put into splitting hairs to come up with winner(s) when there clearly wasn't.

Since most users do like the numbers, and subjective performance would be more controversial (the cure would be worse than the disease), I suggest the rule of thumb be 10% as the minimum for declaring a difference notable.

If toms take my suggestions, its hardware reviews will be 2.19% more enjoyable than other hardware sites.

 

[Aegean BM]

[citation][nom]daglesj[/nom]Wow is it 2004/the DDR era again?.No.Just buy on price and what looks coolest in your motherboard. Stop wasting your life doing RAM reviews.[/citation]

Yes, I agree that the memory tests didn't show a clear winner. I'm still glad Thomas did the tests even if they confirmed what I already thought. I admit I was curious if low voltage RAM was somehow different.

However, just because I wasn't impressed with a clear winner in this batch of hopefuls doesn't mean that there aren't some losers I could buy. I don't think ALL RAM is the same.