This quite often is the DRAM manufacturers taking lower binned chips and pushing them with additional voltage to run at a higher 1600 freq
In other words....exactly what they do with 2400, 2133, 1866....the higher binned chips sell for more money and are priced higher, the user gets to decide if they wanna pay more for faster speeds, lower CAS or better voltage ratings. If the cost is the same for lower voltage, buy it ..... if the cost is the same for lower CAS, buy it..... if the cost is the same for 2133 CAS 9 1.65 volt as 1600 CAS 9 1.5v buy the danged 2133. But if you are budget limited and are getting the 1.65 volt 1600 for a great price, it's strictly a budget decision .... what's more important to the buyer, the money they put back in their pocket versus the real world (as opposed to imagined) impact of the extra voltage ? That's an individual decision.
And that's kinda backwards .... lets remember that when DDR3 broke, as the old post from above shows, 1.65v was by far the most common voltage for XMP at 1600 back then.....so they are not "pushing voltages up", but bringing them down as improved yields since years ago allows them to. Those that can run better than the original design point are rated for lower voltages.
It's pretty much hard to find modules that are rated at 1.65 these days ... but when ya do, it's not always because you didn't get one of the top binned modules.
I couldn't see spending $90 for these 1.65 volt CAS 9 DDR3-1600s .... but at $60, for a person on a tight budget .... in the proverbial heartbeat.
http://www.newegg.com/Product/Product.aspx?Item=N82E16820104344
It was not long ago that 1.65v was the only option for 1600 CAS 7, then we saw 1.55 and now 1.50 ...... again while time gives us better stuff, the passage of time alone doesn't make the older stuff something we **need** to run out and upgrade. If it was fine when we bought it, it's still fine today.
2 years ago, most of the 1600 line would have been 1.65. Now today, all things being equal, it would be as silly to buy 1.65 v 1600 at same price as it would to just as silly to buy 1600 1.5v when 2133 1.6 v was same price. But if someone is selling an old set that is 1.65, or the rare set that comes off the line needing a voltage boost and is offering it at a big discount, if it was fine 2 years ago, it should still be fine today.
Let's actually look at Intel's compatibility list and see how many there are at each voltage
1.25 volts = 2
1.5 volts = 155
1.6 volts = 7
1.65 volts = 78
That's 65 %..... So in 2 years time, the number of 1.5 volt modules went from constituting 31 % of the Intel's list and now it's doubled to 65% ... Like anything else, you pay more for better stuff. Like anything else, production lines today produce better yields of the higher performance stuff then they did when DDR3 1st came out. But how many of them can you actually find today ? Many of those are no longer in production because what came off that line in April 2012 has been superseded by what comes off today.
As yields have improved so much now that the 1600 set that comes off the line today rarely needs 1.65 volts. Yeay ! But that does not change the viability of what came off the line 2 years ago...... if you can save $30 on a tight budget build, why is the tradeoff for lower performance / lower quality any different for RAM than it is for any other PC component ?
The fact remains, the platform was designed to handle 1.65 volts and more. In 2012, everyone but the misinformed (by that I mean those yelling "Intel voids warranty at 1.65v) found 1.65 volts perfectly fine and it's a much more recent development that 1866 became more available at 1.5 volts. Now even 2133 is down to 1.6v and 1.55 for many models. So how did the passage of 2 years time change whether 1.65v modules were acceptable ? Certainly 1.5 is better, so is 1.35 and 1.25 .... but what is the real world impact of the 3 issues raised:
1. Warranty - 1.65 volts are still today 1/3 of Intel's certified compatibility lists, none have come off....just that so many newer upgraded' models have been added. Back in the Sandy Bridge era, 1.5 volt DDR 1600 was in the minority. Let's remember just as far back as 2012, 2/3 of the sticks on the list and on the market were 1.65 volt rated....and back in 2012, I saw no posts about having to run them up at 1.7 / 1.75 volts to reach their XMP profiles. I certainly have never had to break 1.65 volts on any module ever to run at 1600. Only time I have ever needed to break 1.65 (most times BIOSs would if left on Auto) was with 2400 modules and then only when CPU was substantially overclocked. So zero real world impact here.
2. Overhead - What is the number of regular users here, I'm talking about the peeps who come in and ask questions, not the small % of diehards, go beyond XMP and manually set their speeds and timings ? Only these people will be affected by the loss of headroom. Not a whole lot of those .... and those are the people who invest in DDR2400, invest in better cooling and have no issue paying a bit more for better stuff. Those looking to save $30 on "clearance" deals are not in that group either. So no real world impact here as those that are overclocking RAM are not looking at DDR3 1600....why would they buy 1600 and OC it when they could buy 2133 at same price ?
3. Heat - yes we all know heat is bad .... but real word impacts ? There is simply no way to support this argument with regard to RAM at 1.65 volts. Heat, in this instance, is not the proverbial red herring but the red whale. The Hyper 212 gets recommended here 100s of times a day but yet it delivers 7 - 10C less cooling than a Noctua, Phanteks or Silver Arrow. So why not the "Heat is death message" there ? It's simply a budget decision. Ya get what ya pay for.
Again, let's go to real world numbers. (CPU Multiplier / Cache Multiplier / DRAM Speed / DRAM Voltage = AVG of 4 Max Core temps (HWiNFO64) after 2 hour stress test)....Note ambient temps were not consistent, so appropriate adjustments were made so apples and apples comparisons could be made
42/42/1600/1.65 = 58.75 (warm night)
42/42/2400/1.70 = 59.00
43/43/1600/1.65 = 57.00
43/43/2400/1.70 = 57.00
44/44/1600/1.65 = 58.25
44/44/2400/1.70 = 59.00
45/45/1600/1.65 = 64.25
45/45/2400/1.70 = 64.25
46/46/1600/1.65 = 71.75
46/46/2400/1.70 = 71.50
* 1.65 generated errors at 46/46/2400 so I upped it to 1.7 ..... never did go back and try 1.69 or 1.68 etc as saw no advantage to putting in the T&E....prolly cuda got 1.65 at 42 and 43 multipliers.
Each of the test pairs were done on different days and therefore subject to different ambient temps. test at each multiplier were both done over a 4.5 hour period (two 2 hour tests, 30 minutes apart) No adjustment was made between days, however where there was a difference between ambient temps which happened on 2 nights (which would be lower in the 2400 tests as they were done later at night), the difference in ambient was added to the 2400 results. For example, the 2400 run at 46 mutiplier had an ambient 3C lower so 3C was added to 68.5 to get 71.5
So yes , it can be said that excess heat is to be avoided but it can't be said that there is an actual significant impact from 1.65v ....even ignoring the 46 drop as an aberration and calling it a tie, that's 1C total over 5 tests or 0.2C average . While an increase of 10C (82C), or some would say 20C (92C) possibly being a potential "big killer" of CPUs, the implication that that 1.65 volts will have such an effect is unsupported and inappropriate ..... If 1.65 volts has no scary heat impact at 2133, 2400 and up .... why is magically "scary" at 1600 and 1866 ? The tests showed 1.7 volts having an average impact of 0.2C Your Tridents are rated at 1.65V; why should anyone using 1600 @ 1.65 be any more afraid than you are of your 1.65v modules ?
So no, 0.2C will not have any real world impact and by no means will it kill anything. As for heating components around the RAM ? No, too small to even be measured. We are talking a 0.3% increase in heat .... that's a 0.01 watt increase in power consumption on a 5 watt RAM stick.....and a 0.4 watt increase in CPU Power consumption....increase ya fan speed by 2 rpm and you have more than compensated.
Again,
all things being equal, the lower voltage stuff is always going to be the better choice..... but it is a marginally better choice and getting 1.65 is certainly not going to have any "real world" detriment to your build.
DDR3 1600 CAS 9 @ 1.5 versus DDR3 1600 CAS 9 @ 1.65 (same price) .... take the 1.50
DDR3 1600 CAS 9 @ 1.5 versus DDR3 1600 CAS 9 @ 1.65 (1/3 off) .... budget decision, nothing scary about either choice
DDR3 1600 CAS 9 @ 1.5 versus DDR3 1600 CAS 7 @ 1.65 .... take the CAS 7 if price premium satisfactory
DDR3 1600 CAS 9 @ 1.5 versus DDR3 2133 CAS 9 @ 1.65 .... take the 2133 if price premium satisfactory
No, I wouldn't buy 16 GB of 1.65 volt 1600s .... I'd do what "most knowledgeable builders and techs" would do (in the real world) and ignore the entire 1600 speed 1.5v category entirely and buy 16GB of the 2133 1.65v modules for $150, the exact same price as the cheapest 1.5v 1600s. I'd also relish that they had Hynix modules so if I ever did wanna OC them, I know I could take them up to 1.9 and higher w/o a care in the world.
However, if someone was on a limited budget and finds both the 1600 1.5V and the 2133 1.65v too expensive for their budget, and could potentially save a few bucks by getting 1.65 1600s (if there's no savings why consider it ?), I certainly could not make a legitimate argument that 1.5 v 1600, 1.65v 2133 or 1.65v Tridents at 2666/2400 is just fine and somehow, by no means that has been supported or quantified, 1.65v 1600 was somehow going to be disastrous.